Semantic Structure in English

studies in functional and structural linguistics 73
Semantic
Structure
in English
Jim Feist
John Benjamins Publishing Company

Semantic Structure in English
Studies in Functional and Structural Linguistics (SFSL)
issn 1385-7916
Taking the broadest and most general definitions of the terms functional and structural,
this series aims to present linguistic and interdisciplinary research that relates language
structure – at any level of analysis from phonology to discourse – to broader functional
considerations, whether cognitive, communicative, pragmatic or sociocultural. Preference
will be given to studies that focus on data from actual discourse, whether speech, writing
or other nonvocal medium.
The series was formerly known as Linguistic & Literary Studies in Eastern Europe (LLSEE).
For an overview of all books published in this series, please see
https://benjamins.com/catalog/sfsl
Founding Editor Honorary Editors

John Odmark Eva Hajičová Petr Sgall
Charles University Charles University
General Editors
Yishai Tobin Bob de Jonge
Ben-Gurion University of the Negev Groningen University
Editorial Board
Alexandra Y. Aikhenvald James A. Matisoff
La Trobe University University of California, Berkeley
Joan L. Bybee Jim Miller
University of New Mexico Emeritus, University of Edinburgh
Ellen Contini-Morava Marianne Mithun
University of Virginia University of California, at Santa Barbara
Nicholas Evans Lawrence J. Raphael
University of Melbourne CUNY and Adelphi University
Victor A. Friedman Olga Mišeska Tomić
University of Chicago Leiden University
Anatoly Liberman Olga T. Yokoyama
University of Minnesota UCLA
Volume 73
by Jim Feist
Semantic Structure
in English
Jim Feist
John Benjamins Publishing Company

Amsterdamâ•›/â•›Philadelphia
TM
The paper used in this publication meets the minimum requirements of
8
theâ•¯American National Standard for Information Sciences – Permanence

of Paper for Printed Library Materials, ansi z39.48-1984.
doi 10.1075/sfsl.73
Cataloging-in-Publication Data available from Library of Congress
isbn 978 90 272 1583 3 (Hb)
isbn 978 90 272 6652 1 (e-book)
© 2016 – John Benjamins B.V.

No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any
other means, without written permission from the publisher.
John Benjamins Publishing Company · https://benjamins.com
Table of contents
Chapter 1
Introduction1
1.1 Goals of the book 1
1.2 Approach 1
1.3 Argument of the book 2
1.4 Plan of the book 5
Chapter 2
Semantic structures in the strata of English 7
2.1 Introduction 7
2.2 Semantic structure in lexis 7
2.2.1 Paradigmatic lexical relations 7
2.2.2 Syntagmatic lexical relations 8
2.2.3 Words without paradigmatic or syntagmatic relations 9
2.3 Semantic structure in morphology 9
2.3.1 Semantic classes 9
2.3.2 Grammatical meaning 10
2.3.3 Dependency and modification 10
2.4 Semantic structure in syntax 10
2.4.1 Semantic structure of clauses: The figure 10
2.4.2 Semantic structure in groups 13
2.5 Semantic structure in phonology 14
2.5.1 Introduction 14
2.5.2 Semantics of tonality 14
2.5.3 Semantics of tone 15
2.5.4 Semantics of tonicity 16
2.5.5 Semantics of rhythm 17
2.5.6 Semantic structure in phonetics 18
2.5.7 Discussion: Semantic structure in phonology 19
2.5.8 Conclusion: Semantics in the phonological stratum 21
2.6 Conclusion: Semantic structure in the strata of English 22
vi Semantic Structure in English
Chapter 3
Basis of semantic structure 25
3.1 Introduction 25
3.2 The intention to speak 25
3.2.1 Main intention 25
3.2.2 Subordinate intentions 26
3.2.3 Discussion 27
3.2.4 Conclusion 27
3.3 Functions, as the formulation of intention 28
3.3.2 Expressive function 28
3.3.3 Interpersonal function 30
3.3.4 Ideational function 31
3.3.5 Conclusion 31
3.4 Constraints on meaning 32
3.4.2 Linguistic constraints 32
3.4.3 Semiotic constraints 33
3.5 Conclusion: Basis of semantic structure 37
Chapter 4
Elements of semantic structure 39
4.1 Introduction 39
4.2 Preliminaries 39
4.2.1 Areas of meaning: Cognitive and linguistic meaning 39
4.2.2 Aspects of meaning 42
4.3 Dimensions of linguistic meaning 43
4.3.2 Quality dimension 43
4.3.3 Intensity dimension 43
4.3.4 Specificity dimension 44
4.3.5 Vagueness dimension 44
4.3.6 Basicness dimension 45
4.3.7 Viewpoint dimension 45
4.3.8 Boundedness dimension 45
4.3.9 Expectedness dimension 46
4.3.10 Salience dimension 47
4.4 Content meaning 47
4.4.2 Descriptive meaning 49
4.4.3 Affective meaning 52
4.4.4 Attitudinal meaning 56
4.4.5 Social meaning 58
4.4.6 Conclusion: Types of content meaning 59
Table of contents vii
4.5 Grammatical meaning 61

4.5.2 Types of grammatical meaning 64
4.5.3 Dimensions of grammatical meaning 65
4.5.4 Discussion 66
4.5.5 Conclusion: Grammatical meaning 67
4.6 Semantic classes 68
4.6.2 Basic classes 69
4.6.3 Discussion: Semantic classes 71
4.6.4 Conclusion: Semantic classes 72
4.7 Uses of meaning 72
4.7.2 Marked use 73
4.7.3 Defining and descriptive uses 73
4.7.4 Literal and figurative uses 76
4.8 Discussion: Elements of semantic structure 77
4.9 Conclusion: Elements of semantic structure 77
Chapter 5
Network structure 79
5.1 Introduction 79
5.1.1 General introduction 79
5.1.2 Introduction to networks 80
5.2 Structure within a word’s meaning 80
5.2.2 Word senses as a structure of meaning types 82
5.2.3 Structure within descriptive senses 87
5.2.4 Structure within non-descriptive senses 92
5.2.5 Discussion: Compositionality of sense structure 93
5.3 Structure among word senses 95
5.3.2 Sense relations: Synonymy 96
5.3.3 Other sense relations 98
5.3.4 Variation in sense structure 99
5.3.5 Conclusion: Structure among word senses 107
5.4 Structure of sublexical meaning 108
5.4.2 Dimensions 109
5.4.3 Elements and their sublexical relations 110
5.4.4 Elements’ external relations 112
5.4.5 Discussion: Structure of sublexical meanings 113
5.4.6 Conclusion: Structure of sublexical meaning 115
viii Semantic Structure in English
5.5 Networks in other strata 116

5.5.1 Networks in clause syntax 116
5.5.2 Networks in group syntax 118
5.5.3 Lexical network: Cohesion 119
5.6 Discussion: Network structures in English 120
5.6.1 Whole utterances as networks 120
5.6.2 Networks in imaginative English 121
5.6.3 Relations among one word’s various senses: Polysemy 121
5.7 Conclusion: Network structures in English 123
Chapter 6
System structure 125
6.1 Introduction 125
6.1.2 Introduction to system structure 126
6.2 System processes 127
6.2.1 Introduction: Grammatical meanings 127
6.2.2 Preliminary process: Obtaining content for the main procedure 128
6.2.3 Processes applying within words and groups 129
6.2.4 Processes applying within figures and figure complexes:
Complementation 134
6.2.5 Processes applying to a whole figure 135
6.2.6 Discussion: System processes 137
6.2.7 Conclusion: Grammatical meaning 138
6.3 System procedures: Using the processes 138
6.3.2 Constructing hierarchic structures 139
6.3.3 Constructing network structures in morphosyntax 140
6.4 Discussion: System structure 140
6.4.1 Details not yet explained 140
6.4.2 Signs used for grammatical meaning 141
6.4.3 Grammatical meanings in figurative and other uses 141
6.4.4 Grammatical meaning as backgrounded meaning 142
6.4.5 Other views of grammatical meaning 143
6.5 Conclusion: System structure 144
Chapter 7
Hierarchic structure (1): Figures 147
7.1.1 Introduction to hierarchies in semantics 147
7.1.2 Introduction to figures 148
Table of contents ix
7.2 Figure structures 149

7.2.1 Introduction: Processes 149
7.2.2 Material-Process structure 150
7.2.3 Mental-Process structures 151
7.2.4 Relational-Process structures 152
7.2.5 Discussion: Figure structures 153
7.2.6 Conclusion: Summary of figure structures 154
7.3 Processes 155
7.3.1 Introduction: Processes 155
7.3.2 Internal structure of Processes 155
7.3.3 Syntagmatic structure of Processes 156
7.3.4 Semantic class: Process 158
7.3.5 Conclusion: Summary of Processes 159
7.4 Participants 159
7.4.2 Internal structure 160
7.4.3 Syntagmatic structure 160
7.4.4 Participant as semantic class 161
7.4.5 Discussion: Participant roles and “semantic roles” 161
7.4.6 Conclusion: Summary of Participants 162
7.5 Circumstances 162
7.5.2 Internal structure of Circumstances 163
7.5.3 Syntagmatic structure of Circumstances 164
7.5.4 Semantic class of Circumstances 166
7.5.5 Discussion: Circumstances 166
7.5.6 Conclusion: Summary of Circumstances 168
7.6 Relations among Participant, Process and Circumstance 168
7.7 Structures larger than the figure 169
7.8 Discussion: Figures 170
7.8.1 Co-ordination and linearity 170
7.8.2 Constructions 171
7.8.3 Ergativity 172
7.8.4 Compositionality in figures 174
7.9 Conclusion: Figures 175
Chapter 8
Hierarchies (2): Groups and senses 177
8.1.2 Introduction to groups 177
x Semantic Structure in English
8.2 Entity groups 178

8.2.2 Classifiers 180
8.2.3 Descriptors 189
8.2.4 Epithets 193
8.2.5 Reinforcers 200
8.2.6 Determiners 203
8.2.7 Postmodifiers 205
8.2.8 Heads 206
8.2.9 Entity group as a unit 211
8.2.10 Discussion of Entity groups 217
8.3 Process groups 222
8.3.2 The Finite 223
8.3.3 Grammatical auxiliaries 224
8.3.4 Modal auxiliaries 225
8.3.5 Negative as modifier 226
8.3.6 Premodifiers 226
8.3.7 Postposed particles as modifiers 231
8.3.8 Heads of Process groups 232
8.3.9 Process groups as units 236
8.3.10 Discussion: Process groups 237
8.3.11 Conclusion: Process groups 239
8.4 Property groups 239
8.4.2 Property groups with Property heads 240
8.4.3 Headless Property groups: Prepositional phrases 242
8.4.4 Conclusion: Property groups 244
8.5 Words and morphemes 245
8.5.1 Internal structure 245
8.5.2 Syntagmatic structure 247
8.5.3 Semantic classes: Content meaning 248
8.5.4 Grammatical meaning in words and morphemes 249
8.5.5 Discussion: Compositionality of word senses 249
8.6 Discussion: Hierarchic structure in groups and senses 250
8.6.1 Language constraints 250
8.6.2 Semantic change 251
8.7 Conclusion: Hierarchic structure in groups and senses 252
Table of contents xi
Chapter 9
Hierarchic structure (3): Information structure 255
9.1.2 Introduction to information structure 255
9.2 Relevance structure 258
9.2.2 Reporting structure 260
9.2.3 Loose structures 261
9.2.4 Topic-Comment structure 262
9.2.5 Conclusion: Relevance structure 265
9.3 Orientation structure: “Theme” 266
9.3.2 Classes of Theme 267
9.3.3 Themes occurring at ranks other than the figure 269
9.3.4 Discussion: Orientation structure 270
9.3.5 Conclusion: Orientation structure 271
9.4 Salience structure: Rheme 272
9.4.2 Salience within an information item 273
9.4.3 Salience of items within an information unit: Rhematic structure 276
9.4.4 Salience of information units in larger units 279
9.4.5 Conclusion: Salience structure 281
9.5 Discussion: Information structure 282
9.5.1 Relation of information structure to the bases and elements
of language 282
9.5.2 Compositionality 283
9.5.3 Case 284
9.5.4 Cohesion 286
9.6 Conclusion: Information structure 287
9.6.1 Summary 287
9.6.2 Conclusions drawn 288
Chapter 10
Other structures 291
10.2 Semantic units with indeterminate structure 291
10.2.2 Ideational function 292
10.2.3 Interpersonal function 292
10.2.4 Expressive function 294
xii Semantic Structure in English
10.3 Semantic units with multifunctional structure 295

10.3.2 Holophrases 295
10.3.3 Ideophones 299
10.3.4 Discussion: Semantic units with multifunctional structure 300
10.3.5 Conclusion: Semantic units with multifunctional structure 301
10.4 Semantic units with field structure 301
10.5 Semantic units with wave structure 304
10.5.2 Wave structure in information 304
10.5.3 Wave structure in aesthetics and emotion 306
10.5.4 Conclusion: Wave structure 306
10.6 Discussion: Other structures 307
10.7 Conclusion: Other structures 307
Chapter 11
Realisation (1): Interpersonal functions 309
11.1.2 Introduction to realisation 310
11.2 Realising the Expressive function 312
11.2.2 Holistic realisation of Expression 312
11.2.3 Phonological realisation of Expression 313
11.2.4 Phonetic realisation of Expression 314
11.2.5 Lexical realisation of Expression 315
11.2.6 Conclusion: Realisation of Expression 315
11.3 Realising emotion and attitude 316
11.3.2 Conveying emotion by phonology and phonetics 316
11.3.3 Conveying emotion by lexis and syntax 317
11.3.4 Conveying attitude 318
11.3.5 Conclusion: Conveying emotion and attitude 319
11.4 Establishing personal and social relations 319
11.4.1 Establishing personal relations 319
11.4.2 Establishing social relations and social status 320
11.4.3 Discussion: Personal and social relations 320
11.4.4 Conclusion: Personal and social relations 321
11.5 Guiding hearers’ use of meaning 321
11.5.1 Introduction: Realisation of grammatical meaning 321
11.5.2 Guiding hearers’ overt response 322
11.5.3 Guiding hearers to syntagmatic structure 323
11.5.4 Guiding hearers to information structure 324
Table of contents xiii
11.5.5 Guiding the hearer’s attitude to content: Irrealis 330

11.5.6 Discussion: Guiding hearers 332
11.5.7 Conclusion: Guiding hearers 333
11.6 Provoking imaginative responses 333
11.6.2 How imaginative responses are provoked 334
11.6.3 Discussion: Provoking imaginative responses 337
11.6.4 Conclusion: Provoking imaginative responses 337
11.7 Discussion: Realisation of interpersonal meanings 338
11.8 Conclusion: Realisation of interpersonal functions 339
Chapter 12
Realisation (2): Ideational function 341
12.1.2 Introduction to ideational realisation 341
12.2 From intentions to words 343
12.2.2 Unmarked realisation into words 344
12.2.3 Marked realisation into words 348
12.2.4 Conclusion: From intentions to words 349
12.3 Syntacticisation (1): From words to groups 350
12.3.2 Grouping the words 351
12.3.3 Structuring the group 352
12.3.4 Ordering the group 356
12.3.5 Signalling the group structure 357
12.3.6 Discussion: Syntacticisation into groups 360
12.3.7 Conclusion: Syntacticisation into groups 360
12.4 Syntacticisation (2): From groups to clauses 361
12.4.2 Structuring the units 361
12.4.3 Ordering clause units 365
12.4.4 Signalling the structure 366
12.4.5 Conclusion: Syntacticisation into clauses 367
12.4.6 Grouping figures in a clause complex 367
12.5 Physical realisation 368
12.5.2 Realisation in sound 369
12.5.3 Realisation in writing 371
12.6 Discussion: Realisation of ideational function 372
12.6.1 How we conceptualise realisation 372
12.6.2 Incongruent realisation: Grammatical metaphor 372
xiv Semantic Structure in English
12.6.3 Construal 373

12.6.4 Statives 373
12.6.5 Domains offering alternative realisation 374
12.6.6 Comparison between morphosyntactic layers 376
12.7 Conclusion: Realisation of ideational meaning 376
12.8 Conclusion: Both forms of meaning realisation in English 378
12.8.1 Relation between the two forms of realisation 378
12.8.2 Sequence of steps 380
12.8.3 Realisation strategies 381
12.8.4 Adequacy of realisation as a semantic concept 382
12.8.5 Constraints on realisation 383
12.8.6 Looking forward 384
Chapter 13
Discussion385
13.1 Compositionality 385
13.1.2 Full compositionality 386
13.1.3 Limited compositionality 387
13.2 Word classes 388
13.2.2 Needlessness and unworkability 388
13.2.3 Lack of explanatory power 389
13.2.4 Support 391
13.2.5 Discussion 392
13.2.6 Conclusion 393
13.3 Prototypes 393
13.3.2 Needlessness of prototype theory in English grammar 394
13.3.3 Unworkability of prototypes in grammar 395
13.3.4 Confusions in acceptance of prototypes 395
13.3.5 Conclusion: Prototypes 398
13.4 Lexicon 399
13.5 Systematisation 401
13.6 Minor topics for discussion 402
13.6.1 Semiotics 402
13.6.2 Distinction between descriptive and referential use 402
13.6.3 Distinction between grammatical and content meaning 403
13.6.4 Distinction between cognitive and linguistic areas of meaning 403
13.6.5 Formalisation of linguistic description 404
13.6.6 Concept of classes 404
13.6.7 Philosophical tradition in linguistics 405
Table of contents xv
Chapter 14
Conclusion407
14.2 Nature of meaning 407
14.2.2 Symptomatic meaning 408
14.2.3 Semiotic meaning 409
14.3 Nature of semantic structure 410
14.3.2 Analytical view of semantic structure (1): Structures of units 410
14.3.3 Analytical view of semantic structure (2): Structures in a medium 412
14.3.4 Functional view of semantic structure 412
14.3.5 Conclusion: Nature of semantic structure 413
14.4 Stratification of the semantic structure 413
14.4.2 Semantics 414
14.4.3 Lexical items 414
14.4.4 Morphology 415
14.4.5 Morphosyntax 415
14.4.6 Information structure 419
14.4.7 Phonology and graphology 419
14.4.8 Phonetics 421
14.4.9 Discussion: Stratification 422
14.4.10 Conclusion: Stratification 422
14.5 Contribution to semantics 425
14.6 Further research 425
References427
Index443
Chapter 1
Introduction
1.1 Goals of the book
The primary goal of this book is to set out the semantic structure of English.
“Semantic” here is contrasted with “syntactic” and “phonological”, and also with
“pragmatic” (concerned with “meaning” which relies on social conventions such as
the Gricean maxims, and on inferences from them). “Semantics” means “to do with
studying what is expressed in language”, which includes “meaning” and “significance”;
that meaning of the term will be refined later in the book. It is not the study of prop-
ositional or truth-conditional meaning alone, but the study of significance and mean-
ings that are “coded” in linguistic items and structures, be they truth-conditional or
not (Hansen 2012, p. 233). “Structure” is also taken widely, having a range of applica-
tions, as it does in syntax. In particular, it includes the abstract sense in which social
structure, for example, is thought of as consisting of class and other relationships; and
it includes the more concrete sense in which we may itemise individual structures.
The study of semantic structure is to be linguistic. That will exclude sociolin-
guistics, conversation analysis, and discourse analysis, and paralinguistic features (e.g.
volume, tone of voice, speed and voice quality). The book also excludes “dynamic”
semantics, dealing with the building of a whole text, utterance by utterance, as in
Discourse Representation Theory; see Kamp and Reyle (2011), for instance. Further,
the study will be limited to English, although my knowledge of French and my read-
ing about other languages indicates that what is true of English semantic structure is
true widely.
Researching this book has led me to change my understanding of language, par-
ticularly in how it carries meaning, and in how meaning is structured in expression.
A secondary goal of the book is therefore to persuade readers to adjust their un-
derstanding if necessary – crucially, perhaps, in how they understand “semantics”
and “syntax”.
1.2 Approach
Approach to language. The book will treat language as it exists in use; meaning will be
taken as contextual, and the argument will be empirical, citing utterances as evidence.
Language is assumed to be functional: each utterance is presumed to have a cogni-
tive, personal or social function, which may be deliberate or only semiconscious; and
2 Semantic Structure in English
language itself is seen as functional in the evolutionary sense of being adapted to a

beneficial result, representation of knowledge being one of the functions – only one
of them. Language is also seen as functional in the further sense that it involves op-
erations, with “input” of some sort, and the “output” of speaker utterances or hearer
response.
The approach to language is also semiotic, rather than logical (to do with think-
ing). That seems necessary to deciding what “meaning” is, since conceptions of it
vary widely. The fundamental sense of “meaning”, according to the Shorter Oxford
English Dictionary (“SOED” hereafter), entails being expressed “by a sentence, word,
dream, symbol…”; that is, meaning is the meaning of a sign. (I have adopted in my
own thinking, as a rule of thumb, the maxim, “No sign, no meaning”.) Further, we
need to be conscious of the nature of linguistic signs; we should not assume that all
signs are symbols, giving full expression or representation of the meaning; the SOED
definition quoted just above says, more fully, that meaning is “…expressed or indi-
cated by a sentence…”; signs may be mere indications of the meaning. Finally, the
semiotic approach reminds us that meaning in its broad sense may include intention;
the SOED definition of the fundamental meaning of “mean” as a verb reads, “Have as
one’s purpose or intention…”.
Approach to studying language. The aim of studying language is taken here to be not
only description but explanation. By analogy, it may be explanatory in studying the
speech organs to consider not only their anatomy, but their function, their mode of
operation, and their development. Similarly, in studying the semantic structure of
speech, it may be explanatory to call on psycholinguistics, neurolinguistics and de-
velopment studies – of both language itself and children’s use of it. The explanation
should, moreover, be comprehensive – complete.
Because there is no consensus on the nature of semantics, we should be particu-
larly careful of assumptions – for example, the incompatible assumptions made by the
various approaches listed by Koptjevskaya-Tamm (2012), in which meaning is either
(1) decomposable, descriptive and absolute, or (2) non-composable and denotational,
or (3) always associated with a certain construction.
I have accordingly tried to avoid commitment to particular theories and systems
of description. The approach taken is in fact close to Systemic Functional Grammar in
syntax, as in Halliday (2014), and to Cruse (2011) in semantics. Readers who distin-
guish between semasiological and onomasiological approaches should note that the
approach here is semasiological – concerned with how language carries meaning, not
with what words assigned to particular meanings.
1.3 Argument of the book
Content of the argument. The nature of syntactic structure will be familiar to readers,
perhaps as a hierarchy represented in tree diagrams, or as dependency structures. The
Chapter 1. Introduction 3
nature of phonological structure is also clear, with a hierarchy of units from intona-
tion groups down through metrical feet to syllables and phonemes. But there is, to
my knowledge, no overall structure of semantics presented in the linguistic literature.
Three uses of “semantic structure” may be noted. To Langacker, it is conceptu-
al structure, without any distinction between concepts and meaning (1987, p. 99);
there is an “unequivocal identification of meaning with concepts”( 2005, p. 164); other
cognitive linguists take much the same view; so does Jackendoff (1972). Croft (2001,
pp. 19–21 in particular) refers to both “semantic structure” and “semantic structures”
in constructions; but such structure is not distinguished from the syntactic structure
which determines the construction; semantics is in effect simply conceptual for him,
also. In also these instances, “semantic structure” is conceptual relations, such as those
between entities and properties predicated of them, or those of a “semantic field”, i.e. a
group of concepts or things in the same cognitive domain. Evans (2009) sees semantic
structure as consisting of cognitive representations which are specialised for use in
language (2009, p. 42), but does not explain what kind of structure it is, or what spe-
cific structures there are; the only distinctions he gives are between concepts such as
MATRIX, DURATION, and EVENT, which are not specialised for language.
Moreover, those discussions of semantic structure do not even include the se-
mantic structure of the relationships among synonyms, antonyms and so on. (Those
relations are not usually thought of as semantic structure, but as “lexical relations”,
for example.) We need an account that integrates those elements into a comprehen-
sive structure, and provides a rationale for the integration. The main argument of the
book, then, is that semantics in English does have structure, and the substance of the
book is an exposition of what that structure is.
Form of the argument. Since there is no generally accepted structure or set of data, the
argument here cannot be simply deductive, or inductive. Instead, it will work from
accepted specific understandings, such as the fact that the relation of modifier to head
is semantic as well as syntactic, and from data introduced for the purpose. It will then
work by inference and generalisation from those bases, using other concepts, used in
some fields or approaches to linguistics but not used universally. Although the con-
clusions may be new, and in some cases contrary to conventional views, it is intended
that both the starting points and the argument from them be all conventional and
acceptable.
Development of the argument. The argument of the book begins, then, from the rela-
tions among synonyms and antonyms. Clearly, they are often related by having the
same conceptual meaning, or by being distinguished by their concepts. However, the
differences are sometimes not conceptual; famous and notorious, for example, both
denote the concept, ‘well-known’; they differ only in that they express different at-
titudes to being well known – favourable and unfavourable. Words like lovely and
horrible generally have no conceptual meaning, but are meaningful: we must allow
for emotive meaning. Those three forms of meaning – conceptual, attitudinal and
emotive – may be grouped as content meanings; but some meaningful words have no
content, as with utter and sheer, in “utter rubbish” and “sheer nonsense”. They inten-
sify the word they modify; that is certainly their significance, although it may not be
what we usually call “meaning”; since it is significance, semantics must account for it.
Those words are “grammatical items” or “functional items”; they contrast with “con-
tent items”; since the latter have content “meaning”, can we not say that the former
have “grammatical meaning”? In the first stage of the argument, then, we see a struc-
ture of types of meaning; and simple, relatively obvious observations lead naturally
to an unfamiliar, and perhaps uncomfortable, conclusion – that there is such a thing
as grammatical meaning (with the relation of meaning to function, and of the con-
nection between grammatical relations and grammatical meaning yet to be clarified).
Further reflection leads to the familiar fact that word meanings have an internal
structure, sometimes thought of as definitions, as in dictionaries, and sometimes as
lists of features, as when the meaning of stallion is represented as HORSE + ADULT +
MALE. Those conceptual elements are used in many words, and are related in various
ways; they must be seen as constituting a network, which has nodes with many links
each. Networks are a major semantic structure in English.
The relation between syntax and semantics has been controversial, but there is
now general agreement that the relation between them is very close, and approxi-
mates interdependence. There is a semantic structure “underlying” the syntax, or
“embodied” by it, or “interpreted” from it; its relations are not those of Subject, verb
and Object, but must parallel them, perhaps as “semantic roles”. However we describe
the elements of the structure, it must have the same form as the syntactic structure,
namely a hierarchy – which constitutes a second major semantic structure in English.
It is now widely agreed that English has “information structure” or “information
packaging”; but its place in grammar is not agreed. However, it seems to be clearly a
matter of content rather than of syntax; we conclude that it either is semantic, or has
an important relation to semantics. Our account of semantic structure must find a
place for it.
That provides a very brief outline of the argument of the book. It begins, I believe,
from familiar and accepted premises; it will end with some conclusions that will be
similarly unsurprising, but with some others that are likely to seem controversial or
even strange. That is particularly because of detail which has been passed over so far.
The following questions to be answered will give the flavour of the detail. What is
the semantic significance of being a Subject? In what different ways can the meaning
of an adjective modify the meaning of its head word? If we allow grammatical mean-
ing, what is its nature, and what is its structure? Can semantics resolve the apparently
syntactic arguments about the difference between Complements and Adjuncts, for
example? Expression of meaning in words is sometimes called “coding”, but what does
it entail beyond a one-to-one conversion, like coding the letter S into Morse code as
dots and dashes? The substance of the book lies in such detail, rather than in the ar-
gument outlined above.
Chapter 1. Introduction 5
The aim of the book, then, is to set out an orderly explanation of the structure
of meaning in English, with those and other questions resolved, and comprehensive
enough to find a place for all that is valid in present semantic thinking.
1.4 Plan of the book
Outline. The book has been planned to make the argument easy to follow, and to be
persuasive. For example, unfamiliar concepts are presented through examples and
through use, with abstract definition delayed until the concept is clear. The important
generalisations are given after the empirical detail on which they are based, rather
than being given first, without context.
Thus, Chapter 2 surveys familiar knowledge of English, looking for relatively ob-
vious instances of semantic structure, building up some of the concepts and technical
terms that will be used throughout later chapters, and – perhaps most important –
bringing out phenomena that need explanation through semantic structure.
Chapters 3 and 4 formalise the concepts introduced in Chapter 2, and add others,
building the foundation for the remainder of the book. Chapter 3 deals with the basis
of semantic structure in the intentions from which meaning rises, and in the func-
tions which it serves. Chapter 4 outlines the elements from which semantic structure
is built, such as the types of meaning (see above), conceptual elements of meaning,
and the dimensions which define them. Chapter 5 discusses the network structure
built up by those elements and dimensions, chiefly in senses and their relations.
Chapter 6 explains the processes which restructure the elements of the network into
the other main semantic structure, the hierarchy which parallels the syntactic hierar-
chy. Chapters 7 and 8 set out the nature of that hierarchic structure, from the clause
level down to the level of senses and their constituents. Chapter 9 sets out the nature
of information structure, similarly. Chapter 10 deals with other, less used structures.
That completes the analytical explanation of semantic structure; it is complemented
in Chapters 11 and 12 by a discussion of the structure of realisation; that is, the way
in which specific messages are formulated out of the vast multi-dimensional network
of potential meaning, formed into words, organised into utterances, and delivered
in speech or writing. Chapter 13 discusses implications from the previous chapters
which are relevant to our understanding of linguistics, but outside the narrow scope
of the book. Chapter 14 summarises and draws conclusions.
Conventions observed. Concepts are printed in SMALL CAPITALS. Meanings are printed
in single quotation marks – ‘……’. Words and phrases being discussed as utterances
are printed in double quotation marks – “……” – whereas words being discussed as
words in the language are printed in italics. That avoids the clumsiness of continually
saying “the word ‘table’” and “the meaning ‘table’,” but can cause its own difficulty
when the reader must keep in mind that “Table is…” (i.e. the word is…) does not
mean the same as “ ‘Table’ is…” (i.e. the meaning is…). Initial capitals are used where
needed to distinguish technical from general terms, as in “The Subject of the sen-
tence…” and “….a subject for investigation”.
Quoted utterances are put on a new line as numbered examples when they are to
be given some discussion, but are otherwise run into the text.
Chapter 2
Semantic structures in the strata of English
2.1 Introduction
Purpose and approach of the chapter. Like Chapter 1, this chapter prepares for the rest
of the book. It examines the strata of English, identifying some semantic structures
to be described fully later, identifying structural and conceptual elements which will
aid that description, and noting some things that need further explanation and some
that have no familiar explanation at all. The approach is exploratory, looking for rep-
resentative instances, leaving comprehensive treatment to later chapters. The purpose
is thus to persuade readers fully that the book is needed, and to indicate what they
can expect in the following chapters. Phonology has been given more space than the
other strata, because its semantic significance is much less known, and to allow later
chapters to draw on its explanations.
Arrangement. The remaining sections deal in turn with the strata of lexis, morpholo-
gy, syntax, and phonology, in relation to semantics. There is no developing argument;
rather, points are made independently, and the conclusion will be general.
2.2 Semantic structure in lexis
2.2.1 Paradigmatic lexical relations
We begin the study of semantic structure in lexis by considering the words that fit into
paradigms of alternatives – sets of synonyms, antonyms and so on. Comparing closely
related lexical items such as sob, whimper, weep, keen, and blubber shows that word
meanings seem to be built up from elements. For example, all those words include ‘to
manifest pain, misery or grief ’ (SOED, on weep); sob adds the element ‘convulsively’,
and whimper adds ‘feebly’ and ‘intermittently’. That analysis shows that weep has a
general sense, and that the other senses are more specific. That is a general pattern;
senses differ on the dimension of “generality” or “specificity”. Dimensions will be im-
portant in setting out semantic structure.
Those conceptual elements do not exhaust the differences between synonyms,
however. We use weep in formal contexts, and blub and blubber in informal ones, in my
judgement at least – SOED does not specify their usage. That is part of the significance
of the words, though it may not feel natural to say that it is part of their “meaning”.
We use whimper and blubber to convey disapproval: attitudinal “meaning” is another

aspect of words’ significance. Finally, to keen has Irish “connotations”, which must be
accounted for somewhere in our description of meaning. Among other relations, each
of the synonyms considered has its antonyms; and weep is a hyponym for the others.
The book will argue that these forms of significance are indeed part of meaning, and
will be referred to as “types” of meaning.
The types and dimensions of meaning make “planes” on which word senses are
related, additional to the conceptual elements. The natural term for such for such a set
of patterns on several planes is “network”, which is one of the fundamental semantic
structures of English.
2.2.2 Syntagmatic lexical relations
2.2.2.1 Grammatical and content items

When we compare words in their syntagmatic arrangement, a more familiar distinc-
tion quickly arises, the distinction between “lexical” or “content” words and mor-
phemes, and “functional” or “grammatical” ones. It is not obvious whether that is a
semantic contrast, or a contrast between semantics and grammar. Nor is its nature
made clear in the literature: lexical and grammatical items are said to constitute class-
es, but the classes overlap, with many words belonging in both. The book will distin-
guish between content meaning and grammatical meaning, as types of meaning. They
need thorough examination; for example, prepositions and conjunctions link chunks
of content, so is grammatical meaning structural? What do content meaning and
grammatical meaning have in common? It will be argued that grammatical meaning
is fundamental to semantic structure.
2.2.2.2 Semantic classes

If we consider repeating patterns, as in such utterances as “Roses are red; violets are
blue; sugar is sweet, and so are you”, it seems obvious that meanings belong to seman-
tic classes – of “entities” (‘roses’, ‘violets’, ‘sugar’) and “properties” (‘red’, ‘blue’, ‘sweet’).
“Events” (‘ran’, ‘broke’, ‘fell’) are an obvious addition. Is and are are not so straightfor-
ward – do we need to posit a relationship class?
2.2.2.3 Semantic classes and cognitive classes

Problems emerge very quickly, however. Relationships are so different from the mem-
bers of the other three classes, which are clearly based on the way we experience the
world directly, that we must enquire whether the classification is valid. Second, it
seems doubtful that there should be semantic classes of kinds of things, such as flow-
ers, grasses, weeds and so on; those are horticultural classes, and part of our knowl-
edge of the world, not of grammar; so we need to resolve the nature of being semantic,
including the relation between the linguistic (e.g. semantic) and cognitive (e.g. hor-
ticultural). Since the classes of things, properties and so on support the classes of
Chapter 2. Semantic structures in the strata of English 9
nouns, adjectives and so on, we will need to examine those as well, if we think they
are relevant to semantics.
2.2.3 Words without paradigmatic or syntagmatic relations
2.2.3.1 Interjections
Some words do not have significant lexical relations of either the paradigmatic or
syntagmatic kind. Interjections are obvious examples: “Heck!”, “Oh,” “Blow!”, and
“Whoopee!” Just as they do not have a place in syntactic structure, they do not, it
seems, have a structure of meaning elements and meaning types, as we saw with
whimper and its synonyms. They seem to stand apart from semantic structure.
2.2.3.2 Holophrases
There is a second group of expressions which are unstructured, namely holophrases.
They are words or other lexical items which have complex but indeterminate mean-
ings. They include children’s earliest linguistic utterances, in the “one word” stage:
“Gone” and “More!”, for example. Adult examples include discourse particles like
well, general purpose rejoinders such as “Cheers!”, the ironic response, “Really?” to
the remark, “I got an A grade”, and the response “Oh, no!” to the remark “I’m stuck
again.” They are not meaningless, but it does not seem possible to give their meaning
a dictionary-style definition.
Both groups of words pose problems which an account of semantic structure
must resolve. Since it is easier to state an intention or function for these unstructured
items than to state a meaning, we should examine the relationships among meaning,
function and intention.
2.3 Semantic structure in morphology
2.3.1 Semantic classes
There are clear patterns in the suffixes which derive new words from base words. The
suffixes -ise/-ize and -en form event senses; -ion and -ness form entity senses, and
-ous and -ful form property senses. That supports the earlier suggestion that semantic
classes underlie semantic structure. But this derivational morphology sets a problem
for the concept of semantic classes. For example, stupefy means ‘to induce stupour’;
the sense seems to include a thing as well as an action. Does the sense belong to the
action class (‘induce’) linguistically, but include a member of the thing class (‘stup-
our’) cognitively? The problems noted in the previous section increase.
2.3.2 Grammatical meaning
The number and gender inflections that form the agreement between Subjects and
Predicators raise other issues. The relation of agreement is taken to be “grammatical”,
rather than semantic; but it does not seem to be meaningless, although any meaning
is certainly not like that of content words. Perhaps agreement is a grammatical item,
with grammatical meaning? That, too, needs resolution.
2.3.3 Dependency and modification
Prefixes such as pre- and sub- had distinct content meaning in the languages from
which they were borrowed, but do not seem to have such meaning in words like pre-
side and substantial. On the other hand, mega- and mini- clearly are meaningful in
present-day English, since they have currency as independent words; un- must be
meaningful for “He was a shy man, but he had to un-shy himself ” to be meaningful.
Presumably then, words such as megastar and mini-rugby have the semantic struc-
ture of modifier and head – to be noted in the next section. It will be hard to resolve
whether the meaning of words with several morphemes is compositional in structure.
Suffixes affect the meaning of the root in different ways. Some change the part of
speech, sometimes adding conceptual meaning as they do so, as with -esque (‘like’);
some change the word grammatically (in tense or plurality, for example); and some
add a scornful or other emotive quality (lordling, princeling). Suffixes, as a single mor-
phological structure, represent several different semantic structures.
2.4 Semantic structure in syntax
2.4.1 Semantic structure of clauses: The figure
2.4.1.1 Constituents of figures

Structural constituents. Following Halliday (2014), the syntactic structure of clauses
is taken here to be that of Subject + Predicator + optional Complement + option-
al Adjunct. As syntax, that structure is taken to concern the order of the units, not
meaning. Part of their function, however, is to realise semantic units, as follows. In
unmarked1 use, the Subject realises whatever “thing” the Predicator is asserted about,
such as the person who carries out the action; Complements realise whatever else is
centrally involved in the predication. Semantically, then, they are participants directly
involved in the event. (Marked uses include the use of there and it as dummy Subjects.)
The Predicator realises an event or relationship, which will be called the “Process”.
1. “Marked” here and elsewhere means “contrasting with standard or rule-bound use”; only
rarely will it mean “carrying a distinctive marking”.
Adjuncts realise meanings which will be called “Circumstances” (“adverbials”, rough-

ly). Thus, there is a semantic structure closely paralleling the syntactic structure, with
units which are of equal status – i.e. interdependent rather than dependent – and
which are components of a larger unit, the “figure”, the approximate equivalent of the
clause. (“Dependent” will be equated with “subordinate”; dependency will not be an
important concept, as it is in dependency grammar.)
Class constituents. The semantic constituents just listed seem to fall naturally into se-
mantic classes, in simple figures at least. The “Participants” either are entities or are
very similar to them; “Processes” are usually events; “Circumstances” are properties in
a broad sense, or a combination of relationship + Participant (“prepositional phrase”).
However, the concept of semantic class will need further modification. First, it must
allow for Predicators like is which are not events. Second, it must allow for the seman-
tic roles in the figure, such as being the actor (as in “Mr Barnett lit a fire”) or being a
term of the predicating relationship (as in “Mr Barnett had been the captain”). Third,
it must allow for changes in class, or at least changes in role, as words are structured
in groups.
I take the substance of these “roles” to be semantic, although it is often treated as
being syntactic: the roles are sometimes called “semantic”, but are often called “gram-
matical relations”, or “thematic roles”, and so on. If we accept that grammar comprises
phonology, syntax, morphology and semantics, then these roles must be semantic.
2.4.1.2 Relationships among figure constituents

Transitive relations. “Mr Barnett lit a fire” means more than ‘lit’ + ‘a fire’ + ‘Mr Barnett’.
For example, it includes the elements that Mr Barnett carried out an action, and that
fire appeared as a result. Those elements are parts of a semantic structure, which is
created, it seems, by the figure’s transitivity. An account of semantic structure must
explain how such structures are generated.
Relations with a copula. Copulas – as in “roses are red” – do not denote any event or
action, or entail any duration in time or “instability in time”. Copulas, then, are quite
distinct from what “verbs” are commonly conceived to be, yet copulas can be “the
verb” in a clause, or the “head of a verbal phrase”. We need an explanation of their
semantics.
Dimensions: specificity. There are several ways of interpreting “Sometimes I’d see
someone I knew on the bus and we’d wave” (Corpus of Contemporary American
English – “COCA” hereafter). The figure can “mean” that I was on the bus, or that
the other person was on the bus, or that both were. Perhaps we should say that the
figure does not specify who was on the bus; specificity would then be an important
dimension here, as well as in the lexical stratum. In any case, an account of semantic
structure must explain why Adjuncts such as “on the bus” pattern so differently from
the other constituents of the figure – with so much variability – and what controls
their interpretation.
2.4.1.3 Significance of group order

Order of constituents. Uttering “He is.” is an assertion; uttering “Is he?” is a question.
Subject-Predicator order signifies asserting; Predicator-Subject order signifies ques-
tioning. The order of syntactic units can thus carry significance, just as words can; and
the difference between word and group carries significance, as we saw in the section
on class constituents. To account properly for semantic structure, we must include
this semiotic issue: if both words and word order are signs that can carry significance,
what is the range of meaningful signs in English?
Asserting and questioning are semantic (to do with significance) rather than syn-
tactic, but in a quite different way from expressing the conceptual content that is as-
serted or questioned. As the common term “speech act” indicates, they are instances
of language as action – or interaction between people – rather than of language as
coded information. Exclamations are not exactly either of those alternatives; rather,
they are expression – expression of feeling. It is likely that these differences in func-
tion and in the intentions behind them affect semantic structure.
We can say that the Predicator-Subject order in “Is he?” sets the interrogative
mood, in the grammatical system; but we also need to say that for the speaker it rep-
resents a speech act, and that for the hearer it is a request for a reply. Utterances, then,
have three aspects, all of which should be allowed for: the speaker’s, the hearer’s, and
the grammatical or “system” aspect. That probably unfamiliar three-way distinction
will be maintained throughout the book.
In speech acts, speakers are guiding the hearer as to the response they want, and
every utterance constitutes a speech act. The significance of language includes guiding
the hearers in understanding the utterance and responding to it.
Order of content. There is one other type of significance which hearers understand
in clauses, for which the terms “Topic”, “Comment”, “Theme”, “Rheme” and perhaps
others are used. (The Theme is the first information, orienting the hearer; the Rheme
develops the rest of the information from that starting point.)The issues go under
the name of “information structure”, “information packaging”, and so on. The va-
riety of terms reflects both different understandings of the structure, and different
understandings of how it relates to “semantics”. This book must attempt to resolve
those issues.
If we pursue for a moment the semiotic issues involved in Topics, we quickly
become puzzled. We can identify “topicalising constructions” that signify a Topic, but
we cannot find linguistic forms that signify Comment. Similarly, there are no linguis-
tic forms to signify degrees of Rhematic importance. (By that, I mean increase of in-
formation value leading to a focal point, usually at the end.) On reflection, we see that
we interpret part of an utterance as Comment because it follows the Topic (in default
structure), and (again by default) interpret the order of information in the Comment
as representing increasing importance. Order of content, as well as order of syntactic
units, carries significance.
Conclusion. Order of words and of content generally is a necessity in speaking and

writing, but we conclude that it has been assigned meaning of its own. We also con-
clude that just as syntactic groups and clauses form a hierarchic structure, so do se-
mantic groups and the figure they constitute.
2.4.2 Semantic structure in groups
2.4.2.1 Groups as part of the semantic hierarchy

Groups consist of modifiers and heads, semantically as well as syntactically. The group
has constituents, then, the modifiers being dependent on the head. That continues
the hierarchic structure we noted in figures, and it continues in the morphological
structure of words, discussed in the previous section. We conclude that hierarchy is a
major structure in English semantics, along with network (§2.2.1 above); analysing it
will be a major task for the book.
2.4.2.2 Modification structure

Modifiers in nominal groups raise several issues for semantic structure. Commonly,
they belong to our proposed semantic class of properties, but premodifiers such as
fake and alleged do not (being alleged is not a property of an alleged criminal); do they
denote a meaning, or carry out a function? Is there a semantic significance to the fixed
order of premodifiers? What is the significance of the fact that words change their po-
sition according to their meaning? (Compare “a high positive predictive value”, “mild
high blood pressure”, and “a typical suburban high street” – the meaning changes as
the word’s position changes.) Where we used to say “patients’ entrance” in the plural
and possessive, we now commonly say “patient entrance”, apparently singular and
not morphologically possessive; what is the semantic significance of that? Even the
maxim that modifiers modify the head of the phrase can apparently be falsified, as
when a tramper reports that “our happy feet danced from boulder to boulder” – it was
the trampers that were happy, not their feet, and “happy” apparently modifies the ‘us’
in “our”.
2.4.2.3 Compositionality
Modification like that of fake and alleged, and like that of happy, in the examples
above, poses considerable problems for explaining the compositionality of semantic
structure, since they do not seem to follow the rules for composing meaning in nom-
inal groups.
2.5 Semantic structure in phonology
2.5.1 Introduction
Basically, phonology is simply the way in which meaning is made audible to hearers.
This section deals with ways in which phonology adds to the meaning of the morpho-
syntax which it makes audible – just as word order adds its own meaning.
The vocal sounds of speech have a number of dimensions, including pitch (both
change of pitch and absolute pitch), volume, speed, duration, and voice quality. Of
those, pitch has been conventionalised (in the tone system especially), and so has
stress (as a combination of pitch, volume and duration). Those dimensions are used
to constitute four major phonological systems (Halliday and Greaves 2008, p. 210).
They are as follows.
– Tonality, “the chunking of words into groups or phrases” (Wharton 2012, p. 567).
– Tonicity, the location of the “pitch accent” or “main stress” in the chunks defined
by tonality (Wharton 2012, p. 567).
– Tone, the type of “melodic” pitch contour on the main stress.
– Rhythm, the “distribution of utterances into feet (metric units) with location of
boundaries” (Halliday and Greaves 2008, p. 210).
All four are here included in “intonation”.

This section will give more detail than the previous section did, because the “se-
mantics” or significance of phonology is much less discussed in the literature.
2.5.2 Semantics of tonality
Tone groups as information units. Tone groups, “tone units” or “intonation units” –
the chunks of words marked off by tonality – are units of information. They are thus
related to clauses, but the differences are important. In writing and in very formal
speech, information units are usually represented by clauses, the clauses being spoken
as intonation units. In most speech, however, clauses and intonation units are not co-
extensive: there may be several clauses in one tone group, or several tone groups in the
one clause. For example, the speaker of example (1) made three tone groups out of the
single main clause (with an embedded clause). The tone groups are divided by slashes.
(1) “Well I would just like to say / [0.5 second pause] that for the first time in my
life / I’ve been in hospital.” (Couper-Kuhlen 1993, p. 45)
The three tone groups signal that the speaker felt she had three pieces of information
to convey.
A tone group may consist of a single phrase or even a single word; in informal
speech especially, “Yes” and “Six o’clock” are quite acceptable as tone groups, because
speakers and hearers consider them complete as items of information. Tonality works
in information structure.
Phonological paragraphs. The tonality system also creates phonological units consist-
ing of several tone groups; each such unit creates a semantic structure consisting of
several information units (Tench 1996, pp. 23–24). It is the functional equivalent in
speech of a paragraph in writing, so may be called a “paratone” (Tench 1996, p. 24,
citing work by Brown). In unmarked form, paratones begin on a high pitch, both on
the initial syllable and on the initial tone group. The pitch falls fairly steadily to the
end. In formal style, it is often followed by a pause.
Affective meaning of tonality. An unusual degree of rise or fall in the overall pitch con-
tour conveys an affective meaning; that is, it conveys emotion or attitude. Examples
are “I see!”, uttered as a sigh, a groan, or as an indignant exclamation.
2.5.3 Semantics of tone
Introduction. Phonetically, the system of tone is made up of seven tones (Halliday

and Greaves 2008, p. 45), each being a pattern of change of pitch. There are five “sim-
ple” or basic tones: tone 1 is a fall; tone 2 is a rise; tone 3 is level; tone 4 is a fall and
rise; tone 5 is a rise and fall. There are also two compound tones. Cruttenden (1997,
§3.7) gives a similar analysis into seven “nuclear tones”. There is a natural significance
of falling pitch as completeness or finality, and of rising pitch as incompleteness or
lack of finality – that is, that the speaker’s speech intention is complete or incomplete
(Halliday and Greaves 2008, p. 112).
The tone system’s main significance is that it defines which speech act is intended
(assertion, question, command or exclamation) for the information unit marked off
by tonality.
Basic significance of the tones. According to Halliday and Greaves (2008), the basic
significance of the commonest tones is as follows.
– Tone 1, the falling tone, is a default tone used for tone groups whose speech func-
tion is marked clearly by the syntax, so it has no significance of its own.
– Tone 2, the rising tone, marks tone groups as questions.
– Tone 3, the level tone gives the feature of neutrality or lack of commitment; it can
be taken as meaning “I decline to choose between the alternatives.” Specifically, it
can show that the speaker is suggesting rather than asserting, or that the speaker
is agreeing with the interlocutor, rather than asserting something of his or her
own, and so on.
– Tone 4, the fall-rise combination of certainty and uncertainty, is used to give a
feature of reservation; it means that the statement seems certain but is not so in
fact: “There’s a ‘but’ to this!”
– Tone 5, rise-fall, signals that the tone group is an exclamation.
Other significance conveyed by tone. Tone is also important for conveying emotive and
attitudinal meaning. Speakers use the rising tone’s sense of incompleteness in high-
rising tone 2 to express an attitude or “feature” of questioning, or of challenging the
interlocutor’s assertion (Halliday and Greaves 2008). The compound tone 1 followed
by 3 is used with a negative imperative such as “Don’t do that!”, to convey a feeling of
urgency, or an emotion of anger (Halliday and Greaves 2008, p. 119).
Cruttenden (1997) notes that tonal variation is often a marker of style, and that it
is sometimes an index of the speaker’s class and sex. One well studied example is the
“high rising tone” or “uptalk”, which is a rising tone ending a statement which in itself
seems complete. At first (in the 1970s), it characterised the speech of young Austral-
ians, especially females and those in state schools; later, it spread to other places and
social groups (Cruttenden 1997, §5.2). Thus tone sometimes conveys social mean-
ing – social identity, in this instance.
Tone 3 can be used to signify that the speaker’s speech act is not finished yet, the
finality of the questioning or asserting being still to come. It can be taken to mean
“Please wait”. Thus it carries a kind of grammatical meaning, guiding the hearer’s re-
sponse. Another kind of grammatical meaning is conveyed by some marked variants
of the tones, that of emphasising or intensifying the content of a word. A fairly famil-
iar example is the use of marked pitch, which is a phonological alternative to morpho-
logical intensification, e.g. bigger or very big, and lexical intensification, e.g. huge. The
effect is to shift the sense along its intensity dimension. “I’m taking the train this time”
can be spoken with tone 1; the falling tone spreads over all of “train this time”, and
there is a single significant stress (on “train”). It could, however, be spoken with the
falling tone (tone 1) restricted to “train”, and with a level tone (tone 3), with secondary
stress on “time” (Halliday and Greaves 2008, p. 45; it is a “compound tone”, “tone 1,3”).
The semantic difference is that there is now a secondary focus in the tone group; that
makes “this time” a subordinate assertion within the main assertion, because fall in
tone after the focus on “train” marks that information as unimportant – which is part
of information structure. The compound tone guides the hearer’s interpretation, as a
grammatical meaning.
2.5.4 Semantics of tonicity
Focus. Like tonality and tone, tonicity works primarily for information structure. The
main or “tonic” stress marks the end of the information unit, and highlights the most
important part of the information, the focus.
Rhematic structure. As noted above in §2.4.1, the Rheme of the information unit con-
sists of a sequence of peaks of importance. Those peaks are signalled by the phrase
stresses, culminating in the focus, as in example (1), above: “…say that for the first
time in my life, I’ve been in hospital.” (Couper-Kuhlen 1993, p. 94. The underlined
stresses, most of them in marked position, are as recorded by Couper-Kuhlen.)
Relation to other strata. Tonicity works with other strata, too. It works with the lexis,
especially with focusing words such as alone, only, especially, and even, in marking
what is important; and it works with emotive or otherwise forceful words. It also
works with syntax, since the order of postmodifying groups and of Adjuncts helps
in defining the importance of information. Semantic structure has the strata of lan-
guage sometimes working together, reinforcing each other, but sometimes working
as alternatives.
2.5.5 Semantics of rhythm
Basis of rhythm. Speech is naturally rhythmical to some extent, because of its basis in
human physiology, but it gains semantic significance when speakers make it regular
enough to be noticeable. Chunks of text, usually several feet long, are “rhythmised”
(Couper-Kuhlen 1993, p. 98), by the weakening or strengthening of some stressed syl-
lables, the speeding up of some syllables, and the use of pauses, until there is a pattern
of beats at approximately even time intervals. The evenly spaced beats may be word
stresses, phrase stresses, or even sentence stresses.
The passage in Diagram 1 is from a radio interview; the woman speaking is shar-
ing the excitement she felt when she visited Withington. Key to the diagram: the high-
er the column of X’s, the greater the stress. The horizontal line marks sections that
were rhythmised.
x x x x
x x x x x
———————————————— ——————————— ————————————— ————————————
x x x x x x
x x x x x x x x x x x x x x x
Withington was a name on the map [.4 sec.] Yes, because living in
x x
x x
x x x x x x x
—————————— ————————— —————
x x x x x x x x x x x
x x x x x x x x x x x x x x x x x
North Manchester I knew nothing at all about Withington” . . .
Diagram 1. Structure of rhythm (Couper-Kuhlen 1993, pp. 95–96)

Significance of rhythm. Conveying emotive meaning is a familiar use for rhythm; for
example, see Quirk et al. (1985, p. 1598) on its use for irritation and sarcasm. It can be
seen in the brief passage about Withington just quoted, conveying excitement – add-
ing to the effect of the marked stress, on the first Withington and on map, for example.
Note that the rhythm as delivered was not what we might expect from the written text:
speakers adjust the “natural” rhythm to their purpose.
The significance of rhythm is often that it heightens meaning represented in other
ways, by stimulating the imagination and rousing excitement accordingly. That use is
so well known from oratory and poetry that it does not need illustration.
As well as the meanings listed, rhythm often has a grammatical meaning which
builds information structure. The rhythm is reduced at the beginning of an utterance:
unstressed syllables are spoken with greater speed and reduced enunciation. The in-
troductory phrase, “There’s a dreadful…”, may even be pronounced as ‘Th-dreadful”
(Cruttenden 1997, p. 21). Second, rhythm is the main marking of Rhematic structure
(see Chapter 10). Less important content is spoken with lighter and quicker rhythm.
An example from Halliday will illustrate the point.
(2) “In this job, Anne, we are working with silver. Now silver needs to have love”.
(Halliday 2014, §3.5)
In the second sentence, only the word love was important, so “needs to have”, the first
part of the Rheme, was spoken with light, quick rhythm.
2.5.6 Semantic structure in phonetics
Introduction. Speaking is a physical activity, using organs that have other functions.
We vocalise on the out-breath, which seems to be the basis for tone groups, with the
in-breath separating them; perhaps the syllabic structure of words is partly derived
from rhythmic mouth movements as in chewing. When we speak more forcefully, we
use extra force from the lungs; states of emotional tension go with states of physical
tension, and expression of those states uses tense speech muscles. This section argues
that the physical features of speech, and of articulation especially, often have expres-
sive significance – that phonetics, as well phonology, can have semantic significance.
Meaningful phonetics in English. Differences in regional and social dialects consist to

an important extent of differences in the vowel quality, and other phonetic features.
Phonetics is thus meaningful in conveying to the hearer the regional or social back-
ground of the speaker, as noted in §2.5.3.
There is a much wider and perhaps more important issue here, however. Ono-
matopoeia is accepted universally as an example of meaningful phonetics. However,
its nature as an imitation of natural sound is a widely misunderstood, I believe. For
example, its value is often attacked on the grounds that onomatopoeic expressions
vary from language to language, and even within a language; thus English has both
“bow wow” and “woof woof ”, and French has the equivalent of “wa wa”. But the imi-
tation lies, not in the phonemes, but in the phonetic structure – in the sudden onset
of the syllables, in the vowels being made at the back of the mouth, and the repeated
strong beats; and the effect lies, not in those phonetic abstractions, but in the artic-
ulation: we articulate the sounds abruptly and forcefully in the back of the mouth,
imitating dogs’ articulation.
That physical nature of speaking the words explains why so many interjections
and swear words use plosive /b/ and /d/ (blow, bloody, blast, damn), or guttural /k/
or /g/ (God! Christ!). It explains odd little things like adding /p/ to yes and no, in
yep and nope: closing your mouth to finish the word adds finality. All that being so,
it is natural for us to give a little more emphasis to the sounds in our words: to give
a stronger outburst at the beginning of brutal and please, and to lengthen the vowel
in please. Emphasising stop consonants leads to giving them extra aspiration. Con-
sequently, “Oh T[h]om!” expresses exasperation, the phonetic value having emotive
meaning; and “They spent b[h]illions on it” carries intensification – the phonetic value
has grammatical meaning, and is an alternative to using an intensifying word, such
as the grammatical item, very. Extra aspiration was used in the passage in Diagram 1,
“Withington was a name on the map. Y[h]es. Bec[h]ause living in North Manches-
ter…” (Couper-Kuhlen 1993, p. 95).
Conclusion: semantic structure of the phonetic stratum. This use of the phonetic value
of English sounds, independent of their phonemic value, makes an important con-
tribution to the semantics of English, although it is sharply reduced in formal style.
See also Halliday and Greaves (2008, p. 169) on “phonaesthesia”, and the literature on
“sound symbolism”.
2.5.7 Discussion: Semantic structure in phonology
2.5.7.1 Wave-like structures

As noted by Halliday (2014), many phonological structures are like waves. In their
phonetic pattern, tonality, tonicity and tone all consist of rise and fall of pitch or stress,
and rhythm consists of alternation of stress level; in all of them, the levels are relative,
with no level or status being absolute. Semantically, the meaning or effect is also a rise
and fall – particularly of importance to the information structure, but also of feeling.
That structure is quite different from the hierarchic one created by syntax, which has
well-defined units, which are either equal or unequal absolutely, with a definite num-
ber of levels. The method by which meaning is conveyed is also quite different: the
pattern of pitch or tone, as rise and fall, images the pattern of meaning, just as, on a
map, the meandering of the line for a river images the meanderings of the real river;
the meaning is conveyed iconically, not symbolically.
2.5.7.2 Field-like structures

The significance of a fronted /e/ or rolled /r/, and of any other socially marked phonet-
ic feature, is not that the one word or phrase is spoken by a person from a particular
social group; the significance is that the whole utterance, or the whole text, is spo-
ken by such a person. The feature creates a “field” which affects the whole utterance,
as a magnetic or gravitational field affects everything in it. That structure is radical-
ly different from both the hierarchies and the waves studied so far. As with waves,
the structure has been noted before, by Halliday (2014), building on work by Pike.
Halliday distinguishes meaning as field from meaning as units or “particles” (building
a hierarchy), and from meaning as wave, making an analogy with electromagnetic
energy, which occurs as particles, waves and fields. Waves and fields are also basic
structures which will be studied in the rest of the book.
2.5.7.3 Basicness
Children use phonology expressively in the first weeks of life, gradually differentiat-
ing their cries to the point where adults can distinguish a cry of hunger from a cry of
pain. Rise and fall structures are thus more basic than hierarchic structures, and the
expressive function is more basic than the function of representing information con-
ceptually. That explains various things, such as the fact that phonology can override
syntax, as noted above, and the fact that we can make meaningful utterances such as a
prolonged “A-a-a-h” with phonology but no syntax, whereas we cannot speak syntax
without phonology. Basicness is a semantic dimension – one of some importance.
2.5.7.4 Compositionality in phonological semantics

This section discusses the much debated issue of compositionality: whether the re-
lations among sense elements construct the whole sense by regular processes, as
(6+2) × (2+1) constructs 36. (Applied to words, the issue is sometimes referred to
as “decompositionality”.) The issue is not important to the argument of the book, be-
cause, in the view of meaning as functional and being structured as a network, there
is no reason why meaning should be fully compositional (as will become clear later in
the section), and as mathematically stated meaning should be. I discuss it because it
is important in the literature – there is, for example, a whole large handbook on com-
positionality alone (Werning et al. 2012) – and because the discussion will illumine
semantic structure from another point of view.
Superficially, the phonology of English is compositional: phonemes combine into
a syllable, syllables combine into feet, and so on. Semantically, there is very little com-
positionality in the significance of phonological patterns. In its prime function, tonal-
ity creates a unitary tone group; similarly, tonicity creates a single tonic stress; tone,
as a pattern, is a unitary structure, and rhythm has a unified overall effect. But there
is some compositionality in the way that secondary stresses make up Rhematic struc-
ture, in the compounding of tones, and in the modification of tones by secondary
tones (Halliday and Greaves 2008). That leaves the role of compositionality in English
still less clear than it is in syntactic semantics: it will need study.
2.5.8 Conclusion: Semantics in the phonological stratum
2.5.8.1 Structure
The relations of phonology to syntax and lexis are complex. In syntax, words, groups
and clauses make a distinctive and unified structure, and build a similar unified se-
mantic structure. Phonology, however, has no such unity or consistency; its units are
relatively independent in nature and function, remaining close to their origin, where-
as lexis and syntax are highly developed, specialised and integrated. Whereas syntax
organises and complements the meanings represented by words, phonologically ex-
pressed meanings may reinforce a meaning already expressed in lexis or syntax, or be
an alternative for it, or override it.
2.5.8.2 Meaning
Like syntax, phonology is a means for making word meaning transmissible to hear-
ers, but – again like syntax – it adds meaning to what is represented in the words.
Whereas lexis is focused on conceptual content, and syntax is focused on grammati-
cal meanings concerned with structuring the conceptual content, phonology is con-
cerned mainly with conveying feeling and attitude, signalling information structure,
and very rarely with conceptual content. In some marked uses, it has a meaning which
no other stratum can convey.
This discussion of phonology has confirmed what the discussion of morphosyn-
tax suggested: that information structure is a regular and systematic part of semantics
in English – or at least of its significance if not of “semantics” – and that grammat-
ical meaning is so also, being wide-ranging and of many types. The discussion has
brought out an even wider range of significance carried by language, extending, in
rhythm for example, to what is sometimes an effect, rather than a definable meaning.
2.5.8.3 Semiosis
Phonology is close to the physicality of language and its evolutionary origin; stress
and tone, especially in their marked uses, have clear similarities to shrieks and other
involuntary vocalisations, where volume and high pitch, for example, carry signifi-
cance as “indexes” of the states that cause them, as smoke is an index of fire. In a com-
parable way, certain phonological patterns are indexes of the social group to which
the speaker belongs, as we noted in the §2.5.3. In several places, it has seemed appro-
priate to describe phonological features as signals – like monkeys’ danger signals, and
birds’ territorial calls. The unmarked uses of the four phonological systems, however,
are conventionalised and formed into paradigms; so they are symbolic. (They are not
arbitrary, however, being motivated in the ways described in the section on phonetic
meaning.) We see that meaning is carried by different kinds of sign, which is likely to
affect the nature of meaning and its structure.
2.5.8.4 Nature of language

It is impossible to see tone, tonality, tonicity and rhythm as primarily concerned with
symbolising conceptual meanings. Here, language is not representation, or the for-
mulation of knowledge. It is a matter of what the speaker is doing to affect the hearer.
From the speaker’s point of view, it is action, and from the hearer’s point of view it is
responding to the speaker. It is hard to formulate phonology or its significance as a
system, from a strictly linguistic viewpoint. Since phonology is basic, we should per-
haps conclude that the basic nature and function of language is action.
2.6 Conclusion: Semantic structure in the strata of English
Structures. The chapter has identified several semantic structures in English. Words
form a network structure. Morphology and syntax build a different structure, with
constituent units arranged in a hierarchical structure. Phonology is quite different
from the other strata in having semantic elements of different kinds; it does not as a
whole fit any particular structure; and it also creates field and wave structures which
carry meaning, and which are quite different from networks and hierarchies. (The
chapter has not discussed the obvious contribution of graphology to the transmission
of meaning.) The higher strata do not simply facilitate the expression of meaning car-
ried by lower strata, but carry additional meaning.
As well as those structures, which need to be set out fully, we have noted several
structural issues which need special study. There are binary relations of various sorts,
constituting structure per se, as an abstract quality: structure, distinguished from
(specific) structures. There are unstructured units, such as holophrases and interjec-
tions. The nature of information structure is not clear (yet); nor is its relation to other
structures.
Useful concepts. The chapter has introduced several concepts which will be developed
and applied further in the following chapters. Examples include types of meaning,
such as conceptual and emotive; dimensions of meaning, such as basicness and spec-
ificity, and semantic classes. It is useful to approach meaning from both functional
and structural approaches, but also from semiotics. There are several important dis-
tinctions to be made: between different types of significance (such as content and
grammatical meaning, the effect of an utterance, and what it shows of the speaker);
between the cognitive and the linguistic in meaning; and between the speaker, hear-
er and system points of view or “aspects”. Use, as in marked and unmarked use, is
distinct from structure, but closely related to it. Those concepts will be developed in
Chapters 3 and 4, and applied in later chapters.
Further analysis needed. While the chapter has reached those useful conclusions, sev-
eral issues have been raised which are not very clear in the literature on semantics.
Specific issues include, for example, the relation between transitive figures and copula
figures, the number and nature of Processes, and the mechanisms that make meaning
compositional. They will be discussed in Chapters 6 to 10.
More general issues include the following. What are the relationships among in-
tention, function and specific meanings? What is the relation of surface meaning to
underlying knowledge? Since syntax and phonology “realise” meaning so differently,
how can realisation be characterised adequately? What is the range of semiotic signs
in English, and how do they affect semantic structure? Finally, as to language itself:
we have seen that language is both interaction and transmission of information, so
what else is it? Most of those issues will be given preliminary discussion in Chapters 3
and 4; most will be amplified in Chapters 6 to 10, and realisation will be discussed in
Chapters 11 and 12.
Looking ahead. The next two chapters will develop distinctions and concepts intro-
duced here, and some others, to provide a full basis for the discussion of semantic
structure in the chapters that follow.
Chapter 3
Basis of semantic structure
3.1 Introduction
This chapter is intended to set out the basis on which the semantic structure of Eng-
lish rests, developing points from the last chapter. It will be complemented by the next
chapter, which outlines the elements of semantic structure. The two chapters will then
form the foundation for the rest of the book.
Section 2 deals with intentions, which are the ultimate foundation of our utter-
ances, and Section 3 deals with the language functions which begin the differentiation
of the intentions into linguistic form. Section 4 deals with linguistic and semiotic
constraints which hinder or help the realisation of the intentions in semantic form.
Semantics is usually presented from the “bottom” up, as consisting of conceptual
units which are organised into propositional form – semantics is seen as a combina-
tion of lexical semantics and argument structure. The approach here is from the top
(intentions) down (to senses), primarily because semantics is better understood that
way, and secondly to highlight the importance of this complementary approach in
making clear facts about semantic structure which are not evident in the bottom-up
approach.
3.2 The intention to speak
3.2.1 Main intention
Semantic structure is based in various ways on the intentions we have when we speak;
compare Levelt (1989). Some of our main intentions have been formalised and gram-
maticised for us in advance. That is, they are given grammatical not lexical form:
declarative, imperative, interrogative and exclamatory syntax, for example, create se-
mantic structures which formulate the intentions to state, command and question.
Stating and questioning both relate to conveying information, which is general-
ly taken as the dominant or even exclusive function of language, and as controlling
its structure. That standard, philosophical tradition of linguistics is illustrated by
Wierzbicka (1996), who regards meaning as consisting merely of “conceptual struc-
tures encoded in language” (1996, p. 8).
To those concerned with language as it is used in living, that is unduly restric-
tive. The very existence of the imperative implies that language is often action;
speakers intend to affect hearers’ behaviour, through the action of speaking. Clark
(1996, pp. 148–149) goes further, regarding meaning as “joint action”. The linguistic
literature shows language in other guises. Bloom included not only the response but
the situation that prompted it, as the meaning (1933, p. 158); that fits children’s earli-
est use of language, and what many linguists take to be the evolutionary origin of lan-
guage. Malinowski (1923, p. 315) saw some language as social “communion”, without
any more specific intention, and often without content. It may have other social func-
tions, ranging from entertainment to enhancing the speaker’s social status. For many
linguists, meaning and functionality are equivalent, as for McGregor (1997, p. 30),
and Croft (2004, p. 28), who defines linguistic symbols as pairs of “form and meaning
or function” (emphasis added). Moreover, among the various main intentions, the
expressive or exclamatory intention is basic in individual development (in infants’
cries), and if language developed from primate communication, then expression or
warning was presumably basic evolutionarily.
Meaning intentions, as represented by those grammaticised in syntax, grade in
their degree of differentiation of the speech situation. The expressive or exclamatory
intention involves only the speaker; the intention to command involves speaker and
hearer; the intention to convey information involves speaker as grammatical first per-
son, hearer as second person, and things talked about as third person. The intentions
themselves become more distinct accordingly – more conscious and deliberate.
3.2.2 Subordinate intentions
When we speak to carry out our main intention, we inevitably carry out some sub-
ordinate, less conscious intentions as well. We are forced into some of them by the
nature of English. For example, English requires that we inflect verbs for tense, and
that we specify nominals for number and definiteness, thereby helping the hearer’s
interpretation. Guiding the hearer is thus a subordinate intention whenever we speak;
it will be given a good deal of attention in what follows, partly because of its inherent
importance, and partly because it has been inadequately studied in the past. Again,
both vocabulary and syntax have levels of formality; even if we avoid words and con-
structions that are markedly formal or informal, we will still set a level of formality by
the complexity of sentence structure – or by the lack of sentences. Moreover, English
provides that, unless we take other measures, the end of the utterance will be most
salient; the result is that showing the relative salience of different parts of the content
is an almost unavoidable secondary intention.
Other subordinate intentions are optional, and more deliberate. We adopt such
intentions as being clear in pronunciation and syntax to ensure success of our main
intention. We are likely to conform to social expectations that apply to everything we
do, such as being polite, or that are specifically linguistic, such as making clear the re-
liability of what we say, in the semantics of modality. We may intend to be discursive,
Chapter 3. Basis of semantic structure 27
which results in a descriptive semantic and syntactic structure, or we may intend to

do no more than identify referents, which results in a spare semantic structure – or
even in semantic emptiness through the use of deixis and naming. We often have
more personal intentions, such as being vivid or expressive. These intentions include
“motivations,” as in work by Silverstein (as cited by Du Bois 1987, p. 849), Croft (e.g.
2003, Chapter 3), and Haiman (1983) on competing motivations such as economy
and forcefulness, and Haspelmath (2007) on “extravagance” as a motivation.
3.2.3 Discussion
The relevance of intentions to semantic structure is not always obvious. It is only

occasionally that they are rendered explicit, as in “I want you to leave,” and “I’d like
to know whether you’re ready”. Imperative and interrogative mood make intentions
central; but the intentions are then implicit in grammatical structure, and without
expression in individual words. Intentions may be carried out indirectly, as with re-
quests which amount to commands but are expressed as questions – “indirect speech
acts.” Intentions are often expressed incidentally; for example, a subordinate intention
to qualify one’s certainty in making an assertion can be expressed by the intonation
contour or by a modal auxiliary such as might. The intention to express disapproval
can be conveyed by choice of synonym, as with “He was quite stubborn about it”,
rather than firm. Finally, intentions may be rendered without grammatical structure
or semantic content, as in “Hello!” and “Goodbye!”, whose significance is no more
than their intention.
The importance of intention in speech is shown in neurolinguistic research by
Reverberi et al. (2015). They showed, in studying bilinguals, that forming the inten-
tion to speak in a particular language uses an area of the brain distinct from the area
for actually speaking the language, an area associated with a high degree of cognitive
control of action.
3.2.4 Conclusion
The semantic structure of English to be set out in the following chapters must provide
for the realisation of all kinds of speaker intentions, and for their realisation in differ-
ent ways, as discussed in the previous subsections. The effect on structure will some-
times be obvious, as in social formulae and extremely formal or informal semantics,
but will often be much more subtle.
3.3 Functions, as the formulation of intention
3.3.1 Introduction
The primary intentions of giving information and explaining things are naturally em-
bodied in similar forms of expression, different from the forms natural to entertain-
ing, boasting and sharing feelings. Similarly, the secondary intentions of guiding the
hearer take yet another set of forms of expression. The intentions thus group together
in “functions,” which mediate the formulation of intention into utterances. Just as
utterances commonly arise from both major and minor intentions, so do they com-
monly serve two or even three functions at once.
I classify the functions into the Expressive, interpersonal and ideational. That is
based on the “metafunctions” of Systemic Functional Grammar (see Halliday 2014,
§1.3.5), but excludes the Textual function, and adds the Expressive one. It effective-
ly coincides with those of Lyons (1977, p. 50): the expressive, the social, and the
descriptive.
Some language intentions and functions will not be considered here or later, since
the book as a whole is not considering pragmatic and sociological issues, as noted
in Chapter 1. They include such social functions as opening verbal interaction, and
verifying that communication channels are operating (Couper-Kuhlen, 1993, p. 231,
citing work by Schegloff).
3.3.2 Expressive function
By “Expressive” function, I mean the vocalisation of affect, seen most clearly in grunts
of effort and cries of pain, pleasure or amusement. Put simply, affect is undifferen-
tiated emotive response to a situation; more precisely, it can be seen as the psychic
state that exists on the dimensions of pleasure-displeasure, tension-relaxation and
calm-excitement. (See Matthews et al. 2003, p. 89, citing Wundt.) It does not come
from cognition, but from a different “faculty” or “module” of the mind.
Considered narrowly, affect is “unmonitored, purely physiologically determined
externalisation” of mental states (Couper-Kuhlen 1986, Chapter 10, §1). Considered
more broadly, it includes a range of states, varying in the degree to which people’s
affective responses are processed by their cognitive and value systems. So the term
covers several things: undifferentiated tension which is discharged in action or ges-
ture (rather than expressed in language); generalised feeling, such as pleasure and
dislike (expressed in such words as horrible, nasty, nice); and specific emotions such
as jealousy, envy and resentment. (Compare Taylor’s distinction (1980, p. 385) be-
tween unsophisticated emotions such as anger and joy, and sophisticated ones such as
shame, humiliation and guilt.)
In common understanding, expression of affect serves to release the tension

or displeasure, leaving us feeling calmer. According to recent research (Swee and
Schirmer 2015), crying out when we are hurt is not only an evolved signal of dan-
ger, but reduces pain, because the effort required interferes with processing the pain
signals. Expression is the most basic of the metafunctions in that evolutionary way,
and in several others: it develops first in children; it relates to the speaker alone – no
hearer is presupposed; it does not involve the higher mental faculties of judgement
and intellect; it requires very little linguistic ability, or none; it is reflexive to some
degree, rather than deliberate.
Grunts and cries are sublinguistic; but some such vocalisations of affect have be-
come conventionalised, as ha ha for laughter, whew and bah (OED), and so on. Those
interjections come from the Expressive function alone; they may make up the whole
utterance, and may be spoken without any other utterance; more often, perhaps, they
are part of a longer utterances, or alternate with them. Often, speakers combine the
Expressive intention with a second intention, that of showing those present how they
are feeling – in the interpersonal function. That pattern of combining the Expressive
function with the interpersonal and ideational functions occurs in many forms other
than interjections. For example, we mix expletives into descriptive statements; we use
expressive marked intonation to release our feelings while making factual statements;
we lengthen vowels and aspirate initial consonants. The Expressive function thus mo-
tivates a good deal of emotive language, along with the interpersonal function to be
discussed in the next section. (Paralinguistic features commonly serve the same func-
tion, as do many gestures, especially those that have no conventionalised meaning.)
This function has been relatively little discussed, and seldom appears in lists
of language functions, although linguistic recognition of it goes back at least to
Wittgenstein, who noted that expression “escapes” from the speaker (as cited by
Lascaratou 2007, p. 24). It is supported by Wharton (2012), who suggests a spectrum
ranging from the gag reflex, through “ugh” (with gh representing an aspirated gut-
tural sound), and the interjection “Yuck!”, to yucky/yuckier/yuckiest, which are now
standard words (2012, p. 582). It is supported also by Wray (2002, p. 64), who gives a
similar gradation. It is supported indirectly by Cruse (2011, §3.4.1), who notes that an
expression such as gosh “expresses an emotional state”, and that it has no truth value
but has “validity” as a representation of the speaker’s current state. He compares such
utterances to a cat’s purr and a baby’s cry.
While Expressive vocalisations such as cries and laughter are not linguistic and
are outside the scope of semantics, words such as yucky and the use of features such
as intonation and aspiration for the Expressive function are equally clearly within the
scope of linguistics. The range and nature of English semantics will be misunderstood
if we ignore them. However, as will be clear from the examples given so far, they
rarely have a structure, being unitary; so they will receive less attention than the other
functions.
3.3.3 Interpersonal function
3.3.3.1 Introduction
Most use of language is functionally less basic than Expressive language, since
it involves a hearer as well as a speaker, and its function is social, or interpersonal
(Halliday 2014). Language, said Firth (1957, p. 181), is “part of the social process.”
Hanks (2013, p. 409) says that “the fundamental purpose of using words is to facilitate
human social behaviour.”
There are many social functions of language, so I divide this section in three,
following distinctions made in Functional Discourse Grammar as set out by Van de
Velde (2007).
3.3.3.2 Interpersonal action

As the literature on “speech acts” has made clear, all utterances constitute an action
on the speaker’s part, such as stating, asking, commanding or exclaiming. Those ac-
tions call for a corresponding action on the hearer’s part, of accepting the statement,
answering the question, obeying the command, or sharing the feeling expressed in the
exclamation. There is thus a two-way transaction, as in a commercial transaction, in
which there is both buying and selling. The interaction follows conventions that are
linguistic but are set socially, by the speech community; that social dimension extends
from linguistics as conceived here, into the field of conversation analysis and dis-
course analysis – where interaction becomes joint action, like ballroom dancing or us-
ing a two-handed saw (Garrod and Pickering 2007, §26.1). This interpersonal-action
function also includes the speaker’s intention to guide the hearer, in such ways as
using articles to guide the hearer in identifying referents, setting up Topic-Comment
structure to guide structuring the narrative or argument, and foregrounding some
things while backgrounding others to guide evaluation of significance.
3.3.3.3 Personal relationships

Speakers often set up a personal relationship with hearers, inviting them to share feel-
ing, or to adopt and share the attitude expressed. Here, there is not so much an action
on the hearers’ part, as a response, and there are no set conventions for hearers to
follow. This is the realm of emotive wording, of the subtleties of intonation, and so on.
3.3.3.4 Social relations

Intentions such as setting or enhancing social status and establishing group, class or
regional identity belong in the third class of interpersonal function, that of social
relations. They are realised in such forms as the syntax and diction that set the degree
of formality, and in slang and dialect wording. This function is guided by socially set
convention, in such matters as how a person of higher status should speak to one of
lower status.
I include here what Givón (1993) calls the aesthetic function, and what Marchand
(1969, p. 427), among others, refers to as being playful. It includes being funny or
otherwise entertaining, playing with words, and creating works of art. It is the realm
of the imaginative use of words and syntactic structures, in figures of speech for ex-
ample. Sometimes, any Expressive and informational value there may be in the text
is subordinate to the imaginative function, but in other texts the aesthetic element is
like decoration in a functional room. Utterances in the personal-relationship function
sometimes work this way, but I include it here as social because the roles of entertain-
er and audience are set socially.
3.3.4 Ideational function
In the ideational function, we offer information about the world, but do not try to
change the state of the world, as we do in the interpersonal function (by inducing
the hearer to act); we embody our construal of experience, both personal experience
and experience mediated through learning. It involves not only speaker and hearer –
expressed in the grammatical first and second persons – but also the things talked
about – in the grammatical third person. It is less basic than the other functions in
that respect, in being the last to develop in the child, and in requiring integration of
language not only with feeling and interpersonal relationships but with perception
and intellect.
It is the realm of statement, description, reference, and truth value in propositions
stated. It uses transitivity, dependent clauses and the other structures of syntax fully,
but hardly needs intonation, for example. The nature of its relationship to cognition is
controversial, but it is certainly close; and we may note Barsalou’s statements, report-
ing other research, that “control of conceptual representation via language appears to
be central” to representing “non-present situations” (2012, p. 244), and “conceptual
processing relies on varying the mixture of the linguistic and conceptual systems”
(2012, p. 252). (The relationship between ideation and cognition will be discussed
below, especially in §4.2.2 and §4.7.3.)
3.3.5 Conclusion
It is a truism that the function of language is communication; but that formulation

is so vague as to be useless for our purpose. When it means “conveying information
about the world”, as it often seems to, it is quite inadequate and in fact misleading, in
limiting language to the ideational function.
There is considerable variation in the number and identity of functions given by
various authors; but there is, I believe, general agreement that language is functional,
since linguists who do not mention the fact do not deny it, but apparently assume it,
and it has been discussed since at least Wegener, writing in the 1880’s (Geeraerts 2010,
p. 21). I claim validity for the threefold account given here for three reasons. First, it
provides powerful explanation of many things in language, as the rest of the book
should show. Second, it correlates with the three fundamentals of the speech situation
and of the grammar of predication: the Expressive function correlates with the speak-
er and the first person; interpersonal function correlates with the hearer and the sec-
ond person; the ideational function correlates with what is spoken about and the third
person. Third, it matches the stages of human development. However, it is proposed
here only as a useful formulation that gives coherence to the exposition to follow, not
being intended as a semantic structure, or as part of the argument for other structures.
3.4 Constraints on meaning
3.4.1 Introduction
Since meaning in English is conveyed by linguistic signs, the meaning and its struc-
ture are constrained by what the signs can convey – by the linguistic and semiotic
systems. This section deals with constraints in the nature of those systems.
3.4.2 Linguistic constraints
Where speaker abilities provide an internal constraint on the meaning expressed, the
nature of language provides an external one. That nature may be expressed as fol-
lows. (This view follows most thinking in linguistics, but will be qualified in the next
section.)
Our intentions to speak can cover the whole range of human experience – in prin-
ciple, there is nothing that we cannot speak about; but meanings must be expressed
in a single medium, whose nature constrains the structure of what is expressed. Our
experience, as a mental representation, is embodied in the multidimensional nerve
network of the brain; but the meanings that express it must be embodied in a linear
flow of speech or writing.
Psycholinguistic research to be presented later shows that meanings as we produce
them in the brain consist of a continuum of excitation of the neurons. But because
language consists of units (such as phonemes, morphemes and syntactic groups), the
continuum must be reduced to a flow of units. A further consequence is that mean-
ings which represent scalar concepts, and feelings that occur in varying degrees, have
no congruent form of expression in lexis or syntax, but do have it in phonology, in
which the various systems are all gradient.
Since speech is produced as a single flow, there is no way of representing directly
any parallel mental processes, as in the combination of ideational and interperson-
al functions, or simultaneous intentions. Such parallel things must be represented
indirectly: they must be expressed successively, with some indication to the hearer of
how to restructure them, such as co-ordinating conjunctions.
There are, however, some important features of language which weaken those
constraints. Although speech as sound or writing seems one-dimensional, being lin-
ear, English has found a way of making the embodied meaning two-dimensional, just
as hierarchies have a horizontal and a vertical dimension. For example, consecutive
physical words operate as a “horizontal” dimension, and the abstract group that they
constitute operates in a “vertical” dimension, creating syntactic structure. English
provides ways for the speaker to signal to the hearer that the linear structure must
be restructured into those two dimensions; for example, it assigns some words the
syntactic status of modifiers and others the status of heads, defining their place in
the hierarchic group. Those syntactic devices, and the availability of different types of
meaning in one word, provide ways in which parallel lines of meaning can be indi-
cated indirectly. Those linguistic devices are aided by our mental capacity for holding
chunks in working memory while we restructure them, effectively limiting the line-
arity of language. Above all, the enormous range of expression that English now has,
outlined above in section 2 on functions, empowers us by providing a great range of
semantic structures, to mitigate the linguistic constraints.
Finally, language relies on conventions – tacit agreement among its speakers –
which require certain forms and structures. For English, that includes the require-
ment that verbs inflect for person and number, for example: certain meanings must
be included.
We conclude that the nature of English constrains meanings and their structure,
both as to what speakers can mean, and as to what they must mean.
3.4.3 Semiotic constraints
Semiotics constrains the subtlety, range and structure of the meaning we can convey;
yet it has been given little attention in most schools of linguistics. I devote consid-
erable space to it in the chapters to follow, and accordingly I set out fully here the
semiotic understanding I will use.
I follow the familiar tradition of Peirce (1931–1958) in identifying indexes, icons
and symbols as important types of sign. Indexes have a natural and typically causal
linkage between the “signifier” and the “signified”, as with a footprint being a sign of
someone’s recent presence. With icons, the linkage is through resemblance; icons may
work by imaging e.g. in sketches, or work diagrammatically e.g. in maps, or work
analogically e.g. in metaphors. With symbols, the linkage is conventional (based on
agreement among the users), and arbitrary in the sense that the sign could be re-
placed by a quite different one, as the word tree could be replaced by arbor (Saussure
1915). That taxonomy of sign types is useful; but it needs to be supplemented, and, as
Nuckolls asserts (1999, p. 228), “clear-cut types are practically non-existent”, and
Peirce stressed that there are no pure symbols (as cited by Kravchenko, 2007, p. 655).
3.4.3.2 Semiotic processes

There is a preliminary issue to be dealt with, however, before we deal with the sign
types listed above. As Cruse notes (2011, §16.2.3), there are three ways for a speaker
to identify a referent: naming, pointing (deixis, in this context), and describing. The
difference is semiotic, and is a matter of the mental process used, not of the sign type,
since all three use a conventionalised sign. The difference in process goes with impor-
tant differences in semantic structure. Describing – with words or other signs carry-
ing their own content – provides complex structure. Naming provides the hearer with
a labelling word, but not with any content, and therefore without any semantic struc-
ture. That applies to proper nouns, and, as in the tradition associated with Kripke, to
some uses of common nouns. Pointing with a demonstrative (“that book”) provides
neither name nor content; pointing with a pronoun (“He…”) provides no content
directly, but provides it indirectly through the noun to which the pronoun points.
The processes can be seen as each having a unique function, but speakers very
frequently have a choice between them. We may identify a certain person by name
with “Joe Bloggs”, by deixis with “that man over there”, or by description with “the tall
man with the overcoat”. We may choose to give an instruction by using names, as with
“Put the nut on the bolt”; we may give it descriptively, as with “Put the small hexag-
onal bit on the long threaded bit”; or we may give it deictically, as with “Put that on
there.” Consequently, the processes will often be referred to hereafter as the semiotic
“strategies”.
3.4.3.3 Semiotic levels: Types of linguistic sign

Introduction. Meaning is conveyed in language on different levels of distinctness. The
meanings usually studied in semantics are quite distinct, and their signifiers are easily
recognized. I will call such signifiers “signs”; they denote concepts or something else
equally distinct, such as a definable feeling, or a grammatical meaning. On a lower
level of distinctness are “signals” (as in monkeys’ warning calls); they indicate situa-
tions or “states of affairs”. On the lowest level are “markers”, whose significance is not
well-defined at all; to gain sufficient distinctness, a marker needs something further –
additional markers creating a pattern, an accompanying symbol, or indications in
the context. Those definitions involve two elements: first, the semiotic level of what is
signified (the degree of distinctness); and second, the semiotic nature of the relation-
ship between signifier and signified – that of signs’ one-to-one denotation, signals’
one-to-many indication, and markers’ associative and many-to-many effectiveness.
Signs. Indexes that most would regard as fully linguistic are relatively rare, and gen-
erally not clear-cut. However, the conventionalised vocalisations such as “Ow!”, men-
tioned in the section on the Expressive function, are indexes; uttered spontaneously
when we are alone or otherwise not intending to address anyone, they constitute an
index of pain. Many uses of such interjections are only partly indexical, since we add
to the spontaneous impulse a deliberate intention of letting hearers know how we feel.
(We also use nonlinguistic indexes, such as facial expression, gesture, posture and so
on, conveying such states as boredom and nervousness, and paralinguistic features
such as voice quality.) Indexical meaning will be of some importance in what follows.
Icons, the second type of sign, have been studied by some linguists, such as
Haiman (1983) and Croft (2003). An example is Caesar’s “I came. I saw. I conquered”,
where the order of the sentences is an icon of the order of events. I will not often
describe meanings as iconic, since I believe that many apparently iconic meanings in
English are in fact symbolic. For example, the premodifiers that are closest to the head
are usually words that are “closest in meaning” to it, and thus seem to be iconic; but
the generalisation has a number of exceptions, so the position is in fact conventional
and symbolic, although conceivably iconic in origin. On the other hand, much of
phonology carries its meaning iconically, as we saw in Chapter 2, §2.5.7.1. There is a
general reason why iconicity is little used in language: effective iconicity, as in maps
and wiring diagrams works in two dimensions, whereas individual linguistic signs
must work in only one dimension.
Symbolic meaning will be by far the commonest studied here. In language, there
is characteristically a range of meaning elements for a sign, with some being evoked
only rarely, some frequently, and some always; for discussion of the last-mentioned
“invariant” meaning, see Sign-Based Analysis, as in Tobin (1990) for example, and
“sign-based theory” as in Reid (1991). It is worth distinguishing a second meaning of
“symbol,” in which the symbol denotes not only the thing signified but also things as-
sociated with it. In mediaeval literature, “the holy grail” had a straightforward referen-
tial meaning at first, but came to be a symbol in this second sense; similarly, “frontier”
has gained symbolic meanings in the USA, and “(British) bulldog” is symbolic for
many Britons. Such general symbols have a more complex semantic structure, which
will not be examined further, being social rather than linguistic.
Many symbols have a marked use, in addition to their regular or unmarked use.
Inversion of Subject and Predicator in unmarked use signifies the interrogative; in
marked use, as in “Came the dawn,” it is a signal of literary tone. The indicator of
its new significance is that in the context the inversion can not have its usual signif-
icance – a fact which hearers must work out before they can work out the intended
significance; there is a convention that conventions maybe overturned. That compli-
cation seems typical of the semantics of English: over the centuries, users have devised
more and more ways of assigning meaning to a limited number of signs.
Signals. I have added signals to the traditional classification of signifiers because ex-
plaining the semiotics of language needs that concept. It is taken from evolutionary
biology, and denotes behaviour that conveys information, and that has been shaped
by natural selection for that function (Wharton 2012, p. 571). Examples include hon-
ey bees’ dance and bullfrogs’ calls (Wharton 2012, p. 572), or a dog’s angry growl.
In nature, signals elicit a response; the dog’s growl will elicit aggression or fear. In
humans, a toddler’s raising its arms or saying “up” will elicit sympathy and the action
of picking the child up. Rhythm can usually be seen appropriately as a signal of excite-
ment, bringing a response of corresponding excitement.
In language, however, the response commonly includes conceptualisation – a
meaning – as well as the affective and physical response. With signals, that meaning
is holophrastic, as with the toddler’s “Up”, being (for the hearer) an understanding
of the situation that prompts it; perhaps we should say that it has various potential
meanings, and no specific actual meaning. Although that makes signals distinct from
signs, it will be convenient at times to treat them as a fourth type of sign, using that
word in a looser sense.
Certain intonation patterns signal a warning; emphatic phonology in the utter-
ance of a statement like “I will” signals a promise. I suggest that the concept of sig-
nalling gives a better account of those expressions than the concept of “speech acts”.
Markers. Markers, like signals, have been added to the semiotic taxonomy simply be-
cause they are needed to explain language. They serve to define signs or signals. For
example, full stops are markers which mark off sentences, which are signs, but may
mark off abbreviations. Capital letters mark both sentence beginnings and proper
nouns. In writing, spaces mark off words. As those examples illustrate, markers are
often ambivalent; and as noted in the introduction, the hearer often needs several
markers to be sure of their significance – they are cumulative. Moreover, markers are
sometimes missing where we might expect them. Although they are unlike signals in
being conventional rather than motivated, they are like signals in having an “effect” or
a “significance”, rather than a “meaning”.
Discussion. It was said above that the types of sign are not always clear-cut. An ex-
ample worth noting is that the types of sign can coincide. Indecent and blasphemous
words are often used in the Expressive function (“Oh shit!”, “Christ!”), and are taken
by those who overhear them as indexical signs of the speaker’s state; but they usually
retain for the hearer their original significance, so they work as symbolic signs as well.
Similarly, different semiotic processes can coincide. An acronym like “NP” is natu-
rally read as a name, with the reader going directly to its reference without bringing
descriptive concepts to mind, but it can be read descriptively, by mentally expanding
it to “Nominal Phrase”, with its constituent concepts made fairly conscious.
Arbitrariness has been much discussed in the literature on linguistic semiotics,
but will be given little attention here. Symbols may fairly be characterised as arbi-
trary, but some elements of motivation apply even there, as with grammatical items
being short and content items being longer. Fischer and Nänny (2001) gives useful
discussion; see also Halliday and Greaves (2008, p. 168). Indexes are obviously fully
motivated, and signals are fairly fully so. Arbitrariness is irrelevant to our concerns.
The discussion above assumes that “Meaning implies choice” (Cruse 2011, §4.3).
That is why individual phonemes and letters are in principle meaningless – we cannot
choose between “meaning” and “meening”. Signs occur in paradigms.
3.4.3.4 Forms of linguistic sign

The most familiar linguistic signs are words, but they come in many other forms, as
the preceding discussion has shown incidentally. To understand semantic structure,
we need to be aware of all kinds of meaning, and therefore of all forms of sign.
Signs may be overt or covert. Overt signs include free and bound morphemes,
punctuation, and phonological features such as stress and pitch, for example. An overt
sign can include covert signs. For example, although the conventional doctrine of
double articulation (cf. Hockett 1960, p. 90) denies it, individual phonemes within
a morpheme can be signs of expressive meaning, as I showed in Chapter 2 (§2.5.6).
The letter or phoneme is overtly part of the morpheme as sign; its expressive meaning
is covert.
The most distinct covert signs are arrangements of overt signs. That includes the
order of words or groups, as when the inversion of Subject and Predicator signals
the interrogative. Position can also be a sign, as with the position of premodifiers
and determiners in nominal groups, which is a sign of their being subordinate and
of being modifiers. (Some linguists would include constructions here, as meaningful
arrangements of words; I do not, primarily because I do not consider “construction”
to be a useful precise term – though I will continue to use it in its traditional general
meaning.) The absence of a sign can be a sign, as when the absence of a Subject signals
imperative mood. Those signs are quite abstract, where words are concrete.
The strata of English have characteristic forms of sign. In general, morphology
and lexis use segmental signs (phonemes, morphemes, words); syntax uses the ar-
rangement of segments (word order); phonetics uses features of the segments (pho-
nemes, or morphemes perhaps). Phonology uses suprasegmental signs. All of those
strata, and most of those forms, have unmarked and marked uses of their signs.
As forms of sign, we have noted: the overt and the covert; arrangement, position
and absence of overt signs; segmental and suprasegmental signs; and unmarked and
marked forms of those signs.
3.4.3.5 Conclusion: Semiotic constraints

We should keep these types and forms of sign in mind in the following chapters; se-
mantics and semantic structure will look rather different from the simplified seman-
tics we see when thinking only of words and clauses. It is not an easy task, however,
since the types and forms are not used very systematically.
3.5 Conclusion: Basis of semantic structure
This chapter has set out features of language and of English in particular that are basic
to semantic structure. Most basic are our intentions in using language, which flow
from all areas of life and all parts of the mind. As well as the basic intention, there are
secondary intentions, some of them independent of the basic one, and some entailed
in carrying out that basic one. Our intention is further differentiated into linguistic
functions, which have been formulated here as the Expressive, the interpersonal and
the ideational functions.
Our use of the functions is constrained by the nature of language and its semi-
otic system. The chief linguistic constraint is the linear structure of speech, which
semantic structure must overcome. The main semiotic constraint is that the multidi-
mensional continuum of meaning must be represented by discrete physical signs, in
speaking and writing. Counterbalancing those constraints is the enormous range and
flexibility of expression that English has developed.
Those intentions and functions, subject to those constraints, constitute the basis
of semantic structure in English. The next chapter will extend it to the elements that
are used in the structure.
Chapter 4
Elements of semantic structure
4.1 Introduction
This chapter’s purpose and place in the book. This chapter is the second of two that will
provide a basis for the rest of the book. Like the previous one, it develops points made
in Chapter 2, but introduces others in order to make the basis more complete. The
previous chapter presented the several bases of semantics as being speakers’ inten-
tion to speak, the functions their utterance will serve, and the linguistic and semiotic
constraints that apply to utterances. This chapter considers the elements of semantics
which are available to speakers in structuring what they say. It therefore starts afresh –
from the opposite starting point, that is. Instead of working from the most general, the
speech intention, the chapter starts from the most particular, the basic elements from
which semantic structure is built up. That reflects the fact that when we start talking
or writing, the first thing we say is one particular word, not our general intention.
Outline of the chapter. The main content of the chapter deals with types of meaning,
the dimensions along which the types vary, and classes of meanings. There is a prelim-
inary issue to be dealt with, however: it is thought by linguists and lay people alike that
an important part of what we call “meaning” is knowledge, independent of language.
If that is so, we must distinguish that cognitive meaning from what is linguistic, since
it is the intention of this book to deal with semantics as a linguistic matter, not to
study knowledge as such. That preliminary issue is dealt with in the next section, §4.2;
then follow sections on the dimensions of meaning (§4.3), types of content meaning
(§4.4) and of grammatical meaning (§4.5), semantic classes (§4.6), and the uses of
meaning (§4.7). The chapter will thus incidentally make much clearer what is meant
by “meaning” and “semantic” in the book.
4.2 Preliminaries
4.2.1 Areas of meaning: Cognitive and linguistic meaning
Introduction. Before considering the elements of semantic structure, we must consid-

er the distinction I will make between cognitive and linguistic meaning, because the
elements of structure are different in those two domains.
Other linguists’ views. It is now almost universally agreed (Ferreira and Slevc (2007,
§27.5.1) that the conceptual systems of knowledge and language are distinct but
closely related, although until recently a number of linguists thought otherwise (§1.3
above); but different scholars have held a wide variety of views on the issue. Some
have upheld a strict distinction, asserting that there is a “watertight border” between
linguistic and extra-linguistic or “encyclopaedic” knowledge (Mel’cuk 2012, p. 89).
Some seem to eliminate the distinction, asserting that language and knowledge are
both conceptual and cognitive (e.g. Langacker 2005, p. 164). Others have held that
they form two levels or aspects of meaning (e.g. Bierwisch and Schreuder 1992), or
have concepts from knowledge used within language (e.g. Hudson and Holmes 2000).
According to Cruse (2011, §5.1.2), some have held that grammatical meaning is lin-
guistic, and that content meaning is cognitive.
View taken here. The view taken here, like that of Hudson and Holmes (2000), is that
there is indeed a distinction between cognitive and linguistic meaning, but that they
overlap in one “area” of meaning. (I am taking “cognitive” to mean ‘to do with knowl-
edge’ – knowledge of the world – although it sometimes seems to mean ‘to do with
any mental activity’). By analogy, the post office’s distribution system and your system
for collecting your mail can be distinguished, but your letterbox is in an overlap area.
The need for a distinction seems clear from the following facts about English.
Grammatical items such as the articles, pronouns and demonstratives are meaningful
but do not convey any knowledge. Pig and swine are different linguistically, although
they are the same cognitively, since they denote the same type of animal. Proper
names like Plato are meaningful in some way, but who Plato was is not part of the
language; they are the converse of grammatical items, having cognitive meaning but
no significant linguistic meaning. Active and passive forms of a sentence convey the
same information, but are different in their significance linguistically.
On the other hand, consider SOED’s “meaning” of the noun, horse: “A solid-hoofed
perissodactyl ungulate mammal, Equus caballus, having a short coat and long mane
and tail, native to central Asia….” Short, native and central denote meanings (‘short’,
‘native’, ‘central’) that must be part of the English language, and they must equally
denote concepts (SHORT, NATIVE, CENTRAL) that are used in knowledge of the world.
I concur, then, with what I take to be the widely held view that linguistic mean-
ing (e.g. ‘short’) and cognitive knowledge (e.g. SHORT) are areas which can be dis-
tinguished, and which overlap in that both use concepts which make up the area of
overlap. Semantic structure mediates between conceptual structure and expression in
morphosyntax and phonology; its study needs both “cognitive semantics” and “lin-
guistic semantics”.
Explanation of the distinction. The principle is simple, but the application is complex,
because meanings vary in their degree of dependence on cognitive concepts, and in
the way the concepts are used. The issues are important in semantic structure; accord-
ingly I give more explanation here.
Chapter 4. Elements of semantic structure 41
The relationship between the cognitive and the linguistic can best be understood,
I believe, through psycholinguistic research on language learning, as reported by
Gentner and Boroditsky (2001). When children begin to learn English, they already
grasp the nature of some things, such as cats and teddy bears. Words like cat and teddy
are learned very easily, since they simply name things that are already understood; the
word is attached to an existing concept; at that stage, language learning depends on
knowledge. Terms such as sister are more difficult to learn, since they are relational.
Words such as walk and run, drop and fall, are more difficult still. The child must learn
precisely which qualities of movement belong with walk and which belong with run.
That depends on greater facility with language than learning concrete nouns does. At
the extreme, learning grammatical determiners such as the, and conjunctions such
as and, does not depend at all on the child’s knowledge of the world; it is impossible
without a good grasp of language. Gentner and Boroditsky (2001) call the initial situ-
ation “cognitive dominance” in learning, and the last-mentioned situation “linguistic
dominance.” But adult use of words is more complex: if we think as nonverbally as we
can about iron and copper as chemical elements, we use the concept, IRON, cognitive-
ly; but in “a will of iron,” that concept combines with concepts such as HARD, DURA-
BLE and DETERMINED to form a meaning that it is dominantly linguistic, because our
knowledge of the world does not include any objects definable as hard and durable
and determined. (In “a will of iron”, a and of are purely linguistic.) Concepts may then
be thought of as having roles in both cognition and language, just as people may have
roles as employee, citizen and parent, without being defined by any of them.
On that view, a large majority of word meanings will have elements which have
both cognitive and linguistic roles; but it will be helpful to be able to distinguish be-
tween meanings which are so heavily cognitive that their linguistic element may be
ignored for the purpose in hand, and those which are equally strongly linguistic. That
distinction is sometimes made on the criterion of being grammatically relevant (e.g.
Fortescue 2010, p. 182) or having a “reflex” in the grammar. That works reasonably
well, but is not wholly satisfactory. For example, the distinction between argon and
xenon is cognitive not linguistic, although it has a linguistic reflex in the difference
between the word forms.
Therefore, for items with clear linguistic and cognitive elements, I will use the
criterion that an item is linguistic if it has an alternative which is a different construal
of the same intention. (That puts aside items whose significance is wholly linguistic, in
a way that will be made clear in the section on types of meaning.) Thus the distinction
between ‘argon’ and ‘ xenon’ is cognitive since there is no alternative construal into
language by another form – no other words for them; but the distinction between
‘sodium chloride’ and ‘salt’ is linguistic – they are alternative linguistic forms for the
same cognitive entity.
Construal. The concept of construal, used in the last paragraph, will be important –
central, indeed – to this book, one of whose themes is that meaning is constantly
adapted to function and context, not produced mechanically out of knowledge. It
consists of shaping something into a new form which includes the original but tran-
scends it. Its linguistic and psycholinguistic nature will become clear through exam-
ples; its neurolinguistic reality is supported by Zaidel et al. (1988, p. 71), who say that
the two hemispheres of the brain often represent the same input differently, applying
different strategies to it, such as inhibition of one side’s contribution to processing.
4.2.2 Aspects of meaning
There is a second preliminary issue to be made clear before the elements of structure
are outlined: aspects of meaning, introduced in §2.4.1.3.
The description of a house will vary, according to whether it is described from
the front, from the side, or from above; the three “aspects” are complementary, and all
are needed for a good description. Similarly, meaning has three aspects – three ways
of looking at the meaning of an utterance – which will be used from here on: those of
the speaker, the system, and the hearer. Speaker meaning is what the speaker intends
to convey. System meaning is the meaning of what is said, according to the system of
English; that is, the conventional meaning of the words, as given in a dictionary, and
the significance of the syntax as a standard system; it is the meaning studied by most
varieties of semantics. It mediates between speaker and hearer, and is appealed to
when there is misunderstanding (“Don’t you mean militate, not mitigate?”). Hearer
meaning is ideally the same as speaker meaning, in general. However, it may prop-
erly differ by extending a little further, as with allusions and metaphors intended to
be evocative; and it should extend further in the significance of speech acts, where
the hearer should respond with the appropriate action such as accepting the state-
ment or replying to the question. For hearer and speaker aspect, see also Van Valin
(2013, §4.7).
Hearer meaning, as a process, has two phases (Sanford 2008). First, the hearers
build meaning compositionally from the words, syntax and phonology of the utter-
ance, working almost mechanically. Second, they modify that compositional meaning
according to their interpretation of the hearer’s intentions, as with irony, allusions,
and metaphors. The second phase is what Wharton (2012, §28.4), for example, calls
the “inferential phase” of processing. Murphy (1988, p. 531) describes two phases, in
similar terms. Hsu et al. (2015) give neurolinguistic support for the phases (dividing
the first phase in two). What I am excluding as pragmatics is a third phase, calling on
non-linguistic knowledge.
4.3 Dimensions of linguistic meaning
4.3.1 Introduction
This section deals with linguistic meaning; that is, it does not deal with cognitive
meaning (see §4.2.1 above), although the points made may perhaps apply to cognitive
meaning.
The concept of dimension was used in Chapter 2, to explain differences in mean-
ing: sob and whimper, for example, were said to differ from weep on the dimension of
the generality. That dimension, and the dimension of degree or intensity will be quite
familiar to readers, if not by that name. They appear implicitly in the study of seman-
tic change, where words are said to “broaden”, which is change in generality. Just as
the dimensions of length, breadth and depth define a solid object, so do the semantic
dimensions define sense elements.
The dimensions apply to meanings carried by words, syntax and phonology, but
they will be illustrated largely from word meanings, since the distinctions are clearer
there, and it is to word meanings that the distinctions will be most often applied. The
first group of dimensions apply to sense elements considered individually; the last two
dimensions apply both to the distinction between different senses, and to different
uses of the same sense.
This section will not define the dimensions abstractly, but ostensively, pointing
out examples. Also, it draws heavily on Cruse (2011); I know of no other work that
accounts for these elements of meaning at all thoroughly.
4.3.2 Quality dimension
In non-technical language, the quality dimension provides the kind of meaning. It

is what makes the difference between blue and yellow, and honest and intelligent; be-
tween buy and rent; between house, flat, road and star. This dimension applies to both
cognitive and linguistic elements of semantics, in a fairly straightforward way. The
quality differences among the names for birds and animals, for example, are primarily
cognitive; the differences in meaning among fortunate, lucky, and serendipitous are
primarily linguistic.
4.3.3 Intensity dimension
Difference in intensity is difference in degree, or position along a scale. The mean-

ings ‘unpleasant’, ‘nasty’ and ‘loathsome’ differ in intensity, while being the same in
quality. Other examples are the meanings of small, tiny and minute, of small, smaller
and smallest, and of sore, painful, and excruciating. It is significant that all of those
examples are quality meanings, expressed by adjectives, since we do not usually con-
ceptualise real-life things and happenings as differing in intensity.
4.3.4 Specificity dimension
Specificity is what makes the difference between collie (most specific), dog, and animal
(most general), and between chaste (specific) and virtuous (general). Cruse (2011,
§10.3.3) distinguishes between type specificity, as in the examples just given, and part
specificity, as in the difference between toe, foot, and leg. (I am putting aside his third
kind of specificity, that of “intensity”). Taxonomies set up a scale of specificity, as do
“meronymies” such as ‘arm’, with ‘forearm’, ‘elbow’, and ‘wrist’ as parts.
The importance of specificity extends far beyond lexis, however. A major reason
for modifying a nominal, for example, is to render the meaning of the head more spe-
cific, as in “hot water” and “reliable contractor”; the whole phrase, “a reliable contrac-
tor”, is more specific than the phrase “a contractor”. Inflectional morphology makes
the number and tense of the stem specific; sometimes a whole clause or sentence
serves to specify another syntactic unit, or (shifting it in the other direction on the
scale) to generalise it.
The psycholinguistic reality of this dimension is supported by Levelt (1993),
and Flores d’Arcais and Schreuder (1987, pp. 153–154). Fortescue’s neurolinguistic
account (2009) describes percepts as having “affordances” (that is, features which
“afford” ways of beginning the conceptualisation process) as having links to the bun-
dle of the neurons representing the word at different levels of the cortex. The more
general the percept or concept is, the higher is the level at which it joins the neuron
bundle.
4.3.5 Vagueness dimension
What is meant by “vagueness” can be made clear by specifying its two types, as fol-
lows. Vagueness can consist of (a) being ill-defined; for example, if chair is defined
as “seat having legs and a back”, then it is ill-defined or vague as to having arms. It
can also consist of (b) having lax application; for example, line is lax in application in
being applied to an uneven row of people, as well as to a geometric line. The opposite
of being vague is being precise.
Like specificity, vagueness is a dimension of groups and clauses. The light verb
construction is useful because it lets the speaker leave the specific part of the expres-
sion to the end of the phrase, where it is salient – “I’ll take a shower”, “He went for a
walk” and so on. Often, a good question is vague, and a good answer to it is precise, as
in example (1), from a television interview about fishing at sea.
(1) Interviewer: “Gets a bit hairy, does it?”

Fisherman: “Yeah, pretty rough sometimes.”
(Television New Zealand, “Country Calendar”, July 2013)
4.3.6 Basicness dimension
Basicness here is being primary in the mind’s system; and it applies in several ways.
Cognitively, basic categories are presented directly by experience, are learned first, and
are used to develop other categories. They are relatively concrete, but are not always
at the bottom of their taxonomy. The concept DOG is more basic than ANIMAL but is
also more basic than COLLIE. Linguistically, meanings are more basic if they are used
to understand other meanings: ‘red’ is more basic than both ‘crimson’ and ‘coloured’;
‘slow’ is more basic than ‘slowly’; and in general nouns are more basic than verbs
(Gentner and Boroditsky, 2001), and both are more basic than prepositions and con-
junctions. See also Geeraerts (2010, p. 200). Basic things are generally learned earlier.
Dimensional basicness is distinct from basicness in children’s development and
in evolution, although they are related. It is also distinct from whether a meaning is
a “prime” – a meaning that can not be analysed into semantic elements. It is also dis-
tinct from logical basicness, in which a concept is more basic if it is presupposed; that
sort of basicness will not be an issue here.
4.3.7 Viewpoint dimension
Deictic meanings presuppose a viewpoint, that of the “deictic centre” – the time and
place of the speech event, and its participants. ‘This” and ‘that’ assume a ‘here’, a point
from which things are construed as near or distant. Other examples are ‘now’, ‘I’ and
‘you’. Some prepositional meanings assume a viewpoint also: ‘in front of ’ and ‘to
the right of ’ depend on where the speaker is viewing the scene from. A few verbal
meanings assume viewpoint, as when “He came to London” is preferred to “He went
to London”. Verbal aspect, as in the distinction between “he has come” and “he had
come”, depends on viewpoint. The term might perhaps be extended to include the
difference between ‘A accompanied B’ and ‘B accompanied A’; but I will not extend it
as far as the novelist’s “viewpoint”.
4.3.8 Boundedness dimension
Our most basic concepts are abstracted from perception; they are shapes and move-
ments, represented in speech as concrete nouns and verbs. The shapes have outlines:
they are “bounded”; Rijkhoff (2002) describes them as having the feature “Shape”.
The movements are bounded by their start and end points in time. A new stage of
abstraction removes those bounds, forming substance meanings like ‘stone’, formed
from knowledge of many stones, and ‘water’ from experiences like those of drinking
water and seeing puddles. Those meanings are still bounded in space – their instances
always have length and breadth, though not shape; those bounds are removed in the
next stage of abstraction, producing meanings like ‘stoniness’, and ‘liquid’. Meanings
based on perceptual happenings lose their start-point and end-point bounds; for ex-
ample, “They climbed Everest successfully” has both a start and an end, but “They’re
out climbing at the moment” has neither, and is unbounded. Property meanings like
‘red’ and ‘big’ are unbounded when they are treated as gradable, but bounded when
they are not gradable.
The boundedness dimension has more obvious significance in its other guise as
the concrete-abstract scale. The writer of example (2) preferred senses that are not
fully bounded.
(2) “This does represent a clear case of dismemberment of the body and removing
of tissues for consumption.” (Forensic anthropologist reporting evidence
of cannibalism in Jamestown, USA. New Zealand Herald, May 3rd 2013, A22)
The content of (1) could have been expressed as, “Clearly, the body was cut up and
parts were taken for food.”
4.3.9 Expectedness dimension
Cruse (2011, §3.2) illustrates expectedness from the meaning of dog (when used de-
scriptively). Of the many concepts that make up the cognitive meaning, ANIMAL is
necessarily part of the meaning, ABLE TO BARK is an expected part, BROWN is possible,
ABLE TO SING is unexpected, and OF THE FISH FAMILY is impossible. Possible elements
are often excluded by the context, or simply not evoked by it; they include concepts
from “frames” and “scenarios” evoked by a word. Even expected elements may be ex-
cluded by the use. This is how the concept of collocation (Firth 1957, p. 194ff.; Hanks
2013) is accounted for in this book: collocated words are part of the context which
controls the evocation of the expected and possible elements.
These distinctions are also made (in different terms) by Burnley (1992, p. 466).
The basic distinction goes back at least to Paul (1880, cited Geeraerts 2010, pp. 14–
15), who distinguished “usual” meanings from those that are “occasional” i.e. occur-
ring only in specific uses. “Connotations” has often been used for possible meaning
elements.
There is considerable psycholinguistic support for the distinction: Schwanen-
flugel (1991, p. 246) says that there has been a “general movement” in recent studies
towards such a view of meaning; presumably, degree of expectedness is (high) degree
of entrenchment, and (low) degree or accessibility (Flores d’Arcais and Schreuder
1987, p. 158); possible meanings will become actual meanings when the necessary
threshold of activation is reached. Our everyday experience of conceptual knowledge
shows the same pattern: the concept BASKETBALL will regularly bring ROUND to mind,
but not FLOATS. In neurolinguistics, Fortescue (2009, p. 141) notes that the full mean-
ing of a word is activated in the neuron bundles to the degree needed in the context.
Possible meanings are surprisingly extensive. Spivey et al. (2005, pp. 252–253)
cite interesting research which showed that English speakers regularly associated a
smoothly undulating horizontal line with gold, a zigzag line with silver, and a line with
square “saw-teeth” with iron; that shows that even line shapes are possible meanings
of the names of those metals. Interesting (but disputed) research by Jung et al. (2014)
shows that hurricanes with feminine names caused significantly more deaths than
comparable ones with masculine names, because their possible meanings involve less
risk and danger, so people protect themselves less when an approaching hurricane has
been given a feminine name.
4.3.10 Salience dimension
Salience is the degree to which a meaning element stands out from the mental back-
ground, or is “foregrounded”, making hearers more aware of it. Elements are salient
if they are contrasted with another element, or evoked by the linguistic context or
a referent which inhibits other possible elements. For example: sense <1> of woolly
(SOED) is “Of the nature, texture or appearance of wool”; in “woolly hat”, the ‘nature’
element of the sense is salient; in “woolly caterpillar”, the texture element is salient,
and in “woolly sky”, the appearance is salient.
4.4 Content meaning
4.4.1 Introduction
Content meaning. The term “content meaning” is to be understood through a double

distinction. First, it is the significance carried by the descriptive semiotic strategy, as
an alternative to the naming and pointing strategies (see Chapter 3, §3.4.3.2); it there-
fore excludes the meaning of deictics and names (proper nouns and similar uses of
common nouns). Second, it is distinguished from grammatical meaning, as set out in
the following paragraphs. It is thus “descriptive” in the general sense of describe <1>
“Give a detailed or graphic account of.” (Angle-bracketed numbers such as “<1>” will
hereafter indicate senses in SOED.)
The nature of content meaning, as in the familiar distinction between “grammat-
ical” or “functional” items and “content” or “lexical” ones, is never clear in the liter-
ature, in my judgement. The reason, I suggest, is that linguists assume that English
is made up of word classes, and that the grammatical/content distinction is therefore
one of classes. That immediately becomes problematic, because conjunctions and
many prepositions have content as well as being grammatical e.g. “to town” versus
“from town”. A solution is then sought in the concept of prototypes, or in defining
the classes in another way, by whether they are “open” or “closed” – “solutions” which
make the problem worse. The distinction becomes unproblematic, however, when
we see it as a distinction between meaning types, since a word can carry two types of
meaning at once.
That allows a sharp and potentially categorical distinction, but its nature still
needs to be clarified. An analogy with the mathematical concept of operator is useful
here. In maths, the meaning of one type of sign (the signs for multiplication, addi-
tion, and so on) operates on the meaning of the other type (numbers, and so on). In
language, grammatical meanings are those that operate on content meanings. Thus,
articles operate on nominal-group headwords, specifying their reference; inflections
for tense operate on the stem of the verb; prepositions operate on the groups of words
they link. In “The cat sat on the mat,” on has a grammatical meaning in linking “the
mat” to the rest of the clause, and a content meaning of location; it is not in two word
classes at once, but has two functions (linking, and conveying a concept), and two
types of meaning accordingly.
That leaves “content” not well explained; that is done below, through discussion of
its subtypes – descriptive, affective, and so on. Stated briefly, however, content is the
sort of meaning that includes descriptive, conceptual elements, and feeling, and atti-
tude; it is linguistic meaning which is not grammatical. The discussion will be illus-
trated largely from the meaning of words, in which the distinctions are relatively easy
to see; but the types of meaning are carried by forms in other strata, also. Phonological
tone and tonality carry feeling and attitude, for example; some syntactic structures
carry formality (e.g. the elision in I’ve); some affixes carry descriptive meaning, and
inflectional suffixes carry grammatical meaning.
Types of content. Types of meaning, like dimensions, are implicit in linguistic accounts
of semantic change. For example, knave and churl have undergone “pejoration”; that
is, while the conceptual meaning of ‘young man’ and ‘man’ has remained, the feeling
and the attitude the words express has changed. Put in the terms to be used here, the
descriptive type of meaning has remained unchanged, while the emotive and attitudi-
nal types of meaning have changed.
A word history will illustrate the nature of meaning types more thoroughly. The
adjective lousy was derived in Middle English from the old English noun, louse. (The
words are given in their modern spelling.) It meant, in SOED’s sense <1>, “… infested
with lice” – a concrete and specific meaning. Evidently, speakers and hearers gradu-
ally associated the word more and more strongly with a feeling of revulsion, until the
word had two layers of significance: (1) the descriptive meaning, ‘infested with lice’,
and (2) the feeling, which was now conventionalised as a standard part of what the
word signified. By the Late Middle English period, the word was being used with its
emotive significance only – <2> “Vile, contemptible” – and applied to things that have
no connection with lice: Thackeray wrote “I’ve been trying to write today & [sic] only
squeezed out one lousy page” (SOED quotation). In this use, lousy has only one layer
of significance – one type of meaning – the emotive.
In the 19th century, the original meaning, “infested with lice,” was applied meta-
phorically to other things that could be seen as “infested”: “Were not shipping mag-
nates lousy with shekels?” (SOED quotation). That meaning became increasingly
slangy, and dropped out of standard use, so it became <3> “Teeming (with) …; slang.”
In that use, the word again has two types of meaning, but in a different combination:
(1) the abstract conceptual meaning, ‘teeming’, and (2) the slang value – the effect
by which the speaker sets up a feeling of solidarity within a social group, such as the
working class, teenagers or criminals.
In that history, we see three layers or types of meaning, which are independ-
ent enough to develop separately, and to survive while others disappear: descriptive
meaning, emotive meaning, and social meaning in the form of slang. The contempo-
rary significance of the types is that we can explain many of the difference between
synonyms only in terms of those types. For example, wicked, cool and neat differ from
their synonyms good and excellent in having slang meaning, and differ from each
other in the nature of the slang.
The types have been recognised by some linguists, as will be shown in the follow-
ing sections, but are quite outside several strong semantic traditions, which restrict
their consideration to descriptive meaning. They are poorly dealt with by dictionaries.
SOED sometimes does not account for them, and sometimes uses conceptual words
to state emotive significance, as with “vile, contemptible” for lousy, as if the emotive
meaning were conceptual. It gives the social-group significance as a usage note, not as
significance, as with “slang” for lousy.
4.4.2 Descriptive meaning
4.4.2.1 Characterisation
Descriptive meaning is “descriptive” in SOED’s sense <2>: “… not expressing feel-
ings or valuation.” It is a type of content meaning; as noted in the previous section, it
is distinct from affective, attitudinal and social meaning, which are discussed in the
following sections. It is the prime constituent of information. It is the sort of mean-
ing that determines whether a statement can be judged true or false; exclamations
and emotive statements such as “It’s horrible” cannot be so judged. It can be negated
and questioned. In response to “There’s a horrible bloke at the door”, hearers may
reasonably question whether the person is a man not a woman (questioning the de-
scriptive meaning of bloke), and whether he is at the door not the window; but they
cannot reasonably question the meaning of there, a and the, nor whether it is a ‘man’
(formal) not a ‘bloke’ (informal). Descriptive meaning is objective in being not sim-
ply an expression of the speaker’s state, as emotive and attitudinal meaning are, and
it is “displaced” in having relevance outside the immediate speech situation – man
and door are meaningful to observers across the street whereas there and the are not.
(Those points are all from Cruse 2011, §10.2. Cf. Lyons 1977, pp. 50–51.) It is the sort
of meaning that is conspicuously absent from names. Since descriptive meaning con-
stitutes information, it serves the ideational function.
4.4.2.2 Relation to cognition

Descriptive meaning may usefully be divided into two subtypes which correspond to
types of cognition, as follows. (a) Perceptual meaning is close to perception, either to
sense perception or to perception of the mind’s own state. The underlined words in
example (3) are perceptual.
(3) “A young boy watched in fear as his little brother’s heart stopped after he
choked on a sandwich.” (New Zealand Herald, August 2, 2013, page A1)
“Perceptual” is roughly equivalent to “concrete”. Perceptual meaning corresponds to

Cruse’s “basic meaning” and the meaning of “observation vocabulary” (§4.3.3.1).
(b) Conceptual meaning is general and abstract, being relatively remote from
perception, as in ‘elementary’, ‘capable’, and ‘correct’. (This use is therefore distinct
from the broad cognitive linguistics use of “conceptual”, which covers all meaning.)
Barsalou (2012, §1.1) notes that perception is image-like, and that concepts “inter-
pret” those image-like representations.
Cognitively, there is a gradient between percepts and concepts. In “his little broth-
er” (example 3), brother has a concrete element (‘physical being’), and a conceptual
element as well (the family relationship); and little has primarily the abstract element
of ‘younger’, and secondarily perceptual associations of size. At the perceptual end of
the gradient, hearers may well have specific images from their own experience come
to mind as part of their hearer meaning of an expression – memory of their own ex-
perience of choking, in example (3), perhaps. System meaning excludes images that
are personal to the hearer, but includes conventionalised image schemas. “Concept”
as used in the book is consequently vague as to degree of abstraction; it is contrasted
mostly with “affect” and “attitude”, sometimes with “percept”. It is vague in other ways
as well, so it is not used in this book as a technical term; it will be given some further
discussion in §13.3.
The perceptual-conceptual scale is relevant to grammaticalisation: that process
moves meanings along it. The content of the prepositions at, on, and by once had only
the concrete elements of place, and have gained abstract meanings of time; will as
‘desire’ and have as ‘possess’ have become more abstract as auxiliaries; very has lost its
descriptive meaning, ‘true’, in becoming an intensifier.
The cognitive-linguistic distinction applies to descriptive meaning in the basis
for making inferences. For example, the definition of the meaning of conscious entails
‘living’, so inference from “He is conscious” to “He is a living being” is linguistically
based; whereas inferring “He is at 51° north and 0° west” from “He is in London” is an
inference from world knowledge.
4.4.2.3 Psycholinguistic support

Schreuder and Flores d’Arcais (1989) give psycholinguistic support for the reality of
this type of meaning. They illustrate word structure from words for a physical object
or substance, for example coffee. The central meaning has three parts: a minimal node
representing the bare concept and acting as the access point for the word; perceptual
elements such as colour and taste are linked to it; and and so are conceptual elements
such as the object’s origin and usefulness. (That account will be amplified in Chap-
ter 5, §5.2.3.)
The account of descriptive meaning given above has emphasised that it is based
in experience, which is construed into concrete and then abstract meaning. That is
supported by Barsalou: “The semantics of natural language are closely related to the
human conceptual system. Although lexical meanings are not identical to concepts,
the two have much in common and influence each other extensively” (2012, p. 244);
and concepts are abstractions from percepts (2005).
Pulvermüller (1999) and Fortescue (2009) support that neurolinguistically. Pul-
vermüller describes abstract and concrete words as represented differently in the
brain, with words for visual or action meanings including neurons in the visual or
motor parts of the cortex, for example. Pulvermüller (2002, p. 89) notes that generic
words such as animal have slower response times than concrete ones such as doe.
To Fortescue, the bundles of neurons which process conceptual words are linked by
affordances to related percepts. The left hemisphere is better at abstract and function
words than the right hemisphere is (Zaidel et al. 1988, p. 72).
To MacWhinney (2005, pp. 200–201), “declarative facts” about bananas (e.g. they
are rich in potassium and vitamin E) are “decoupled” from “the primary notion of a
banana that we derive from direct embodied perception.”
The specificity-generality dimension, which ranges from the concrete to the ab-
stract, is thus important in descriptive meaning.
4.4.2.4 Expression of descriptive meaning in the various strata

Descriptive meaning is in part expressed in strata other than lexis. It is expressed mor-
phologically in derivational prefixes such as pro- and contra-, centi- and mega- . In-
flectional affixes commonly have a conceptual element in their significance, although
they are sometimes taken to be simply grammatical, and thereby without descriptive
meaning. For example: the -es suffix in “fashionable dresses” signifies plurality, and
PLURALITY is a concept. Similarly, there are concepts in the significance of syntactic
structure. The significance of Subject-Object structures like “A stranger paid her $12
fare” includes the understanding that the Subject entity is ‘agent’, and that the object
entity is ‘affected’; that entails abstract concepts (AGENCY, AFFECTEDNESS), and there-
fore descriptive meaning. Phonology is apparently exceptional – which is natural if it
began as no more than the schematisation of the sound used to realise meaning.
4.4.2.5 Dimensions
Descriptive meaning has all of the dimensions; the examples given in the section
on the dimensions were all from descriptive meaning. I will illustrate here only the
boundedness dimension – descriptive meaning is the only type to which it applies.
The meaning ‘stone’ in a stone is fully bounded, which constitutes being a “count
noun” sense. Stone as in “made of stone” entails physical dimensions such as weight,
but not shape; it is less fully bounded, and has a “mass noun” sense. Stoniness is fully
unbounded, being an abstract sense. Boundedness is important, but less obviously
so, in uses such as “There were stone-throwing incidents”; although that entails many
stones, the word is singular, so that the specifics of how many stones were thrown,
their size, and so on are removed from the possible meaning elements of the word: it
is unbounded.
Boundedness also constitutes the distinction between such event senses as “He
mowed the lawn, and then trimmed the hedges,” where ‘mowed’ has an end bound,
and “He was mowing the lawn for hours,” where ‘was mowing’ does not.
Boundedness, then, is the dimension that distinguishes the aspectual features of
event words (Aktionsarten), and the features of nominal classes such as count and
mass nouns (“Seinsarten” to Rijkhoff 2002).
4.4.3 Affective meaning
4.4.3.1 Characterisation of affective meaning

Affective meaning is the expression of affect, which is undifferentiated emotive re-
sponse to a situation (see Chapter 3, §3.3.2). I emphasise that affect is expressed, not
represented or symbolised; expression of affect is quite different from statement of
it, which is cognitive, and serves the ideational function. In “Get out!”, for example,
anger is expressed, but in “I’m angry” it is stated. (I therefore exclude the meaning
of “emotion words” such as fear, joy, and surprise, which name feelings, not express
them.) Affective expression is often secondary to the speaker’s main intention; indeed,
speakers are sometimes unaware of it.
Affective meaning is derived from internal perception, which includes physical
perception such as balance and bodily stance and attitude, thirst, and perception of
affective states. The systems of the brain and mind that process the perceptions are
specific to particular modalities, such as affect, motivation (e.g. hunger), cognitive
states (e.g. disbelief) and conceptual knowledge (Barsalou 2012, §1), just like external
perceptions such as hearing and sight.
Some writers have described those different modalities of perception as being
processed in different “modules”, as if they were self-contained; modern research
seems to show that that was an exaggeration, but it made a useful emphasis. Many
writers, especially those in the philosophical tradition and in some formal traditions,
have written of meaning as if it were uniform – consisting of concepts alone. For ex-
ample, they have acknowledged emotion only in the form of “emotion words” such
as love and anger; but those conceptual words have descriptive meaning, not emotive
meaning. Emotive and other affective meaning is feeling; we “ex-press” the feeling –
push it out – for our own benefit, or for the hearer to share or react to.
I will distinguish three broad forms of affective meaning. The basic, undifferenti-
ated form appears in the Expressive function, as in non-verbal or semi-verbal excla-
mations. The emotive form appears in standard words, such as frightful, horrible and
lovely, having affective meaning only, and disastrous, beautiful and delicious, usually
having both affective and descriptive meaning. The imaginative form is the state of
excitement and stimulated imagination, expressed in humour, storytelling and liter-
ature. We make distinctions among emotions in everyday life, but I do not make any
for emotive meaning. That is partly because, as argued by Fillenbaum and Rapoport
(1971, p. 209), there are too many possible criteria for the distinctions to be reliable,
and partly because I have not observed any differences in form of expression for dif-
ferent types of emotion.
4.4.3.2 Support from linguistics

Although it has been ignored by many writers on semantics, affective meaning
has been recognised from as far back as Erdmann (working in 1910, according to
Geeraerts 2010, p. 19), with his concept of “Gefülswert”. It has been supported by
Fillenbaum and Rapoport (1971), Leech (1974, p. 26 – “affective meaning”), Lyons
(1977, p. 50 – “expressive meaning”), Adamson (1999, p. 573 – the encoding of “emo-
tions and evaluations”), and by Cruse (2011, §10.4.1 – part of “expressive meaning”),
and Gibbs et al. (2012). Tucker (2002, p. 53) says that “verbal semantics rests on a
foundation of affective evaluation”.
4.4.3.3 Psycholinguistic support

The standard psychological understanding of human experience, as set out by
Matthews et al. (2003, p. 85), is that it consists of three domains: affect, cognition and
motivation. Affect consists of moods and “subjective arousal states”, which are the
substance of affective meaning; descriptive meaning arises from cognition. Psycho-
linguistically, it is now generally accepted that affective meaning has two dimensions,
those of intensity and of being positive or negative in quality (Citron et al. 2014).
Neurolinguistically, affect is typically based in the right hemisphere (Brownell
1988, pp. 28–29; Tucker 2002), whereas cognition and descriptive meaning are typ-
ically based in the left hemisphere. Specifically, it is based chiefly in the amygdala
and the midbrain (Pulvermuller 2002, §5.3.1; Hsu et al. 2015). It is independent of
“neutral” i.e. descriptive meaning (Chen 2004); it is so strongly represented in the
brain that activation can be predicted from the presence of emotive words in a text
(Hsu et al. 2015); it is processed faster than descriptive meaning, and in distinct stages
(Citron et al. 2014).
This shows forcefully, I suggest, that affective meaning is real, and should not be
ignored in semantics, as it has been almost universally.
4.4.3.4 Aspects of affective meaning

There is an important distinction to be made in the hearer aspect (see §4.1) of affective
language. Feelings like admiration and moods like humour are intended to be shared,
and, all being well, the hearer’s response will match the speaker’s feeling. When a
speaker in a discussion said, “I saw a beautiful little girl in a wheelchair waiting on the
sidewalk” (Corpus of Contemporary American English; “COCA” hereafter), hearers
were to share the admiration. But emotive meanings like anger bring a converse re-
sponse such as resentment, rather than a sharing of it. When a speaker said, “You’re
gonna pay for this, you little shit” (COCA; fiction), the hearer is intended to respond
to the affect, not share it, feeling fear, for example, not anger; in sexual harassment, the
speaker’s affect and the hearer’s response are very different. The hearer aspect of the
meaning resembles an effect rather than something symbolised; see the discussion of
intention in Chapter 3, §2.1. Chen (2004) supports this distinction.
A parallel distinction occurs in the speaker aspect: the basic forms of affect are
expressed, directly and spontaneously; but moods such as humour are created in the
hearer’s mind, rather than issuing in a stream from the speaker’s mind. The nature of
the language used varies accordingly.
4.4.3.5 Realisation of affective meaning

The most obvious expression of affective meaning is in lexis: emotive words, as not-
ed just above (horrible, nasty, nice), and in the interjections discussed in Chapter 3
(§3.3.2). Some words have descriptive meaning, with their emotive meaning being
activated only in certain contexts e.g. dangerous. Some words gain affective meaning
when used metaphorically or ironically. Example (4) illustrates those points.
(4) “The way Westminster is going, [an environmental) party could also end up
with real power. But Mrs Bennett’s outfit is parochial and recidivist, a big fish
in a muddy, left-wing puddle, and that is not much use to anyone…”
(The Economist, January 3rd 2015, p. 42)
Parochial and puddle have emotive meanings in SOED; muddy gains it here by met-
aphor; left-wing gains it by sarcasm; and recidivist gains it through extension of its
application.
It has been noted above that phonology is used little in English for descriptive
meaning; it is correspondingly used heavily for affective meaning. Chapter 2 recorded
the use of both primary and secondary tone for expressing affect (§2.5.3). Rhythm, al-
literation and other devices of sound symbolism, in both poetry and prose, rouse such
affective states as imaginative excitement and aesthetic pleasure, as well as emotions.
Even syntax, which may seem too abstract to carry affect, sometimes expresses
it; for example, exclamatory structure has this as its main purpose, and rhetorical
questions often express scorn, indignation and so on. It is the marked nature of that
use that makes it emotive (the anomaly of the question form when no answer is al-
lowed for). Indeed, markedness itself seems to have affective purpose. For example,
“Gloves will be worn” seems to be a quite neutral statement; but when it was spoken
by a headmistress, with marked emphasis and slowness, with markedly flat tonality,
to a school of girls who were refusing to wear gloves, it was not only a command, but
strongly affective.
The realisation of affective meaning contributes to some sets of antonyms and
synonyms. In some instances, there is a pair of words for the same concept but op-
posed feeling, as with slim and skinny; they are synonyms descriptively, but antonyms
emotively. Other pairs have opposed feeling based on opposed concepts; ugly and
beautiful are antonyms in both respects. Often there is a contrast between emotive
and neutral words: “child behaviour” / “puerile behaviour” and “pig behaviour” / “pig-
gish behaviour”; and we have “obliging” as favourably emotive, “sycophantic” as unfa-
vourably emotive, and “co-operative” as neutral.
4.4.3.6 Dimensions of affective meaning

The semantic dimensions apply fairly fully to affective meaning. As noted just above,
on psycholinguistic support, affective meaning is on the dimension of positive-
negative. On the quality dimension, the descriptive meaning ‘of large size’ can be
coloured by admiration in whopper, or revulsion in monster. An action can be colour-
ed by scorn with over-mighty, and by respect with powerful as in “You need a powerful
engine”.
Affect is gradable, on the intensity dimension. Expression of enjoyment is mild
in enjoyable, stronger in wonderful and stronger still in marvellous. Alternatively, the
feeling can be graded syntactically by modification with very and totally, and phono-
logically by stress and tonality. On the generality dimension, dislike is general in smile
unpleasantly, and specific in simper and smirk; and general liking in laugh has specific
alternatives in chuckle, chortle, gurgle, coo and giggle – and for specific dislike, we have
guffaw, snigger and snicker. Enjoyable, wonderful and marvellous are all vague; precise
favourable alternatives for enjoyable include memorable, entertaining and friendly.
4.4.3.7 Affective meaning as content

It may seem odd to rate affective meaning as content, since content seems to require
concepts. The general reason for including it in content is that it is part of what we
want to convey to hearers, not part of the guidance we give hearers for constructing
that meaning; content, as what is operated on, contrasts with grammatical meaning,
which does the operating. More technically, grammatical operators, such as intensi-
fiers like very and the negative not, can operate on affective senses, which therefore
must have content meaning. That explanation applies also to attitudinal meaning and
social meaning, in the following sections.
4.4.4 Attitudinal meaning
4.4.4.1 Characterisation of attitudinal meaning

Attitude is mental orientation to events, especially future ones; it shows in our expec-
tations, and particularly in our patterns of behaviour. It has a basis in affect, which
motivates it, but it also has a basis in cognition and conation, since we set our attitudes
by reflecting on affective experience and relating it to our goals and values.
Words with attitudinal meaning often come in antonymous pairs, as with affective
words, often using negative prefixes such as un- and dis-, as in honourable/dishonoura-
ble. Thus we can divide attitudinal meaning into the favourable and the unfavourable.
As with affect, we often make finer distinctions in our thinking, but they do not seem
to be correlated with linguistic distinctions; however, it will sometimes be helpful to
follow SOED in distinguishing jocular and facetious attitudes from others.
Attitudinal meaning is realised lexically, as above, but also phonologically, in
much the same way as with affective meaning. For example, “He isn’t coming, you
see?” with a high falling tone conveys an expectant attitude; the same expression with
a low rising tone expresses a stronger attitude (Cruttenden 1997, p. 89). It is expressed
syntactically, also, in positive and negative question tags: the tag in “You’re coming,
aren’t you?” reduces the certainty expressed by the unqualified statement, “You’re
coming” (Mithun 2012). Attitude is a vital part of politeness expressions such as sir
and madame, where the favourable linguistic meaning helps establish and maintain
social relationships.
Attitudinal meaning is often evoked in context, without being a necessary part of
the word’s meaning. In example (5), target has favourable meaning, but in example (6)
it has unfavourable meaning.
(5) “His proposals … of targeting the relief to those who have been unemployed
for a long period of time will help all of us….” (COCA)
(6) “The Transport Agency has been accused of targeting students with … fare
rises.” (newspaper report)
In those examples, there is to me an affective element also; the distinction between

affective and attitudinal meaning is not categorical.
As with emotive meaning, there is often a double contrast. The favourable expres-
sion “a pensive mood” contrasts with the unfavourable “a distracted mood”, but also
with neutral “a thoughtful mood”; favourable modern contrasts with unfavourable
new-fangled, and with neutral new; favourable biddable contrasts with unfavourable
servile and with neutral compliant. The game of inventing “irregular verbs” illustrates
the double contrast also: “I am firm; you are stubborn; he is pig-headed.”
Attitudinal meaning is the substance of discriminatory language. “Men who put
themselves forward at work are ‘assertive’, women who do the same are more often
‘pushy’ or ‘bossy’; men are ‘persistent’ whereas women are ‘nagging’; men are ‘frus-
trated’, women ‘upset’” (The Economist, February 13th, 2016, 80).
Lay people’s awareness of the distinction is further illustrated in example (7); its
writer must have been conscious of the attitudinal contrasts in the pairs of words.
(7) “Forcefulness can become bullying; decisiveness can turn into pig-headed-
ness; niceness can develop into indecision.”
(The Economist, June 8th 2013, 60)
Linguists, however, have seldom recognised attitudinal meaning as a type of meaning.

One linguist who does deal with it is Van Linden (2012, §12.1). I have not found any
direct discussion of it in the psycholinguistic or neurolinguistic literature; however, it
seems to be supported by one of the distinctions mentioned in §4.4.3.3 above, since
the positive/negative dimension (“valence”) fits the linguistic characterisation of atti-
tudinal meaning given just above.
4.4.4.2 Dimensions of attitudinal meaning

The dimensions of attitudinal meaning are much less fully developed than those of
descriptive and affective meeting. It does not seem to vary on the intensity dimen-
sion; we speak of things as either acceptable or unacceptable, without construing
our attitude as a matter of degree – as reflected in the positive/negative morphology.
Attitudinal meaning thus shares with emotive meaning the positivity dimension of
positive-negative.
Attitudinal words are quite often graded; but it is not attitudinal meaning itself
that is graded, I believe. It is sometimes the descriptive sense: in “Those trousers are
still around, but they’re mostly very skinny”, the descriptive narrowness element of
skinny is intensified, not attitude. Alternatively, the grading may be of affective mean-
ing which accompanies or replaces the attitudinal meaning, as in “Non-dairy milks
… make very acceptable soups,” and in the invented example, “Those trousers look
very skinny on you.”
Similarly, there may seem to be variation between general and specific attitudinal
meanings, as with bad (general) and illegal, immoral and crude (specific), all applied
to behaviour. However, the variation there seems to be in the specificity of the field
where the terms are applied, not in specificity of the attitude. Basicness and viewpoint
do not seem to apply either.
The contrast in dimensional structure between attitudinal meaning and descrip-
tive meaning is thus very great. There are also important differences between attitudi-
nal and affective meaning, although they are similar in being non-conceptual and in
serving the interpersonal function. English is much less fully developed in some areas
of semantics than in others.
4.4.5 Social meaning
4.4.5.1 Characterisation
Social meaning is what a word expresses of the social situation in which it is being
used. Two types may be distinguished, following Cruse (2011): (a) dialect meaning,
including geographic, historical, and social group variation in language – both so-
cial class and occupational group; (b) register, including field i.e. the subject of the
utterance, such as law or science, mode i.e. spoken or written, and style i.e. degree
of formality and individually chosen variation in language. (Register is social to the
extent that it depends on the relationship that the speaker or writer is setting up with
the audience.) The subtypes overlap a little; for example, being somewhat informal is
a characteristic of being spoken.
Social meaning is most easily understood from the hearer aspect, in forms rang-
ing from the hearer’s perception of the speaker’s background to the hearer’s sympa-
thetic response to the speaker’s informality. The social class and regional significance
is generally indexical, as smoke is an index of fire; it comes from a secondary inten-
tion such as fitting in with the social situation, and is usually not a deliberate part
of speaker meaning. (Informality and slanginess are usually more deliberate and of
more conscious importance to the speaker.) It is thus rather different from the other
meaning types: it is not a direct product of a mental faculty, such as perception, but an
outcome of cultural learning.
Within dialect meaning, geographic meaning can be illustrated by Chinook, föhn
and nor’wester, all of which mean ‘a warm dry wind coming down from mountains’,
but in different parts of the world. Eftsoons (‘soon’) has historical meaning. Wee is di-
alectal (from Scots) in contrast with standard English small. Within register meaning,
guy is colloquial in contrast with standard man; bestarred is literary where starry is
standard; alleged is legal where claimed and asserted are standard.
Words may carry several of these subtypes at once, as with prang (‘crash’), kite
(‘aircraft’), and in the drink (‘in the sea’), which are geographic (British), historical
(World War II), and social (upper-class, and of pilots as an occupational class). Like
the other types of meaning, it can stand alone as the whole meaning of the word;
examples include terms of address, social formulae such as greetings, voguish slang
words such as like in “I was like, ‘Get real, man’,” and fillers such as um and er. A word’s
social meaning can vary with social class: a New Zealand hunter who broke his leg in
the mountains praised his rescuers as “real good buggers”.
4.4.5.2 Realisation of social meaning

The examples I have given so far are all lexical, since words provide the clearest illus-
tration. But social meaning is carried also by grammatical paradigms (I am/I be, you/
youse) and choice among allophones (e.g. variants of /r/) as in regional dialects, and
spelling (e.g. “bovver” as against “bother”). Elisions such as he’d and marked range of
phonological pitch are informal.
4.4.5.3 Dimensions
There is limited variation on the dimensions of social meaning. A dialect may be
taken as British, in general, or as Yorkshire specifically; the field may be technical in
general (acceleration and reciprocal), or specifically nautical (halyard and sheet for
‘rope’). Only formality seems to vary on the intensity dimension.
4.4.5.4 Support: Social meaning

Social meaning is what Cruse (2011) calls “evoked” meaning. It has been less often
recognised than affective meaning; but it is recognised by Leech (1974, p. 26 – “stylis-
tic meaning”), Lyons (1977, p. 50 – “social meaning”), and Halliday (1977, pp. 200–
201). It is characterised in dictionaries by such terms as “taboo,” “racially offensive,”
“coarse,” “euphemism” and “slang”. Although often missed by linguists, it is a major
issue for sociolinguists and sociologists; see, for example, Ochs (1992).
Barsalou (2012, §3.6) gives psycholinguistic support. He recognises social and
cultural cognition, which differs from the intellectual cognition used in decision-
making, reasoning and problem solving, and expressed in descriptive meaning. It is
used in specifically social and cultural situations, and is the cause of behaviour, mental
state and self-concept, not the cause of decisions and logical conclusions. For those
reasons, and since it is transmitted differently, it seems to correlate closely with social
meaning.
4.4.6 Conclusion: Types of content meaning
This section provides general support for the analysis of meaning into types, to com-
plement the support already cited for specific types.
Support from history. The semantic history of English content words shows a very
regular pattern of change of meaning type, which can be illustrated with the adjective,
gross.
The original meaning of gross, taken from French in Middle English times, was
<1> “Thick, stout, massive, bulky”, applying to physical things. In Late Middle English,
that sense was generalised to all physical things, becoming partly abstract, and thus
forming the new, additional sense, “Of conspicuous magnitude”. From that, a wholly
abstract sense developed, “plain, evident”, and then “obvious”. While those changes in
descriptive meaning were occurring slowly, gross acquired emotive and social mean-
ings, as lousy did (§4.4.1). As the sense ‘stout’ was applied to people whom speakers
regarded with disapproval as being too stout, the disapproving attitude became more
and more strongly associated with the word, until, by mid-16th century, the disap-
proval had become part of the meaning, as <3> “Overfed”. By the same process of
association, the word then gradually acquired the emotive meaning represented as
“Repulsively fat” (part of <3>.) The related subsense, “big-bodied, burly”, gradually
Grammatical meaning Intensifying
Social meaning Dialectal
Emotive meaning Dislike Dislike

Un- Un-
favourable
Attitudinal meaning favourable favourable
‘Larger ‘Larger ‘Larger
Descriptive meaning than than than
expected’ expected’ expected’
<1> <3> <3> <3> <4>

Sense “Thick” “Bloated” “Overfed” “Repulsively “As intensive”
fat”
Diagram 1. Gross: changes in types of meaning
dropped out of use in standard English, but was kept in some dialects; thus it now has
the social meaning of being dialectal.
The development, as changes in types of meaning, is shown visually in Diagram 1,
where the history is represented in a modified bar graph as the addition of layers of
meaning. The historical development is read across the page, with the senses and pe-
riods stated at the bottom of the diagram; the meaning types are then represented by
the boxes – the columns of boxes are intended to resemble a bar graph. In the boxes,
descriptive elements are paraphrased from the SOED senses given at the bottom of
the diagram, to show the continuity underlying SOED’s synonyms; non-descriptive
meanings are given without quotation marks. (Grammatical meaning has been add-
ed, in anticipation of §4.5.1.2 below.)
I emphasise that the pattern is a standard one: concrete descriptive meanings
are the basis for abstract ones; attitudinal, emotive, and social meanings develop in
turn from them. The only exceptions I am aware of are words that are borrowed with
abstract or emotive meaning; and where we know their history, we know that they
developed in the same way in the original language – they are not truly exceptional.
Support from authorities. There is now a considerable tradition in British linguistics

for understanding word semantics as a composite of meaning types. It includes Leech
(1974), Lyons (1977) and Halliday (1977, 2014), Ferris (1983), and Murphy (2003).
Wray (2002) includes all of the types of meaning given here, notably attitudinal mean-
ing in “lexicon III”, and grammatical meaning in “lexicon I”. (Her “lexicon V” approx-
imates the Expression function, discussed in Chapter 3.)
Barsalou (2012) gives general psycholinguistic support. First, as already cited
in §4.4.2.2 above, he notes (§3.4) that “the semantics of natural language are closely
related to the human conceptual system;” and (§1.1) that knowledge is believed to
be grounded in “modality specific systems”; that is, grounded in such modalities or
modules as visual, tactile and other types of perception, as forms of experience. I am
extending that to other forms of experience, the internal forms such as emotional,
attitudinal and social experience, which are beyond Barsalou’s concern in the arti-
cle being quoted, and using the statement to support the analysis of meaning into
modality-specific types. Barsalou goes on to say (§1.1) that “selective attention” is
central to creating knowledge. If we assume that selective attention applies in creating
meaning as in creating knowledge, then we see why meanings become more and more
remote from perceptual verisimilitude as they are extended to wider and wider uses.
Asp (2013) gives psycholinguistic support. She says that the research literature
has a robust finding that discourse processing normally engages not only the language
networks, but also the memory network, the affect network, and executive control of
action, affect and cognition.
This view of meaning as having many types disagrees with the view, held by
many cognitive linguists, which accepts “the unequivocal identification of meaning
with conceptualization i.e. the cognitive activity constituting our apprehension of the
world” (Langacker 2005). Those linguists seem to have developed that view through
over-reacting to “the many varieties of formal semantics based on truth conditions”
(Langacker 1987a, p. 5), in spite of their concern with “language behaviour” and “psy-
chological plausibility” (Langacker 1987a, p. 6).
I emphasise that it is the identification of meaning with conceptualisation that
is rejected; the cognitive linguists’ discussion of conceptual meaning has much to
offer. So, too, does work in formal semantics and Functional Discourse Grammar
(Hengeveld and Mackenzie 2008): their variables correspond to content meaning, and
their operators correspond to grammatical meaning. There is a powerful case, then,
that semanticists can agree on something like the comprehensive structure given here,
allowing all to follow their specialities.
All that is needed for that agreement is acceptance of the view that language is
adapted to its use in all parts of our lives. Use in our intellectual life needs descriptive
meaning; use in our emotive life needs affective and attitudinal meaning; use in our
social life needs social meaning.
4.5 Grammatical meaning
4.5.1 Introduction
4.5.1.1 Description
Grammatical meanings were introduced, in the introduction to types of content
meaning (§4.4.1), as meanings that operate on content meanings. That dry definition
gives little sense of grammatical meaning’s practical reality, and especially of its vital
importance in language. Let us consider, then, some examples. Example (8) is a sen-
tence from a newspaper with the words arranged in alphabetical order. Example (9)
is the sentence in correct order, but with the “grammatical items” removed. Exam-
ple (10) is the original sentence.
(8) “A and cafes comprise convenience environs of outlets restaurants retail street
surrounding the variety.”
(9) “Surrounding street environ comprise variety restaurant, convenience retail
outlet cafe.”
(10) “The surrounding street environs comprise a variety of restaurants, conveni-
ence retail outlets and cafes.”
(New Zealand Herald, August 10th 2013, page C1)
We can reasonably say that example (8) is meaningless as well as ungrammatical, al-
though it has all of its grammatical items included. Example (9) is a much more mean-
ingful, because better structured, although it has none of its grammatical items. The
examples show the importance of grammatical meaning; more strikingly, example (8)
shows that the grammatical items on their own give relatively little of the grammatical
meaning, partly because they must be in their correct position, and partly because a
good deal of grammatical meaning is given by the syntax, not by lexical items.
Let us consider the ungrammaticality and effective meaninglessness of exam-
ple (8) from the hearer and reader aspects. Readers need something to tell them that
environs is the headword of the Subject, and that surrounding and street are subordi-
nate to it, as modifiers; that comprise introduces the predication about the Subject,
and so on. It is the position of surrounding and street before environs that tells the
reader they are modifiers. It is the existence and position of the modifiers and of the
determiner the that tell the reader that the first four words comprise a nominal group,
and that environs is its head.
The grammatical meanings thus guide readers in how to construct the full mean-
ing from its constituents. Understanding the sentence may be compared with cooking
from a recipe: the words are merely a list of ingredients; the reader needs instructions
in how to process them. Grammatical meanings are instructions from the speaker to
the hearer, or – from the hearer’s point of view – procedures to be followed. In the
sentence about street environs, the procedures include: “(1) Modify ‘environs’ with
‘surrounding’ and ‘street’. (2) Make that group the Subject.” For speakers, grammat-
ical meanings are the corresponding set of instructions, which they must give to set
out the procedures. The mathematical analogy is even closer. “(2+3) × (4+5)” means
“(1) Add two and three. (2) Add four and five. (3) Multiply those totals.”
In traditional linguistics, grammatical meaning is dealt with by assigning utter-
ances and parts of them to abstract grammatical categories, such as subject, declar-
ative, and third person. That accounts for only part of the truth – not the whole of
grammatical meaning, but only its system aspect. For example, the instruction “Mod-
ify ‘environs’ with ‘surrounding’ and ‘street’,” and the corresponding hearer procedure,
are reduced to a single concept, “modifier”; “Make that group the Subject” and its
corresponding procedure are reduced to “nominative”.
This understanding of grammatical meaning is not merely one possible interpre-
tation of meaning: it is a necessity, because of the nature of language, as discussed in
Chapter 3, on constraints. Since language is linear, but the speaker’s meaning is not,
there are necessarily procedures for reconstructing that non-linear meaning as we lis-
ten and read; it is a truism that meaning is compositional. We conclude that grammat-
ical meaning is essential to English, and that it consists of operations or procedures
that compose the content of an utterance. A full account of its nature and forms is
essential for an understanding of semantics, and particularly for semantic structure.
4.5.1.2 Development of grammatical meaning

The presentation of grammatical meaning so far may suggest that it is totally different
from content meaning. In fact, it is a natural and regular development from certain
wholly abstract meanings, in the process of grammaticalisation. It can be illustrated
from the history of the adjective, gross, as follows, building on the account given above
in §4.4.6.
The abstract subsense, “… plain, evident”, was intensified as “… obvious”, i.e.
‘very evident’. In the late 16th century, that intensification became the function of the
word, rather than its sense; that is, it was used to intensify the head word, while still
adding its own descriptive meaning, ‘obvious’. Thus it had sense <4>, “As an inten-
sive: flagrant, glaring…”, as in “… made a gross fool of himself ” (SOED’s example);
in that use, the sense combines descriptive and grammatical meanings. Finally, sense
<4> lost the descriptive element, ‘flagrant, glaring’; in this use, it has only the gram-
matical meaning of intensification, as in “There must be gross inefficiency on board”
(SOED’s example); it is a sheer intensifier or ‘Reinforcer”, like very. In a word, gross
has grammaticalised.
We can now fit the grammaticalisation changes into the sequence of changes pro-
ducing sense <3>, in Diagram 2. The grammaticalisation is shown by arrows. (The
layout replicates the layout of Diagram 1 with affective and social meanings omitted
for clarity, and space left to suggest where the other developing senses fit.)
[Intensi- [Intensi-
Grammatical meaning fication] fication]
Descriptive
meaning
‘Conspicuous
in size; [No descriptive
‘Stout’ ‘Obvious’
…evident’ meaning ]
<1a> <1a> <3> <3> <3> <4> <4.>

“Thick, “Of conspicuous “Overfed” “Repulsively “Repulsively “As intensive: “As intensive:
stout, magnitude; fat” fat” Flagrant’ Complete’
massive.” plain, evident”
M..E. Mid 16th C. Late 16th C. Later period
Diagram 2. Grammaticalisation of gross

4.5.1.3 Types of sign that realise grammatical meaning

We have just seen that the signs which carry grammatical meaning are not only lexical
(“grammatical items”) but also syntactic, such as inversion of Subject and Predicator,
which indicates the interrogative. There are also phonological signs, such as the rising
tone which also indicates the interrogative. Further, grammatical meaning is some-
times carried by a combination of signs, not a single sign; for example, in speech not
specifies negation, but when marked stress specifies which content is to be negated,
both signs are necessary for understanding the meaning.
4.5.1.4 Looking forward

The rest of the section lists types of grammatical meaning in §4.5.2, and discusses its
dimensions in §4.5.3. It will be amplified in Chapter 6, on the system structure built
up by grammatical meanings.
4.5.2 Types of grammatical meaning
Functions of grammatical meaning. Grammatical meanings have two main tasks.

First, they relate the content items to each other – adjectives to nouns and adverbs
to verbs, for example, and nominal groups to verbal groups. Second, they adjust the
content that is related – reinforcing it (“extremely slowly”, for example), or qualifying
it (modals such as possibly and ought, for example). This section sets out specific tasks
within each of those two broad ones.
Formulation of the meanings. I have formulated these grammatical meanings so as to

bring out their significance for hearers – as implicit instructions from the speaker,
which the hearer is to carry out. They would be expressed rather differently if the
focus were on grammatical statuses such as tense and transitivity, or on abstract for-
mulations such as argument structure – in the system aspect.
Qualification/coverage. As just noted, grammatical meanings have different functions;

there are in fact a number of types; which operate at different levels of language –
word, group and clause for example. Consequently this introduction to them will be
kept simple, giving no more than the basic types. They are to be explained more fully
in Chapter 6, on system structure.
Main types of grammatical meaning. One broad type specifies that units are to be relat-
ed as co-ordinate, that is, of equal status; its primary use is in setting up the structure
of clauses, and of their semantic equivalents, figures.
In its hearer aspect, it may be paraphrased as, “Relate these two items as co-
ordinate”. The other type specifies that units are to be related as head and dependent:
“Relate these two items with one subordinate to the other.” Its primary use is in setting
up group structure.
The specific forms taken by the subordinating type specify what kind of relation
to the head the subordinate item is to have. The main forms are as follows. (1) “Add
the content of this modifier to the content of the head.” In “She stealthily crept into
the house,” the reader must add the property meaning ‘stealthily’ to the event mean-
ing ‘crept’. (2) Inflections carry a meaning like, “Specify the number/tense/etc. of the
stem as plural/past/etc.” In “books”, the hearer must specify the number of the sense
‘book’ as plural. (3) Intensifying modifiers adjust the intensity dimension of the head.
In “very large”, the instruction is, “Intensify the sense of large.” (4) The preceding pro-
cedures assume that the content to be acted on is in the head; but for pronouns, that
is not so, the content being in the previous clause, normally. In “She saw two seagulls.
They…”, the reader’s procedure for they is, “Go back, find the antecedent, and incorpo-
rate the meaning (‘seagulls’ in this instance) into the new clause.” This meaning copes
with words which use the semiotic process of pointing; the preceding meanings deal
with words using the descriptive process (see Chapter 3, §3.4.3.2). With deictics such
as “over there” and “that book”, the content to be acted on is to be found in the physical
or linguistic context.
The final specific form of subordinating meaning to be mentioned here – (5) in
the list – is noteworthy for its difference in kind from the others: the meaning of
prepositions that introduce dependent prepositional phrases. (a) Among the general-
purpose prepositions, of is notoriously vague. For example, the relationship to be as-
signed in “ a litre of milk” is part-whole; in “that book of yours”, it is possession; in
“Lady of Spain”, it is origin. The relations are not specified by the preposition at all,
but must be inferred by the hearer, working from context and using cognitive as well
as linguistic knowledge. The meaning seems to be no more that, “Relate these two
entities in some way or other.” It is thus purely associational, without any element
of logical relationship. It is basic developmentally, and is what Gil (2005, §2.3) calls
“the association operator”, which is the only grammatical meaning in the most basic
of languages. (b) Other prepositions, such as on and under, carry the same signifi-
cance that the Entities must be related, but obviously also specify the location; they
carry that conceptual element as content meaning, additional to their grammatical
meaning.
4.5.3 Dimensions of grammatical meaning
Grammatical meanings vary on some but not all of the dimensions set out above in
§4.3. The first important dimension of variation is that of vagueness. As just noted,
prepositions such as on, under, after and between are fairly precise as to the temporal
or spatial relationship to be assigned to the following content, where others are not.
Focusers like only and just are vague as to which item of content is to be focused; in
speech, the hearer must use the location of stress as a second grammatical sign to
complement the focuser; compare §4.5.1.2 above on not.
The other important dimension of variation is that of point of view. Deictics rely
on the speech situation as the viewpoint for both temporal deixis (now, then) and
spatial deixis (here, there). In a looser sense of “viewpoint”, it could be said that the
whole system of grammatical meaning relies on it. Content meaning is “displaced” in
not relying on the speech situation; but grammatical meaning is inherently the speak-
er’s guidance to the hearer, and is therefore bound to the viewpoints of speaker and
hearer, which are central to the speech situation.
4.5.4 Discussion
Support. My description of grammatical meaning as procedural is based on an old

and widely accepted distinction, that between “knowing that” and “knowing how”,
between factual or “declarative” knowledge, and “procedural knowledge”. There is also
substantial support for it in the semantic literature. Knowledge of grammar is some-
times acknowledged to be procedural knowledge, as by Bybee (2002). One group of
linguists, represented by Blakemore (2002), uses “procedural meaning” for certain
kinds of pragmatic meaning, using the concept presented here but not applying it as
widely; others include Janssen (2007, p. 353) and various writers in Escandell-Vidal
(2011); Fretheim (2011) applies it a little more widely than they do. More direct sup-
port comes from: Fawcett (2000, p. 34), Rijkhoff (2008, p. 798), and Wharton (2012,
§28.4). A number of writers characterise it as giving an instruction to the hearer:
Diessel (2006), Kirsner (2011) and Pietroski (2012, p. 135). Lamb (2004) gives general
support. To Croft (2001, p. 19) meaning represents “all of the conventionalised aspects
of a construction’s function.” Allan (1986, p. 387) prefigured the concept, saying that
constituent order is “an indicator of scope relations”. Those linguists do not, howev-
er, present grammatical meaning as often carrying content meaning which it uses in
carrying out the function, as precise prepositions do (§4.5.2 above). One linguist who
does see that is Mel’cuk (2012, p. 183). Cognitive linguists such as Langacker give
only a weak account of grammatical meaning; Langacker (2008, p. 3), for example,
acknowledges that “the elements of grammar – like vocabulary items – have meaning
in their own right”, but gives no explanation.
There is psycholinguistic support for the distinction between content and gram-
matical meaning. In aphasia, for example, different types of failure occur for “closed
class” and “open class” items (Pulvermüller 2002, p. 115). There are distinct sub-
processes for the two types (Ferreira and Slevc 2007, §27.1.2). Neurolinguistically,
there are separate neuronal circuits for the two (Pulvermüller and Knoblauch 2009).
Pulvermüller (2002, p. 117) notes that grammatical meanings sometimes carry con-
cepts, as when tense meanings carry the concept of time. Fortescue (2009) has gram-
matical “templates” which complement word templates.
Related ideas. Some closely related conceptions of grammatical meaning are worth
noting here. Pietroski’s inclusion of “fetch” in “instructions to fetch and combine
concepts” (2012, p. 135, cited just above) implies that even content words carry a
grammatical meaning, that of instructing the hearer to fetch the concept from mem-
ory, as a preliminary to combining it; for example, in “He’s an honest man”, honest
signifies “Fetch the concept HONESTY from memory”; honest’s syntactic position spec-
ifies “Combine HONEST with MAN as modifier”. I accept that; but building it in to all
my explanations will make them cumbersome.
4.5.5 Conclusion: Grammatical meaning
Summary. We have noted two broad types of grammatical meaning: coordinating,

and subordinating. Each of those takes various specific forms. We have seen that
grammatical meaning is crucial to the meaning of utterances in English, and that it
is much more widespread than is generally acknowledged. It occurs not only in the
so-called “grammatical items”, but is the significance of many other overt signs, as in
intonation, and of covert signs such as word order. (“Grammatical items” is therefore
a misleading term.)
From the speaker’s point of view, grammatical meaning consists of the instruc-
tions given to the hearer to guide construction of the meaning of the text. From the
hearer aspect, it consists of procedures to be followed, according to the speaker’s in-
structions; they enable him or her to (1) establish the hierarchical or “vertical” re-
lationships among content meanings, such as main-subordinate and head-modifier;
and (2) to relate co-ordinated items’ meanings on the “horizontal” dimension. From
the system aspect, grammatical meaning is the specification of grammatical status
and relationship, such as subordinate status from the vertical relationships, and co-
ordinate status from the horizontal relationships. The meanings apply at all ranks –
from the sentence down to the word. (That may not be very clear, perhaps, from this
section, but it will become clear in later chapters, particularly in Chapter 6, which
extends the description of grammatical meaning greatly.) The linguistic meaning thus
constructed, in the first phase of comprehension, can then be related to cognitive
understanding, in the second phase. (The real-life process of comprehension is not
so schematic, of course; the processes interact, and may work cyclically even within
a single phrase.)
That summary covers what has been set out in this section; there is more to be
said about grammatical meaning, especially in Chapter 6.
Further conclusion. We can draw a further conclusion. Treating the distinction as

one of classes – a closed class of “grammatical items” and an open class of “content
items” – requires the classes to be gradient, and be “prototypes”, because many con-
junctions and prepositions are partly grammatical and partly content in their sig-
nificance. The resulting confusion is needless, since treating the difference as one of
function provides a distinction that is clear and categorical, with one “item” able to
have two or more functions. (This is one instance of what will be a recurring issue:
traditional linguistics’ predilection for treating linguistic phenomena as things, not as
events or functions for example, and then putting them into classes.)
Final comment. It is striking, I believe, that transformational/generative grammar is

now including grammatical meaning in its representations, having previously exclud-
ed meaning from the generation and structure of utterances (restricting it to inter-
pretation). For example, Muysken (2008, p. 16) says that “functional categories” have
as functions “connecting and organising parts of information”, and “structuring the
clause”. Since formal semantics also accepts it, in the form of operator functions, and
many forms of functional grammar also accept it, we can conclude that its acceptance
is likely to become a consensus view.
4.6 Semantic classes
4.6.1 Introduction
The senses that have been described so far as having types, dimensions and so on have
been illustrated with words, as if they could be equated with word meanings, as stated
in a dictionary. That has been a simplification for the sake of clarity; they are generally
elements of word meaning, although they are sometimes realised in a word of their
own; ‘red’, for example, is realised in red and it is an element in scarlet. These senses
are accordingly “sublexical”, rather than lexical.
They can be grouped into semantic classes according to linguistic properties
they share. In principle, senses of all types can be put into classes; but here “semantic
classes” will be limited to content senses. Grammatical meaning will be described as
having “types”, since those meanings represent functions or procedures rather than
entities that can be classed. Within content meaning, only descriptive meaning will
be discussed; the reason, given in a previous section, is that emotive, attitudinal and
social meaning do not have subtypes distinct enough to be analysed and classified.
I emphasise that the term “classes” is potentially misleading: the senses being
grouped occur in a mental network, so the classes do not have inherent boundaries;
further, the senses are formed by construal and are subject to reconstrual. I emphasise
also that the classes are linguistic, defined by linguistic properties, not cognitive class-
es or concepts. That is, a semantic class is correlated with some other class or structure
in linguistic expression. For example, one subclass will be formed by the distinction
between animate and inanimate, since that controls the use of he/she versus it. Verbs
of existence, seeing, thinking and so on do not constitute semantic classes (see Biber
et al. 1999, p. 364); nor do nouns of activity or of speech act (see Dixon 2005, §3.1):
those alleged classes have no demonstrated relation to structures in expression.
Further, these classes are not word classes. They are independent of words’ syn-
tactic and phonological properties. More important, they are classes of semantic use,
not of any fixed semantic entities as stored in the mind; a certain cluster of meaning
elements may be used in one class on one occasion, and in a different class on anoth-
er. Indeed, I use the term “class” reluctantly, because of the associations it has in the
linguistic literature, and because “use” or “function” would in part be more accurate.
However, classes as constituents seem to be “required by linguistic theory to explain
how the meaning of a complex category is determined by the meaning of the expres-
sions constituting it” (Gillon 2005, p. 184, referring to “semantic categories”); class is a
familiar term; and use and function have several senses here already.
4.6.2 Basic classes
Being simply semantic, senses and semantic classes do not have morphological, syn-
tactic or phonological properties. As noted above, they do not correlate one-to-one
with particular words; for example, the sense ‘mammal’ correlates with cat and dog
as well as mammal. Their nature is therefore to be inferred from the linguistic units
and structures they relate to, and from what we know of the psychology of language.
The analysis in §4.4 above showed that although word senses are units to the ex-
tent that they are associated with a single word form, they have constituent meaning
elements. It is those elements which are being classified here. Their classes are being
called “basic”, to distinguish them from the classes to which word senses belong.
The basic classes posited here are entity, event, and property. It is assumed that we
perceive actions as changes over time, as “happenings”; and that we perceive shapes as
continuing over time, as “things”; things and happenings are thus the basic percepts.
Comparison between different things and happenings produces “qualities”, as a less
basic third type of percept. Qualities, things and happenings are thus the cognitive
bases from which strictly linguistic semantic classes are developed.
That set of classes is supported as the basic set by Croft (2007, p. 340). They are
supported from psycholinguistics by Lamb (1999, on the visual system specifical-
ly), Barsalou and Wiemer-Hastings (2004), and Hanson and Hanson (2005, p. 131).
Pulvermuller does not deal with those specific classes, but says that word classes are
processed in different parts of the brain.
4.6.2.2 Entities
Senses of the entity class are based on cognitive things, and consist primarily of con-
cepts which have the feature of extension. The most basic entities are formed from
perceptual objects, physical things, with extension in space, such as people, plants and
the sun; they exist in the spatial dimensions of length, breadth and depth; they seem
to consist of an image schema. That basic nature has been developed to extension in
time, e.g. ‘hour’ and ‘week’ (without the sense of time passing, which would make it
an event), and expanded further to extension over a group, e.g. ‘population’, ‘heap’,
and so on.
What makes a sublexical entity part-way linguistic is potential strong linkage to
function as head of a group that can act as Subject or Complement, and to the mor-
phology of possession and plurality; those links form its “bond”. The bond or valency
of a chemical element is its potential to combine with other elements – with certain
elements only, and in a set ratio; the mechanism for it is the sharing of electrons. Sim-
ilarly, the bond of a linguistic element is its potential or actual link to certain others
in a semantic and syntactic relationship. Entities may have weaker links to functions
other than being head of a group; and they usually have many conceptual links to
other concepts which are not distinctively linguistic either. For example, the concept
SLEEP, considered alone, is simply cognitive; it is linguistic to the extent that we con-
ceptualise it with its link to nominal or verbal function.
When an entity is used in speech, its bond is activated, and the potential relations
constituting it become actual. By default, it is realised as head of a nominal group, and
typically as a “noun”; but it may become a modifier, as “noun” e.g. “the America team”,
or as “adjective” e.g. “the American team”. The sublinguistic sense then becomes a
fully linguistic word meaning; in doing so, it changes class becoming an Entity (with
a capital E).
4.6.2.3 Events
Events are based on cognitive percepts which feature the passage of time, a form of
extension; they exist in the time dimension. The most basic events are formed cogni-
tively from perceptual happenings, in which we are aware of no more than the pres-
ence of change – change in position, size, physical or mental state and so on. Examples
include FLY, FALL, HIT, and CUT.
Percepts of happenings usually have a potential link to function as head of a Pred-
icator group, but may have other links. For example, perceptions of people and ani-
mals sleeping are the cognitive basis for the concept SLEEP, which thus has links to the
memory of those percepts, and thus to functioning as an event in Predicator groups;
SLEEP also has links to object percepts (periods of sleep), which are the basis for the
use of the sense, ‘a sleep’, as an entity. (They are also the basis for a still more abstract
use of the concept, underlying the abstract noun, sleep.) Events’ bond, then, is their
link to “verbal” function, and to taking the morphology of tense and aspect.
4.6.2.4 Properties
The class of properties is the class of basic sense elements which consist of cognitive
qualities (concepts like RED, LOUD, HONEST, PAINFUL), and a potential link or bond to
modifying function. Although the fact is not important for our purpose, we may note
that properties are evidently derived from entities and events, by abstraction. Per-
ceiving many instances of red objects, for example, may lead to the formation of the
property concept, RED. That is supported by Givón (1970), and implicitly by Barsalou
(1999, §3), who presents the basic conceptual system underlying language as consist-
ing of events and entities, with other conceptual elements developed from them.
Properties default to being modifiers in verbal or nominal groups – “adjectives”,
“adverbs” and so on. There cannot be a distinction between adjectival and adverbial
types at this level, because these senses are independent of syntactic function and
morphological form, which they take on only when a speaker decides on the structure
in which they will be realised. (It will be argued in §8.2.4.1 that there is in fact no se-
mantic distinction between them, even at that level – the similarity between then can
be seen in the fact that senses such as ‘fast’ and ‘big’ can be realised, in fast and big, as
either adverb or adjective.)
4.6.3 Discussion: Semantic classes
4.6.3.1 Relations
There is a noteworthy absence from my list of semantic classes, namely relations. En-
tities, classes and properties must stand in some relation to each other, so where are
the relations that relate them? In network terms, where are the links between the
nodes? The answer to the question – a little complex – is as follows. The relations
of transitivity for example, which bond the major elements in figures, are built into
the semantic structure by the hearer, in response to the transitive construction, i.e.
the arrangement of Subject, transitive Predicator and so on; the construction car-
ries meaning – a grammatical meaning – which implicitly specifies the relations. The
relations between figures are typically indicated by conjunctions; their grammatical
meaning specifies whether the figures are to be co-ordinated or subordinated, and
their content meaning specifies the logical relation, such as place (in where), time (in
later) and cause (in because). Other relations are conveyed by grammatical meaning
too, as explained in §4.5. (Many of the relations are thus covert, while some are partly
overt, as in conjunctions.)
Those conceptual meanings, such as place, time and cause, belong to the property
class: a thing’s location is one of its properties; so is an event’s time, and a situation’s
cause.
Thus, there may be a cognitive class of relational concepts, but there is no seman-
tic content class of relations.
4.6.3.2 Linguistic and psycholinguistic status of the classes

The classes have been portrayed above as being semi-linguistic or sublexical. That
status is further supported by the following facts. First, the classes are interdepend-
ent with syntax and morphology; for example, entity senses such as ‘wine’ can be
realised as either mass nouns or count nouns; so they must, in their basic form, be
neither, gaining that status in use, as their boundedness dimension is specified. Re-
search shows that children learn the use of the count/mass distinction separately from
learning words. The analysis of conceptual elements (“features”) in word senses sug-
gests that they, and the senses they form, are indeterminate as to semantic class. For
example, Miller (2014, p. 11) analyses the verb coil as “put (a rope etc.) into a coil”;
but instead of the entity ‘coil’, that analysis could equally well have a property, ‘coiled’,
(“make a rope etc. coiled”) or perhaps an image schema of the coiled state – the anal-
ysis of these features in the form of words is notorious for its problems and lack of
consensus. Finally, property senses such as ‘slow’ have gradable and non-gradable
uses, and verbs such as ‘climb’ can be used as activities or achievements; again, the
basic senses must be distinguished from the corresponding word meanings.
That discussion shows not only that there is a sublexical level of senses, but also
that senses’ dimensions are determined only when used in an utterance. They are
accordingly substantially different in semantic structure, and the nature of their se-
mantic class is different also. Those classes other than the basic ones introduced here
will be described in later chapters, since the description will rely on material yet to be
introduced.
Pulvermuller (1999) ascribes different “word classes” to different parts of the
brain; he does not define word classes, but it is clear that he intends content classes
such as those “referring to actions”, which are my semantic classes.
4.6.4 Conclusion: Semantic classes
The semantic classes which are basic elements of semantic structure are entity, event
and property. They are close in structure to corresponding cognitive classes, since
they utilise the cognitive concepts of extension, change and quality, which define
them as set out in the preceding sections. Those concepts are summarised in Table 1,
where “Yes” means that the concept identified at the head of the column is present.
Table 1. Defining features of semantic classes

Extension Change Quality
Entity Yes No No
Event Yes Yes No
Property No No Yes
Semantic classes differ from cognitive classes in having potential links (“bonds”) to
words and to syntactic structures. Later chapters will show how they are used in words
and syntax, becoming part of larger structures and of more complex semantic classes.
4.7 Uses of meaning
4.7.1 Introduction
This section deals with three pairs of uses of meaning, such as figurative and literal
use. They are the result of different speaker intentions, and consist of systematic mod-
ifications of the meaning types discussed so far, the modification being carried out by
the linguistic structure in which they occur, or by the content of the context.
4.7.2 Marked use
The rules of language, established by convention, are often broken in ways that have
become accepted by a secondary convention. The rule for sentence stress, for exam-
ple, requires it to be at the end of the intonation group; but placing it earlier is an
established way of making a contrast, in “marked use”. We have noted in Chapter 2
(§2.2.5.5), the marked use of rhythm, giving an utterance an emotive effect that un-
marked rhythm does not. Semantics has marked uses, similarly, such as the following:
the ironic use of words; neologisms (e.g. “disload”, “he had to un-shy himself ”); re-
vivals (e.g. “perpetuate” as a modern intransitive use); and use of proper names de-
scriptively, as in “a very Jarmusch vampire movie”, where a director’ s name becomes
a descriptive word. The distinction is important, because some statements about se-
mantic structure will apply in unmarked use, but not in marked use.
4.7.3 Defining and descriptive uses
Introduction. This section deals with a broad distinction between two strategies in
using language: to narrow meanings down, and to expand meaning. The two uses take
various forms, and have several functions; the narrowing-down use, for example, is
used to define meaning, or restrict detail, or identify a referent. (Consequently, “de-
fining”, “restrictive” and “identifying” (or “referential”) are loosely synonymous here;
so are “descriptive” and “non-restrictive”.)
Restrictive and non-restrictive use in modification. The most familiar occurrence of

the distinction is between restrictive and non-restrictive adjectival clauses. In non-re-
strictive clauses, the hearer must apply the content to the noun as extra information,
as in “Tom, who had been waiting, was…”; the speaker’s intention is to describe. In
restrictive clauses, such as “The main who had been waiting was…”, the speaker’s in-
tention is to define the referent. The distinction applies also in nominal modification,
compare “I’d like a big black one” (descriptive intention), and “I’ll buy that big black
one” (intention to define). It explains such apparent anomalies such as “An impatient
President Obama”, which is acceptable descriptively, but not restrictively. It also helps
to explicate the difference between a phrasal construction and a lexical item with two
or more words. For example, “white wine”, “short story” and “natural law” are phrasal
if the modifier is intended descriptively; but if the modifier is used restrictively, the
pair of words becomes a technical term, and can properly be called a lexical item.
Leffel et al. (2014) found neurolinguistic differences in the processing of restrictive
and non-restrictive modifiers.
Referential and descriptive use of words. There are several other manifestations of this
distinction. An important one is the distinction between referential and descriptive
uses of words considered on their own. The nature of referential use is seen clearly
in proper nouns, such as Canada and other names of places, people, books, and so
on. On hearing the name, our minds may go directly from the central bare node (see
§4.4.2.3) to identifying the referent, bypassing the perceptual and conceptual elements
such as ‘maple leaf on the flag’, for Canada; the word in that use will thus clearly have
no descriptive meaning. (The neurolinguistic mechanism is presumably that the links
to the perceptual and conceptual elements (§4.4.2.3) may receive some stimulus, but
do not reach the threshold for the elements themselves to be activated; that is, not
enough neurons are stimulated (Pulvermuller 2002, p. 101).) If, instead, our minds go
to the referent indirectly, via an evocation of those descriptive elements, we then real-
ise that they are outside the speaker’s intention, and exclude them from the meaning
of the text being built in the “mental model”, though they may remain available to be
called on. Speakers can have “lots of relevant information without explicitly represent-
ing it during cognitive processing.” (Horgan, 2012, p. 557.) We sometimes have alter-
native terms, a referential term and a descriptive one; examples are “Grover’s disease”
and “transient dermosis”, “cats’ eyes” and “raised reflectorised pavement markers”.
Many words have both referential and descriptive uses. For example, after a ref-
erence to a married couple, “the man” would be referential, identifying which person
of the two is intended; possible elements such as ‘aggressive’ and ‘strong’, and even
‘adult’, will not be treated as additions to the mental model. In “Come on! Be a man!”,
however, man is used descriptively, and hearers will treat elements like ‘aggressive’
and ‘strong’ as relevant. That is shown in the lexicography: man <1> is “A human be-
ing”, but man <3b> is “An adult male eminently endowed with manly qualities.” Some
words, however, are designed, as it were, for only one use. Deictics and pronouns
are referential, in unmarked use, like proper nouns. Expressions like thingummy and
what’s-his-name can hardly be used any other way – they have no content meaning.
Words with natural descriptive use include many property words, such as foolish and
meticulously, and the related entity words such as fool. One of the values of acronyms
such as “PC” is that we can take them either way: we can bring ‘personal’ and ‘com-
puter’ to mind and build up a complex meaning, or we can go straight to identifying
the appliance referred to, without bothering with those abstractions.
This distinction explains why we often use co-referential words in the Subject and
the Complement, as in “The man’s a fool!”; and the difference between “It was not a
picnic” (referential) and “It was no picnic” (descriptive). It also explains the surface
contradiction in “Platitudes are not platitudes while they are being tested in the fire
of personal experience” (attributed to Samuel Johnson”, and in “a rather un-American
American citizen” (invented phrase), since “un-American” is used descriptively, but
“American” is used referentially. The two uses are the two “roles” for noun phrases
given by Rijkhoff (2002, pp. 227–229).
Used referentially, common nouns become “names”. In “Cats are mammals,” cats
is used as a name; apart from its referential function, it is semantically empty; indeed,
if it were being used with its descriptive meaning, the sentence would be redundant;
since the descriptive meaning of cats includes ‘mammal’. We have seen, in the last
chapter (§3.4.3), that naming is a distinct semiotic process from describing. In this
defining or referential use, words fit the description which Lyons (1977, pp. 197–206)
gives for proper names: they “differ from other noun phrases in achieving […] ref-
erence independently of the semantic characteristics of the words out of which they
may be constructed.” That use of name is in the philosophical tradition of Kripke (see
Hanks 2013, p. 339, for example); but it is very misleading to think of names as a class,
since even proper names can be used descriptively (§4.7.2): rather, we should think of
referential use of words.
The referential use of words is further illustrated in example (11), which is from
the transcript of the conversation between an air-traffic controller and the pilot of
“Flight 1549”, who successfully crash-landed his plane in the Hudson River, after hit-
ting birds.
(11) [Flight 1549] “Maintain one five thousand, Cactus 1549.”

[Controller] “Cactus, turn left, heading two seven zero.”
(Reported in New Zealand Herald, February 7th 2009, A21)
Example (12) follows on immediately; in the emergency, the pilot lost his carefully
learnt referential use, and reverted to descriptive use.
(12) [Flight 1549] “Ah, this is Cactus 1549, we hit birds, we lost thrust in both
engines, we’re turning back to LaGuardia.” (Source as above)
The distinction affects the extent to which possible meanings are evoked in a particu-
lar use, but is not a distinction between evoking all such meanings and evoking none.
Consider examples (13, (14) and (15).
(13) “The two-year-old tabby was horrifically injured… The cat would have to the
put down.” (British National Corpus – “BNC” hereafter; news report)
(14) “You draw back your front foot into a cat stance, withdrawing your body from
danger.” (BNC; description of karate)
(15) “There was once a cat called Nutmeg.” (BNC; children’s story)
In example (13), tabby is used descriptively; cat is used referentially, adding no de-
scriptive elements at all to those that have been already evoked by “two-year-old
tabby”. Example (14) is used very descriptively; a visual image, and the qualities of
alertness, nimbleness, and many others are all relevant. In (15), cat is used referen-
tially, but some descriptive elements, such as ‘pet’ and ‘furry’, must be invoked for
the reference to be effective; they are actualised in that use, the elements of alertness,
nimbleness and so on remaining potential, available for later actualisation.
Example (16) is a short text which illustrates descriptive use in several ways.
(16) “A burning fug of teargas mingled with the acrid black smoke of blazing tyres
hung over the streets of Kiev as protesters and police battled for control of the
city’s historic centre.” (New Zealand Herald, January 24th 2014, A21)
Fug, acrid, blazing and historic are generally used only for a descriptive purpose; burn-
ing, mingled and battled become descriptive in this context. That enhances the de-
scriptive nature of the content.
Other works that are compatible with this account include Bolinger (1986,
p. 103), Coates (2000), Anderson (2003), Bauer (2004), and Cruse (2011).
Discussion: defining and descriptive uses. The understanding of uses given here partial-
ly reconciles the opposing views in each of two long-running controversies. When a
word is used referentially, its meaning is not decomposable, because it has no descrip-
tive elements. But when the word is used descriptively, the meaning is decomposable.
It thus has an “atomic” or “holistic” meaning in some uses, but not in others. See the
discussion in Wunderlich (2012), for example.
This understanding can also be seen as a possible reconciliation between polyse-
my and monosemy. If we choose to define the word meanings by their central refer-
ential node, then they are monosemous; if we define them with descriptive elements
included, then many are likely to be polysemous. (Polysemy involves other issues, and
will be discussed further in a later section.)
Switching between defining and descriptive uses is a powerful mechanism for
producing new senses of words, as in using “a radio” for “a radio set”, “a television” for
“a television set”, and so on. Radio changed from descriptive use as a modifier (with
the conceptual meaning, ‘for communicating by radio-frequency waves’) to typically
referential use without those concepts as necessary meaning. The process continues
actively, as with a recent television advertisement for cars costing “$24,990 + on-
roads”, i.e. on-road costs, with “on-road costs” still more recently reduced to “ORC”.
Brownell (1988) gives neurolinguistic support from research showing that in ref-
erential use the descriptive elements of meaning are inhibited by the left hemisphere
of the brain, but remain available in the right hemisphere if wanted later.
4.7.4 Literal and figurative uses
The issues involved in literary and figurative language are both complex and highly
controversial. This book cannot deal with them fully, but should touch on them, be-
cause they involve a couple of important issues in semantic structure. I will assume
that there are distinct literal and figurative uses of words, which some linguists deny,
with the expectation that discussion in later chapters will vindicate the assumption by
showing that the two uses come from distinct intentions and lead to distinct semantic
structures. Brownell (1988, p. 29) gives evidence that the processing of metaphorical
meanings is distinct from that of literal meanings.
4.8 Discussion: Elements of semantic structure
The analysis in this chapter has provided an answer to one of Cruse’s list of unsolved
problems, which he believes to be the most fundamental question of all: “How does
language connect with things and events in the world around us? How does the whole
system work?” (2011, p. 449). The answer to the first question has been basically sim-
ple, and now has general agreement among functional linguists of all kinds, psycho-
linguists and neurolinguists: language connects with the world around us through
meanings, because they use the perceptions in our cognitive, affective and other fac-
ulties – experiences which are responses to the inner and outer worlds. I believe that
we have now reached the point where there is no dissent; however, not all linguists
deal with the issue.
As to the question of how the system works: the general issue must be left to
psychology; but linguistics has a partial answer, which is that the system works as the
operation of grammatical meanings on content meanings (to be amplified in Chap-
ters 6 and 12), except for language in the Expressive function, which is immediate and
without workings (to be amplified in Chapter 11).
4.9 Conclusion: Elements of semantic structure
Summary. The chapter has described the resources that speakers have available in
fulfilling their intentions and carrying out the functions surveyed in the previous
chapter. Speakers use content meaning, which has as subtypes descriptive, emotive,
attitudinal, and social meaning; the descriptive meanings have concepts as elements;
the other subtypes have comparable elements; those elements are defined by their
various dimensions. Speakers use grammatical meanings to relate and structure the
content; from the speaker’s point of view, those meanings amount to instructions;
they are procedures when seen from the hearer’s point of view, and grammatical sta-
tus or features from the system point of view.
The content elements can be grouped into semantic “classes”, namely entities,
events and properties, according to their potential bonds; a member of a particular
class may consist of a single element, or of a cluster of them. All types and forms of
meaning have different uses, such as the defining and the descriptive.
Looking back, and forward. The chapter has complemented Chapter 3 in giving a com-
prehensive, although not fully detailed, explanation of meaning and of what is meant
by “semantic”. Together, they provide the foundation on which the rest of the book
is built. The elements outlined in this chapter are used in various combinations to
build the semantic structures of English, which are set out in the following chapters.
The most basic structure is the network, which is described in the next chapter. The
basic elements of that formal structure are reworked and developed by an operational
structure, described in Chapter 6, into hierarchies – another form of fully developed
formal structure – which are described in the following Chapters, 7, 8 and 9. Chap-
ter 10 describes less frequent and less fully developed structures.
Chapter 5
Network structure
5.1 Introduction
5.1.1 General introduction
This chapter applies the approach and concepts outlined in the previous chapters to
the first major type of semantic structure in English, that of networks. It is treated
first because it applies primarily to words and their senses, which are commonly tak-
en to be fundamental to all linguistic structure. The main purpose is to explain the
network nature of relationships among word senses, and to show that senses have in-
ternal structure which is also a network. A secondary purpose is to describe semantic
networks that appear in the higher layers of the language. (The other major type of
semantic structure, to be dealt with in later chapters, is that of hierarchies.)
Section 5.2 describes the semantic structure within words – of lexical items, more
strictly – as used in utterances. Section 5.3 deals with the structure that exists among
word senses, consisting of relations such as synonymy. Section 5.4 deals with the se-
mantic structure of senses in the “mental lexicon”, which are simpler than the senses
of words when inflected and used in a context. Section 5.5 deals with semantic net-
works that occur within syntactic structures, which are normally hierarchic. Discus-
sion and conclusion sections follow.
The coverage of this chapter is more limited than may be suggested by the previ-
ous paragraph. Just as there are different kinds of meaning, as discussed in Chapter 4
(§4.3), so there are different kinds of sense; not all of them are related as parts of a
network. Specifically, a grammatical meaning relates only to the content meaning it
operates on, not to several senses and not on differing dimensions, as is characteristic
of semantic networks. Secondly, nondescriptive meanings lack the internal complex-
ity that goes with the conceptual nature of descriptive meanings; consequently, their
relations are largely those of units in a hierarchy, not the interrelationships of a net-
work. Holophrastic expressions do not need discussion either, because their lack of
determinate structure prevents them from having determinate relations with other
expressions. Finally, idioms are given little direct attention, being regarded simply as
lexical items.
5.1.2 Introduction to networks
In general use as an abstract term, “network” denotes “a system of interconnected

or intercommunicating immaterial things” (SOED). It has nodes which are terms of
relationships; the links are the relations. Hierarchies, including those represented in
tree structures, are similar, but have simple horizontal and vertical connections, not
interconnections, and have a “top” and a “bottom”. Networks have neither top nor
bottom, and have no set entry or exit points. Network structure is fundamentally
different from constituent structure, with its parts and wholes.
All of that is true of the networks in language to be described here. There are sev-
eral extra points to note, however. Semantic networks have several kinds of relation-
ship: both co-ordinating and subordinating, both logical and associative, and so on.
Also, “node” and “link” are relative terms; for example, when we discuss the structure
of a complex sentence, we treat each clause as a node, but when we discuss each clause,
we treat it as a network with its own nodes and links. Next, learning in a network is
seldom a matter of adding new units; it is usually building new connections between
existing nodes, strengthening existing connections, or adjusting the thresholds which
control their activation (Lamb 1999). Finally, while the interconnections are always
present, the intercommunicating is not. Neurally, the networks exist at all times, since
they consist of nerves; but linguistically, they exist only during use, for as long as ac-
tivation lasts; when we generalise about it – as “the semantic structure of English”, for
example – we are describing an abstraction. To describe a linguistic network as if it
were a sum of parts – simply compositional – would be seriously misleading; meaning
lies in connectivity (Lamb, 1999, p. 241), not in a series of units.
General support for semantics as a network has come most clearly from Lamb
(e.g. 1999) and others in his school of thinking, such as the writers in Lockwood and
others (2000). Other support is in Cruse (2011, §8.1.3), which does not use the term
“network”, but presents sense relations as situated on three axes, the paradigmatic,
syntagmatic and derivational. Lyons (1968, 1977) also presents meaning as basically
relational in a way that is tantamount to a network. According to Langacker, as quoted
by Inchaurralde (2000, p. 104): “The entity designated by a symbolic unit can […] be
thought of as a point of access to a network. The semantic value of a symbolic unit
is given by the open-ended set of relations… in which this access node participates.”
5.2 Structure within a word’s meaning
5.2.1 Introduction
5.2.1.1 General
This section deals with the structure of the meaning of words, or, strictly, meaning of
lexical units, including idioms as well as words. It is assumed that lexical units often
have several senses, semantic units of the kind identified by successive numbers in
Chapter 5. Network structure 81
SOED entries, for example. These are senses of words as used in utterances – lexical
senses. As noted in the previous chapter, those senses are distinct from sublexical
senses, which are treated in a following section. A word with both descriptive and
emotive meaning (for example) is treated as having a combination of two senses, since
the different meaning types have different types of structure.
Grammatical meanings are not discussed here; they do not have constituent
structure, and are here distinguished from senses. The syntagmatic relations between
word senses, which structure groups, will be dealt with in the later chapters on figure
and group structure, rather than being treated here as constituents of word senses.
That goes with treating hierarchic structure separately from network structure, for the
sake of clear exposition.
5.2.1.2 Problems in identifying senses

I define a sense as the content (descriptive or nondescriptive) by which people repre-
sent the meaning of the word; that is, the image schemas, concepts and so on which
come to people’s minds as the meaning of a word they encounter. (That is based on
Barsalou’s definition (1992b, p. 31) of concepts.) But as we begin to think closely
about senses, difficulties arise. It is commonly assumed, in the literature on polysemy
for example, that senses, like words, are quite distinct – we can count them, after all.
If they are distinct, senses are units, and there must be fairly well defined boundaries
between them.
However, the nature of their unity is elusive It is natural to think of it as arising
from the things they denote in the real world: one thing, one sense. If we look for
the unity by taking the sense as a single concept, we face the problem that we want
to define the concept, but that entails other concepts, and so on indefinitely – or in
a circle. If we try to work by linguistic senses instead of cognitive concepts – on the
principle of one sense to one linguistic referent, for example – we have the problem
illustrated by serial verbs. Compare “let go” and “release”, “let fall” and “drop”, as to
whether each pair has one sense or two; and in many languages, asking your dog to
“fetch” the newspaper would be expressed as asking him to “go pick-up carry bring
give” the newspaper: the apparent unity of the real event denoted by a sense is arbi-
trary. However, we can see unity in the singleness of a sense’s syntagmatic function,
and in its unifying central node (see below, §5.2.3).
There is also a problem with boundaries. Elements or features cannot be in two
senses at once, since that would require copies of the feature (a copy in one sense
and another copy in the other sense), which would have to be identical, which would
require verification from an original, and so on indefinitely. The spatial metaphor
entailed by “contain”, “in”, and so on is clearly unsatisfactory. See Cruse (2000) for
further discussion of the issues.
We must accept that sense elements make a network, without boundary lines, and
without senses defined by a unique combination of elements. That network view has
psycholinguistic support from Fortescue (2009, 2010), for example, based on work
by Pulvermüller in particular. Senses are represented in the brain by linked bundles
of neurons, which form columns in the cortex. The minimum linguistic unit is repre-
sented by a group of physical columns, with multiple links to other columns, so that
each minimal unit can be accessed by many others. A minimal sense is “represented”
by a mini-network of those minimal cortical units, with a “word column” as the focal
centre of the pattern. Activation of a linked unit is not automatic; the stimulation
must reach a threshold strength, and certain forms of stimulus will inhibit the con-
nection, not activate it. The sense, then, is a gestalt–like pattern of activation. It resem-
bles the figures displayed by a digital watch. The different segments of the display can
be activated in different combinations, to make “1”, “2” and so on; it is the pattern of
activation, not the segments themselves, that constitutes the meaningful figure.
From the “unity” and “boundaries” point of view, a sense is like an island which
is a peak in a submerged mountain range; the peak makes a distinct unit, and the
island seems isolated, but the base overlaps those of other mountains. Accordingly, it
is appropriate to use SOED’s numbered senses as starting points for discussion, since
their sharpness of distinction from other numbered senses has some theoretical jus-
tification, and their variety of subsenses separated by commas and semicolons gives a
useful indication of their radiating network links.
5.2.2 Word senses as a structure of meaning types
5.2.2.1 Synchronic structure

As indicated in the chapter on elements of meaning, words commonly have senses of
different meaning types. For example, brave as adjective is given in SOED as follows.
(I have adjusted graphology and abbreviations a little.)
<1> “Courageous, daring; intrepid; able to face and withstand danger or pain. L15”
<2> “Splendid; spectacular; showy; handsome. Now literary. M16”
<3> “Of excellent quality, admirable; fine, highly pleasing. Frequently an exclama-
tion of approval. Now archaic or dialectal. L16”
Sense <1> has descriptive meaning, with concepts such as COURAGEOUS, WITHSTAND,
DANGER, and PAIN. Sense <2> has a semantically more complex structure: the attitudi-
nal meaning of approval and the emotive meaning of pleasure (implicit in the choice
of synonyms); the social meaning of being literary (stated explicitly); and descriptive
meaning, with the concept STRIKING IN APPEARANCE, which I take to be common to
the various affective synonyms. Sense <3> has the same attitudinal meaning and emo-
tive meaning as sense <2>; it has the social meanings of being dialectical in some uses,
and archaic in other use. I believe that it now has no descriptive meaning, although
the synonyms given by SOED suggest that it does; its lack of descriptive meaning is
clear from the Oxford English Dictionary (“OED”), which notes that in this sense the
word is used “loosely, as a general epithet of admiration or praise”, as in Shakespeare’s
“O that’s a braue man, hee writes braue verses, speakes braue words” (As You Like It
1623). SOED’s note, “Frequently an exclamation of approval” shows that sense <3>
brave <3> (iii)

brave <2>
brave <1>
Descriptive Attitudinal Emotive
meaning meaning meaning
(‘high quality’) (approval) (pleasure)
Social
meaning
(archaism)
Grammatical
meaning
(exclamation)
Diagram 1. Meaning types in the three senses of brave
also often has grammatical meaning, consisting of the instruction to the hearer that
the one word is to be taken as a complete utterance, a “speech act” of exclamation.
The word, with its three senses, is semantically a network: there are many dif-
ferent elements of meaning, related more or less closely (in various ways, not in a
continuum), and with some overlap, as in the occurrence of EXCELLENCE, explicit-
ly or implicitly, in all three senses; the different meaning types constitute different
dimensions of the network. The broad structure of that network – its structure of
meaning types – can be set out usefully as a semantic map (Haspelmath 2003), as in
Diagram 1.1 It sets out what is sometimes called “conceptual space”.
The map is to be understood as follows. Each box represents one sense of the
word, with its contents representing the constituent meaning types; the sense is iden-
tified in italics just above the box. Thus, brave <1> is shown as having only descriptive
meaning; brave <2> shares descriptive meaning with brave <1>, and shares attitudi-
nal, emotive and social meaning with brave <3>, but excludes grammatical meaning.
(Other features of semantic maps will appear in later sections.) For sense <3>, the
third subsense is represented – the exclamation of approval – and labelled with “(iii)”.
(The social meanings are simplified in the diagram.) The network nature could be
emphasised by linking all the senses with lines; but that has not been done, to keep
the diagram clear.
1. To Haspelmath, semantic maps are only of grammatical meaning, not of content.

5.2.2.2 Diachronic development of the structure

There is no obvious reason why sense <1> of brave is only descriptive, while sens-
es <2> and <3> have affective and social meaning as well. That apparently arbitrary
connection between meaning types can be understood through history – as can the
sequence of the three senses in SOED.
The most common way in which word meaning types have developed is that a
sense begins with descriptive meaning and later acquires other types of meaning.
Speakers associate with the word the feeling they have when using it, or the feeling
they have for the referent; that association becomes conventionalised as standard for
that use, and the feeling becomes established as a sense. The same process applies to
attitudinal meaning, which often develops at about the same time as emotive mean-
ing. Social meaning evidently is also acquired by association, but the association is
with the situation in which we hear the word, rather than with our own use of it.
‘Admirable’ is an archaic or dialectal element of brave <3> because we meet it only in
old literature or as an element in certain dialects.
I will illustrate the historical development of word meaning a little further (ex-
panding the analysis in Chapter 4, §4.4.1), to highlight the importance and semi-
autonomy of the different types of meaning, and to demonstrate the relevance of the
analysis of semantic structure given in this book to the understanding of word history.
The etymology of brave given by SOED begins with the Greek barbaros, which meant
‘non-Greek, foreign’. Latin took it over, then Italian, and Spanish, in which language
it meant ‘savage’, and by extension ‘courageous’. Then French “borrowed” it, and then
English, in the 15th century, in sense <1>. By the 16th century it was being used in-
creasingly with its quality dimension changed by loss of ‘courage’ and addition of ‘in
appearance’ (“showy, handsome” in SOED’s words); those changes shifted it towards
the general end of that dimension. There was evidently a social consensus on approv-
ing what looks good, so approving attitude became a standard association with the
descriptive meaning. Brave had thus acquired sense <2>, which I will paraphrase as
‘of good quality in appearance’, while sense <1> remained in use (‘of good quality in
courage’). By the 17th century, that new sense was often used with the ‘good’ element
shifted towards the vague end of the specificity dimension: the descriptive meaning
was simply ‘of good quality’. The attitudinal meaning of approval had strengthened in
this use, which added the affective meaning of pleasure (“highly pleasing” in SOED);
the word was thus often used in the Expressive function (“Frequently an exclama-
tion”). Brave now had sense <3>. In the centuries since then, this sense has dropped
out of standard use, remaining current only in certain dialects; therefore, it has the
social meaning of being either archaic or dialectal. (Uses as noun and adverb devel-
oped from the adjective quickly, in the 16th century; the verb was borrowed directly
from French.)
Those developments are represented in Diagram 2. It uses the conventions of the
diagrams in Chapter 4, but adds arrows to highlight historical change, and adds text
between the boxes to explain the changes.
Feeling is added
[Pleasure]
to attitude
Approving attitude
is added. [Approval] [Approval]
Of good Descriptive meaning [No

‘In courage’ is replaced Of good
quality, weakens descriptive
by ‘in appearance’. quality,
in courage and disappears elements.]
in appearance
<1> <2> <3>

“Courageous, “Showy, “Highly pleasing”
daring.” handsome”
15th C. Early 16th C.[?]. Late 16th C.
Diagram 2. Historical changes in types of meaning, in brave
Descriptive Attitudinal Emotive

meaning meaning meaning
Social
meaning
Grammatical
meaning
Diagram 3. Historical development of meaning types
The pattern of semantic development is shown in Diagram 3, which is a dia-

chronic version of the semantic map given in Diagram 1. It has the same entries, in
the same pattern; but there are no boxes, because the map is not illustrating different
senses. The lines link entries that are related historically, as well as being similar, and
the arrowheads indicate the direction of change.
5.2.2.3 Independence of the meaning types

Words’ affective, attitudinal and social meanings are relatively independent of the de-
scriptive meaning, which is evident in several ways. The types occur on their own:
brave <1> has descriptive meaning alone; brave <3> as exclamation consists of af-
fective meaning alone, as do frightful, horrible, lovely, nice and so on; hello and please
have social meaning alone. When types do combine, the link between them is linguis-
tically arbitrary; ‘intelligent’, as descriptive meaning, combines with favourable attitu-
dinal meaning in quick-witted and shrewd, but combines with unfavourable meaning
in crafty and cunning.
There is one way, however, in which a meaning type may be dependent on anoth-
er, seen in the historical development of word senses. Attitudinal meanings develop
from abstract senses, not from concrete ones, because it is only about abstract quali-
ties that there is enough social consensus to establish an attitude as a standard part of
a word sense. Similarly, affective meaning seems to depend on a previous attitudinal
meaning.
5.2.2.4 Discussion: Word meanings as a structure of meaning types

The word meanings given in dictionaries such as SOED seem to present no more
than the descriptive meaning, since they are nearly all worded in the same abstract
descriptive way. (I am putting aside here the distinction between linguistic meaning
and cognitive or “encyclopaedic” meaning.) However, the definitions do testify to the
analysis of meaning I have been giving. First, synonyms are given carefully and sys-
tematically, evidently to provide words with the same descriptive meaning but differ-
ent affective or attitudinal meaning; that shared descriptive meaning (never identified
directly, unfortunately) represents the necessary meaning (Chapter 4, §4.3.9). Social
meaning is not given descriptively as meaning, but is sometimes indicated by tags,
such as “coarse slang”. (SOED gives very few; COBUILD (2001) gives many more.)
Grammatical meaning is given as part of speech, which does not give the nature of
that meaning clearly, but at least it is done regularly. Lexicographical practice thus
confirms the reality and importance of what I have called meaning types, although
it accounts for them differently. One implication of the argument of this book is that
lexicographers should change their treatment of meaning.
The distinction between descriptive, emotive and attitudinal meanings as struc-
tural elements is supported neurolinguistically by Fortescue (2010, pp. 195–197). As
noted in §5.2.1.2, words are embodied in the brain by bundles of neurons forming
columns which run vertically in the cortex. The neurons have branches or “affor-
dances” running horizontally from different levels of the column; the neurons for
descriptive meaning run from the mid-level to lower and middle layers of the cortex,
and those for grammatical meanings run from the top of the column to the upper lay-
er of the cortex and to greater distances. At the bottom of the column are affordances
that lead downwards to the limbic system of the brain, serving emotive and attitudinal
functions. (Fortescue does not use my terms for meaning types.)
5.2.2.5 Conclusion: Word meanings as a structure of meaning types

The senses of different types – descriptive, affective, attitudinal and social – are par-
tially interdependent; but the nature of the relationships varies, and their occurrence
in particular lexical meanings is partly arbitrary. This part of the network is quite
loosely structured.
5.2.3 Structure within descriptive senses
5.2.3.1 Basic structure of descriptive meaning

In discussing the types of meaning in Chapter 4, I described descriptive meanings as
being structured in three parts, with a central node, percepts linked to it, and concepts
linking them to wider meaning. Schreuder and Flores d’Arcais (1989) explain that
structure as follows, using the meaning of coffee as their example. The meaning ‘coffee’
has, first, a minimal node, which by itself has no content. That content is provided by
“perceptual elements”, such as colour, smell, and taste (using the concrete concepts
BLACK, LIQUID, etc.), and by “functional elements” (using abstract concepts giving
such information as how coffee is produced and where it is grown). Those elements
of the network are activated together or independently in different general situations,
such as those of experience, thought, and speech, and in different speech situations,
such those of referential and descriptive use (Chapter 4, §4.7.3). What Schreuder and
Flores d’Arcais call the central “node” can also be thought of as a link, since it has
function, not content of its own.
This structure explains much of the difference between the different senses of a
word. For example, fetter <1> is “A chain… for the feet…” and fetter <2> is “Anything
that confines…”; the word is linked to the perceptual elements ‘thing’ and ‘chain’, and
to the functional elements ‘confine the feet’; sense <1> activates them all, but sense
<2> activates only the functional ones.
5.2.3.2 Structure of descriptive elements

To Schreuder and Flores d’Arcais (1989), the minimal central node of a sense is the
access point for the whole sense in comprehension, and for thinking of the word out
of context; it links the word form with meaning, and thence with knowledge. When
it is activated out of context, all of the various perceptual and functional elements
become accessible. When it is activated in context, their activation is selective, not
automatic. In descriptive use (e.g. “It ain’t got the real coffee taste” – BNC), many of
the elements will be accessed, and brought into the hearer meaning. In referential
use, when a word is used simply as a name for the referent (e.g. “Pass me the coffee”),
all of the descriptive and linguistic nodes are bypassed – the outer nodes – with the
referent in cognition being accessed instead, rather as pronouns and gesturing make
us bypass descriptive meaning, accessing the referent in the previous sentence or the
physical situation.
Accessing the sense in production – that is, when we find a word to express the
meaning being formulated – is quite different. Psycholinguistic studies of priming
(e.g. Mcnamara 2003) show that word senses can be accessed from a related meaning
of any type, and common experience confirms that. We can bring a word and its sense
to mind in searching for an antonym or synonym, for a more or less general word, or
for other technical senses. (Access can also be through nondescriptive elements, such
as slang value or affective meaning.)
According to Barsalou and Wiemer-Hastings (2005, pp. 135–136), there is also a

difference in what is accessed by concrete and abstract uses. Abstract uses of a word
activate situational elements primarily, such as the function and location of use; con-
crete uses focus on the object or action. (I am extending a little their general linkage of
concepts in thought with concepts in meaning.) Fortescue (2010), as cited in §5.2.2.4
above, supports that neurolinguistically. For instance, the relatively abstract use of cof-
fee in example (1) will activate ‘infusion from the seeds of a shrub’ (from coffee <1>),
but the more concrete use in (2) will activate ‘a drink’ (also from coffee <1>).
(1) “Do not make coffee with boiling water.” (BNC)

(2) “I am woken up gently with a strong cup of black coffee.” (BNC)
I suggest that the unitary nature of the referential use explains why meanings are very
often treated as representing a single concept, although dictionary definitions (which
give the meaning of descriptive uses) present them as conceptually complex.
5.2.3.3 Structure in the quality dimension

Introduction
In Chapter 4 (§4.3), descriptive meaning was said to have various dimensions: quality,
degree/intensity, generality, specificity and so on. The relevance of those dimensions
to word meanings as we usually conceive them can be illustrated from the SOED
definition of brave <1>, cited in §5.2.2.1; “Courageous, daring; intrepid; able to face
and withstand danger or pain.” The substance of the definition consists of quality
elements: ‘courageous’, ‘daring’, ‘intrepid’, and so on. As in most SOED entries, the
other dimensions are not treated explicitly, being left to readers’ understanding of
the synonyms given. This section deals with the quality dimension, the others being
treated in the next section.
The constituent concepts (“conceptual elements”, strictly) in brave <1>, and their
relations, can be set out in a semantic map, as in Diagram 4. The diagram uses the
following elements: ‘undaunted’, ‘confident’, ‘enterprising’, ‘not well considered’, ‘taking
risks’, and ‘unforced’. Those words for the concepts have been derived, using SOED,
to give non-circular definitions for the words used in the SOED definitions of brave
<1>, as follows.
– Courageous: ‘undaunted’. (That is, “able to face and withstand danger or pain”;
courageous means ‘undaunted’ + formal social meaning.)
– Intrepid: ‘undaunted’ and ‘confident’.
– Bold: ‘Intrepid’ and ‘enterprising’, and ‘taking risks’.
– Daring: synonymous with bold on this qualitative dimension; differing on the
degree dimension, being marked as higher in degree.
– Adventurous: ‘bold’, the risks taken being ‘not well considered’.
– Audacious: ‘bold’, the risks taken being ‘unforced’ (i.e. not forced on the person
acting).
UNDAUNTED
co-occurrence UNFORCED
property
TAKING
cause / effect
RISKS
cause / property
effect
NOT WELL
CONFIDENT
CONSIDERED
cause / effect
ENTERPRISING
Diagram 4. Semantic map of brave and its synonyms; (a) descriptive sense elements
and relations
‘Undaunted’ is the necessary sense element in brave <1>, the element that is always
present, the minimal meaning – the “invariant” meaning in the some schools of lin-
guistics. The full list of the elements makes up the possible meanings.
Diagram 4 maps those conceptual elements, showing them as nodes, in upper-
case letters. The conceptual relationships between them are represented as links, in
lower-case letters. (As noted previously, networked concepts acting as links in one
use may be more salient and act as nodes in another sense.) Thus, on the left of the
diagram, confidence and being undaunted are represented as related by cause and
effect; on the right, being unforced is represented as a property of the risks taken. (The
relations given are intended as common links between the concepts; in particular uses
of ‘undaunted’, ‘confident and so on, the relations may well be different.)
brave <1>
audacious
adventurous
daring / bold
intrepid UNFORCED
courageous
UNDAUNTED TAKING
RISKS
NOT WELL
CONFIDENT
CONSIDERED
ENTERPRISING
Diagram 5. Semantic map of brave and its synonyms; (b) whole senses
Diagram 5 shows how the network making up sense <1> operates in particu-
lar uses of the word. Use of brave to mean ‘courageous’ will evoke the concept UN-
DAUNTED. Use to mean ‘intrepid’ will evoke UNDAUNTED and CONFIDENT (related as
in Diagram 4). The biggest box, labelled “brave <1>” represents the whole of sense
<1> – that is, all of the possible elements.
5.2.3.4 Structure in other dimensions

The dominant structure of descriptive meaning in use is that of its various dimen-
sions. By analogy, the structure of physical objects is defined by the length, breadth
and height dimensions; and vowels are defined by position on the frontedness and
openness dimensions.
For example, to take words for SIZE that might be applied to a house, enormous
is higher on the degree or intensity dimension than huge, which is higher on that
dimension than big. Big, roomy and 300 m² are equal in degree (let us say), but differ
on the specificity dimension, since 300 m² is quite specific, whereas large is vague, and
big and enormous are somewhere in the middle of the scale. Finally, on the generality
dimension, roomy and 300 m² are particular to houses, whereas big, large and enor-
mous are general in application. The meanings are thus defined by their position on
their various dimensions.
That analysis can be formalised in the kind of formula often used in semantics, as
follows. The distinctive elements are underlined, to highlight the differences between
the synonyms. (The senses are still those of the words when applied to a house.) The
formula for enormous is to be read as follows: “Of the words denoting size, enormous
denotes ‘size’ in a very high degree, and denotes it with medium specificity (that is,
specifying length and breadth but not depth), and denotes it with low particularity.”
– Enormous = ‘size’ (very high degree); (medium specificity (length + breadth));

(low particularity).
– Big = ‘size’ (medium degree); (medium specificity (length + breadth)); (low
particularity).
– Roomy = ‘size’ (high degree); (high specificity (length + breadth + height)); (high
particularity (housing)).
Those formulations are intended to be read as either (1) feature lists e.g. enormous is
‘size’ with the features very high degree, medium specificity, and low particularity; or
(2) function statements e.g. the size denoted by enormous is a function of enormous’s
very high degree, medium specificity and low particularity, just as the volume of a
brick is a function of its length, height and breadth. The formalisation may be carried
further, with numerical values such as “4” for very high degree, “3” for high degree,
and so on; the meaning may then be diagrammed with those values as a spatial di-
mensions, as in Diagram 6.
When a word is used in various domains, SOED defines the various uses sepa-
rately and deliberately. For example, it defines the military, clothing and sports uses of
right (noun1, branch II) as follows:
Specificity
Specificity Specificity
Partic- Partic-
SIZE SIZE Partic-
ularity ularity
ularity SIZE
Degree Degree
enormous big
Degree
roomy
Diagram 6. Formalisation of aspectual structure: enormous, big and roomy
– <9> “Military. The right wing of an army.”

– <11> “A shoe etc. for the right foot; a glove etc. for the right-hand.”
– <12> “Football & hockey etc. (The position of) a player who….”
Those senses differ only on the specificity dimension, so could be formalised as fol-
lows: <9> ‘Right’ (specific (military)); <11> ‘Right’ (specific (clothing)); <1> ‘Right’
(specific (sports)). Those definitions would not be very useful in a dictionary, of
course.
The expectedness dimension is particularly important in structuring descriptive
meaning. SOED’s entry for stiff <11> illustrates that well.
<a> “Of a wind: severe, strong; esp. blowing steadily with a moderate force.”
<c> “Of a drink, esp. of spirits: strong, potent. Also, of generous quantity.”
In subsense <a>, ‘severe’ and ‘strong’ are alternative necessary elements; “esp.” indi-
cates that ‘blowing steadily…’ is expected. In <c>, “esp.” indicates that the whole of
the subsense is expected to apply to spirituous liquors, and not others; “strong” and
“potent” are likely meanings; “also” indicates that “of generous quantity” is a possible
but unlikely meaning. “Of a wind” and “Of a drink” set particularity dimensions – ra-
diating, as it were, from the core meaning of stiff (which is ‘rigid’, given in branch I of
the entry for stiff); the expectedness dimension can then be pictured as specifying the
distance out on those radiating lines.
Accounting for distinctions in the semantic structure of events requires two di-
mensions not discussed so far. Duration is needed in specifying the difference be-
tween achievements such as ‘arrived’, which have no duration, and the other aspectual
types, which do have duration. In the conventional account of aspect, change is need-
ed as a dimension, to distinguish between states and other aspectual types. But in this
account, change is a cognitive dimension, and is relevant sublinguistically in sense
elements; but it is not a linguistic dimension of Events and Processes, as will be ex-
plained more fully in Chapters 7 and 8.
This section has not dealt with the basicness and boundedness dimensions, be-
cause they have been illustrated previously, and will recur later in the book.
5.2.3.5 Subtypes of descriptive meaning

In some circumstances, abstract property words are distinct from concrete ones, in
being gradable. We can grade “hot water” into “very hot water” or “hotter water”, but
cannot grade “running water” into “very running water”. Hot is gradable, being ab-
stract, but running is not gradable, being concrete. The distinction obviously applies
in morphology, but it is basically semantic; that is shown by the fact that some words
can have both gradable and non-gradable senses; “a very slow trip” is acceptable, but
not “a very slow zone”, for a road-speed zone. That distinction is important for struc-
ture in nominal groups (Chapter 8).
Fortescue (2010) supports the distinction neurolinguistically. As noted in §5.2.2.4
above, he reports that abstract and concrete meanings have distinct neuronal rep-
resentations: auditory, visual, spatial and motor meaning elements all have their own
affordances, lower on the cortical columns than the abstract ones.
5.2.3.6 Conclusion: Structure within descriptive senses

The descriptive elements in a sense constitute a basically conceptual network, with
patterns of elements on each of the many dimensions.
5.2.4 Structure within non-descriptive senses
Affective senses
Affective meanings are structured on the degree dimension, as well as by the kind
of emotion; to dislike, to hate and to loathe increase in intensity; delicious is stronger
than delightful (delicious <1> is “Highly delightful”). The other dimensions sometimes
seem to apply; but the difference seems to be in the descriptive meaning, and not in
the emotive meaning. For example, delicious (when applied to food) is more precise
than delightful, I believe, through having the descriptive element ‘in flavour’.
Attitudinal senses
Although I am usually distinguishing only between favourable and unfavourable at-
titudes, there are other distinctions which are significant in some circumstances. As
well as those two general attitudes, expressions may be characterised by more specific
ones, such as irony, sarcasm and making jokes. For the general attitude of joking,
there are the more particular attitudes of jocularity and facetiousness, as noted by
some dictionary entries. Expression of attitude often seems to differ in degree, but
I believe that to consist of variation in emotive meaning: to “take a very favourable
view” is to have a simply favourable attitude and a strong degree of feeling.
Social senses
The social meanings that consist of marking the speaker’s regional or social back-
ground as American, upper class and so on are unitary, not varying on any dimen-
sions. Other social meanings, however, vary in degree: words are informal to varying
degrees; dictionaries mark words in degrees of acceptability, as “not in polite use”,

“indecent” or “obscene”. Words may be technical in general, or nautical, military and
so on, in particular.
Expressive senses
Senses that are Expressive (in the strict sense, that is) have not been dealt with so far
in this chapter. Senses such as those of heck, oh, and blast it, when spoken spontane-
ously to no audience, have no structure. They are simple, consisting only of a quantum
of the affect expressed. However, as noted previously, expressive interjections some-
times lose some of their Expressive force, gaining structure by the addition of specific
emotive and attitudinal meaning. Blasphemies and obscenities obviously gain social
meaning as well.
Conclusion: Structure within nondescriptive senses

The nondescriptive senses are structured quite differently from the descriptive ones.
They typically have no internal structure of sense elements like the concepts of de-
scriptive meaning, and they relate directly to the nonlinguistic mental module which
they express, not through the mediating area shared by language and cognition.
5.2.5 Discussion: Compositionality of sense structure
The discussion here will concern descriptive senses only, since the others do not have
elements which could be integrated into a unitary sense, and the loose combination of
meanings of different types into a word sense is obviously compositional.
Considerations supporting compositionality of senses

It is intuitively obvious that senses are compositional to an significant extent: the
nature of dictionary entries, shows that, for example. Further, it is entailed in the
analysis of meaning into types and dimensions given in Chapter 4, and networks are
“composed” of nodes and links.
Linguistically, as Bierwisch and Schreuder argue (1993, §4.1), acceptance of com-
positionality is necessary for three reasons: to explain errors in the choice among
closely related words, to explain sense relations among words, especially synonyms,
and to explain lexical access. For example, we can access chair through thinking about
furniture, or about sitting down, or about four-legged objects; so those concepts must
be in some way part of the meaning of chair. I find those arguments powerful.
Compositionality is also entailed in the analysis of meaning into dimensions, as
in Section 2.3 above. It is apt for accounting for the descriptive use of words – extra
sense elements are invoked or inhibited according to what is described. It is neces-
sary to explain the relationship between linguistic semantics and cognitive semantics,
since word senses regularly differ from the corresponding concept by having extra
sense elements, or by lacking some.
Compositionality works particularly well in the cognitive domain, as with for-

mulae. Technical terms constructed morphologically work in full compositionality, as
with technical names, such as sodium chloride and sodium chlorate, and pittosporum
longiflora as a tree name; so do other terms with a strict definition such as primate.
Finally, there is psycholinguistic support for compositionality. Bierwisch and
Schreuder (1993) report that verbs are processed faster and remembered better, if
structurally complex (being more determinate helps). Pickering and Branigan (1998)
show that a word can be primed (i.e. activated) by a different one; the various words
must have a common element.
Considerations opposing compositionality of senses

Decomposition cannot account satisfactorily for relational senses, for example ‘ex-
ceeds’ and ‘father of ’. The relationship of father and son cannot be reduced to the two
constituents, father + son, or to three constituents, father + son + relationship. This
difficulty is more serious than it may appear to be, for two reasons. First, the whole of
meaning is a network, made up of relations. Second, the difficulty applies to “features”
of meaning, as Ferris points out (1983, §5.5). For example: on the featural view of
meaning, ‘ewe’ consists of the features ‘sheep’ + ‘adult’ + ‘female’; but we understand
the sense only when we establish linkage to experience or learned knowledge of the
animal (the ‘sheep’ feature), and to the significance of animals’ maturing (‘adult’) and
to sexuality (‘female’). When those linkages are made, the result is not compositional
but relational. Since most senses have “features”, most (and perhaps all) senses are
relational. Therefore, although features are in themselves compositional elements,
senses cannot be fundamentally compositional.
Three more specific points may be made. The sense of words used with purely
referential function is simple, as explained previously, and therefore not composition-
al. Levelt (1993, p. 7) makes a similar argument; see also Roelofs (1993). Second, it
seems impossible to account for associative meaning elements compositionally either,
as noted by Engelberg (2011, pp. 129–130): associations such as ‘red rag’ with ‘bull’
derive from knowledge of the world or from personal experience, and many associa-
tions vary from social group to social group; consequently, they are linguistically arbi-
trary, which weakens their compositionality. Some associated elements, however, are
linguistic to the extent that they are linked to particular words, not their synonyms,
and therefore not to the referent; for example, sails and hull associate linguistically
with painter and sheet – all having the social meaning, ‘nautical’ – but the association
of painter and sheet with rope is cognitive. Third, figurative uses seem to be not fully
compositional.
Discussion: Compositionality of senses

I believe that scholars’ lack of agreement on the extent of compositionality is due, to
an important extent, to their failure to consider some of the issues outlined above. In
particular, proponents of unitary meaning, such as Fodor and Lepore (2002), seem
to be thinking of words in referential use, where meaning is indeed unitary, and
ignoring other uses. Proponents of fully compositional senses seem to think only of
words wholly reliant on descriptive meaning, and therefore susceptible of definition,
which are naturally compositional.
Conclusion: Compositionality of senses

We conclude that senses are compositional to an important extent, and that the extent
is greater than is usually thought, since the analysis into types and dimensions of
meaning shows systematic construction of senses. On the other hand, there are sharp
limits to the compositionality, and it applies to the more superficial layers of meaning,
such as combining prefixes and roots.
Karmiloff-Smith (1979) notes that children decompose the meanings of complex
symbols into 1:1 symbols until they have mastered the meaning, then re-compose
them. Given the alternative processing routes that networks provide, they will retain
both patterns: word senses are likely to be both compositional and non-composition-
al, for different people, and for the same person on different occasions.
5.3 Structure among word senses
5.3.1 Introduction
The last section dealt with structure within an individual word sense; this one deals
with relations among two or more such senses – often referred to as “sense relations”
or “lexical relations”. The relations dealt with here are the paradigmatic ones of synon-
ymy, antonymy and so on; syntagmatic relations such as modification are dealt with
in Chapters 7 and 8, on hierarchic structure, since the relation of modifiers to head is
by its nature hierarchic.
I begin indirectly by recalling the scale between cognitive and linguistic domi-
nance, set out in Chapter 4 (§4.2.1), on areas of meaning. To Gentner and Boroditsky
(2001), the scale is a developmental one. Children’s word learning is at first dominated
by cognition, as they learn words that name things and concepts they already know –
cat, cup, bed, perhaps; relational terms such as aunt are more difficult to learn, relying
on knowledge of both language and the world; grammatical meanings come last, re-
lying almost entirely on grasping language. The scale is also relevant synchronically,
to semantic structure. Applying the scale to senses rather than words, and restricting
it for the moment to concrete senses, the scale has purely referential senses such as
those denoting people and places at the cognitive extreme. Entity senses such as ‘dog’
and ‘spoon’ come next, being strongly cognitive but carrying descriptive elements.
In the middle come event senses such as ‘skate’ and ‘enter’, and property senses like
‘red’ and ‘big’. Many spatial senses are strongly linguistic, since they rely on linguistic
distinctions between cognitively similar situations, as with ‘over’ / ‘above’, and ‘in’ /
‘within’. At the linguistic extreme come senses such as those of pronouns, of and and
the, and grammatical auxiliary verbs; their senses may be defined by their place in the
linguistic paradigm.
The relation between words that are wholly cognitive such as London and Bir-
mingham is that they denote different places; they have no relationship in linguistic
semantics. Similarly, the “synonymy” of ‘ice’ and ‘steam’, the “hyponymy” of ‘reptile’
and ‘dinosaur’, and the “antonymy” of ‘positive’ and ‘negative’ (of electric charges) are
all cognitive; they will not be studied here, though they are often treated as if they
were linguistic.
There are other problems in studying “synonymy” and so on. The terms are usu-
ally applied to words; but words have so many senses, and such varied ones, that to
generalise about one word’s relations is often impossible. Next, senses vary in different
contexts, according to the frame or scenario into which they fit, so that a sense which
is usually treated as synonymous with another may be contrasted with it in another
use; for example, snake <2> and serpent <2b> both denote a treacherous or deceitful
person, so can be synonymous; but serpent can evoke ‘malicious’ or allusion to the
Biblical story of Adam and Eve, thereby contrasting with snake.
Antonyms, synonyms, hyponyms and so on thus form indefinite classes, not
categories. (That view of sense relations is supported by Geeraerts (2010, p. 89) and
Murphy (2003, p. 11), for example, and it fits the treatment given by Cruse (2011).)
I will not study them as categories, since my concern is with the nature of semantic
structure, not with the nature of particular items or classes of item; and I use the terms
to relate my discussion to other writers’ discussions, not as technical terms.
5.3.2 Sense relations: Synonymy
Introduction
Synonymy will be understood here as follows. Two senses are synonymous if their
similarities are more important than the differences, for the purpose in hand. It is
assumed that there are no pairs of words that are identical in meaning, and that even
the sense of a single word is never the same in another use. (Even if repeated imme-
diately, and therefore in the same context, it will change: if the answer “No,” becomes
“No, no”, the second occurrence of no will be either stronger or weaker.) Synonymy is
structured by identity and difference in the areas, types, and dimensions of meaning,
as set out in Chapter 4.
Areas of meaning
Synonymy varies according to the area concerned, cognitive or linguistic. Cognitive-
ly, the words water, ice, steam, and H2O are synonymous in referential use; ice and
steam (and in some contexts, water) may have extra concepts defining their state.
Linguistically, they are distinct: H2O is technical in social meaning, ice and steam are
semi-technical, and water is general.
Frames and scenarios – structures of cognitive meaning – also structure synony-

my. For example, bill is an approximate synonym for invoice in the scenario for dining
in a restaurant, but not in other scenarios. Light has different synonyms according
to its various frames: in the soil frame, it means ‘friable’ (sense <6>); in the clouds
frame, it means ‘fleecy’ (also <6>); it occurs in other frames as ‘digestible’, ‘graceful’,
and ‘unstressed’.
Types of meaning
If we take synonyms for woman, such as lady, dame, bitch and so on, we see that they
share their descriptive meaning, but differ in the type of meaning they combine with
it, giving speakers a very wide choice of what they can convey. The speaker can invoke
derogatory affective meaning with lady, female or broad, or jocular meaning with
dame, wench or female. (The ratings of type are all from SOED.) The speaker can also
use lady to express a courteous, honorific, respectful, or ironically respectful attitude.
Speakers can identify themselves as North American with broad or dame, as Scots
with lass, or Australasian with sheila. Readers will identify the writer as belonging to
the past if they meet wench, or some uses of maid, or (once again) lady. Speakers can
set an informal tone with hen if they are Scots, or (yet once more) with lady; they can
make the tone poetic with maid, dame or lass, or slangy with sheila, dame or broad.
(I have not used slut as an example, because SOED gives it no rating at all for usage or
level, although to me it is the most powerful of all affective words for ‘woman’.)
Dimensions of meaning
Synonyms are also distinguished by their position on the various dimensions of the
descriptive meaning type.
On the generality dimension, many of the words for ‘woman’ discussed in the last
paragraph can be used in a general sense, denoting the concept ‘woman’, no more.
Doll in its sense <3b> is more particular: ‘a woman who is pretty but unintelligent or
frivolous’. Matron <1>, however, is extremely particular: ‘a married woman, especially
characterised by dignity, staid discreet behaviour, and plump, motherly appearance’.
Cool and frigid differ on the intensity dimension; boiling and hot, applied to water,
differ on the vagueness dimension. Bread and loaf denote the same food, but differ on
the boundedness dimension.
Salience distinguishes even the closest synonyms. For example, brave, bold, cou-
rageous, valiant and intrepid are used by SOED to give the meaning of the others; they
share most of their descriptive sense elements. But bold is unique in having ‘enterpris-
ing’ as an element, which is accordingly more salient in it than the other elements;
intrepid is distinct in featuring ‘fearless’; valiant has ‘on the field of battle’ as salient.
Courageous seems to be different from the others in having no unique feature.
Expectedness is important also, as with leaden, sombre and dull. Obviously, dull
has the conceptual meaning ‘dull’, as a necessary element. Some senses of leaden in-
voke it as an expected element, because its sense <3c> is “lacking animation…”, which
equates with “…not lively…”, from the relevant sense of dull. Sombre can evoke ‘dull’
as a possible element, in sense <2>, “…oppressively solemn or sober”.
Discussion: Synonymy
The distinctions within the sense of synonyms provide a powerful source of expres-
siveness in English. Consider example (3).
(3) “In the war led by Margaret Thatcher’s government – against the left, the
trade unions, the post-war consensus – her side was crushingly, devastatingly,
humiliatingly victorious.” (New Zealand Herald, April 19th 2013, page A25)
The second and third of the underlined words are clearly intended to add extra mean-
ing, so the reader must find something extra in their minor meanings, although all
three mean ‘overwhelmingly’. Devastatingly must invoke its elements, ‘lay waste’, ‘rav-
age’ and ‘make wretched’. Humiliatingly is normally a weaker word than the others;
but, because of the climactic structure, must be assigned a stronger meaning through
intensification of its affective value. (It then has the interesting implicature that, for
Thatcher’s enemies, being humiliated was worse than being crushed and devastated.
We may note also, in anticipation of later discussion, that the elements that supply the
force of the last phrase are evoked partly by the rhythm and the sound symbolism.)
Conclusion: Synonymy
Discussions of synonyms usually present them as being related only in their descrip-
tive elements. But from this discussion we conclude that synonyms are related also
through the areas they call on (cognitive or linguistic), and the types and the dimen-
sions of meaning they use. The areas, types and dimensions constitute layers of the
semantic network.
5.3.3 Other sense relations
Antonymy
The term “antonymy” is more unsatisfactory than “synonymy”, for its range of mean-
ing and the overextension of its use. See Cruse (2011, Chapter 9) for a useful discus-
sion. Cruse rightly insists that we need a more general term, “incompatible,” to cover
not only antonymous senses, but also senses denoting things on the same level of a
taxonomy, senses denoting parts of the same whole, and hyponymous senses. Howev-
er, since Cruse (2011) and Lyons (1977) provide a good coverage, and the issues that
concern us here are unaffected, I will keep to the familiar concept of antonymy.
As argued in §5.3.1, many relationships usually thought of as antonymous are
cognitive rather than linguistic; they will not be discussed here. There are, however,
some linguistic antonyms, distinct from cognitive ones.
The linguistic distinctions between synonyms also apply to antonyms. Well wa-
tered and fruitful are antonyms for arid when it means ‘parched’ or ‘barren’, of ground
(sense <2> of arid); but interesting is an antonym for arid when it means ‘uninterest-
ing’ (sense <3>). (For dry and wet in their referential senses, the opposition lies in the
conceptual area shared by language and knowledge.)
The words execute, lynch, slay, kill, murder and so on have synonymous senses,
but they are distinguished from each other by antonymous descriptive elements. Each
means ‘deprive of life’; but execute includes ‘by process of law’ whereas lynch is oppo-
site in including ‘without process of law’. Similarly, some of them are distinguished
by antonymous meaning types: slay in its usual sense is literary, archaic or jocular,
in social meaning; in that respect, it is opposite to murder, which is a standard word.
(Slay also has some distinct minor sense elements.)
Hyponymy, and so on
Cruse (2011) discusses several other relationships. Hyponymy, for example, is well-
known; its hierarchies include taxonomies – from tableware down to teaspoon – and
“meronomies” of parts – from body down to fingernail; see §5.4.3 below for further
discussion. “Chains” include haze, mist, fog, peasouper. “Grids” includes words relat-
ed in two ways at once, as with sex and species, in man / woman // ram / ewe. Some
authors discuss “semantic” or “lexical” “fields”, such as those of colour and minerals.
I will not discuss them, since the distinctively linguistic elements involved are dealt
with above.
5.3.4 Variation in sense structure
Section 5.3.2.4 above showed variation in the salience of different sense elements, and
in selection among the expected and possible sense elements. That leads to the issue
of what controls the variation in the structure of senses, when words are used in utter-
ances. That variation and what controls it is the subject of this section.
The issue is a substantial one, for the following reasons. We assume that English
words are markedly polysemous, and we know that those numerous senses vary in
context. Moreover, it is intuitively clear that speakers do not vary the meaning with
deliberate thought, and that the contextual variation is not signalled in the utterance.
There is therefore a great deal of variation, but no clear explanation of it.
Several things make the issues to be discussed here important, as well as extensive.
First, the issues affect the long-running and unresolved controversies about polysemy
and monosemy. Second, they relate to the complexity of the account of sense elements
given in dictionaries. In SOED, set verb1 has branches I to XI, with 79 numbered
senses, and 54 lettered subsenses, which are further divided into subsenses separat-
ed by semicolons, and those are divided into sense elements separated by commas.
Moreover, that complexity illustrates some fundamental features of semantic struc-
ture: the hierarchic nature of the sense variations enters into the network, since each
of the 79 numbered senses and of the multiple subsenses has its own set of antonyms,
synonyms and so on; and remarkably, all that variation of elements can occur without
the core of the sense being destroyed, since all of the senses of set verb1 are variations
of ‘put in place’. (Compare Ruhl (2002), arguing that the verb break is monosemous,
and the emphasis made by Sign-Based Linguistics (e.g. Reid and others, 2002) that
every sign has an invariant meaning.) Finally, from the hearer’s point of view, the var-
iation is totally controlled by linguistic and situational context: the meaning exists in
context; any meaning stated without context is an unreal abstraction.
The previous paragraph describes senses from the system aspect; from the other
aspects, the situation is very different. It seems extremely unlikely that hearers who
could understand any of the SOED’s variations on set have all of the variations in-
dexed in their mind, explicitly and hierarchically, as learned units. Rather, they will
have a much more limited set, which they use to infer the relevant exact meaning in
context. Similarly, it is intuitively clear that speakers do not have such exact meanings
in mind consciously, and presumably even their unconscious formulation is com-
monly what linguists would regard as underspecified. Hence the need for a semantic
system by which hearers can cope with sense variation.
The section will treat variation in sense structure as having three forms: select-
ing among existing possible elements (subsection .2), specifying vague elements
(subsection .3), and changing existing elements to something different in nature
(subsection .4).
5.3.4.2 Selecting among possible elements

The expectedness dimension in use
Expected elements are those which the hearer expects, from his knowledge of the
sense, to be invoked when the word is used. Speakers, however, having a meaning of
their own which they want to invoke, may not follow hearers’ expectations. I will il-
lustrate the extent to which those expectations are fulfilled, and how the expectedness
dimension operates as words are used, from flare as an intransitive verb, using the
literal sense, given as example (4).
(4) “Burn with a spreading, unsteady flame; blaze or glow (as) with flame”.
(SOED, <4>)
I take the two phrases, “Burn with… flame” and “Blaze… with flame”, as defining a
single sense, united by the synonyms “burn”, “blaze” and “glow”. (The variation in
wording seems intended to suggest the variety of possible elements; “blaze” suggests
‘up’ and ‘high’.) Taken that way, the sense structure on the expectedness dimension
seems to be as follows. (In the words below for the sense elements (in single quotation
marks), differences in part of speech and semantic class are not significant; mostly, the
distinction between linguistic and cognitive meaning is ignored also.)
– Necessary elements, shared by the two phrases for flare <4>: ‘burning’ and ‘flame’
(stated explicitly); ‘shining’ (implied by ‘burning’ and ‘flame’).
– Expected elements: ‘spreading’, ‘unsteady’ (mentioned explicitly, but the dif-

ference between the two alternatives shows them to be not necessary); ‘bright’
(seemingly necessary, but inhibited by “glow”).
– Likely elements: ‘brief ’ (implied by “blaze”).
– Possible elements: ‘suddenness’ and ‘speed’ (implied by “blaze”); ‘upward move-
ment’ (suggested by “blaze”); ‘not very bright’ (implied by “glow”).
That distribution of elements within the descriptive meaning of flare is illustrated in

examples (5a–f). They are taken from COCA; they were the first half dozen instances
of “flared” in this sense, when the corpus was consulted on 10th March, 2014.
(5) a. “…a flamb[é] dessert that flared up nicely when Vern … touched a match
to it.”
b. “The fire beneath the burners flared high for an instant.”
c. “Lanterns and small fires flared.” (Description of an outdoor scene.)
d. “Way out there in space that speck of light flared one more time and went
out.”
e. “At that very second, the lights flared on. We all blinked in the onslaught
of brightness.”
f. “A light flared to life in the other room.”
The distribution of the elements is analysed in Table 1; the columns represent the ex-
amples, (a) to (f); the rows represent the sense elements; “Y” stands for “Yes”, meaning
“Yes, this sense element is present.” Thus the first row below the headings indicates
that the necessary element ‘shine’ occurs in all of the examples, (a) to (f). The mor-
phological form given (e.g. “shine” or “shining”) is fairly arbitrary; as will be argued
later, these fundamental meaning elements are not equivalent to words. The table
shows each element’s level of expectedness. ‘Shine’, ‘burn’ and ‘flame’ are necessary
Table 1. Sense elements of flare, in the sample sentences

Level of expectedness Sense element (a) (b) (c) (d) (e) (f)
Necessary ‘shine’ Y Y Y Y Y Y
‘burn’ Y Y Y (Y) (Y) (Y)
‘flame’ Y Y Y (Y) (Y) (Y)
Expected ‘spreading’ Y
‘bright’ Y Y Y Y Y Y
‘unsteady’ Y
Likely ‘brief ’ Y Y Y Y
Possible ‘not bright’
‘sudden’ Y Y
‘speed’ Y
‘upward’ Y Y
elements in flare, and occur in all uses. ‘Spreading’, ‘bright’, and ‘unsteady’ are expect-
ed, since they distinguish flare from synonyms like shine and burn.
The column for example (c) thus represents the analysis that in “Lanterns…
flared”, flared has the necessary elements ‘shine’, ‘burn’ and ‘flame’, and the expected
element ‘bright’, and the likely element ‘brief ’; and it indicates that the other elements
are not invoked. Thus “Lanterns… flared,” means that the lanterns shone as if burning
briefly with a bright flame. (‘Not bright’ is not invoked by any of the uses in this ran-
dom selection of quotations.)
The rows for ‘burning’ and ‘flame’ show the “Y” for “Present” in brackets, for
examples (d), (e) and (f). That is because those examples refer to lights, which do not
burn or have flames – the sense varies between the lights frame and the flames frame;
in the real situation depicted, those elements were not present, but they are present
in the linguistic meaning by metaphor, which is indicated in the definition by “as” in
“glow (as) with flame”.
Flare also illustrates again the odd fact that words can have incompatible ele-
ments: flare as ‘blaze’ and ‘burn’ entails the expected element, ‘bright’; but as ‘glow’, it
entails the merely possible element, ‘not bright’.
The possible sense elements of flare form a network, which can be diagrammed
as a semantic map, as in Diagrams 7 and 8. Diagram 7 shows the elements with some
of their relationships; for example, the top of the map represents the relationships that
FLAMES BURN, and that they SHINE as they do so, and the left-hand side of the map
indicates that the FLAME is often SPREADING and UNSTEADY. (In users’ mental net-
work, there are other links, of course, as between BRIGHT and SUDDEN, and SUDDEN
and SPEED, for example.)
Diagram 8 is a semantic map showing how the uses in (5) variously invoke the
elements of flare’s semantic network. Thus the box labelled “(a)” at the top left corner
represents the sense of flare in (5a), “…a flamb[é] dessert that flared up nicely…”; it
adds the conceptual elements SPREADING, UNSTEADY, SUDDEN, and UPWARD to the
basic elements of BURNING BRIEFLY with a BRIGHT FLAME. Note that the combination
of elements for the uses does not always correspond to a subsense of the word as
FLAME
Necessary
elements
SHINE BURN
Expected SPREADING
elements BRIGHT
UNSTEADY
Likely BRIEF
elements
Possible
elements NOT BRIGHT SUDDEN SPEED UPWARD
Diagram 7. Flare, as network of sense elements

(a)
(b)
(e) and (f)
(c) and (d)
Necessary FLAME
elements SHINE BURN
Expected SPREADING BRIGHT

elements UNSTEADY
Likely
BRIEF
elements
Possible
NOT BRIGHT SUDDEN SPEED UPWARD
elements
Diagram 8. Sense elements of flare in uses (a) to (f)
defined by SOED; dictionaries are generalisations across uses, and cannot account for
every use individually.
I suggest that this discussion of a small sample of uses shows the following. The
analysis into degrees of expectedness works well, giving a useful analysis that fits the
instances, and fits our intuitive understanding of the words and of English seman-
tics. The descriptive meaning of words does have a “core” of necessary elements, with
expected and merely likely elements that may be thought of as more and more pe-
ripheral, and attached to the core more and more loosely – they are less and less
entrenched, to put it psycholinguistically. I suggest that that vindicates the analysis
set out in Chapter 4. On the more immediate issue of variation in sense structure, we
conclude that an important part of the contextual variation of senses in English con-
sists of selecting among sense elements of varying degrees of expectedness.
Salience dimension
Selecting among various elements often shifts their degree of salience, as well as omit-
ting or including them. That can be illustrated from the uses of flare given in the
previous section. My analysis – necessarily subjective – is as follows.
Use (a) is “…a flamb[é] dessert that flared up nicely…”; the analysis in Table 1
lists almost all the elements as being invoked. Their relative salience is controlled by
our knowledge of flambé desserts, and by the context word nicely, which inhibits any
potential negative elements. ‘Bright’ seems moderately salient, because desserts are
usually prepared indoors, with a dim background, and ‘spreading’ is also moderately
salient because flambé flames spread across the pan in an unusual way. ‘Brief ’ seems
Table 2. Sense elements of flared on the salience dimension

Use Most salient Next most sali-
element(s) ent element(s)
(a) “…a flamb[é] dessert that flared up nicely…” ‘bright’ ‘brief ’ ‘up’
‘spreading’ ‘unsteady’
(b) “The fire beneath the burners flared high…” ‘flame’ ‘up’ ‘brief ’
‘bright’ ‘sudden’
(d) “…in space, that speck of light flared one more time…” ‘bright’ ‘not very
‘brief ’ bright’
(e) “At that very second, the lights flared on. ‘sudden’ ‘speed’
We all blinked…” ‘bright’
salient, because brevity is distinctive of such flaring, but less salient than ‘bright’, be-
cause we expect the flare to be brief. ‘Up’ is very salient because stated explicitly; in
itself, it is not salient in flare. Other elements, such as ‘unsteady’ and ‘speed’, are not
salient at all.
I will not analyse other examples fully, but summarise my analyses in Table 2, for
readers to compare with their own. Readers will doubtless differ in their judgement of
the examples; but I trust that the examples do demonstrate that senses are structured
in this way.
5.3.4.3 Specifying vague elements

What makes specification necessary
Some degree of vagueness is very common in the senses of English words, partly
because they vary on the general scale of vagueness-specificity, and partly because
the other semantic dimensions are also scales, with the position of the sense on that
scale being often not fully determinate in the mental lexicon. For example, ‘wine’ as
a sublexical sense (not as the meaning of wine in an utterance) is not specified on the
boundedness dimension; in the utterance, “Have some wine?”, it becomes specified –
as unbounded, being a mass noun; but in “He prefers French wines” it is specified as
bounded, being a count noun; hearers must determine that variation as they hear the
utterance. Similarly, the degree dimension is often specified by the context, not being
inherent in the sense. Thus specification of the vague is very common, and important,
in English lexical semantic structure.
Cognitive methods of specification

Some specification is achieved cognitively: compared with its use in “a small plum”,
small in “a small pineapple” denotes a greater size, because of our general knowledge.
More subtly, we invoke different frames and scenarios. The sense of flare, in the ex-
amples given above, varied according to whether the word invoked the scenario of
flames flaring or of artificial lights flaring. The sense of door may be specified by the
scenarios of opening a door or of going through a door, or by the frame of door +

handle + hinges.
We should note that just as the mini-network which constitutes sense <4> of flare
is part of the slightly larger small network of all four senses of flare, and of the whole
network of English semantics, so the frame or scenario invoked by a single sense is
part of a cognitive network extending indefinitely through our everyday and scientific
knowledge.
Linguistic methods of specification

The sense can also be specified linguistically. Linguistic methods include lexical
choice – between hyponyms and their superordinate terms, for example, as with blue
and azure. Syntactic methods include modification. ‘Blue’ can be specified syntacti-
cally by “sky blue” or “deep-sea blue”, instead of lexically by such words as azure. The
Subject of a clause often specifies the sense of the verb. In the flared examples (§5.3.4),
for instance, the Subject “star” specified that ‘flared’ would not include ‘bright’, but
“lanterns” did specify ‘bright’; “flambé’d dessert” specified ‘spreading’. Adverbials
function similarly, as with “to life” and “that very second”, used with “flared”; they ap-
parently add information, but are in fact activating information available in the sense
of the verb as possible sense elements.
Transitivity structure often specifies the structure of a verbal sense, as in
example (6).
(6) “The enlisted people had already eaten, so we gave them the rest of the day
off.” (COCA)
Since there is no Object in that intransitive use, the thing eaten and such elements as
‘consume’ are inhibited.
5.3.4.4 Changing elements

Changing the sense by changing domains
Speakers can use context to vary senses more profoundly than by simply specifying
what is vague, or invoking the less likely sense elements. They can also invoke relative-
ly unlikely elements. Syntactic modification is a particularly powerful way of doing
that, when the modifier and its head come from different domains, such as different
fields of reference, or from the abstract and the concrete domains, as in example (7).
(7) “12 Years a Slave [a film] … is at times sublime, but hampered by an austere
didacticism.” (New Zealand Herald, February 6th 2014)
Didacticism is abstract, although applied to a film; austere is personal and fairly con-
crete; accordingly, neither word applies straightforwardly to its referent. Each is en-
riched accordingly, with an image of the director as a stern schoolteacher emerging,
although not entailed in any of the individual word senses.
Changing the sense by metaphor

Consider the metaphor in example (8), selected to relate to examples given in §5.3.4.2.
(8) “Blue flowers and red flowers flared in the roadside hedges.”
(K. Mansfield, cited in SOED entry on flare, verb)
Appreciating the descriptive effect of Mansfield’s description of flowers requires us

to do several things with flared. First, we must apply to the flowers its established
meaning, part of sense <4> “blaze… with flame”– the blue and red flowers blaze with
flame – the literal meaning. Second, we must realise that flowers do not blaze, and
seek an appropriate meaning for blaze, which does apply. To do so, we relate that lin-
guistic meaning to the situation, and derive the cognitive meaning that the flowers are
bright, contrasting strongly with the background, and spread irregularly (paralleling
the suddenness, brightness and irregularity which are likely elements of flare <4>) – a
figurative meaning. Finally, we should enjoy the aptness and detail of the imagination
that produced the metaphor in Mansfield’s mind and our own. Those elements con-
stitute the metaphor.
In terms of types and dimensions of meaning, metaphors like this have con-
crete (and sometimes abstract) descriptive elements applying to the referential object
and the analogical one, in parallel but different ways; in this instance, the metaphor,
“flowers flared”, has colours and irregular spacing referring to the flowers, and sud-
den flames and irregular occurrence, referring to flaring. The metaphors also have an
affective meaning; it may be imaginative excitement (as here), or an emotion (anger,
hate), or some other affect, perhaps.
Now flare <4>, “blaze… with flame”, is derived from flare <1>, ‘spread out (one’s
hair)’, from the 16th century. That derivation is metaphoric, but sense <4> is not met-
aphoric: it does not require the parallel between literal and figurative interpretations,
nor a response to the imaginative relationship between them. President is metaphor-
ic (or metonymic) in origin – the person who sits down (Latin sedere) first (Latin
prae-) – but is not a metaphor. The leg of a table is likewise metaphoric but not a
metaphor. Metaphors are not metaphors after they die. That apparent discursus is
intended to show the falsity of the common assumption that many established senses
are metaphors. The established sense with its single-step application is only the first
of the several elements essential to metaphor. The others (reconstrual of the literal
meaning to construct a figurative one, and enjoyment of the apt parallel between the
two) are evoked in the hearer’s mind, by the speaker’s construction of the context. (See
also Barsalou 1999, §3.4.1).
Changing the sense by other means

There are many other means by which a speaker can change a meaning. They include
providing a slightly discrepant antonym or synonym for the word whose sense is to be
changed, and specifying the cause or result of what the word denotes; but there seems
to be no value in listing or classifying them. Example (9) illustrates changing the sense
of semantics, by specifying an instance of the behaviour denoted. The situation is that

Collins was a cabinet minister accused of conflict of interest in giving an “endorse-
ment” of a brand of milk, and the writer is criticising the prime minister.
(9) “The Prime Minister resorted to semantics when he let Collins off the hook
by deeming her effusive praises of the milk … as merely a ‘promotion’ rather
than an endorsement.” (New Zealand Herald, 15th March 2014, p. A22)
The sense of semantics changes from ‘the relation between words and meanings’ to
‘using words in a clever, dishonest and self-serving way’.
Fairly obviously, meaning may even be reversed, by irony, as in example (10).
(10) [A certain political commentator] “is about as silky smooth as sandpaper.”

(The Economist, 23rd January 2016, p. 51)
5.3.4.5 Conclusion: Variation of structure in context

This section has shown that the makeup of a particular sense varies with the occasion
of use; in particular, it may become more fully specified, and some elements of it may
be altered or even replaced. That highlights the fundamental nature of linguistic net-
works: they are patterns of potentiality; particular nodes and links may be activated
on one occasion and not on others. For the hearer, there are certain points at which
entry to the network can be made (the links to morphosyntactic words); but for the
speaker, there seems to be no limit to the ways the network can be accessed, according
to the situation and the intention. In these ways, linguistic networks are quite different
from hierarchic structures, with their top and bottom points for access, and their fixed
constituents.
5.3.5 Conclusion: Structure among word senses
The section has dealt briefly with the familiar structure of sense relations such as syn-
onymy, and more fully with contextual variation in structure. It has illustrated further
what is meant by the “semantic network” of English. The contextual variation of sense
structure is dependent on senses’ links to other senses through their various meaning
types and dimensions; and, since they affect meaning, group and clause syntax are
part of the network, along with lexis. Also, the variation works rather differently in the
networks of cognitive and linguistic semantics, with the cognitive domain controlled
by frames and scenarios, for example.
This section has described passages which provide items of information which
hearers must relate – a little like the pictures that are created when we join the dots
provided. The words in a passage often have many sense elements, which are dots that
can be joined; the context provides the clues as to which of them the hearer must join.
(Compare Elman (2009, pp. 573–574), on language as cues.)
5.4 Structure of sublexical meaning
5.4.1 Introduction
Sections 5.2 and 5.3 dealt with lexical senses – the senses of words as used in utteranc-
es. This section deals with sublexical senses, those that are “below” word senses, and
intermediate between them and cognitive concepts. The senses dealt with are those of
descriptive meaning, except that nondescriptive meaning is dealt with briefly in the
discussion section; the reason is that nondescriptive meaning has very little internal
structure.
Since they are below the level of words, they have in general no direct expression
in language, so that their very existence as elements of language may be questioned.
However, a theory of language must posit them, since the senses described so far are
specified for tense, number, comparison and so on; they must exist, as in the com-
monly assumed mental “lexicon”, in the basic form from which the specified forms
are derived. Moreover, there are in fact direct expressions of such senses – devel-
oped relatively recently in the history of English – which provide us with direct evi-
dence, which will be set out in §5.4.2. Their existence as meanings distinct from word
meanings is supported by linguists such as be Bierwisch and Schreuder (1992, p. 24).
Finally, there is psycholinguistic evidence for them. Pickering and Branigan (1998)
show that when verbal meanings are first activated before use (“primed”), they do
not have tense, aspect or number; moreover, words prime a sublexical meaning that
can be realised by various words, not only the word that primed it: we must distin-
guish between sublexical meanings and the word meanings that realise them. See also
Pulvermüller (1999, p. 275).
The main points to be established in this section are that in contrast with word
senses, sublexical senses are virtually dimensionless, and that, although they are lin-
guistic, conceptual elements are in the overlap area between the cognitive and the lin-
guistic described in Chapter 4 (§4.2), being perhaps more like cognitive concepts than
like linguistic forms. Related to dimensionlessness is the further point that, in general,
there are only simple sense elements here; word senses in utterances are combinations
and development of these elements; those senses exist here only as potential patterns
of elements. (I say “patterns of elements”, since it seems unlikely, although possible,
that some uttered senses correspond to a single sublexical element.) Nevertheless, I
will refer to “senses” as a shortcut for “patterns of elements corresponding to those of
lexical items”.
It is difficult to discuss this area of meaning, because of their difference in na-
ture from word senses; representing them in words seems very likely to misrepresent
them. Extra care is needed from reader and writer.
5.4.2 Dimensions
Necessarily, sublexical senses are like lexical senses in differing on basicness and qual-
ity dimensions: sublexical RED, YELLOW and BOOK differ as lexical senses ‘red’, ‘yellow’
and ‘book’ do. However, on the other dimensions, which control their semantic na-
ture and behaviour, they are quite different.
Boundedness dimension
The fundamental nature of the familiar distinction between proper, mass and count
nouns is that their meanings differ on the boundedness dimension. The meanings of
proper nouns are fully bounded, to the extent that they denote unique individuals.
Count nouns denote entities with an outline (which is what makes them countable);
they are bounded, although not fully so, not denoting unique entities. Mass nouns
denote entities that are less bounded, having no outline; but they are bounded to the
extent that they exist in space and time, and have other physical properties such as
weight. Some abstract nouns, such as population and transport, often denote entities
that have a direct connection with physical dimensions; others are progressively more
abstract, through generosity to existence. There are several familiar variations on that
simple scale, such as the use of words like wine as either mass noun or count noun,
and the use of common nouns generically, as in “The tiger is…”; nouns like “a flash”
are bounded in time, rather than in space. We should also note that adjectival senses
formed from nouns are commonly less bounded than the noun sense; e.g. watery is
less bounded than water.
For our purpose, the crucial fact is that word senses are sometimes reduced to a
state of unboundedness, in everyday use. Hospital signs labelling a “patient entrance”
or “doctor parking” apply to many patients and many doctors, yet the nouns are sin-
gular. The effect of “patient entrance” is that we conceptualise the meaning, ‘patient’,
without envisaging individuals or even a group; the bare concept, PATIENT, is used to
label a type of entrance, as if it were a property. It is “de-individuated”. A parallel pro-
cess occurs with verbs: finite verbs, bounded in time and aspect, lose the finite-time
specification as participles, as in “burning wood” and “burnt wood”; they can then
lose aspect as well, and become unbounded, as in “the burn time of the rocket” – the
time for which it burns – and “He started up a swim school”.
This use of reduced forms – reduced semantically, morphologically, and in syn-
tactic potential – has become increasingly common in the last century, but it goes
back further, with the loss in the 19th century of forms like “a geographical lecture”
and “a mathematical master”, just as “artistic director” has recently given way to “art
director”. Many compounds are even older instances of the same process, as with rip-
saw, graveyard and pickpocket (someone who picks pockets).
These senses are so reduced as to resemble purely cognitive concepts, but they
are linguistic, in that they have relationships to other linguistic units. I take it that
these uses indicate the nature of sublexical senses: sublexically, there is no difference
between ‘patient’ and ‘patients’, between ‘stone’ as object and ‘stone’ as substance, be-
tween ‘burn’, ‘burning’, and ‘burnt’, or even a between ‘burn’ as a verb and ‘burn’ as
noun. There are only the simple and invariable units, PATIENT, STONE and BURN.
Other dimensions
In the sentence, “Chainbridge House is a five doctor practice” (BNC), the reduction
of form, from “doctors” to “doctor ”, prevents the sense from denoting particular in-
dividuals; yet it is not general, denoting doctors as a class. The sense is neither gen-
eral nor particular; that dimension does not apply to it. We conclude that sublexical
CIRCLE is neither general nor particular also. Nor is it specified on the vagueness di-
mension: the word circle is vague in general use, being applicable to mathematically
precise circles and to irregular ones as when people stand “in a circle”; and in “a yellow
chiffon circle skirt” (BNC), there is no referent for “circle”; CIRCLE must be neither
vague nor precise.
Conclusion
The argument has been that the reduced senses illustrated above are in fact reduced
from their standard lexical structure to the structure that underlies them, sublexical
semantic structure. We conclude that sublexical senses are in general dimensionless,
making them quite distinct from word senses. (They are “reduced” from the morpho-
syntactic point of view; from the semantic point of view, it is perhaps better to say that
morphosyntactic use “expands” or “develops” them.)
5.4.3 Elements and their sublexical relations
My city has a number of “cycle advance stop boxes”. They are areas marked on the
road surface at traffic lights; cyclists may overtake the waiting cars and wait there, to
be first away when the lights go green. We may understand “cycle… boxes” as mean-
ing that the boxes are for cycles or for cyclists; and we may take it that the cyclists
advance (event), or that cycles will be in advance (abstract entity), and that the cyclists
stop (event) or come to a stop (entity). The fine piece of bureaucrats’ jargon is multiply
unspecified for word and semantic class; that is, CYCLE, ADVANCE and STOP have no
defined relation to a semantic class or morphological pattern. The corresponding sub-
lexical elements must be bare conceptual meanings, and be assigned a semantic class
and morphological form only when a speaker decides to use them in an utterance.
Thus, there is presumably a basic form which underlies all derived forms. For
example cajolery and cajole both arise from a root, CAJOLE. That in turn is a complex
sense derived from simpler ones, since the word cajole means: “Persuade or prevail
upon (a person) by delusive flattery, specious promises, etc.” (SOED). The sublexical
sense accordingly includes the following conceptual elements: PERSUADE, PREVAIL
UPON, PERSON, METHOD, PROMISE, BE SPECIOUS, FLATTER, DELUDE. That formulation
has most of those words in the same form (that of an event word) to minimise the
intrusion of parts of speech into the account, which cannot be avoided in any formu-
lation in words. It interprets the grammatical meaning represented in by as inserting
the concept of METHOD into the sense; and it assumes that there is a root sense for
every word sense, since there must be a link node between every word sense and the
roots that comprise it.
Indicating the relationships among those elements is difficult. It is not immedi-
ately clear whether they are cognitive, which is suggested by their being quite abstract
concepts, or linguistic, which is suggested by their having modifying or transitive
relations, for example. I do not know of any thorough account of how they exist in
thought, and the partial accounts are often confusing in using a linguistic assertion,
“ISA” ( = “is a”), for a property. It is reasonable to represent conceptual relations by
words to the extent that we can name the relations, using words; but putting more
words into the meaning produces more items to be related – which threatens an in-
finite regress. It is not clear whether there is a relation between prevail and upon, or
whether the two words denote a unit. The problem is fundamental, and insoluble:
meanings and other forms of thought are patterns of relations, not a collection of
items with relations between them, as the network metaphor can naturally suggest. As
noted previously, meaning is connectivity (Lamb 1999).
Giving up on full accuracy, then, we could perhaps represent the sublexical
meaning CAJOLE with the elements PERSUADE, PREVAIL UPON, PERSON, and so on, in a
network diagram of the kind given in Diagram 7 and 8 in §5.3.4.2 above. On the other
hand, we should keep in mind the point, illustrated with serial verbs in §5.2.1.2, that
the sublexical structure may be quite different from the structure of our dictionary
definitions. We conclude that there must be a structure of sublexical elements, but
that we cannot be sure of its nature.
Just as particular concepts are shared with cognition, so must frame structure
be shared – of the sort set out by Barsalou (1992), for example. The complex concept
CAR, for instance, can profitably be thought of as structured in attributes, with each
type or instance of cars having a value for each attribute; for the attribute FUEL, the
value would be PETROL or DIESEL. The attributes are subject to constraints; in the con-
cept of TRAVEL, for example, SPEED and TIME TAKEN constrain each other.
Each of those conceptual elements has links to other word forms, of course, and
to elements which are conceptually similar and contrasting, and so on. In that very
extensive network, the pattern of elements that makes up each word meaning is po-
tential, rather than real; it exists only when it is activated. As a potential for activation,
it can be identified in the way discussed above in §5.2.1.2: it has identity as a pattern,
and in the salient element which provides the link to the lexical form.
We conclude that, while some sublexical meanings are simple, others are quite
complex. The elements are of different types (attributes and values, for instance), and
have many connections, in different directions, forming a very large network.
5.4.4 Elements’ external relations
Distinction: Cognitive versus linguistic relations

Being in the area of overlap between language and cognition, sublexical meanings
relate to the elements of both of those mental functions, that is, to word meanings and
to concepts. The relations do not include those of synonymy and antonymy, which are
relations between word forms. (Synonymy consists of shared elements; there cannot
be elements shared among the shared elements in an indefinite regression.) Since the
function of this area is to process the language we hear and read, and to formulate
thought into language as we speak and write, it must handle complex word sens-
es, breaking them down or interpreting them in comprehension into the forms of
thought. Similarly, the overlapping concepts in groups of words like distinct, discrete,
differentiated, and distinguishable must be resolved.
The thought relations include those of hyponymy, which I take to be primarily
cognitive rather than linguistic, as in biological taxonomies. (An example of a lin-
guistic hyponymy is that of WRIST, FOREARM and ELBOW for ARM; that must be partly
linguistic, since it is specific to English.) These sense elements have other cognitive re-
lations: other scientific relations, general intellectual relations such as time and causa-
tion, and relations established in everyday experience, such as association and a basis
in image schemas and images.
Cognitive and linguistic dominance

Gentner and Boroditsky’s distinction (2001) between cognitive and linguistic dom-
inance (see §4.2.1) applies crucially here. Babies have concepts for their carer and
everyday objects before they start learning language; the sublexical meanings for
them are derived from those pre-existent concepts; but meanings such as CAJOLE- and
its constituents are largely dependent on the relations with other linguistic elements,
and only partly dependent on knowledge.
Relation to linguistic categories

We have seen that words such as patient and doctor are often used without morpho-
logical inflection, and with adjectival function, even though they are typical “nouns”.
Therefore, their morphology and syntax cannot be constitutive parts of their sublexi-
cal forms; they are distinct elements, related by only weak links. We should conclude
similarly that words like flash, which alternate between noun and verb with negligible
difference in descriptive meaning, also have sublexical correlates which are related
somewhat distantly to morphological and syntactic properties, without having them
inherently – that is, without having entrenched links to morphology and syntax.
The same argument applies to semantic class. The sublexical meanings for words
like wine and cheese, which have both count and mass uses, cannot belong inherently
to the class of countable entities or the class of mass entities; event words that alternate
indifferently between transitive and intransitive uses cannot be inherently one or the
other. That applies not only to such linguistically determined uses, but also to the use
of concepts close to perception, such as DARK, RAIN and RED, which are readily real-
ised in either property or entity senses, or even (for RAIN) as events. There are, on the
other hand, concepts such as SELF or PERSON which seem to be used only as entities –
at the other end of a scale; but that shows only that the class link is deeply entrenched
for a few words, not that the semantic class is inherent in sublexical forms in general.
Conclusion: External relations of sublexical elements

Sublexical senses, forming a semantic network as we saw in the last section, are related
to semantic classes, syntactic functions, and morphological patterns, which are also
related to each other, in another form of linguistic network; the relations are potential,
becoming actual only in use.
5.4.5 Discussion: Structure of sublexical meanings
Nondescriptive senses
Non-descriptive senses fit the description given so far in this section to a lesser and
varying extent. Attitudinal senses are usually attached to words with descriptive sens-
es, but they do have some linguistic reflexes of their own, particularly in positive and
negative prefixes such as un- and dis- (e.g. unattractive, displeasing), and in suffix-
es such as -able. (Pleasure denotes a concept, but pleasurable expresses an attitude.)
Moreover, we need to distinguish between jocular and facetious attitudes in word
uses, so the distinction is certainly linguistic.
Social senses are similar, in that such meanings as ‘literary’ and ‘archaic’ are pure-
ly linguistic, and must have some sublexical linguistic form, since they enter into the
full meaning of various words. Affective senses, however, do not seem to have specific
sublexical forms. Rather, it seems that those meanings are expressed more naturally
in the phonology, through tonality and stress. When they are expressed lexically, they
are not based on cognitive concepts, so there is no need for a mediating level: the
expression seems to be direct from the affective faculty into words, without any form
at the semantic sublexical level.
As noted in the discussion of word senses, the nondescriptive meanings have
only loose links to descriptive meanings: they may occur on their own, and a de-
scriptive meaning may occur with opposite nondescriptive meanings, as with bloke
(‘man’ + informality) and gentleman (‘man’ + formality), and the various nondescrip-
tive meanings given for lady in §5.3.2. The descriptive and nondescriptive are so dis-
tinct that it is accurate to describe them as senses – full senses – not as “elements” or
“aspects” of the word meaning. Thus frumpy, which is said to mean ‘dowdy and old-
fashioned’ (SOED <2>), should be said to combine three senses: a descriptive sense,
‘not in current fashion’, a colloquial social sense, and unfavourable attitudinal sense.
(Dictionaries typically do not make this clear, concealing the nondescriptive senses in
all-purpose synonyms like dowdy, instead of differentiating them.)
Grammatical meanings
There seems to be no need to posit any grammatical meanings at the sublexical level;
as with the social meanings, there is no need for that level, as they are not based on
cognitive concepts. It should be remembered, however, that some grammatical mean-
ings involve the introduction of concepts into the context. Thus the plural inflection
signifies ‘Modify the root of this word, using the concept PLURALITY’. The grammat-
ical meaning of the inflection in “He walks”, and the meaning of I and you, entail
assigning the grammatical concept, PERSON, and the last two assign the everyday con-
cepts of SPEAKER and HEARER, which lie beneath it. Those concepts are sublexical.
Gaps
There are some apparent “gaps” in the pattern of sublexical senses. The first kind of
gap is with names. It is clear with proper names that there can be no descriptive sub-
lexical meaning, just as there is no description in the word meaning; but with names
for everyday objects, used purely referentially, there is no content at this level, either.
As noted previously, with “Give me that thingummy,” we go straight to the referent in
cognition using deixis not description as the semiotic process; when we are familiar
with ordinary names such as the names of physical objects, we are likely to do the
same, as with “Give me that trunnion / lever / grommet.” We may on some occasions
bring descriptive elements or an image to mind, but the system meaning does not
include them.
Another instance of what may seem to be a gap occurs in “argument structure”.
KILL, EAT and so on have fixed relations in thought; but the number of arguments
in utterances varies, so that the argument structure of utterances with the words kill
and eat is a linguistic matter, not a cognitive one. If the arguments were inherent in
the “verb”, then it would be “a full sentence in disguise”, says Lohndal (2012, p. 177).
Nor is it a cognitive matter sublexically; at this level, KILL and EAT must have links to
both transitive and intransitive syntax; which link will be activated is not determined
by conceptual meaning, but by information-structure intentions, as indicated by the
number of Participants or the position of focus in the utterance. Linguistic argument
structure can not be set at this level, although many linguists seem to say so – perhaps
they are thinking of logical structure. (See the chapter on hierarchic structure for
further discussion.)
Holophrastic expressions presumably use sublexical elements, but without follow-
ing the standard patterns of word senses. A child’s utterance of “Gone!”, for example,
must have sense elements associated with the adult word, gone; in the speaker’s mean-
ing, it may on occasion have no others, but more often has an affective element of
triumph or complaint. With an adult’s holophrastic “Help!”, however, there are other
meanings such as the speaker’s fear and sense of danger – and the specific concept HELP.
Sublexical senses not in the word list

To denote various sorts of horse, we have stallion, mare, filly, colt and foal. But for
various sorts of bear, we seem to have only bear and cub; there seem to be gaps in the
paradigm of bear senses, corresponding to stallion, mare, filly and colt. Nevertheless, it
is possible to articulate what belongs in the paradigm’s gaps – “male bear”, “male bear
cub”, and so on – so those senses are present, sublexically, in some form. That suggests,
further, that ‘stallion’, ‘mare’ and so on do not exist as sublexical senses, and that the
relevant sublexical senses are ‘adult’, ‘male’, ‘horse’ and so on.
Other noteworthy examples lie in our use of prepositions; in has different senses
in “a toy in a box”, “tea in a cup”, and “a flower in a vase”, but the distinctions are not
lexicalised. A final type of sense without a word is illustrated by “crepitate”. It means
<2> “Entomology. Of (bombardier) beetle: eject a pungent fluid with a sudden sharp
report”. For those who have experienced crepitation, there will be a more specific
experience than is denoted by pungent and sudden sharp report; the lexicographer
makes the definition indefinite, with “a pungent…”, because there is no word more
specific than pungent or sudden sharp report. There are sublexical senses, then, in
many places where there are no lexical items; but there is no apparent way of deter-
mining their form.
Support
There is quite a long history of work that supports my contention that sublexical senses
are not inherently related to semantic, syntactic or morphological classes. Hoffmann,
writing in 1903 (Hengeveld and others 2004, p. 540) said that the Indian language
Mundari has “functional plasticity” in that roots used as both nouns and verbs could
not be classed as being themselves nouns or verbs. Hopper and Thompson (1984)
asserted that words, cross linguistically, are “acategorial” – without category until such
a status is forced on them by use in discourse; they discussed individuation, and the
term “de-individuation” is theirs (1984, p. 714). More recently, a good deal of typo-
logical work has described roots in a number of languages as being “pre-categorial”, as
with Samoan, as described by Mosel and Hovdhaugen (1992, p. 73); see also Stekauer
and others (2012) surveying languages of the world. Hengeveld and others (2004)
discuss “flexible lexemes”. Haspelmath (2007) asserts that pre-established categories
do not exist. Lehman (2008) says that in some languages categorisation is carried out,
not at the level of the root or even of the stem, but at the level of the “syntagm”. Miller
(2014) supports the claim in a generative grammar approach.
5.4.6 Conclusion: Structure of sublexical meaning
Linguistic nature of sublexical meanings

The foregoing description of sublexical meaning as being at least partly shared be-
tween language and thought, and its emphasis on concepts, may suggest that this area
of meaning is not characteristically linguistic. That is because the discussion has treat-
ed the meanings as content; but, as emphasised previously, the reality of meaning is
not static – mere potential for meaning – but activation in processing. The processing
here is the formulation of meaning, and its transmission for realisation in speech and
writing; if this area were cognitive, the processing would be thought. Moreover, the
nondescriptive elements are clearly not cognitive.
Relation to other areas of language

The term “sublexical” may suggest that these meanings are related only to lexical
items – to word meanings – but that is not so. For example, the sublexical meaning
CAUSE can be expressed in the bound morpheme -ise, in the grammatical item be-
cause, or in the transitive construction (since the agent’s having an effect is causing
change). Again, the emotive sense, ‘surprise’, is related to the exclamatory word wow,
to exclamatory constructions such as “What a…!”, and to certain phonological pat-
terns; alternatively, the feeling may be intellectualised, and represented in descriptive
meaning by words like surprising. Chapters 11 and 12 will expand on the various ways
in which elements are realised.
5.5 Networks in other strata
5.5.1 Networks in clause syntax
5.5.1.1 Networks created by Adjuncts

Clause syntax is a linear co-ordination of units; it creates a semantic structure of units
with equal status – Participants, Process and Circumstances – which are together
subordinate elements of the figure; the standard structure of figures is thus hierar-
chic. However, there are some syntactic structures that sometimes create semantic
networks. The most frequent such network structure is created by Adjuncts. It can
be illustrated by discussion of a sentence used by Quirk and others (1972), given as
example (11).
(11) “Bitterly, he buried his children.” (Quirk and others 1972, §8.42)
We will use the invented variations given in example (12).
(12) a. “He buried his children bitterly.”

b. “He buried his children in their best clothes.”
c. “He buried his children without help.”
We interpret (a) to indicate, though not to state, that the man was bitter – that “bit-
terly” applies to the Subject entity. We interpret (b) to indicate that the children were
in their best clothes when buried; “in their best clothes” applies to the Complement
entity. We interpret (c) to indicate that the action of burying was done without help;
“without help” applies to the event denoted by the Predicator.
Those instances provide a puzzle if we take clause meaning to flow from pure
syntax (formal relations of morphosyntactic units), since the syntax, so defined, is the
same in all three sentences. The distinction between linguistic and cognitive meaning,
however, provides a straightforward analysis: the meanings given above are accurate
for cognitive meaning, but not for the linguistic meaning, which is underspecified;
having determined the linguistic meaning in the first stage of comprehension, and
been left uncertain, we resolve the problem in the cognitive stage. The linguistic
meaning flows from the syntax; the cognitive meaning or “interpretation” uses world
knowledge as well.
The semantics of figures, then, comes with potential links from the Adjunct to
the Subject, the Predicator, and the Complement; those links are realised various-
ly, in various semantic contexts. In variant (b), “He buried his children in their best
clothes,” the Adjunct, “in their best clothes,” is related in cognitive semantics simply
to the children. In variant (c), “He buried his children without help,” the Adjunct
“without help” is related simply to the act of burial. Both sentences have regular, but
different, hierarchic semantic structures. (They have a hierarchic structure syntacti-
cally also, but a slightly different one.)
Example (9) is different, however: “Bitterly, he buried his children.” The detach-
ment of the Adjunct from the rest of the clause provides for a freer construal of its
relationships. When we envisage the situation, we take it that the man was bitter, that
the burial was done in a bitter manner, and that the fact that the dead were his chil-
dren contributed to the bitterness. There are multiple links; BITTER is not just linked
once, to the rest of the figure as a unit, but to each of its units independently. There is
a network of relationships.
5.5.1.2 Networks created by other constructions

In a structure very similar to that of the Adjuncts just discussed, Complements some-
times create a minor network. For instance, in example (13), we link ‘notorious’ cog-
nitively to ‘Lord Palmerston’ as well as to ‘rake’.
(13) “Lord Palmerston was a notorious rake.” (BNC)
That kind of construction often predicates not only the Complement as a whole, but
the elements within it individually, also.
Expressions like “to baby-sit” often create a minor network also. In example (14),
the morpheme baby is co-referential with the Norcross kids, we must link sat to both
of them.
(14) “She babysat the Norcross kids once in a while.” (COCA)
There are many such situations where connections are made, as we interpret language,
beyond the strictly syntactic connections; they grade off into pragmatic implicatures.
“a good thick slice of French bread
APPROVAL
INTENSIFICATION
HAVING ENOUGH [THICKNESS]
Diagram 9. Modification relations in “a good thick slice of French bread”
‘slice of French bread’

[Approval]
‘thick’
[Intensification]
‘having enough’
Diagram 10. Relations of sense elements in “a good thick slice of French bread”
5.5.2 Networks in group syntax
5.5.2.1 Networks created by nominal premodifiers

Nominal premodifiers create networks of meaning in several slightly different ways.
One is illustrated in example (15).
(15) “Tuck it inside a good thick slice of French bread.” (COCA)
“Good” is linked to “slice of French bread” as a modifier, meaning ‘having enough [of
the relevant quality]’. It is also linked to “thick”, in two different ways: it intensifies it,
meaning ‘quite [thick]’, and it conveys the speaker’s approval of the thickness. Those
relations are shown in Diagram 9.
In illustrating the relations among word senses, Diagram 9 simplifies the seman-
tic structure. The relations of the sense elements are shown more fully in Diagram 10.
Relations of elements within one word are shown with a dashed line, and relations
of elements in different words are shown with a solid line. The network nature of the
structure is now apparent.
A different type of network in successive premodifiers is shown in example (16).
(16) “Look at those splendid old electric trains with pantographs.”

(Halliday 2014, §6.2)
Syntactically, the structure of premodifiers and head is as follows: [splendid [old [elec-
tric trains]]]. That is, old modifies electric trains, not trains alone, and splendid mod-
ifies old electric trains, syntactically. The sentence was used descriptively – admiring
the trains. When we grasp the speaker’s understanding, we see that the trains were
splendid in being old, and splendid as electric trains, not as a steam trains. The rela-
tions are not simply linear, from beginning to end of the phrase.
Vendler (1968, p. 132) noted that while “brave young man” and “brave old man”
are acceptable, “brave blond man” is not. Similarly, he noted, “good tall woman” and
“considerate fat girl” are not acceptable. The explanation, I suggest, is that, because
brave would modify blond man, and therefore in some manner modify blond in par-
ticular, meaningfulness requires a link between the two modifiers directly. In “brave
young man”, we unconsciously invoke a possible meaning such as ‘bold’ for both brave
and young, which links them; in “brave old man”, we invoke a link such as ‘enduring’
for both brave and old. In these instances, there are subconscious sublexical links that
do not appear in the surface meaning.
Another variant of the modification network is given in example (17). It ends a
report on the effect of weather on northern European market prices.
(17) “All the same, their findings may bring some cheer to the long, cold, dark
Nordic winters.” (The Economist, June 14th 2014, 65)
Long, cold, and dark have natural associations with each other; here, they are linked
also through the fact that the length of winter contributes to coldness and darkness;
Nordic, as against northern, is also linked to those qualities; and through being used
normally in cultural contexts, it evokes how those cultures deal with coldness, dark-
ness, and winter. Creating such a network of sense elements is a regular function of
cumulative modifiers.
5.5.3 Lexical network: Cohesion
Texts in English are to some degree consistent and connected in content – they are co-
herent; they also have some degree of consistency and connectedness in expression –
they are cohesive (Fetzer and Speyer 2012). That cohesion takes two forms relevant
to our discussion (Fetzer and Speyer 2012). In grammatical cohesion, reference links
items, especially through pronouns; substitution emphasises the linking and bind-
ing by varying its expression; ellipsis emphasises the linkage of an item indirectly by
not expressing it at all; conjunction connects material implicitly by juxtaposition and
explicitly in the use of conjunctions. In lexical cohesion, connectedness comes from
semantic relations like synonymy and antonymy, and consistency comes from con-
sistent use of particular meaning types or dimensions, for example.
In speech, consistency in phonological pattern, e.g. consistently reduced pitch
range, will also give cohesion, though neither Fetzer and Speyer (2012) nor the stand-
ard texts on cohesion such as Halliday and Hasan (1976) and Halliday (2014) discuss it.
The contrast between two short texts will illustrate the nature and effect of cohe-
sion – examples (18) and (19).
(18) “It is useful stuff, concrete, but it does have drawbacks. One of the biggest is
that it is not as weatherproof as the stone it often substitutes for. Salt and ice
routinely turn microscopic fractures in its fabric into gaping holes. These let
water soak in. That, in time, can cause the structure to fail.”
(Economist, June 1st 2013, Technology Quarterly, page 4)
(19) “The cabins are scattered throughout a thick and lush flower garden, and the
rooms are subdued, rustic and sophisticated: wood, bamboo and colourful
textiles combine to make charming, cosy hideaways.”
(Ecuador and the Galapagos Islands, Lonely Planet, 9th edition, 2012)
The two texts are cohesive, and in different ways, in their sentence length and the
construction of nominal groups; the first uses ellipsis (e.g. “the biggest __ is…”) and
frequent pronominal reference, but the second does not. Lexically, the first is simple
and neutral throughout, like its grammar, lightened by a few touches of descriptive
and personal wording (“gaping”, “useful stuff ”); with those exceptions, words are used
referentially. The second has heavy use of modifiers, which are regularly emotive and
attitudinal, and formal almost to the point of being literary; words are used descrip-
tively. In the first, the nouns and almost all of the adjectives are consistently concrete.
In the second, the sense element ‘refinement’ links lush, subdued, sophisticated and
hideaways; favourable meaning links lush, rustic, sophisticated, charming and cosy,
for example.
Asp (2013, p. 177) shows its neurolinguistic reality, reporting that it is handled by
the posterior parietal areas.
Cohesion creates a network of grammatical and lexical meaning, in various
dimensions. It is particularly noteworthy for its transcendence of syntax, creating
patterns of meaning that spread far beyond individual sentences, and that are quite
different from syntactic patterns.
5.6 Discussion: Network structures in English
5.6.1 Whole utterances as networks
Mel’cuk (2012) represents the meaning of utterances, and potentially the meaning of
whole texts, as a single network. Diagram 11 represents Mel’cuk’s illustration (2012,
p. 181) of the network for the English sentence, “David is shaving (himself) with a
new razor.” (It excludes the tense and aspect of the verb, and number and definiteness
of the nouns.)
In the diagram, word nodes are represented by small circles, and are labelled;
relations are represented by lines, with arrowheads representing the direction of de-
pendency; numbers indicate types of dependency relation.
I believe Mel’cuk’s network account to be seriously misleading in implying that
the dependency does not create a hierarchy (as discussed in Chapters 8 and 9); but
‘shave’
3
‘razor’
1 2
1
‘David’ ‘new’
Diagram 11. Mel’cuk’s structure of “David is shaving (himself) with a new razor”
it is valuable in showing that utterances are fundamentally semantic structures, that

syntax creates them, and that they are relational. (It is also limited by its dealing ex-
clusively with conceptual meanings.)
5.6.2 Networks in imaginative English
Imaginative English sets up extra links between senses, creating often complex net-
works. Consider example (20), which is part of a description of Cleopatra, the en-
chanting queen of ancient Egypt. Shakespeare’s sentence (from Antony and Cleopatra,
act I scene ii.) has been set out as if it were a specimen in typological linguistics, to
highlight the structure.
(20) “Age cannot wither her, nor custom stale her infinite variety.”
Subj. Pred. Comp. Subj. Pred. Comp.
The parallelism of the two simple transitive clauses helps create semantic links be-
tween them. It relates ‘age’ and ‘custom’ in the two Subjects, ‘wither’ and ‘stale’ in
the two Predicators, and her in the two Complements. The concepts of DECLINE and
ADVERSE EFFECT link ‘age’, ‘custom’, ‘wither’ and ‘stale’; ‘infinite variety’ stands out
against them all.
That analysis deals only with lexis and basic syntax. The personification of age
and custom extends the network of sense elements much further, especially in the
situational context of the play.
5.6.3 Relations among one word’s various senses: Polysemy
Introduction
Polysemy, as a property of words, is not relevant here directly, since we are studying
the structure of meaning (and how it is expressed), not expressions (and what they
mean). But the meanings of a polysemous word constitute a network of inter-related
senses, and the conclusions reached so far have implications for polysemy that are
worth noting.
Linguistic argument
SOED records the two senses, ‘A soft wool prepared from… certain plants’ and ‘Any
substance prepared for burning next to the skin [as therapy]’ as senses of the same
word: moxa is polysemous. The reason for linking such disparate senses is histori-
cal – they derive from the same Japanese word. Conversely, SOED gives “Of water
etc.: clear, translucent” and “Especially of water: clear…; translucent” as the senses of
different words – because the Old English word scir has produced both shire, with the
first of those senses, and sheer, with the second. For the study of present-day English
semantics, we should take “word” with a semantic definition, not a historical one. A
“polysemous word” like moxa will then be two homonyms.
If we accept that sublexical senses are not specified on any dimension, but be-
come vague or specific, bounded and bounded, and so on, according to context in use
(as argued above in §5.3.4), then it is correct for SOED to give the adjective good the
indefinite (linguistic) sense, “Having (enough of) the appropriate qualities” (branch I,
sense <1>). In “Give me a good knife – this one’s blunt,” good comes to mean ‘sharp’
(in cognitive semantics, when the linguistic sense is applied to the context); ‘sharp’ is
not a meaning of the lexical item good; the word in branch I is monosemous. However,
SOED goes on to give a series of putatively distinct senses for different circumstances,
such as <1b> “Of food and drink:… fresh” and <1c> “Of soil: fertile”, although it does
not give “Of knives: sharp”. As Fretheim (2011) points out, the difference between
“senses” such as <1b> and <1c> is outside language, residing in our practical knowl-
edge of food and soil. From the analysis of meaning given in this book, it follows that
words are much less polysemous than dictionaries such as SOED allow.
A further argument concerns the distinctness of sense boundaries The belief in
polysemy entails accepting that senses are countable; that relies on regarding them as
distinct. That comes with the philological and lexicographical traditions in linguistics,
which make the assumption that semantics, like syntax, consists of constituents built
up into a larger constituent – all with a clear boundary. But we have established, I
believe, that meaning in English is fundamentally a network, as in the semantic maps
used in this chapter; we have seen (§5.2.1.2 above) that the boundaries between senses
realised by the same word are only weakly distinct. We have seen also (§5.3.4 above)
that even established senses vary in context, reconstrual being a fundamental process
in treatment of meaning; specific words and uses of words “draw boundaries” on the
semantic maps, as in §5.2.3 and §5.3.4; compare Haspelmath (2003). All that argues
that, to an important extent, senses are not distinct and countable, and that words are
not very polysemous.
Psycholinguistic view
Psycholinguistic research gives us an insight of a different kind. It has shown that peo-
ple vary in whether a particular precise sense is “stored” in a “polysemous” structure,
or generated afresh in each use from a single stored sense – a “monosemous” process.
It has also shown that individual speakers vary in this respect from time to time; they
call on the ready-made meaning on some occasions, but construct it from a base
meaning on other occasions.
More profoundly, the psycholinguistic understanding that senses are networks
falsifies the assumption that there are integral stored senses: each sense is a pattern of
activation, so when it is not being activated in use, it is merely a potential, without ex-
istence as an entity. As patterns of activity, then, senses may be compared to a horse’s
trotting, cantering and galloping: where does a horse store its trotting when it is not
using that pattern of activity?
Conclusion
We conclude that English has some degree of polysemy, in principle; that it is less
polysemous than most linguists have thought, and much less so than is assumed by
dictionaries such as SOED.
5.7 Conclusion: Network structures in English
The chapter has described semantic networks on several levels. On the lowest lev-
el, sublexical descriptive senses consist of a mini-network of elements, connected by
conceptual or “logical” links, and of certain nondescriptive elements. The senses of
words in use specify those descriptive sense elements, combine them with nonde-
scriptive elements, structure them all dimensionally, and combine them with each
other, in larger networks. Those lexical senses are in turn related in the familiar net-
work patterns of synonymy, antonymy and so on.
The chapter has also described networks occurring at group and clause level,
where the structure is normally hierarchic. Some have structures like lexical relations,
and some have complex patterns of modification. The final pattern was that of cohe-
sion, which is not like any other; it is a more diverse and diffuse network, which can
extend over a whole text.
The main features of networks in semantics are as follows: they have no domi-
nant relation (or “dimension”), such as dependence; they have several dimensions;
units such as senses exist within the network as a cluster of links and nodes, with no
well-defined boundary; and “nodes” and “links” are relative terms, since a semantic
element thought of as a link in one context can be thought of as a node in another.
Networks are built up from the elements set out in the previous chapter; they are
the first of the types of semantic structure to be considered, with other types consid-
ered in the following chapters.
Chapter 6
System structure
6.1 Introduction
Purpose. The purpose of the chapter is to explain the next major type of semantic
structure English – after networks in the previous chapter – system structure. A sys-
tem here is a set of objects organised for a special purpose, as components of an inter-
dependent or interconnecting assembly (based on SOED system <1c>). The system we
are concerned with here is grammatical meaning, the semantic correlate of the gram-
matical system of English (with “semantic” and “grammatical” in traditional senses),
each building structure.
Coverage. The coverage of material is simplified in several respects, in this chapter.

First, some of the detail of system structure must be dealt with in the later chapters
on realisation of meaning, so are given little space here. Second, what I formulate as
a single procedure could be more fully analysed into several procedures. For exam-
ple, my formulation for nominal modifiers, “Add this meaning to that of the head”
could be broken down into “(1) Relate this meaning to [the relevant word]; (2) make
it subordinate in status, as a modifier to the head; (3) add this meaning to the head
meaning, as a property”. Third, I assume an idealised structure of utterances, in which
all utterances are complete clauses, for example. Also, to have a complete account of
how grammatical meanings operate on other meanings, we should have an account of
how the morphosyntactic or content units are identified. For example, tonic accents
are markers of tone groups; a capital letter marks the beginning of a sentence, and a
full stop will mark its end. I take the use of those markers to be familiar to readers, and
the matter is more syntactic than semantic; so it will not be discussed fully.
The chapter does not attempt to be exhaustive. In particular, the list of grammat-
ical meanings is not intended to be exhaustive – which may not be possible anyway.
It does not deal directly with rankshifting or “embedding”, which could be regarded
as carrying a separate grammatical meaning, such as “Shift this item to a lower syn-
tactic and semantic rank”; it is assumed that a dependent relative clause, for exam-
ple, is one form of modifier, with the whole clause being the subordinate unit. Nor
does the chapter deal fully with the scope of grammatical meanings, which I take to
be syntactic, rather than semantic. Finally, the sequence of the processes is left for
psycholinguistics to present accurately; any sequences implied here are intended for
conceptual clarity.
There is a quite different kind of significance in English, which could be regard-
ed as grammatical meaning, but which is not discussed here. Hawkins (1994), with
his concept of “early immediate constituent analysis”, shows that in processing what
we hear or read, we take a relative pronoun, for example, as signifying that a rel-
ative clause is beginning. That kind of significance is different semiotically from a
grammatical meaning, in two ways. First, it is conveyed by a signal not a sign; that
is, we have learned that relative pronouns indicate fairly reliably, “Relative clause is
beginning”, but that is not the meaning of relative pronouns, established as part of the
system of English grammar. Second, that significance does not amount to a speaker
instruction, or a hearer procedure. It therefore falls outside the scope of this book.
Organisation of the chapter. The chapter builds on the material in Chapter 4, especially
on §4.5 on grammatical meaning. Section 6.2 details the processes that build semantic
structures, arranging them by the level at which they apply, from senses and groups
upwards; the processes of subordination, coordination, and complementation may in
principle apply at any level. Section 6.3 presents the system procedures that use those
processes to build complex semantic structures. Section 6.4 gives discussion, and 6.5
concludes.
6.1.2 Introduction to system structure
The structure being built includes “grammatical relations” such as modification and
complementation, and grammatical statuses such as being head. Relations and status-
es are in themselves abstract and static; system structure includes them in dynamic
processes – “assembly” in the definition’s term, or “composition” in a more familiar
linguistic term. We must include the dynamics, as well as static structure, because we
are considering language in use, as speaking and understanding.
The vagueness and variability of “grammar” make it not very helpful; the account
given so far of how grammatical meaning uses content to build structure makes the
traditional senses of “syntax” and “semantics” unhelpful also, since they imply distinct
strata which do not interact. The reader is thus being asked to put aside the traditional
use of those terms in the meantime, to evaluate the explanations given in this and the
following chapters, and to reconsider the nature of syntax and semantics after that.
The grammatical meanings which provide the processes are those outlined in
Chapter 4, §4.5. Their operation may be illustrated simply from “those very large loss-
es”. (The steps in the process are given here in the order of the words, not in processing
order.) The grammatical word those requires the process of making the reference of the
whole phrase definite, that is, of relating the phrase to the context. Very adjusts large,
strengthening its degree. The intensified meaning of large is added to the meaning of
losses. The inflection -es requires the grammatical number of ‘loss’ to be specified as
Chapter 6. System structure 127
plural. Losses is established as head. Those processes construct the hierarchic semantic
structure of the group.
The system is like a computer system, with input, processing and output; it is like
the human nutrition system, with input of food, processing of the food, and output
of energy, the function being support of all activities. In this semantic system, simple
semantic structures are the input, the procedures conveyed by grammatical meanings
are the processes, and complex semantic structures are the output.
This type of structure contrasts with network structure (the input), dealt with
in the last chapter, and with hierarchic structure (the output), to be dealt with in the
following ones. It is not a structure of unitary constituents of the same type, as a wall is
a structure of bricks. As with all systems, it is a structure of diverse elements – chiefly
entities, procedures operating on them , and rules controlling the operations – and
the elements are not constituents that make up the whole by addition.
The description above overlaps what is sometimes regarded as syntax. The asser-
tion being made here is that semantics works through syntax; the function of syntax
is to express or even create meaning, rather than to arrange it. The grammatical mean-
ings illustrated so far, and most others, operate on the meaning of syntactic units,
setting up a semantic structure that combines components in a hierarchic or “tree”
structure, closely resembling that of syntax. Thus, modifiers such as large operate on
the meaning of the head of a syntactic group; meanings like those of or and because
operate on the meaning of a syntactic group or syntactic clause. However, grammati-
cal meanings, such as “Relate these items as Topic and Comment,” operate on units of
content – the information – not on the syntactic units; system structure thus works on
content as well as on syntax, and builds information structure as well as propositional
figure structure.
6.2 System processes
6.2.1 Introduction: Grammatical meanings
Summary from Chapter 4. In Chapter 4, a grammatical meaning was defined as mean-

ing (any sort of significance of a linguistic sign) operating on a content meaning. Such
meanings can be seen as the speaker’s instruction for constructing meaning from con-
tent words and so on, or as the hearer’s procedure for it (as in “Add the meaning of
large to the meaning of losses”), or as the status (such as head or modifier) which the
process assigns to meanings operated on. By relating words to each other, and making
them part of a group, they build a larger structure; they operate on units of any size –
words, groups, and figures. Grammatical meanings, also, can be stated in a relatively
general form, as in “Combine these words in a dependency relationship”, or in a more
specific form, as in “Combine these words, with the first adding to the meaning of the
second,” or “… with the first adjusting the meaning of the second”, and so on. (Note
that no attempt is made here to stipulate the form of the meanings; the formulations
given are representations of subconscious mental processes, which do not have verbal
form.) Grammatical meanings apply in syntax – to words, groups and clauses – and in
semantics – to their meanings. The book describes the resultant semantic structures,
but not the resultant syntactic structures.
Concepts entailed in grammatical meanings. As hearers construct the meaning of “The

government has a broad mandate”, for example, they follow a grammatical meaning
implicit in the syntax, which indicates that the meaning of broad must be applied
to the meaning of mandate; the grammatical meaning simply combines two content
meanings, without adding any content. As noted previously, however, some gram-
matical meanings do require the addition of content. For example, the meaning of
wanted includes the concepts PAST and WANT; -ed adds its own descriptive concept,
PAST, to WANT. Grammatical meanings must be regarded as often combining gram-
matical and descriptive meaning in their structure; “grammatical item” is a potentially
misleading term.
Those facts are important: they explain the almost universal confusion about the
grammatical/lexical distinction, to an important degree. Scholars have seen concep-
tual meaning in grammatical items – especially in prepositions – but failed to distin-
guish between concepts that are part of a grammatical function, and concepts that are
part of content function. That in turn has contributed to the concept of grammatical
and lexical “items”, and the failure to understand that the issue is to be understood
through the function of items, not through items as such, or the class they belong to.
6.2.2 Preliminary process: Obtaining content for the main procedure
The main processes to be described require content to operate on, as the word “large”
in the example above needed the content of “losses” to operate on. But in “the big one”,
for example, there is no content in “one” for “big” to operate on. There are two prelim-
inary processes, then, for obtaining content.
The first procedure can be illustrated from example (1).
(1) “After the spinach has cooled completely, pack it together tightly.”
(From a recipe book)
Before the meaning of “pack it together” can be constructed, the content for “it” must
be retrieved from its antecedent, “spinach”. The same process is needed for all pro-
forms, as in “…and John did too,” and “I think so.” The pro-forms, then, signify, “Ob-
tain the content for this word from its antecedent.”
The second procedure is needed for deictics, in slightly different forms. With
spatial deixis, hearers must often understand the place intended by interpreting the
accompanying gesture; with temporal deixis, they understand it from the time when
the utterance occurs. The procedure may be generalised as “Obtain the content by
reference to the speech situation.”
Both processes require the hearer to move out of the figure being constructed,
obtaining material from elsewhere in the speaker’s network of meaning.
6.2.3 Processes applying within words and groups
6.2.3.1 Subordinating content

Introduction. The general subordinating meaning can be formulated as, “Make this
unit subordinate to the head of its larger unit,” or “Make X subordinate to Y”. Subor-
dinating meanings almost always occur in a more specific form, which specifies the
nature of the dependency on the head – or of the nature of the modification, to put it
in other words. They are itemised in the sections that follow.
The grammatical meanings for obtaining content are clearly signalled, by a word –
a pro-form or deictic – but the relationship of head and subordinate is not signalled
so simply. In “With a sharp knife, trim off the ragged edges”, the procedure of making
the meanings ‘sharp’ and ‘ragged’ subordinate is not signalled by a grammatical item,
but by syntactic position.
The rest of this section describes the main processes of making meaning
subordinate.
1. Add to the head’s meaning. The most familiar of the subordinating meanings is,
“Add the content of this item to the content of the head,” or “Add X to Y.”
It is illustrated in several ways in example (2).
(2) “The main ingredient of this sauce is tomatillos, which are small, round, light
green, slightly sour fruit with a distinctive flavour.” (From a recipe book)
At the level of submodification within the group, the sense of light in the second clause
is added to the sense of green, and ‘slightly’ is added to ‘sour’. At the modification level,
the senses of small, round, light green, and slightly sour are added to the sense of fruit;
the sense of with a distinctive flavour is also added to the sense of fruit. At the figure
level, the whole of the “which…” figure is added to the first clause.
Which acts as an unambiguous symbol signifying that the meaning of its clause
must be added to that of the previous one; but there are no such single signs carrying
the “Add…” meaning for premodifiers and submodifiers. For example, light can be a
modifier or submodifier or head; its function as submodifier is signalled by preceding
green (a modifier), in combination with the absence of a comma after it (compare “a
light, green, slightly…”). The submodifying function of slightly in “slightly sour fruit”
is marked more simply and clearly by its ly suffix. Emotive and attitudinal meaning
may be added – not only descriptive meaning – as in diminutive suffixes: lassie, ring-
let, princeling, babykins.
More specific forms of this meaning occur in nominal groups. If the tomatillos in
example (2) were described as “delicious green Mexican fruit”, each of the premodifi-
ers would add meaning of a different kind, as will be explained in the section on nom-
inal groups in Chapter 8 (§2). In information structure, the Comment can be seen as
carrying a version of this meaning: “Add this content to the Topic.”
2. Specify the head’s meaning. The -s on tomatillos in example (2) specifies that the
word is plural. I take it that having grammatical number is part of the significance
of tomatillo in its context, entailed in its being a count noun; so the inflection is
specifying that non-specific part of the head’s meaning. Tense inflections specify the
tense “feature” of the head, similarly. The instruction to the hearer thus varies with
the situation, but may be formulated as, “Specify the [number/tense/etc.] of the head
as [plural/past/etc.],” or “Specify the X of Y as Z.” As noted above, this grammatical
meaning routinely entails the inclusion of a conceptual meaning such as PAST, adding
to the content. The signs for such grammatical specification are not only distinct mor-
phemes such as -ed and -er, but also the abstract fact that an internal vowel has been
changed, as in rode, and mice.
Property words (such as big and grotesque, stealthily and honestly) are subject to
this procedure, being specified for degree of comparison: bigger, more grotesque, and
so on. So are appositional phrases such as “Alistair Brown, the CEO of the company,
…”, since they specify the reference; in other uses, they simply add to the head’s mean-
ing, descriptively.
3. Adjust the head’s meaning in the manner indicated. We have seen the “adjusting”
meaning – “Adjust Y in the manner indicated” – applied to strengthening and weak-
ening the meaning of the head, as with “very large” and “rather large”, and with the
premodifiers utter and mere. Another application of this meaning is in some forms
of modality. In “the alleged culprit”, the status of ‘culprit’ as a piece of information
about the referent is adjusted epistemically – by being reduced from fact to opinion.
The modal auxiliaries in “He must come” and “He may come” similarly adjust the
epistemic status of ‘come’. (Modal meaning is not always expressed in a grammatical
meaning; in “Possibly, he will come,” it is expressed lexically, as content meaning.)
Derivational suffixes commonly adjust the head’s semantic class, often adding a con-
cept to the meaning of the head as well: lengthen adjusts length into being an Event,
adding the concept CAUSE.
This meaning has important uses in information structure. By putting a word in
focus position, at the end of the clause for example, speakers guide hearers into ad-
justing upwards the informational status of its meaning. Focussing use of only and just
also carries this significance. Even use of a pronoun instead of a full nominal group
can be a marker of this meaning. Consider example (3).
(3) “Remove the papery husks from the tomatillos and rinse. Place them in a
medium saucepan.” (From a recipe book)
Reducing tomatillos to them indicates that the referents are already familiar to readers
as Topic; it instructs readers: “Adjust the importance of this item downwards.”
Whereas the adding procedure involves explicit content – contained in a modify-
ing content word – and the specifying procedure involves an implicit concept – con-
tained in the grammatical morpheme – this procedure involves no extra content; it
only operates on the content of the head.
4. Determine the meaning of the head. The grammatical meaning, “Determine the
meaning of the head,” or “Determine Y”, has a range of effects, according to linguistic
context, including limiting the reference of the group, and guiding hearers in identi-
fying the referent. It may be done deictically, with demonstratives, or quantitatively,
with numerals, or abstractly, with articles. The head here is the whole group, except
for the determiner itself.
5. Associate the meaning with the head’s meaning. The types of subordination discussed
so far have given a precise relation between the two meanings, but some modifiers are
quite vague as to the relation to be set up. One common such situation occurs with
nouns used to modify another noun. In “coma baby” and “beer baby” (attested ex-
amples, from newspaper headlines), the relevant relations were, “baby who was born
while the mother was in a coma”, and “baby who was found lying in a pool of beer”.
The reader was to infer the relations from the content of the news report. The linguis-
tic meaning carried by the modifying construction was “Associate the content of this
item with that of the head, in whatever way the context seems to justify,” or “Associate
X with Y”. Another common occurrence of this loose associative meaning is with the
genitive of; the cognitive meaning which the reader must supply may be possessive or
partitive, and so on. (This meaning is the “association operator” of Gil (2005, p. 352),
which gives detail about loose association, and assigns general typological signifi-
cance to it; cf. Chapter 4 (§4.5.2.)
This procedure also has uses in information structure. The Theme of a figure is
merely associated with the Rheme (Chapter 9, §3 and §4). We may also see it in the re-
lation of social and affective meanings to descriptive meaning within a complex sense,
as noted in Chapter 5 (§5.2.2.3). Nondescriptive sense elements are simply associated
with the descriptive meaning, as with disapproval being associated with ‘yellow’ when
it means ‘cowardly’; and that sense was American dialectal use, but is so no longer.
This procedure requires content to be related to the head, but fails to provide the
nature of the relation.
6. Negate the head’s meaning. The nature of negation and the way it operates in English
are complex matters; this section will be restricted to basics. Negative prefixes, as in
unhappy, disapproving and irreducible, instruct the hearer: “Negate the meaning of the
head.” Not does the same for a single word in “a not unpleasant experience”, and of
course not can negate most or all of a clause. Other items that signify the procedure of
negating include only, in many uses, and negative polarity words such as any.
7. Make this the head. The meaning “Make this the head,” applying to morphemes
within a word, to words, clauses and so on, can be seen as a default meaning, applying
to grammatical units if there is no signal requiring subordination, as with one-word
phrases such as “Grass | is | green”. It is also indicated by markers such as the presence
of a determiner or grammatical auxiliary.
Discussion. Most of the meanings discussed in this section operate on, or “take scope
over”, a single semantic unit of the same type – a word is made subordinate to a word,
or a morpheme to a morpheme. Exceptions to that include rankshift or “embedding”,
as when a group or clause is made subordinate to the head of a group. The subor-
dinating procedure applies to prefixes within a word, word senses within a group,
and to some structures in the phonological stratum, as when one tone group is made
subordinate to another.
The subordinating grammatical meanings grade in basicness. Their relationships
may be shown in a semantic map, as in Diagram 1, with lines linking those that seem
to have a natural connection. The horizontal dimension of the diagram represents the
order in which I believe these meanings develop in childhood, and the order in which
they seem likely to have developed historically. (Note, for example, the basic role that
Gil (2005) gives association in typology, and note that determiners were very late to
develop in Old English – long after adjectives developed their various uses.) The ver-
tical dimension represents conceptual basicness – whether a concept is fundamental,
or to be explained in terms of other concepts. (Note that negation here is specifically
in modification; it is much more basic in its role of denial in the whole sentence.)
Diagram 2 is a reworking of the semantic map in Diagram 1, to show how the
grammatical meanings combine in the modification of Entity groups (roughly, “nom-
inal phrases”), Event groups (roughly, “verb phrases”), and the Property groups
(roughly, “adjectival and adverbial phrases”). The groups are represented by rec-
tangles, and identified at top left in italics. Property groups, shown by the smallest
Negate
More
Adjust
conceptually
Specify basic
Add
less
Associate Determine
More developmentally basic less
Diagram 1. Relationships among grammatical meanings

Marked Event groups, + sublexical sernses
Negate
Entity group
Event group
Property group
Adjust
Specify
Add
Associate Determine
Diagram 2. Relationships among grammatical meanings for different semantic units
rectangle, allow only specification and adjustment (specification of degree by -er and
-est for example, and adjustment of intensity with very.) Event groups allow a slightly
wider range, allowing addition of meaning, as well as specification and adjustment.
Entity groups allow a very wide range of subordinating meanings, as shown by the
examples given so far. “Marked verbal groups” denotes examples like “he didn’t buy
the bike – he stole it”, where the negative modifies ‘buy’ alone, not ‘buy the bike’. An
example of a sublexical sense is ‘happy’, which can become ‘unhappy’, ‘happier’, ‘very
happy’, ‘ecstatic’ and so on.
This account covers the types of modification discussed in formal semantics such
as that of Kamp and Partee (1995), and that of some other writers. Their “intersective”
modification is the additive type here; some of their “subsective” types are in the ad-
justment type here; and their “privative” qualification is negation. Their account omits
the remaining types discussed here.
6.2.3.2 Co-ordinating content

Introductory. By definition, co-ordinating is relating items of equal status; it is also
in fact relating items of the same structure. The commonest elements that carry this
meaning are the conjunctions and and or; but it is carried also by the commas and
phonological rising tone that substitute for and in lists, as in “Ministry of Business,
Innovation and Employment”. The speaker’s implicit instruction is, “Relate the two
items as equals in status and function,” or “Relate X and Y as coordinates.” The scope
of a co-ordinating meaning is defined by the syntactic unit following it: word, group
or clause.
The procedure can be carried out on modifiers of the same head, as in “red and
black balloons”, or on heads within the same group, as in “Jack and Jill went up the hill”.
Some compounds co-ordinate morphemes, as in “fighter-bomber” and “clock/radio”.
Forms of co-ordination meaning. The nature of the relation of equality is made more
specific by the content of the linking word; among conjunctions, or signifies ‘alter-
native’; but signifies ‘oppositeness’ of some kind, and signifies ‘addition’, and so on.
Among prepositions, the on in “on reclaimed land” signifies location, and in the
phrase, “in the 1920s”, in signifies temporal relation, although in other contexts it
signifies spatial relation. (Again we see items with grammatical meaning contributing
content meaning as well.) Just as adverbial clauses have traditionally been classified
as those of time, place, manner and so on, so we could here subclassify coordina-
tion meanings as those of similarity, oppositeness and so on; but since the distinc-
tion is a cognitive conceptual one, not a linguistic one, that subclassification is not
appropriate here.
Occasionally, the nature of the co-ordinating relation is not made specific. Com-
mas in writing, and the corresponding intonation in speech, may be used where the
link is to be logical, or temporal, or descriptive; the link is unspecified, resembling the
underspecification of the association link discussed in the last section. The conjunc-
tion and is also sometimes non-specific, as when it is used to join events, effectively
meaning ‘then’, and when it is used as in “Touch that and I’ll belt you!”, where it effec-
tively means ‘if ’. In those cases, and has the linguistic meaning ‘addition’ as usual, but
hearers interpret it cognitively as ‘if ’.
6.2.4 Processes applying within figures and figure complexes:

Complementation
Complementation is like co-ordination in relating items of equal status and struc-

ture; but where coordination relates items with the same function, complementation
relates items with different functions; the complementation meaning, formulated
generally, is “Relate the items as equals in status but of different function”. In syntax,
the clause constituents – Subject, Predicator, Complement and Adjunct – are com-
plementary; that is, they are of equal status, but with different functions. Similarly,
the clause’s semantic constituents – Participant(s), Process and Circumstance – are in
complementary relation.
Strikingly, there is frequently no overt signal of complementation – no grammat-
ical inflection or conjunction that carries a relevant grammatical meaning or acts as
a marker. The covert markers are in fact familiar, of course; they include the semantic
class of the group heads, the order of groups, and agreement between subject and
predicate. However, for Adjuncts and some Complements, prepositions are a signal:
in “It was built on reclaimed land in the 1920s,” the prepositions signify that their
phrases complement “It was built,” as Adjuncts, just as the Entity group “reclaimed
land” complements “on”. (But prepositions in postmodifying groups such as “the

house on the hill” make their nominal groups subordinate, not complementary.)
Clearly, the relation between the actor Participant and the Process is significantly
different from the relation between the Process and the goal Participant; the gram-
matical meaning can be formulated more specifically to show that: “Relate the two
items as a Process and a goal Participant”, for example. (Being Process and Participant
implies the now unstated “as equals in status but of different function”.) The identity
of the two items to be related is controlled, as before, by the syntactic structure: a
following group is to be related to a preceding one, and so on. The nature of the func-
tion (Participant, Process and so on) is signalled by various things according to the
linguistic context; being a Process is indicated by the syntax of being a verbal group,
and the morphology of inflection and agreement, for example.
As noted just above, complementation occurs mostly in constructing figures;
but it applies occasionally in morphology, where the elements of certain compounds
are Complementary, as in pickpocket, letter-writer and turncoat. We will also meet it
in Chapter 9, on information structure, where Comment complements Topic, and
Rheme in one way complements Theme.
6.2.5 Processes applying to a whole figure
Introductory note. Figures can be co-ordinated and subordinated into complexes in

the ways described in traditional grammar for complex sentences, with conjunctions
signalling the processes, much as described in the preceding sections. No further de-
scription is needed.
6.2.5.1 Processes relating content to the hearer’s knowledge

When hearers have followed the procedures above, and built a semantic structure by
adding meanings to make groups, and by relating the groups in complementary rela-
tion to make figures, the structure still has no practical value for them until they relate
the content to their own knowledge, and gather whether the speaker intends them to
accept the information, or answer the question, and so on.
The first stage, that of relating content to the world of experience and learned
knowledge, has traditionally been thought of as making “acts of reference” – making a
link to “reality”. That understanding has been based on a philosophical tradition, and
is not helpful for linguistic understanding. There must be something that constitutes,
or at least a signals, the “act of reference”. Traditionally, it has been thought to be noun
phrases that refer. That can be only part of the truth. Entity groups (“noun phrases”)
in isolation do not refer; it is an assertion – a complete figure structure – that signals
to readers that they must relate what is being said to their knowledge of the world;
and as hearers, we need a whole utterance to let us understand reference. Figures thus
constitute a level of grammatical meaning, above the level discussed so far, since they
require the procedure of moving out of the linguistic sphere and into the cognitive
one. The meaning may be formulated, “Relate this content to your knowledge.”
While our metaphysics may want us to restrict reality to things – designated by
noun phrases – English as a language does not work that way. Consider example (4),
from a newspaper report of a terrorist holding hostages in a cafe.
(4) “Police made their way to offices at the back of the cafe.”
(New Zealand Herald, December 17th 2014, A3)
As a factual report, the sentence treats the action as the main “reality” to be reported,
in “made their way”– with the minor irony that the definite noun phrase, “their way”,
denotes an abstraction, not a metaphysical reality. “At the back” is also presented as
part of the reality being described, although it is a property of the offices, not an
independently real object. We should put aside the traditional philosophical view of
reference.
6.2.5.2 Processes guiding the hearer’s response

General response. The final stage in the system of composing meaning, following hear-
ers’ relating content to their own knowledge, is that in which speakers guide hearers in
how to respond to the content which has been constructed in the figures and related
to the hearer’s knowledge. Psychologically, there are many possible responses, with
no obvious system in them. But linguistically, there is a paradigm of choices: English
grammar has systematised response along with so much else. It provides for speakers
to expect an answer (to a question), obedience (to a command), affective response (to
an exclamation), or acceptance (of a statement or “assertion”). Three of those forms
are, of course, the “sentence types” or “moods”: interrogative, imperative, and declar-
ative (making a statement or assertion).
I have included exclamation with them, because the three familiar ones are syn-
tactic whereas we are dealing with semantics; and exclamation requires a response
distinct from the others, and has its own forms of expression – the high rise and fall
tone in intonation, and the “WH-word” construction in morphosyntax. Exclamations
are as fundamental as statements and commands; they are not a subclass of any of
them (as affirmations and denials are subclasses of statements); and they arise from
a distinct intention. The set of four make a full functional paradigm, combining both
functions, the ideational and the interpersonal, and coordinating what appear to be
the main faculties of the mind – cognition provides statements and questions, gener-
ally speaking; conation provides commands, and the affective faculty provides affect
in exclamations.
Since those forms of expression guide the hearer in how to deal with content, they
carry grammatical meaning, which may be formulated briefly as follows. For declara-
tive figures, “Accept this;” for interrogative figures, “Reply to this;” for imperative fig-
ures, “Act on this;” for exclamatory figures, “Share this, as feeling.” A full formulation
of those grammatical meanings would include such things as provisional acceptance
in modal statements, and subordinate status in subordinate clauses.
This is the culmination of the system of building semantic structure in English.

It fulfils the intention which begins the processes; and it is the last stage, completing
the process begun by the preliminary processes (§6.2.2), and carried out largely by the
compositional processes (§6.2.3).
These grammatical meanings are commonly known as “speech acts”: “intentional
action[s] performed by means of an utterance” (Levelt 1989, p. 58). The term “act”
is appropriate to the extent that it suggests interaction, which is an integral part of
the speech situation, and fits the concepts of ‘move’ and so on, in discourse analysis.
But “speech act” is an unsatisfactory term. First, it is well known to be vague, with
no consensus having arisen on which acts are to be included. More seriously, it is
too wide: “acts” contrasts with expression of meaning, and entails going outside lan-
guage; but “acts” of reference also do that; qualifying a statement modally is acting,
rather than expressing a meaning; and perhaps deixis and naming are also actions.
The term is based on the misapprehension that language is simply representation of
meaning – rather in the philosophical tradition which analyses language as consisting
of propositions – which makes it difficult to account for assertion. When we accept
that language is fundamentally functional, “acts”, or doing things with language, is as-
sumed, and needs no special place. The element of acting is subsumed by grammatical
meaning, as the speaker’s guidance, and the hearer’s procedure in response.
“Predication” has not been used here, because it is ambivalent between denoting
the act of assertion and being a syntactic term.
Specific responses. The signal given by declarative form that the hearer is expected to
accept the content is given more specific form by what is generally known as “infor-
mation structure” or “information packaging”. The specific forms are in fact what we
have seen already in the sections on coordination and complementation, especially
§6.2.4. The Topic-Comment structure (discussed more fully in Chapter 9) signals that
the Topic section of the content should be treated as the part of the hearer’s knowledge
which the figure is to be about; and signals that the Comment section complements it,
being what the speaker regards as the significant material he or she is saying about the
Topic. Similarly, the Theme-Rheme structure signals that the beginning of the figure,
as Rheme, should be used by hearers to orient themselves to what is to follow, and that
the Theme complements it. The Theme structure signals the degree to which the hear-
er should make most items of information subordinate in importance to the focus.
6.2.6 Discussion: System processes
If we think about the occurrence of the three main processes (subordination, coordi-
nation and complementation), then we see that coordination occurs at all levels, but
that subordination occurs in structuring groups but not figures, and that complemen-
tation occurs in structuring figures but not groups. (I am putting aside figure com-
plexes, considering only the basic semantic structures.) That pattern is very striking; it
seems quite anomalous that such powerful structuring devices should be unavailable
at particular levels. The only possible explanation I can see is that it has evolved evo-
lutionarily, as a valuable way of keeping the three layers of morphosyntax distinct, and
perhaps easier to process through greater simplicity.
6.2.7 Conclusion: Grammatical meaning
This section has listed and described the grammatical meanings of English, present-
ing them as processes in a system, the system which builds the hierarchic and other
semantic structures which will be described in the following chapters, using as input
the sense networks described in the previous chapter.
In summary, the processes may be listed as below.
a. Complementation; it operates on groups in a figure; specifically, it establish-

es them as equal in status, but different in function (Participants, Processes or
Circumstances).
b. Coordination; it operates on most types of unit; it establishes them as equal in
status, and with the same function.
c. Subordination; it operates on words in a group, producing the following forms of
modification: description (adding meaning), specification, adjustment, determi-
nation, association, and negation; it also operates at several other levels.
Grammatical meanings operate at all morphosyntactic levels. At the lower levels, they
simply relate content; at the middle level (the significance of groups in a figure), they
deal with relating content to the hearer’s experience, as things, happenings and cir-
cumstances. At the highest level (figure structure), they deal with structuring content
as hearers relate it to their network of knowledge, and deal with how the speaker
wants them to respond. (That outline of grammatical meanings’ domains of operation
will need refinement after we have dealt with information structure in Chapter 9.)
6.3 System procedures: Using the processes
6.3.1 Introduction
Section 2 described how speakers use grammatical meanings to convey to hearers

what they must do to compose the meaning of the utterance. It presented the sys-
tem of general processes which must be carried out, systematically and in sequence.
This section deals briefly with the particular procedures or “tasks” which use those
processes to build the semantic structure of English utterances. Chapters 7, 8 and 9
will set out the nature of hierarchic semantic structures, and in doing so will note the
procedures carried out in building them; but it will be useful to give here an imme-
diate, very small scale illustration of them next (§6.3.2). The later chapters will not
deal with networks, because the only networks built in syntax are the minor ones al-
ready described at the end of the last chapter. This section will therefore illustrate how
grammatical meanings build those relatively uncommon small networks, in §6.3.3.
6.3.2 Constructing hierarchic structures
Let us consider an example of how grammatical meanings operate to build hierarchic

structures.
A reporter who was summarising a report into the CIA’s efforts to counteract
terrorism after 2001 wished to write that the CIA tried to prevent Congress from
checking what it was doing, and that it took active steps to achieve that goal. What he
wrote is given in example (5).
(5) “The CIA actively impeded or avoided congressional oversight.”

(New Zealand Herald, December 28th 2014, A28)
In planning the sentence, the reporter evidently decided to have the CIA as its topic,
and to formulate the information as how CIA treated Congress; and he selected the
word senses ‘actively’, ‘avoided’, ‘CIA’, ‘congressional’, ‘impeded’ and ‘oversight’. The
reader procedures for constructing the sentence using those senses can be represent-
ed as the grammatical meanings in example (6i–iv). The signs used to symbolise those
grammatical meanings are the words symbolising the senses, and other signs indicat-
ed by bracketed phrases in the following account.
(6) Procedures for constructing example (5).

i. Make ‘CIA’ a head, and determine that head from the context (as indi-
cated by the).
ii. Make ‘impeded’ and ‘avoided’ head words; coordinate them (as indicated
by or) and add ‘actively’ to both heads as modifier (as indicated by its
position in front of them).
iii. Make ‘oversight’ a head, and add ‘congressional’ to it as modifier (as indi-
cated by its being in front).
iv. Relate as complementary the three groups, ‘the CIA’, ‘actively impeded or
avoided’, and ‘congressional oversight’, relating them as agent Participant,
Process and affected Participant (by making them Subject, Predicator and
Complement, as indicated by the morphosyntactic order and agreement).
That series of steps is intended straightforwardly as an illustration of the system struc-

ture in English semantics – the system that builds senses with their internal network
structure into complex hierarchic structures such as the figure and groups in exam-
ple (5). But it is intended also to make a point about the relation between semantics
and syntax. The steps given are essentially identical to those for building the syntax
of the sentence. From the functional point of view adopted in this book, that is so
necessarily: syntax exists for a function, which is to build the meaning – partly by as-
sembling the meanings symbolised by lexical words, and partly by providing further
symbols such as grammatical items (the and or, in the example) and word order.
The analysis in (6) makes a further point. Every word in the utterance is involved
in grammatical meanings; speaking imprecisely, one could say that, in use, every word
has grammatical meaning, not just “grammatical items”. (Strictly, it is sometimes the
word’s position, not the word as such, that carries the grammatical meaning.)
6.3.3 Constructing network structures in morphosyntax
I will demonstrate how grammatical meanings are used to build network structures
in morphosyntax, by using an example from the last chapter (§5.5.1), repeated here
as example (7).
(7) “He buried his children bitterly.” (Quirk et al. 1972, §8.42)
Readers will apply to ‘bitterly’ the standard grammatical meaning for a property sense
standing as a separate group, namely, “Relate this content to the rest of the figure, in a
complementary relation, as Circumstance.” The conceptual network described in the
last chapter will then be built as further meaning, in some such way as this: (i) add
BITTER to the concept of the man, as a property; also (ii) add it to the concept of bur-
ying, as a property. It now has three links, instead of the normal single link.
Another example given in the last chapter (§5.5.2.1) is repeated as (8).
(8) “Tuck it inside a good thick piece of French bread.” (COCA)
The standard reading invokes the grammatical meaning, “Add ‘good’ to ‘bread’ as
modifier,” contributing to the regular hierarchic structure constructed for a sentence.
The network reading will apply the same procedure, of adding senses as modifiers, in
two other ways: ‘good’ will be added to ‘thick’ as an intensifying modifier (‘very good’),
and ‘thick’ will be added to ‘good’ as an explanatory modifier (‘good because thick’).
6.4 Discussion: System structure
6.4.1 Details not yet explained
Some details of the operation of grammatical meanings will be explained in Chap-

ters 7 to 10, because they depend on structure to be explained there. They involve
the internal structure of different semantic classes, which controls what other senses
a word sense can relate to, and therefore what grammatical meanings can operate on
the sense.
6.4.2 Signs used for grammatical meaning
As shown in the preceding sections, the signs carrying grammatical meaning are var-
ied, and their use is complex.
Some of the signs are symbols, in being fully conventionalised, with a set form
and set significance. The section has recorded the following types of symbol for gram-
matical meanings. Some are concrete, namely certain morphemes – free and bound
morphemes with grammatical meaning. Some are semi-abstract – the semantic class
of the content word. Some are fully abstract: position (before or after another content
item, for example); order (of three or more items in complementary relationship, for
example); and intonation pattern (or the corresponding punctuation, in writing).
The signs sometimes operate on their own, as with some grammatical mor-
phemes, such as -ed, but many operate in combination with another sign. We have
just seen that with the. For many modifiers, position before the head combines with
the word’ s own content, as with intensifying modifiers like utter and modal modifiers
like alleged. Semantic class combines with position, since most content words can
be a modifier or a head – in modifier or head position; in the following examples,
position before the head swim goes with several quite different grammatical statuses:
“the swim”, “will swim” and “fish swim”. The signs vary in complexity. The is relative-
ly simple in meaning, signifying definiteness, although the nature of definiteness is
ambivalent. Those, however, signifies not only definiteness, but also deixis, distance
and plurality. Some grammatical signs, such as the pronouns and verbal inflections,
form full and systematic paradigms, but others, such as the articles, form much less
systematic paradigms.
Some of the grammatical signs are not symbols but markers (see Chapter 3, §3.4.3
on semiotics), in being not reliable as indicators. Examples include: to as marker of in-
finitive status; semantic clash; and absence of (other) content items, which will mark
head status in certain circumstances.
We should note one negative fact. “Part of speech” has not appeared in the list of
grammatical signs, although “adjective”, for example, seems to imply that the word
carries the grammatical meaning, “Add the content of this item to the content of the
head” (§6.2.3.1 above). Words are no longer systematic signs of grammatical meaning
as they were in Old English; as we have seen, “adjectives” and “nouns” may be either
modifier or head, according to their position. Second, their nearest equivalent, se-
mantic class, does not have the grammatical functions traditionally attributed to word
class. The issues are among those noted in §6.4.1 above as being yet to be explained;
they will be explored in Chapter 13.
6.4.3 Grammatical meanings in figurative and other uses
Figurative use of language requires reconstrual of the literal meaning. With figures of
speech such as metaphor, that is generally triggered by a semantic clash (Cruse 2011,
§12.2.1). When we read that a cricket team “unleashed their arsenal of skills,” for ex-
ample, the clash between physical, threatening weapons and abstract, admirable skills
forces us to reconstrue the meaning of “arsenal”. The semantic clash is in effect a sign,
with a significance amounting to a grammatical meaning such as “Reconstrue this
utterance figuratively;” but the sign is not a symbol, in the strict sense (see Chapter 3,
§3.4.3).
How hearers distinguish between restrictive and descriptive uses is less well-
defined. In unmarked use, the definite article marks modifiers straightforwardly as
being in restrictive use, as in “The next train is at 6.10.” But in “The elegant, tree-lined
boulevards of Paris”, the premodifiers are descriptive in spite of the definite article.
The reader infers that from the fact that the combination of “the” and “of Paris” identi-
fies the referent completely, determiner and postmodifier acting together, as markers;
the modifiers are therefore presumed to be descriptive. We see that the restrictive/
descriptive distinction, as well as the figurative/literal distinction, is subject to gram-
matical meanings, which do not form a simple pattern.
Most of the instances of grammatical meaning given have applied to descriptive
meaning; it also applies to emotive and attitudinal meaning, as with the negative pre-
fixes of unhappy and disapprove when those words are used affectively. Social mean-
ing is by its nature related by mere association to the relevant descriptive meaning; it
does not appear to be subject to any other grammatical meaning.
6.4.4 Grammatical meaning as backgrounded meaning
General. It has often been noted, especially in the grammaticalisation literature, that
grammatical meaning can be seen as meaning that has been “backgrounded” – re-
duced in salience, or consciousness, in speakers’ and hearers’ minds. There is useful
truth in that view; for example, “He arrived by train” tells us that the event was in the
past, but that information is not salient, because it is expressed grammatically. How-
ever, the view is sharply limited by its assumption that grammatical meaning is simply
conceptual, like PAST and PLURAL, without a procedural element; as procedure, gram-
matical meaning may be below usual consciousness, but it certainly not unimportant.
Grammatical metaphor. Like content meaning, grammatical meaning has a figurative

use (see Chapter 4, §4.7.4), as well as a literal use. Grammatical metaphor is the rep-
resentation of a grammatical form by another which is similar to it, just as standard
or content metaphor is a representation of some content by something else which is
similar to it. (See Chapter 12, §12.6.2, for further detail.)
An important example of it is modality. If we assume that modality is representa-
tion of a speaker’s assessment of probability or obligation in what is said (Halliday
2014, §4.2.2.1), or representation of the speaker’s commitment to the factuality or
desirability of what is said (Levelt 1989, p. 62), then it would be stated straightfor-
wardly and literally (“congruently”) in a figure, such as “I am certain that…” or “That
statement is only hearsay”. Modal auxiliaries represent that form, the figure, as a much
reduced form, a modifier; whole figure is represented by something similar to it

(figures and modifiers are both semantic units); the use constitutes a grammatical
metaphor, functioning to background the comment on factuality or desirability.
Rankshifting is also grammatical metaphor; a restrictive “adjectival” clause treats a
clause as if it were a word, representing a modifying sense by a figure.
6.4.5 Other views of grammatical meaning
Traditional grammar. Traditional grammar dealt with the issues in this chapter only
partially, by using the concepts of open and closed classes, which cannot be defined
strictly and are overlapping, and which are “prototypes”. (Muysken 2008 gives a highly
developed recent view, with nine criteria.) There is no explanation of how the classes
work; and, although grammatical items like the and a/an are clearly signs, there is
no explanation of how they carry meaning. I suggest that there are two fundamental
problems in the approach: the preoccupation with trying to explain things by classi-
fying them, and the failure to find a proper relationship between grammar and mean-
ing. These theorists themselves concede that they do not have a clear explanation.
The resolution of the puzzle is quite simple. Just as the mouth has two functions,
namely eating and speaking, so do words other than names and deictics have two
functions, namely carrying content meaning and carrying grammatical meaning. We
do not propose that organs belong to classes, and hypothesise that the mouth is a
“central” eating organ and a “peripheral” speaking organ; we simply acknowledge that
organs have functions – one or more. So do words.
“Procedural” and “functional” meaning. A few linguists, represented by Blakemore

(2002), describe “procedural meaning”, which is very similar to the grammatical
meaning described here; however, as they define it, it has very limited application,
occurring only in pragmatics, and having little functionality, doing no more than de-
veloping inferences. Harder (1996) describes “functional meaning”; I can accept his
work, but it is theoretical and rather programmatic, so I have not found it useful.
Croft (2001, Ch. 2) describes meaning in general as “functional”. Halliday (e.g. 2014)
seems to be in great agreement with what is given here, but he does not appear to have
dealt with semantics in sufficient detail for a useful comparison to be made. Func-
tional Discourse Grammar, as in Hengeveld and Mackenzie (2008), gives a useful
and closely parallel account of grammatical meaning; it is limited by its restriction of
grammatical meaning to grammatical items or “operators”, without integrating their
functioning with that of syntactic order and so on, which are also signs of grammat-
ical meaning.
Operators. Functional Discourse Grammar, as in Hengeveld and Mackenzie (2008),

and other more formal grammatical theories present grammatical meaning through
the concept of operator. The concept is useful in describing the function of auxiliary
verbs and inflectional morphemes, for example; but the mathematical analogy does
not go far, and soon becomes unhelpful. Premodifiers “operate” on their head, but do
not seem to be “operators”. “Grammatical items” such as prepositions and conjunc-
tions serve as operators, but also commonly carry content. In the interpretation of
ambivalent and ambiguous structures, a choice must be made, and made by a person
not a linguistic operator: in fact, it is the hearer who performs the procedures, and is
the operator, in the instances above, not a word. The concept is unduly simple and
mechanical.
Construction Grammar. Construction Grammar is a response to the same general

perception that motivates the description of grammatical meaning here. Sample views
are that each construction is a conventional symbolic unit (Langacker 1987a, pp. 57–
63); that “the basic units of grammatical representation are symbolic” (Croft 2005,
p. 275); that constructions are “pairings of grammatical form and [a] corresponding
meaning or semantic structure” (Croft 2001, p. 21). Constructions range from syntac-
tic forms such as passive clauses, through morphological forms such as inflected verbs
and nouns e.g. “[Noun-s]”, to lexical items such as this and green; and they range from
the “schematic” to the “substantive” (Croft 2005, p. 275). Consequently, there must be
an indefinitely large number of them; and, since each construction is a conventional
symbolic unit with its own meaning, there must be an indefinitely large number of
meanings. I believe that the account given here is better, in making the meanings
specific, and in showing how the meaning of the construction is built up. It is perhaps
also better in reducing a very large number of constructions to a very limited set of
procedural meanings, structured in a way that matches syntax. (Further critique of
Construction Grammar is given in Chapter 7, §7.8.2.)
“Grammatical relations”, thematic roles”, and so on. “Grammatical relations”, thematic

roles”, and so on deal with grammatical meaning in part, since they denote semantic
structures set up by grammatical meaning. Since they are dependent on the structure
of figures, they are dealt with in Chapter 7.
Semantic relations. The system structure set out in this chapter is an approximate
equivalent of the semantic relations set out by some linguists as the basic structure
of language. Lyons (1977), for example, sees language as the expression of meaning,
and meaning as consisting of semantic relations. The view taken here accepts that, but
extends it: Lyons’s view of the relations is in effect the system aspect alone; we should
allow for the dynamic hearer and speaker aspects, involving processes and procedures
as well as static relations. The same critique applies to dependency grammar.
6.5 Conclusion: System structure
Summary. This chapter has presented a system structure in the semantics of English.
It is a system in being a set of procedures (specified by grammatical meanings) which
operate on input (content, both conceptual and affective, in the networks set out in
the previous chapter), producing output, and in having a particular function. With a
few exceptions, to be dealt with in later chapters, the output consists of figures, which
are structured in ways to be described in the following chapters.
There are three stages – conceptually, that is, rather than in psychological process-
ing. For pro-forms and deictics, the content to be operated on must first be obtained
from elsewhere; in other instances, that preliminary stage is unnecessary. In the sec-
ond stage, the general procedures are either (1) making content meanings co-ordinate
or (2) making them complementary, which establish syntactic and semantic relations
“horizontally” in the hierarchy, and also (3) making one item of content meaning sub-
ordinate to another, “vertically” in the hierarchy. Those processes operate in two ways,
in parallel: they operate on morphosyntactic units (which are small networks, inter-
nally), producing the familiar syntax-like structures of meaning, but also operate on
content units, producing information structure; that distinction in what the processes
operate on is the basis for the distinction to be made occasionally hereafter between
syntactic grammatical meaning and informational grammatical meaning. This chap-
ter has concentrated on the syntactic use, while Chapter 9, on information structure,
deals with the informational use.
In the third stage, the procedures are more general, and at a higher level. They
guide the hearer, not in how to relate content items to each other, but in how to relate
content to existing knowledge (as in “acts of reference” and “information packaging”),
and in how to respond to it – by acceptance, reply and so on – in “speech acts”.
Conclusions drawn. To my knowledge, system structure is not given in any of the

literature. The only procedures recognised are thought of as “processing” – being psy-
chological, not linguistic – and as interpreting structure, not creating it. We have seen,
however, that they are linguistic, because they are specified in the meaning of our ut-
terances – in grammatical morphemes, many syntactic structures, and phonological
patterns.
We should note that a grammatical meaning and its procedure not previously
noted were introduced in §6.2.2; a couple of others will be introduced later; the book
does not assert that the meanings noted make a complete or definitive list.
Looking forward. The next four chapters will consider in detail the structures built as
output by the system considered in this chapter. Chapters 7 and 8 consider the hierar-
chic structure of content built by syntax; Chapter 9 considers information structure,
which is also hierarchic. Chapter 10 deals with other structures.
Chapter 7
Hierarchic structure (1)

Figures
7.1 Introduction
This chapter begins expounding hierarchic semantic structures in English, the third
major type after networks and systems. The hierarchy, which parallels the syntactic
hierarchy closely, is produced by the system set out in Chapter 6, and built up from
the senses set out in Chapter 5, and the elements set out in Chapter 4.
Section 7.2 sets out the figure structure, at the top of the hierarchy. Sections 7.3 to
7.5 describe the constituents of those structures (Participants, Processes and Circum-
stances), and section 7.6 describes the relations among them. Section 7.7 describes
figure complexes (combinations of figures in one sentence). Section 7.8 gives discus-
sion, and Section 7.9 concludes. (Lower levels of the hierarchy are dealt with in the
following chapter.)
7.1.1 Introduction to hierarchies in semantics
The basic structures of language are not hierarchic: the physical reality of language –
the stream of sound – is linear, and the cognitive and affective elements it expresses
form a network. There are substantial reasons why semantic structures (and the mor-
phosyntactic and phonological structures which express them) are hierarchic.
Cognition, of both physical objects and abstractions, uses hierarchic taxono-
mies. Memory operates with them, which facilitates the acquisition and use of large
amounts of information (Kuczaj and Hendry 2003, p. 257). They are particularly
important for working memory, which is crucial to all production and comprehen-
sion of language. (A general item can be held in working memory, with its specifics
available in long-term memory.) Hierarchies are not only very much simpler than
networks, and therefore faster in operation, but also provide two entry points for
processing. The top-down mode provides for speakers to develop their intention to
speak, working through a broad structure of the utterance and further differentiation
into elements; and it provides hearers with the unity and coherence of the complex
whole. The bottom-up mode allows both speakers and hearers to construct complex
utterances from a multitude of pre-existing senses, using the pre-existing grammat-
ical meanings. (The two modes can be combined in each utterance – Lamb 1999.)
Finally and crucially, hierarchy provides a way of building a nonlinear structure from
a linear one, since a segment of the sequence can signal that it is not to be interpreted
sequentially, but as being on a higher or lower level than the preceding or following
segment.
In this chapter and the following one, the highest level is taken to be that of the
figure, just as the highest level of syntax is usually taken to be the clausal sentence,
although in both instances a case can be made for identifying a higher level, such as
paragraph or text.
7.1.2 Introduction to figures
Characterisation and definition. Figures were introduced in Chapter 2 (§2.4.1) as the

semantic equivalent of clauses, which realise them; in a narrative, each figure repre-
sents an action; in expository writing, each figure expresses a proposition. Defined
formally, a figure is a “configuration of a process, participants involved in it and any
attendant circumstances” (Halliday 2014, §5.1.1). The Process is realised by the Pred-
icator; the Participants are realised by the Subject, and by the Complement if there is
another Participant; any Circumstances are realised by an Adjunct. (As noted at the
end of Chapter 1, terms such as “Process” and “Participant” are capitalised to mark
their technical use.) Clauses therefore signify figures; they are signs, with the gram-
matical meaning, “Interpret the content represented here as a figure.”
Note that it is in fact only descriptive or “conceptual” meanings that are thus spec-
ified. English has no articulated hierarchic structure for feelings, attitudes or social
meanings, as some animals and birds do; they use patterns of vocalisation or gesture
or action to establish “pecking order”, for example. But human communication is sys-
tematic only for descriptive meaning, the non-descriptive meanings being realised
incidentally, through intonation or discourse markers, for example.
Relationships. A figure is a relationship; it has terms, i.e. the Participants, and a rela-
tion, i.e. the Process, but those are not mere constituents which make it up additively.
In that, it is like any other relationship, such as friendship or fatherhood. Just as the
term “figure” is needed to distinguish the semantic reality from “clause” as the syntac-
tic reality, the other terms here are needed for the same reason. “Participant” refers
to what is also a “semantic group” (realised in syntax by a nominal group); but using
“Participant” signifies that we are referring to the semantic group in its role in the
figure. Similarly, “Process” denotes the meaning of a verbal group in its role in the
figure. (Items that link figures, usually realised as conjunctions, are not part of the
figure, although they are part of the clause, in most analyses.)
Function. The default function of figures is to operate in the interpersonal metafunc-

tion, creating a unit of interaction, namely a statement, a question, a command or an
exclamation. (In marked function, figures can, for example, act as nominal clauses,
serving as Participants in a larger figure.) Representing a proposition is a secondary
purpose, serving that function in statements and questions.
Chapter 7. Hierarchic structure (1): Figures 149
“Internal structure” and “syntagmatic structure”. To understand semantic structures

such as a figure, we need to understand how word meanings are linked – how appar-
ently independent items become connected in a semantic network and a syntactic
chain, and how linear sequence can be interpreted as hierarchic structure. We need
to understand whether there is a “mortar”, which is added to “stick” the elements
together, or whether there is something in the semantic makeup of the elements that
provides for the linkage, as the nature of electrons provides for linking atoms into
molecules. Discussion of figure elements in this chapter will be divided into a section
on “internal structure” of the atom-like units, and a section on “syntagmatic struc-
ture”, dealing with how that structure functions to create a syntagm out of the units.
However, Processes are realised as groups, and it is the head of the group that deter-
mines its syntagmatic relations with other groups; so the illustration and explanation
of syntagmatic structure will be completed in Chapter 8, which discusses group and
sense structure.
7.2 Figure structures
7.2.1 Introduction: Processes
We think of clauses, in syntax, as having two types of structure, the transitive and the
intransitive; there is a transitive/intransitive distinction in semantics, but it is sec-
ondary to the distinction between types of Process, some of which have transitive
and intransitive variations. Accordingly, figures occur in several different structures,
according to the type of Process. The Process is defined by its place in the figure, not
by anything more concrete such as being a “verb” or denoting an action. That is be-
cause it has evolved under various pressures, which have resulted in a complex set of
characteristics. One pressure, of course, is that of conforming to the old syntax of Sub-
ject + Verb + Object. A second pressure has been that of representing our experience
of the world. Its ancient role was to represent events; but (to take an instance) gram-
maticalisation of be has replaced the function of denoting an event with the copula
function; subsequent grammaticalisation has converted other event meanings into
grammatical and relational meanings, as with ‘have’ and ‘own’. A further and more
recent pressure has been the value of aiding the interpersonal metafunction, which
developed the grammatical meanings signifying predication and assertion (specifi-
cally signalled through the verbal inflection). Those interpersonal functions are now
dominant, controlling the nature of Process, and thus the nature of Predicators: the
Process serves to signal that something is being predicated of the Subject Participant,
and that an assertion is being made; it may once have been definable as denoting an
Event, but that is not so now.
Furthermore, the evolutionary process has in part been one of differentiation.
As noted by Kemmer (1993), situations or “states of affairs”, which are represented
by figures, can be differentiated to varying extents. In expressions like “it’s raining”

there is no differentiation at all – there is the concept, RAIN, which is not specified as
either Participant or Process. Greater differentiation distinguishes Participants from
the Process (e.g. “It rained a November drizzle”); smaller scale elements may also be
differentiated, such as agency and affectedness.
The result of this complex development is that the Process itself is complex; and
because the main development of the interpersonal metafunction has been recent (in
the last 500 years – Halliday 2000), the structure is neither fully paradigmatic nor fully
marked morphosyntactically. Nevertheless, six types may usefully be distinguished,
following Halliday (2014); they form the substance of this section. To emphasise a
point made just above: although the Predicator which realises the Process is usually
referred to as “the verb” or “the verb phrase”, the Process is not characterised by the
“verbal” quality of denoting actions or events or states, or of being unstable in time,
or “unfolding through time” as even Halliday asserts (2014, §5.1.1). That might have
been true of verbs in Old English, but is not true of Predicators in present-day Eng-
lish: we have copulas and other Processes denoting timeless relations.
As given by Halliday (2014, §5.1.2), the six types of Process are material, behav-
ioural, mental, verbal, relational, and existential. Their status as strictly linguistic
classes is open to question. Halliday’s presentation is more cognitive than linguistic:
the types cover the whole of “semiotic space” (2014, §5.1.2), which is the world of ex-
perience; but there are some correlations with morphosyntactic forms, and the types
are real linguistically as a paradigm of semantic alternatives for construing experience.
Only three of the six will be discussed in most of the book, partly to maintain clarity
and brevity, and partly because they overlap in certain respects. (The overlap occurs
between neighbouring types in the list just given.) That leaves some gaps between the
types discussed in this book, so that there are many sentences in the language that do
not fall into one of the classes discussed here.
Each type will be described in terms of the domain it presents and the way the
domain is modelled, and in terms of the roles entailed and how they are related. That
characterisation stands in place of definition, which is not possible, given the incom-
plete differentiation described earlier in this section.
7.2.2 Material-Process structure
The semantic domain of material Processes (Halliday 2014, §5.2.8) is that of the world
outside the mind; these Processes include ‘run’, ‘break’, ‘replace’, ‘melt’, ‘emerge’, ripen’,
‘sell’ and ‘build’, for example. The domain is modelled in terms of doing and happen-
ing: “events” or “actions”, which have an input of energy, are situated in time, and
produce a quantum of change; and it is modelled with phases, which usually include
an initial phase, sometimes a medial phase, and regularly a final phase entailing an
outcome for the Process.
Material Processes entail an Actor Participant, and sometimes have an Undergoer

Participant (Halliday 2014; McGregor 1997; Van Valin 2013, p. 78). The relationship
between Participants and Process is that of transitivity, in which both Participants are
involved in the Process, the Undergoer being affected by change. The Process may
represent physical events with impact, physical change, “force dynamics”, and phys-
ical effects; but it may represent semiabstract events, with causation, and somewhat
abstract effects e.g. ‘Prices slid’; either way, the Process denoted is strongly direction-
al, flowing one way – in time, and often in causation. English extends this semantic
structure still further, representing some quite abstract states of affairs in the same
way, as in “He exceeded his quota”: there is still change, and an outcome; compare
Halliday (§5.2.5).
The transitivity being discussed here is of course semantic, not syntactic; and, al-
though it is based on our experience of the world, it is linguistic, not simply cognitive.
For instance, consider example (1).
(1) “That’s what life-saving is all about. They [i.e. lifesavers] just drop and go.”
(New Zealand Herald, January 23rd 2014, page A3)
The meaning is that lifesavers drop whatever they are doing; the real-world event is
transitive, but it has been reconstrued into intransitivity. Similarly, the construal of
participants generally has an objective basis, but is often subjective: we can say, “He
hurt himself in the foot” or “He hurt his foot”, with alternative Participants for the
same reality. Similarly, the linguistic bond for kill conformed to the transitive cog-
nitive bond for centuries, through 12 SOED senses; its intransitive use began with a
16th century reconstrual, in sense <13> (e.g. “Thou shalt not kill”).
The morphosyntactic correlates of the material Process are that the unmarked
present tense is in the participial form, not the simple form ( e.g. “They are eating
now” not “They eat now”), and that the word expressing the Process is accented, as
content words are (Halliday 2014, §5.7.4, Table 5 (45).)
7.2.3 Mental-Process structures
For mental Processes (Halliday 2014, §5.3), the semantic domain is that of the inner
world of cognition, emotion and perception – Events of feeling, wanting, thinking and
seeing. The Events are modelled as happenings, with no input of energy, and without
phases; they may be conceived with passage of time, but also as instantaneous, or with
the passing of time not relevant as in “He felt sore.” Examples include ‘remember’,
‘think’, ‘worry’, ‘see’, ‘know’, ‘want’ and ‘regret’; and the senses of some words that have
other uses, as in “It strikes me that…”, and “Its significance escaped him”.
There is a Senser Participant, a conscious being which thinks or feels, and some-
times a Phenomenon as a second Participant – a thing or fact which is thought or felt.
The relationship between Participants and Process is commonly the transitive one of
“going over”: the Phenomenon is produced rather than affected. The directionality
can often be reversed. In “Mary liked the gift”, it flows from the Senser to the Phe-
nomenon; but in “The gift pleased Mary,” it flows the other way, from Phenomenon
to Senser; the directionality is subject to the speaker’s construal. Processes of thinking
and speaking commonly represent or denote the same piece of reality in both Process
and Phenomenon. For example, in “The burglar fled”, the Process is perceptually dis-
tinct from the Phenomenon, but in “Searchers now presume that the missing man is
dead,” there is no such distinction: the Process ‘presume’ and the Phenomenon ‘that
the the missing man is dead’ both relate in the real world to the same thing – a pre-
sumption – the Phenomenon is the Process. The Phenomenon “projects” the thinking
or speaking into an abstract domain, of fact or thought. These two types of relation-
ship provide semantic constructions not available in the other Process types.
The morphosyntactic correlates of the mental Process are as follows. For most
Processes, the pro-verb do is available for questions and ellipsis, but it is not available
for mental Processes; it is not idiomatic to say “What John did was know the answer”,
or to probe for “She liked the gift” with the question, “What did Mary do with/to the
gift?”. The unmarked present tense is in the simple form, and the Predicator is accent-
ed. (Halliday 2014, §5.7.4, Table 5 (45).)
7.2.4 Relational-Process structures
Relational Processes (Halliday 2014, §5.4) represent the semantic domain of abstrac-
tion – abstraction from either the inner or the outer world. There are three main types.
The first is that of “intensive” relations, with Carrier and Attribute Participants (as syn-
tactic Subject and Complement respectively); the Attribute either describes the Carri-
er (“It is heavy” or “It is the heavy one”), or identifies it (“In the film, Marlon Brando
played Stanley”). Other examples of intensive relation Processes include “H-o-t spells
hot”, and “This exam represents your last chance”. The second type is that of “pos-
sessive” relations, with Possessor and Possessed Participants; examples include ‘has’,
‘owns’ and ‘belong to’. The third type is that of “circumstantial” relations, again with
Carrier and Attribute participants, in which the Attribute denotes a Circumstance:
‘predate’, ‘outnumber’, ‘contravene’, ‘concern’. The relation may be partly realised in a
Complement or Adjunct; for example, “The story is about a tiger cub” (compare “The
story concerns a tiger cub”), and “Where’s the beef?” (relation of location).
The domain is modelled as being or having. The Process is sometimes placed in
time, especially in the circumstantial type; but there is no passage of time, no change,
no action or other “event”. The words for these Processes are inflected for tense, but
that does not signify passage of time; it makes the clause finite, signalling predication,
and sets current or past relevance in information structure. (Halliday says that there
is always “a process unfolding through time” (2014, §5.1.2), succumbing to the old
fallacy of equating Predicators with verbs and with passage of time.)
These Processes contrast with material and mental Processes in another way. In
those Processes, the element of agency gives one Participant a natural priority, but in
a relational Process neither term has priority: relational Processes can be rendered
abstractly as “X has the property Y”, and equally as “Y is a property of X.”
The morphosyntactic correlates are that the unmarked form of the present tense
is the simple one, and the Predicator is unaccented, as grammatical words are. Com-
pare unaccented “He has a new car” (relational Process) and accented “He bought a
new car” (material Process).
7.2.5 Discussion: Figure structures
7.2.5.1 Reconstrual of Processes

The Process types often provide alternative construals of the same experience, allow-
ing the speaker to express different conceptualisations of the situation, or to choose
among alternative discourse structures. A clear example is provided by Halliday
(2014, §5.1.2). “My head is hurting me” is formulated as a material Process. “I feel
a pain in the head” construes the same experience as a mental Process. “My head
is painful” construes it as an intensive relational Process; “I have a headache” con-
strues it as a possessive relational Process; the unidiomatic expression, “A pain is in
my head,” would express it as a circumstantial relational Process. The experience can
also be construed in two of the types which I have not discussed: “My head hurts” has
a behavioural Process, and “There is a pain in my head” has an existential Process.
Relational Processes can be reconstrued to a different type of relation. For ex-
ample, “The net weight is 10 kg” (intensive relation) can be construed as “It has a net
weight of 10 kg” (possessive relation), or as “The net weight is over the 5 kg limit”
(circumstantial relation). The degree of variation in construal illustrated above with
pain is unusual. Often, the nature of our experience constrains the way we can con-
strue it – “It fell” seems to offer no possibility of construal as a Mental or Relational
Process, for example.
Control of the Process type is often used to manage information structure,
through manipulation of the roles assigned to the people and things referred to, al-
lowing them to be in focus, or to act as Topic. For example, “He exceeded his quota”
uses a material Process; but in “His output exceeded his quota”, the state of affairs has
been reconstrued as a relationship, not an event, with Carrier and Attribute, not Actor
and Undergoer. The reconstrual provides a different Topic, and suits discussion rather
than narrative.
7.2.5.2 The Process types as categories

As noted in §7.2.1 above, the Process types are marked morphosyntactically to some
extent, through their default verb forms, and their phonological accentuation. They
have other characteristics, such as whether they take Property groups as Participants,
as in “You are very lucky.” Those features are markers rather than signs; the distinc-
tions do not create categories (that is, classes with strict definitions), although there is
some morphosyntactic distinction; and, morphosyntactically, the classes are covert.
Semantically, they are real, since they make alternative forms of expression, giving
different conceptualisations of content, and providing realisations of different in-
formation structures. I believe that they are conceptual categories which have been
partly grammaticised, and that they are not (yet) fully distinct as semantic linguistic
categories.
7.2.6 Conclusion: Summary of figure structures
We have examined three types of semantic structure, named for their cognitive se-
mantic domain as material, mental and relational, but having linguistic semantic re-
ality. Material-Process figures have Actors and Undergoers as Participants, and are
semantically as well as syntactically transitive, since the Undergoer is affected by the
Process; Participants and Process are distinct perceptually. Mental-Process figures
have Senser and Phenomenon as Participants; they are transitive syntactically; but
they are only weakly transitive semantically, since the Phenomenon is involved in
the Process but not affected by it, so that a passive form is possible but unidiomatic
to varying degrees; the Phenomenon is not fully distinct from the Process conceptu-
ally. Relational-Process figures typically have Carrier and Attribute as Participants;
the conceptual relations among Participants and Process varying according to the
subtype of figure; they are transitive syntactically, but not semantically; the Process
does not entail any happening, or change, or the passage of time, denoting an abstract
and timeless relation.
That verbal summary can be represented in tables. Table 1 summarises the char-
acteristics of the three types of process.
Table 2 expands what the middle column of Table 1 states, on how the domain is
modelled.
Table 3 picks out from the previous tables the defining qualities of the Process
types.
Table 1. Characteristics of Process types

Type of Domain How the Roles Semantic
Process domain is entailed relationship
modelled between roles
Material World outside the mind Doing or Actor, Strong
happening Undergoer transitivity
Mental Inner world of the mind Happening Senser, Weak
Phenomenon transitivity
Relational Abstraction from both worlds Being or Carrier, No transitivity
having Attribute (etc.)
Table 2. How the Process types model their domains

Type of Process Input of Time Produces Phases?
energy? change?
Located in Passage of
time? time?
Material Yes Yes Yes Yes Yes
Mental No Yes / No Yes / No Yes Yes
Relational No Yes / No No No No
Table 3. Defining qualities of the Process types

Type of Process Input of energy? Produces change? Roles
Material Yes Yes Actor, Undergoer
Mental No Yes Senser, Phenomenon
Relational No No Carrier, Attribute (etc.)
7.3 Processes
7.3.1 Introduction: Processes
We have seen, in §7.1.3 and §7.2.1 above, that Processes are “verb groups” in their role
in the figure, and are accordingly not to be identified with “verbs” as commonly un-
derstood. They are realised as syntactic Predicators, with the ideational functions of
realising the semantic relation between the Subject and any Complement or Adjunct,
and the interpersonal function of establishing predication and assertion.
7.3.2 Internal structure of Processes
We saw in Chapter 4 that meanings are of various types with various dimensions, and
belong to various classes. Of those elements of structure, the crucial one for Processes
is their meaning type – not content type (such as Event or Entity), but grammatical
meaning, which constructs finiteness, and thereby controls the nature and function
of the figure.
Example (2) lacks finiteness.
(2) “On March 12th 2011, the USS Ronald Reagan entering the Sea of Japan on a
humanitarian mission.” (Based on a news report in New Zealand Herald)
Syntactically, the example makes a non-finite clause. Communicatively, it achieves

nothing, because it makes no statement. It becomes communicatively effective if the
finite auxiliary was is inserted before “entering”; or if “entering” becomes “entered”;
the mere string of words becomes an assertion, enabling the writer to make a state-
ment. (Strictly, we should say that it gains the form necessary for an assertion, which
will be made when the form is uttered.) Semantic finiteness is thus crucial to the fig-
ure; it is realised by syntactic finiteness.
Semantic finiteness results from two operations. (See Dimroth et al. 2003, §1, fol-
lowing work by Klein.) “Anchoring” specifies the time and place in which Events are
embedded; “linking” asserts covertly that the Event is so embedded, in fact. In the ex-
ample above, “March 12th 2011” and “Sea of Japan” provide the anchoring; the finite
verb provides the linking, asserting the relationship between the ship as Participant,
entering as the Process, and the date and place as Circumstances. (The term “ground-
ing” is also used for this process, by other linguists.) The Process thus has significance,
carrying general grammatical meaning – “Treat this utterance as an assertion.”
7.3.3 Syntagmatic structure of Processes
Process is one member of a paradigm, the other members of which are Participant
and Circumstance; for each position or “slot” in the figure, we choose one of those
three. Its place in the syntagmatic structure of the figure is created by the links to Par-
ticipant and Circumstance established by its grammatical meaning. Those links vary
with the Process type. The transitivity of material Processes for example, constitutes
an instruction to the hearer to make the links by representing the action as flowing
from the Subject Participant as Actor, through the Process, to the second Participant
as Undergoer.
The links constitute the “bond” in each constituent in the figure. In Chapter 4
(§4.6.2.2), bonds were introduced as potential or actual links to other semantic units
in a semantic and syntactic relationship. We now note that the mechanism for a bond
is linkage between a certain features of a unit’s internal structure with a corresponding
feature in another element. The links in the figure are elements of a complex concept
such as CAUSATION, and consist of such subconcepts as AGENCY in the Actor Partici-
pant, CHANGE in the Predicator, and EFFECT in the Undergoer.
That can be illustrated diagrammatically from example (3), from a report of how
a gunman on a French train was subdued by three bystanders.
(3) “They held him down, beat him up until he passed out and then tied him up
by hands and feet.” (New Zealand Herald, 24th August 2015, A21)
The first sentence of example (3) is represented in Diagram 1, to show the bond-
ing between groups in a material Process figure, through the causation concept. The
agency subconcept in the subject Participant (denoted by “they”) relates to the change
subconcept in the Predicator (“held”), which in turn relates to affectedness in the sec-
ond Participant (denoted by “him”), and to the effect subconcept in the Circumstance
(“down”). The boxes with solid lines represent the one-word groups in the figure; the
They held him down
CAUSATION:
AFFECTED
AGENCY CHANGE
EFFECT
Participant Process Participant Circumstance
Diagram 1. Bonding between groups in a material Process figure
box with the dotted line represents the overall linkage in the figure, causation; and the
arrows represent the details of the bonding.
Bonding allows relationship to only some types of other element, commonly spec-
ified at figure level by semantic class, and allows only certain relations, e.g. transitivity,
or modification. It exists as potential in the semantic network, and becomes actual
in use, when activated. (“Valence” or “valency” is an established term with a similar
meaning, used particularly in “dependency grammar”, but see also Langacker (1987a,
p. 277), and Croft (2001, p. 274). For a modern treatment, see the various authors in
Herbst and Götz-Votteler (2007). I am preferring “bond”, to avoid the entailments that
the linkages are fixed, are inherent in the word, and depend on cognitive linkages.)
The detailed nature of the bond varies with the sense and the context, as well as
Process type. Material Processes commonly have attributes that build up the bonds.
In example (3), “…down on the floor” would provide an explicit value for the implicit
place attribute, giving substance to the EFFECT bond. Relational Processes, however,
lack such attributes, and accordingly cannot take such Circumstances. I take the con-
cept of attribute-value from Barsalou (1992b, p. 32–35), who uses it in explaining the
links between concepts in cognition. Braisby (1990) uses attribute and value as lin-
guistic concepts; the “cognitive schema” of Rickheit and Sichelschmidt (2007, p. 175)
consists of attributes.
Since concepts are the basis for descriptive meaning, (Chapter 4, §4.4.2), I take
descriptive meanings to be linked in the same way. Barsalou describes “frames” such
as that of a driver using a car as being built from “values” and “attributes”; an action
such as operating has the thing operated as an attribute; the engine and the transmis-
sion are two values of that attribute. The concept has been used in linguistics, in the
term “slot”; for example, Murphy (1988, p. 53) says that in “red apple”, “red modifies
the COLOR slot of apple.” In “He operated the transmission clumsily,” the syntagmatic
relation of operated and transmission relies on similar bonds in the internal structure
of the words. Their nature will be elucidated further in Chapter 8, when group struc-
ture is discussed. Miller and Fellbaum (1991) use the concept similarly.
7.3.4 Semantic class: Process
As a type of semantic unit, Processes constitute a semantic class (using “Process” for
the class as well as the figure constituent). Clearly, they are related to events, one of
the three fundamental semantic classes identified in Chapter 4, but the relationship
is complex.
Events, as presented there, are simple and closely related to perception, involv-
ing no more than a conceptualisation of change of some sort; they are not part of a
structure, and thereby not fully linguistic; examples are SEE and FIND. When some
of their potential links to other words are realised and used in a group, they become
at least partially determinate in aspect and tense; and they have the attributes which
make it modifiable in content by adverbs, and modifiable grammatically by auxilia-
ries; examples include ‘saw’, ‘will see’, and ‘has been found’. As fully linguistic entities,
they can be seen as belonging to a distinct semantic class, which, as previously noted
(Chapter 4, §4.4.6.2), I am calling Events, with an uppercase E. When used in a figure,
they gain the roles of relating the Participants and Circumstances, and of carrying the
predication; they should thus be seen as belonging to a further distinct semantic class,
that of Process.
Not all members of the Process class, however, are based on members of the Event
class. Section 7.2.4, on relational-Process structures, makes it clear that those Process-
es do not involve change or the passage of time; they may be classed as “relational”,
while those involving change and time may be classed as “eventive”. Other examples
include some uses of ‘consist (of)’, ‘contain’, ‘depend’, ‘equal’, ‘taste (of)’. They are per-
haps best understood through the concept of linguistic dominance (Gentner and
Boroditsky 2001), as introduced in Chapter 4 (§4.2.1). Taste began, as a borrowing
from French, with a simple transitive meaning, dominated by cognitive experience of
actions (e.g. <5a> as in “She tasted course after course)”; the relational sense, SOED’s
sense <7> (e.g. “It tasted sour”), evolved in the 16th century, understanding of its
meaning being dominated by its linguistic use. The relational sense of equal is also
highly derivative, though in a different way. The word first had a Property sense, ex-
pressed by an adjective; it entered the language from Latin in Late Middle English; it
gained use as a transitive verb in the 16th century; this relational use is its third verb
sense. Those word histories emphasise again that the words that express relational
Processes are very far from the standard concept of verbs, as words that express per-
ceptible happenings. The semantic weakening of concrete happenings into abstract
structuring elements can be illustrated further in a newspaper statement that a train
collision “caused a fire which took seven hours to extinguish”; “took” constructs a re-
lational Process which is almost empty of content, and does not denote a happening.
Other Process meanings, in many uses at least, involve change and occurrence
in time, but do not involve time passing. Examples include ‘adhere’, ‘believe’, ‘desire’,
‘know’, ‘lie’, ‘seem’, and ‘understand’ – the words are commonly called “stative verbs”.
They have the morphosyntactic characteristic that they do not, in unmarked use, oc-
cur in progressive aspect, so that we do not say “He was understanding it”. They are
also restricted in whether they can take manner adverbs (Katz 2008). They are used
in both mental-Process figures and in relational-Process figures. They constitute a
semantic class, in that they have a semantic characteristic and affect morphosyntactic
expression; but the distinction is different from the structural distinction between
eventive and relational classes, and overlaps it.
These issues have been subject to much misunderstanding. Syntactic Predicators
have effectively been equated with morphosyntactic “verbs”, and with the cognitive or
metaphysical notion of time passing, because of philosophical preconceptions, and
because the equation had some validity in Old English and, I presume, in the Classical
languages. For understanding present-day English, the issues should be kept distinct;
the concept of Process aids that, by being neither syntactic nor cognitive nor histor-
ical. “Verb,” “verbal group” and “VP” generally denote Predicators; in that case, we
should call them “Predicators” – the class having eventive, stative and relational sub-
classes. A further difficulty has been that we take head words as representing nodes in
the network; but in cognitive semantics relational Processes are links, although they
are nodes in the syntactic structure.
Auxiliary verbs, both modal and grammatical, do not represent Processes, since
they have grammatical senses, not content ones.
In summary: we have identified Processes as a semantic class, with eventive, rela-
tional and stative Processes as subclasses; they rank above Events hierarchically, being
more complex.
7.3.5 Conclusion: Summary of Processes
Processes are finite or non-finite, semantically as well as morphologically, which con-

trols their possible function in the utterance. They relate to the other constituents of
the figure through their bond, which has a feature of conceptual meaning with an
intrinsic link to Participants, such as cause-effect or attribute-value. (Their linkage to
Circumstances will be considered in Section 5.)
7.4 Participants
7.4.1 Introduction
A Participant is a meaning which can serve as a term of the relation which constitutes
a figure, and thus serve as syntactic Subject or Complement, and as complement of a
preposition in an Adjunct; it is not defined as an entity which takes part in the real-
world event. (Real-world “participants” can be expressed in a Circumstance; see §7.5
below). It may be expressed as a noun, a pronoun, or a nominal clause and so on. In
a nominal clause, it may be very abstract, as in “The enemy’s destruction of the city
was complete”. (In marked use with a dummy Subject, there may be no Subject Partic-
ipant, as in “It is raining,” and “There is a man at the door.”)
The term “Participant” denotes the role played, such as Actor or Goal. The role
is usually carried by a meaning of the semantic class, Entity. That may represent a
person or object; it may be reconstrued from some other semantic class, as in “He had
a swim”. But as illustrated in the previous paragraph, it may represent a short-lived
event; and honesty represents what is cognitively a property (HONEST) which is quite
outside time. It is clear, then, that Participants do not always construe permanence
through time, as even Halliday thinks, like most other linguists – “the experience of
lasting through time and being located in (concrete or abstract) ‘space’ ”; thus they
are “relatively stable through time” (Halliday 2014, §5.1.1). (Being a Property is reg-
ular for the attribute Participant in a relational Process, as with “It turned green,” and
“That was pretty stupid.”)
The Participants being considered here are those listed above, in the general sec-
tion on Process types: Actor and Goal (“The keen basketball player saved the mon-
ey”), Senser and Phenomenon (“I just knew that was what I wanted”, Carrier and
Attribute “The $300 rent is more than enough”. However, that list is not presented as
definitive. First, there are the Participants in the Processes not being considered here
at all, such as behavioural Process. Second, it is possible to formulate subtypes of the
material and mental Processes; and the intensive, possessive and circumstantial Pro-
cesses are subtypes of relational Process; so subtypes of Participant can be identified.
The Participant roles have morphosyntactic relevance, since they affect the syn-
tactic status of the groups that express them, as Subject or Complement, and they
affect the case of pronouns similarly. But they are semantic, defined in conjunction
with the Process types, not syntactic or vaguely “grammatical”.
7.4.2 Internal structure
We have seen that Processes are structured internally to be determinate, through be-
ing finite. Participants are also structured to be determinate; that is, the hearer must
be able to identify the referent, by relating the expression to the situation within the
discourse and to the speaker-hearer speech situation. Syntactically, it is specified with
definite or indefinite reference, and may act as Subject or Complement. Semantically,
being determinate enables the head of a group to take on a role in a figure, e.g. as Par-
ticipant or Process. (The nature of Participants’ being determinate will be set out more
fully in Chapter 8; being definite is the main form of it.)
7.4.3 Syntagmatic structure
The bond of a Participant in a material Process is its potential to act or be acted on

transitively. As noted in §7.3.2, the potential consists of the availability of concepts
such as AGENCY in the Actor Participant and EFFECT in the Undergoer. In a relational
clause, the Participant’s bond consists of such links as INSTANCE OF, as in “That dog is
a Rottweiler”, or PROPERTY OF, as in “Grass is green.”
For the hearer, the link is activated through the syntactic status, Subject or Com-
plement. That status carries the grammatical meaning that the content is to be taken
semantically as Participant, which entails linking it into a figure. That status, the posi-
tion of the nominal (preceding or following the Predicator), and the Process type to-
gether signify the Participant’s role in the figure. For example, a Participant preceding
a material-Process Predicator must be taken as Actor (as in ‘He fell in”); a Property
Participant following a relational-Process predicator must be taken as Attribute (as in
“That pie is superb”).
7.4.4 Participant as semantic class
Just as Processes stand above Events in the hierarchy of semantic classes, so do Par-
ticipants stand above Entities. They are more complex than Entities, consisting of an
Entity and its bond, and being determinate. To put that in network terms: in saying
“Participant” rather than “Entity”, we are referring to a larger area of the network of
total meaning, including the links between the Entity and the Event in the Process,
and any Circumstance.
There are several subclasses, namely the two terms for each of the three Process
relationships, as listed in §7.4.1. (We here begin to be affected by the limitation of the
terminology: Participant is at once a class and a role. Perhaps we can say that they are
members of the class of linguistic entities which take certain roles.)
Participants are like Processes also in being derived from various basic semantic
classes. The basic derivation is from semantic Entities which are derived from percep-
tual entities, but they may be derived from nominalised Events, and from Properties.
That is discussed in the following chapter, in §8.2, on Entity groups.
7.4.5 Discussion: Participant roles and “semantic roles”
The issues dealt with here have been discussed in the past with various terms, includ-
ing “functional roles,” “semantic roles,” “thematic relations,” “thematic roles,” “deep
case” and “theta roles” (see Dowty 1991 and Meyer 1995, for example); and they over-
lap the issues of “argument structure” and “argument linking.”
There are many well-known problems with the traditional “semantic roles” (as I
will refer to those traditional concepts as a group). They include the indeterminacy of
the list (Mel’cuk 2012); the difficulty of defining terms (Dowty 1991); the difficulty of
identifying roles in some common types of sentence (Meyer 1995; Schlesinger 1995);
and the occurrence of arguments that have two different roles at once (as in reflexive
constructions; DeLancey 1987, §3). A less familiar difficulty is the obscurity of what it
is that carries the role – a nominal group, a word, a meaning, the thing denoted by the
meaning, or a concept. Finally, traditional accounts leave it obscure whether the roles
are being defined in semantics, syntax or some undefined “grammar”.
I suggest that the fundamental problem is the confusion between cognitive and
linguistic semantics. In the literature, semantic roles are in fact defined in terms of
the real world, or of knowledge of it (Barsalou 2012, §3.4), although their context im-
plies that they are linguistic. For example, Levelt (1989, p. 71) makes them cognitive
explicitly, as “conceptual structures”. Dowty (1991, p. 572) characterises his “Agent
Proto-Role” through functional involvement, sentience, causing an event, movement,
and existence independent of the event. It is the physical being that has those charac-
teristics; no grammatical entity has sentience or causes an event. There is a grammat-
ical correlate for sentience, in the paradigm of pronouns (distinguishing he/she from
it); but there is no grammatical correlate for the “semantic roles” as he defines them.
The confusion has been made worse, I believe, by the acceptance of “prototypes” as
grammatical categories; the concepts have proliferated, because prototype concepts
cannot be invalidated (because what they mean is neither defined strictly nor delim-
ited in any other way). The cognitive, syntactic and semantic issues, which are com-
bined in the prototype, should be kept separate. (The same issue was discussed with
reference to “argument structure” in §5.4.5.)
The account of Participant roles given here avoids the problems, by basing the
distinctions in grammar, not cognition, and, within grammar, by basing them in se-
mantics not syntax. It lets the roles be well-defined; it provides for a finite number of
them; it explains how the same real-world entity can be presented in different roles;
and it seems to be comprehensive enough to provide for all sorts of Participants in all
sorts of sentence.
7.4.6 Conclusion: Summary of Participants
Participants are typically Entities with a role in the figure specifying their relation
to the Process and other Participants. To carry that role, they must be determinate
(specified for definiteness and number). Their role and bonding link are interdepend-
ent with the Processes; Attribute and Carrier occur only with the relational Process,
for example; and a possessive relational Process such as “He owns a Maserati” takes
POSSESS as the link in the Process, and POSSESSOR, and POSSESSED as the link in the
Participants.
7.5 Circumstances
7.5.1 Introduction
A Circumstance is an element of a figure that is presented as other than the Process

and its Participant(s). Note the emphasis on presentation; something that is directly
involved in a happening (a “participant” in the common understanding), may be re-

construed by the speaker and presented in the utterance as a Circumstance instead, as
in “It was broken by Tom”, and “She gave it to Bill”. It may be expressed as a participial
phrase, an “adverb”, or any other “adverbial”. As with Processes and Participant, being
a Circumstance is a role; but those roles are each regularly realised by a particular
semantic class (Event, Entity), whereas Circumstances are not.
Circumstances have two normal positions: at the end of the clause / intonation
group, as in “He entered the room deliberately,” and at the beginning, as in “Deliber-
ately, he entered…”; but they may occur in other positions. That contrasts with Partic-
ipants and Processes, whose position is constrained semantically by transitivity, and
syntactically by mood. Circumstances are regularly realised as Adjuncts, which have
fewer constraints, leaving the Circumstances relatively free to vary in position; that is
because the bond links of Circumstances are under-specified, leaving their relation-
ship free to be interpreted cognitively; see Chapter 5, §5.5.1.1, and Chapter 6, §6.3.3.
(The speaker’s choice of position depends also on the intended information structure,
to be discussed in Chapter 9.)
Distinguishing Circumstances from other elements of the figure can be difficult;
it will be discussed in §7.5.5, after Circumstances have been discussed more fully.
The rest of this section follows the same plan as previous ones, discussing the
internal and syntagmatic structures in turn.
7.5.2 Internal structure of Circumstances
7.5.2.1 Being determinate

Entity classes quite clearly need to be definite and specified for number, and Event
classes need to be finite, if they are to act as groups in a figure; with Properties, the
need is less obvious. They are made definite by degree of comparison, with -er/est or
more/most for comparative or superlative degree. Since most Circumstances are in
the positive degree, and the positive degree has no inflection, they are determinate
without appearing to be so.
7.5.2.2 Constituency
By default, groups are interpreted as constituents of figures, and accordingly hearers
expect figure constituents to be groups. That is so fairly regularly for Processes and
Participants, but not for Circumstances, which are often realised as single words or
clauses. Those units are taken as rankshifted semantically, just as they are syntactically.
The constituency of prepositional phrases, however, differs between syntax and
semantics. We deem the preposition to be a part of the phrase syntactically, because
it must be moved or deleted with the rest of the phrase. But semantically, the prep-
osition is as closely related to the rest of the figure as it is to the nominal group it
precedes. In “Read our blog for interesting news,” for example, “for” denotes purpose,
which is specified by “news”, but it also relates directly to “read”, since reading entails
a purpose, and the “for…” phrase therefore specifies something about the reading. The
prepositional content meaning (if any) links two blocks of content, and its grammat-
ical meaning instructs the reader to relate those two blocks; the preposition is not a
constituent of either block semantically.
Since nominal groups in Circumstances normally work at group level, not at this
figure level, their semantic structure will be considered in the next chapter.
7.5.3 Syntagmatic structure of Circumstances
7.5.3.1 Linguistic semantic structure

Complementary relation in the figure. As noted previously, Circumstances are comple-
mentary to Participant(s) and the Process; the relationship is no more fully specified
than that. That fact is shown by the variety of internal structures illustrated in the
previous section. Any specification carried out is done, utterance by utterance, by the
structure and meaning of the Circumstance as word, phrase or clause.
Attribute-value relation. The mechanism by which the Circumstance is given its com-
plementary relation is that of attribute and value. Consider example (4), which offers
several instances of the relation. Numbers have been added for reference.
(4) “Two thieves have stolen thousands of dollars worth of groceries [1] from
supermarkets and shops across Auckland [2] by filling trolleys and running
out of the store – past other shoppers and staff trying to stop them – [(2)
continued] to their getaway car.”
(New Zealand Herald, January 15th 2014, page A3)
Circumstance [1], “from supermarkets and shops across Auckland,” provides the val-
ue of the place attribute of ‘steal groceries’. Circumstance [2], “by filling trolleys …”
provides the value of the manner attribute of ‘steal groceries’. (Both [1] and [2] have
rankshifted Circumstances within them – “across Auckland” and “out of the store”.)
Those attributes (place and manner) are characteristic of material Processes.
Construal. The manifold resources supplied by English for reconstruing the happen-
ings we experience are illustrated by the alternative expressions of the same happen-
ing given in the report. The headline read, “Trolley bandits hit shelves,” in big bold
type over four columns. The filling of trolleys, which was construed in example (3) as
a Circumstance, is construed in that headline as a modifier of a Participant (‘trolley’
modifies ‘bandits’) – an extraordinary semantic transformation, treating the method
by which the theft was carried out as if it were a personal quality of two bandits. The
subheadline read, “Supermarkets targeted by two thieves who fill their carts and run
off without paying.” ‘Filling trolleys’, which is elsewhere construed as a circumstance
group and as a property concept, is now construed as a whole figure, filling a subor-
dinate clause. Also, “Supermarkets targeted…” has the supermarkets as a Participant;
“from supermarkets” has them as part of a Circumstance.
7.5.3.2 Cognitive semantic structure

As noted in Chapter 5, in §5.5.1 on networks in syntactic structure, Circumstances
often set up direct links with the Participants or Process in the process of our com-
prehending them, although the grammatical system of English does not specify any
such specific links. The issue has had considerable discussion, going back to Jackendoff
(1972) at least, according to Ernst (2012, p. 34). Quirk et al. (1985, §8.25) discuss “I
typed the letter in the kitchen”, as “subject oriented” (I was in the kitchen), and “I found
the letter in the kitchen” as “object oriented” (the letter was in the kitchen).
Those links are not strictly linguistic, but cognitive. That can be seen clearly in the
example just above, of thieves stealing from supermarkets. There, the events are men-
tioned in chronological order (filling trolleys, running out of the store, reaching their
getaway car), but the linguistic structure is expository (explanation of the manner of
the theft), not narrative. Readers perceive the chronological pattern once they have
understood the expository structure, and apply it in a second phase of comprehension
to their knowledge of the world.
The distinction between the semantic and the cognitive semantic structures also
explains the ambiguity in expressions like “He closed the door quickly” and “Quick-
ly, he closed the door.” Syntactically, the underlined words are the same, being Ad-
juncts. Semantically, they are equally alike: they are Circumstances, complementing
‘he closed the door.’ The sentence does not specify whether it was only the swinging of
the door that was quick – the unmarked interpretation – or the whole action of which
that is only a part. As we apply the sentence to a context, we interpret it cognitively;
we are likely to take “Quickly, he closed the door” as applying to the whole action,
because of the initial position of quickly, and its being separated by a comma – those
formal markings lead us to the marked interpretation.
7.5.3.3 Conclusion: Syntagmatic structure of Circumstances

We conclude that the relation of Circumstances to the rest of the figure is much
more straightforward than is allowed in the literature, which describes the situation
through countervailing tendencies. Generalisations become possible when we make
three distinctions: between basic and marked structures; between discourse struc-
tures and semantic ones; and between the semantic structures which the grammar of
English specifies and the conceptual cognitive structures which hearers develop when
they apply the linguistic meaning to world knowledge.
In basic use, then, Circumstances come at the end of the figure; indeed, that position
marks them as Circumstances. Being complementary to the Participants and the Pro-
cess, their grammatical relationship is a general one, a relation to the predication made
by the figure. The hearer will commonly extend that general relation to specific con-
nections among concepts, relating them to wider cognition, in the cognitive-semantic
phase of interpretation. In marked use, Circumstances can be placed elsewhere for a
discourse purpose, most often at the beginning, but sometimes medially.
Circumstances thus operate compositionally, with grammatical meanings guid-
ing the hearer in how to combine the content into a more complex structure.
7.5.4 Semantic class of Circumstances
Perceptual qualities, such as HEAVY, SLOW and SILENT, when combined with a po-
tential link to linguistic function, form sublexical senses of the basic semantic class,
property (Chapter 4, §4.5.2.3). When they are lexicalised as heavy or weighty, slowly,
and silent or very quietly, and used in a group, their potential link is actualised as
their bond, and they constitute Properties (with uppercase P); that bond is commonly
specific to Events, as in weightily and slowly (“He has slowly regained his strength”),
or to Entities, as in weighty and slow (“a slow race”). When those meanings are used
in a figure as Circumstances, their form does not change, and there is no significant
change in their bond or semantic role: “He rose to his feet slowly”; “He sank to the
ground weightily”. That contrasts with Entities and Events used as figure constituents,
which become determinate and gain quite different roles: those changes are marked
by the change in term, with “Entities” becoming “Participants”, and “Events” becom-
ing “Processes”. The change in Properties does not seem to justify a new term, or the
assertion that they change in semantic class.
The semantic class is independent of their syntactic form. “In silence” expresses
a Property, just as “silently” does, and so does the non-finite clause “without making
a sound”. There do not appear to be any subclasses. The commonly recognised “ad-
verbial classes” of time, place, manner and so on are conceptual classes of content,
just as nouns of time, place and manner would be; so they are cognitive classes, not
linguistic ones.
7.5.5 Discussion: Circumstances
7.5.5.1 Distinguishing Circumstances from other elements

Introduction. Distinguishing Circumstances from other elements is often difficult and
perhaps contentious, as with the “on…” phrases in “It’s on the floor,” “He put it on the
floor,” “It’s on Thursday,” and “On the floor is fine.” The following paragraphs outline
the issues.
Circumstances and modifiers of the Process. The distinction between Circumstances

and modifiers of Processes is as follows. Words modifying the Process come immedi-
ately before it by default (just as words modifying a noun do): “He has seldom arrived
on time,” whereas Circumstances default to the beginning or the end of the clause / in-
tonation group. However, the default rule is quite often overridden for the sake of in-
formation structure. That marked usage is sometimes signalled by marked intonation
or punctuation – e.g. as used for parenthesis – but the hearer is sometimes left dealing
with a word or phrase which is a Process modifier linguistically, i.e. by position, but
which must be reconstrued in cognitive interpretation as a Circumstance, because of
semantic clash, as in “They have politically achieved little.” (The clash lies in the fact
that ‘politically’ does not fit in the types of meaning that can premodify the Predicator
(see Quirk et al. 1972, Chapter 8 – ‘politically’ is not a manner of achieving); “politi-
cally” must be read as a Circumstance.
Circumstances and Participants. Words that appear to be “prepositions” introducing a

Circumstance are in fact sometimes parts of a phrasal verb. “Henry VIII slept in this
bed” may have a Circumstance, “in this bed”, with the Process “slept”; but it may have
a Complement, “this bed”, with the Process “slept in”, meaning ‘occupied’ approxi-
mately. The difference in meaning is cognitive, and is not signalled linguistically, in
writing at least. Prepositional phrases at the end of a clause may be Participants acting
as Complement. In “He lives a lonely life, in Edinburgh,” the prepositional phrase is a
Circumstance. However, in “The others live in Aberdeen, but he lives in Edinburgh,”
the prepositional phrase is a Complement, which acts as the attribute Participant in
a circumstantial relational Process; “lives” acts as a copula. (Not all Complements are
nominals, and not all Participants are nominals.)
Circumstances and modifiers of Participants. Prepositional phrases at the end of the

clause may be modifiers of a Participant, not Circumstances, and the intended seman-
tic structure must sometimes be interpreted through information structure. Consider
the two uses of “at the rear” in example (5).
(5) “Each one has a membrane at the rear. … Some of [the liquid] is evaporated
through the membrane at the rear.”
(Economist, November 30th 2013, Insert p. 9)
In the first sentence quoted, “at the rear” must be taken as giving a piece of new
information, and therefore as standing as a unit, at figure level, and therefore as a
Circumstance. In the second sentence, it is a restrictive modifier of “membrane” – a
lower-level unit, and not a Circumstance.
Circumstances and sentence adverbials. The distinction between Circumstances and

“sentence adverbials” such as “Frankly,…” and “In conclusion,…” is that sentence ad-
verbials relate to a whole figure as a unit, not to one group within it or to the predi-
cation, as Circumstances do. They are discussed later, as a part of clause complexes,
in §7.7.
7.5.5.2 Other views of Circumstances

It follows from the section above and from earlier sections that speakers have sub-
stantial freedom to construe relations within the content into different relations in the
figure. A good deal of confusion has been created, in the literature on “adjuncts”, “ad-
verbials” and so on, from failure to recognise that fact; linguists often identify mean-
ing with real-world information. Even Halliday (2014, §5.6.1) makes this mistake,
in saying that Participants are directly involved in the Process, but Circumstances
are indirectly involved in it, being “associated with” the Process, or “attendant on” it.
Halliday illustrates the distinction with the pair of sentences: “I’ll sail across the
mighty ocean [Circumstance]” and “Little drops of water make the mighty ocean
[Participant]” (2014, §5.6.1); but the involvement argument does not work, because
the ocean is involved in the Process of sailing, because it is disturbed by the ship’s
movement; it is less involved in the Process denoted by making, which means ‘consti-
tute’. In fact, we take ‘across the mighty ocean’ as presenting a Circumstance because
that is signalled by the preposition “across”, and take ‘make the mighty ocean’ as pre-
senting a Participant because ‘the mighty ocean’ is a Complement. The presentation is
in turn determined by the speaker’s intention and construal, not by his or her under-
standing of the real world.
7.5.6 Conclusion: Summary of Circumstances
Internally, Circumstances are determinate, as Processes and Participant are, but are
much more varied in their constituency, being word senses or groups or figures. In
their syntagmatic relations, they are linked to something in the rest of the figure as
the value of an attribute, in complementary relation. They are Properties, by semantic
class. They are highly dependent on speaker construal: real-world things, happenings
and properties can all be construed as part of a Circumstance, and their being com-
plementary is itself a construal. That goes with their being very varied in content and
underlying semantic class, and the under-specification of their relation to the rest of
the figure.
7.6 Relations among Participant, Process and Circumstance
We have seen that hearers compose the semantic structure of a figure according to the
general grammatical meaning, “Relate these semantic units as syntactically equal.”
For co-ordination, the units may be words, groups or clauses, so long as they are of
the same semantic class. The coordination assigns them the same semantic function:
both words will be modifiers or both heads, or both groups will be part of the Subject
Participant, and so on.
Relating the constituents of a figure, in complementation, is significantly dif-
ferent. The relation of complementation holds only between groups, in Participant-
Process-Circumstance structure. Each unit must be semantically complete; and an
Entity group must have the necessary degree of determination, an Event group must
be finite, and a participial phrase must have the complete structure of preposition plus
Entity group. Next, the groups must belong to different and complementary semantic
classes: Entity and Events groups are related as Participant and Process. For a seman-
tic unit to be linked as Circumstance, the requirements are looser: the unit may be
related to the Participant, to the Process, or to the whole situation.
The nature of complementation has often struck linguists as obscure, with much
discussion of the distinction between Complements/Objects and Adjuncts, for
example. The reason, I suggest, is the general acceptance of a philosophical approach

to linguistics in which the explanation for grammatical distinctions is sought in the
nature of the world, so that the explanation has been sought in such issues as whether
the relevant entity is affected (Complements or Objects) or not (Adjuncts). A better
approach is through how we perceive the world (rather than how it really is), and how
we render that perception in language, through Kemmer’s concept (1993) of differ-
entiation, for example. As noted in §4.2.1, situations can be differentiated to differing
extents, giving different construals of the same situation. Elements of the situation
that can be differentiated easily, with the remainder unanalysed, default to being Cir-
cumstances; for example, “It was blowing hard here yesterday” has Circumstances,
without Participants having been differentiated. What have been traditionally distin-
guished as Objects and Complements can then be distinguished by whether their
degree of integration with the situation allows ready differentiation. In “He turned
the page”, the second Participant is clearly differentiated from the Process, and is an
Object in traditional grammar; in “Without light, grass turns yellow,” the second “par-
ticipant”, ‘yellow’, is not fully distinct from the process, and is a Complement. (Turns
here means ‘become different in a quality such as colour’, and ‘turns yellow’ can be
lexicalised as “yellows”.) Mel’cuk (2001, Ch. II, §7) deals with differentiation in his
distinction between what is “unitary” and what is “articulated”.
7.7 Structures larger than the figure
Traditionally, clauses, as the expression of figures, are regarded as constituents of sen-

tences – a higher rank – just as groups are constituents of clauses. However, I accept
Halliday’s assertion (2014) that that analysis is inaccurate, since combining clauses
does not make a unit in a higher rank; that is, it does not make a linguistic structure
of a different kind, with a different function, as is true when groups are combined
into a clause, or words into a group. The sentence is a graphological unit, and some-
times phonological one, not a syntactic or semantic one; the corresponding syntactic
and semantic units are “complexes” – clause complexes and figure complexes. Since
figure complexes involve no fundamentally different structure, and can be dealt with
briefly, they are included here, in this chapter devoted primarily to figure structure.
The structure of figure complexes will be well known to the reader, since it matches
the coordination and subordination structures of complex sentences and compound
sentences, which realise figure complexes very straightforwardly; so very little will be
said about them here.
The structure of figure complexes is signalled by conjunctions and pronouns.
Conjunctions like and, or and but signal that figures must be coordinated; relative
pronouns and conjunctions like because, if and when signal that hearers must subor-
dinate them, in composing the structure. Complementation, the other comparably
general grammatical meaning, does not operate at this level, just as it does not operate
at group level. Like prepositions, the conjunctions and pronouns are links between
units semantically, not the first part of the unit that follows them.
Introductory expressions such as “sentence adverbials” are often elements of the
figure complex, not of a single figure. Consider example (6).
(6) “A year after that[,] it was in profit and was to appoint its chief executive of
the past decade, Mark Dunkerley.”
(New Zealand Herald, December 27 2014, B18)
The introductory phrase, “a year after that”, modifies both main figures – “it was in
profit” and “was to appoint…” – so it is a figure also, although a group in internal
structure. (It is rankshifted to a higher rank.)
Similarly, many “discourse particles” are figures functionally, though not in struc-
ture. Example (7) illustrates that. (The source does not give the intonation, so it is
printed without punctuation.)
(7) “The really infuriating thing for the police is that right the guy at one stage
waded out into the river to like wash off all the blood and everything…”
(Wellington Corpus of Spoken New Zealand English, T_DPC068)
The particles right and like have no place in the semantic structure of the figures in
which they are placed, but operate in the interpersonal metafunction, modifying the
whole utterance; so they are figures, structurally.
7.8 Discussion: Figures
7.8.1 Co-ordination and linearity
Since syntax is linear, and has fillers in slots, the filling of one slot implies that the
next unit will fill the next slot: after a Participant such as “Alistair Brown…,” acting as
Subject, we expect a Predicator, as if there were an implicit signal, “Go on to the next
syntactic and semantic position”; I will symbolise the signal with an arrow. Thus we
read “Alistair Brown…” as “Alistair Brown→…”; and we read a full sentence as “Alistair
Brown → has cut → 150 jobs,” with Actor → Process → Undergoer. That linear structure
can be represented visually, as in Diagram 2.
But in “Alistair Brown, the CEO of the company, has cut 150 jobs,” the group “the
CEO of the company” fills the same position as “Alistair Brown” (syntactic Subject,
Agent Process Undergoer

Alistair Brown has cut 150 jobs
Diagram 2. Unmarked structure – linear

Agent Process Undergoer

Alistair Brown
CEO of the company has cut 150 jobs
Diagram 3. Apposition as marked structure – non-linear
semantic Actor); we do not go forward to the Process, but go back to the position
which “Alistair Brown” occupies. That is represented in Diagram 3.
The structure “doubles back”, with “the CEO of the company” filling the same slot
as “Alistair Brown” does; it has duplicate Actors / Subjects.
The structure is a marked one (breaching the rule that linear order represents dif-
ferent semantic and syntactic positions); so it is marked by an explicit signal – a pair
of commas in this appositional instance, or a pair of rising tones if the sentence were
spoken, or a co-ordinating grammatical item such as and, for co-ordinated figures.
7.8.2 Constructions
A construction is defined by Goldberg (1999, p. 199) as “a pairing of form and func-

tion such that some aspect of the form or some aspect of the function is not strictly
predictable from the C[onstruction]’s component parts.” (“Function” includes what
I call “meaning”.) That unusually specific definition – more specific than those cited
in Chapter 6, §6.4.5 – points to the weakness in the concept. The construction has
by definition some meaning that does not derive from any part of it; but something
signified must have a signifier; and the identity of the signifiers is not clear elsewhere
in Goldberg’s text. In her table of constructions (1999, p. 199), the first construction,
“intransitive motion”, has a Predicator but no Complement; the second, “transitive”,
has a single Complement; the third “resultative”, has both Entity and Property Com-
plements. Goldberg is right in asserting that some aspect of the meaning is not strictly
predictable if we treat the “component parts” as self-contained or isolated units. But
if we treat meaning as made up of relationships (as well as the units), which are spec-
ified by the grammatical meanings and the bonds which are part of the units’ internal
structure, then the meaning of the whole is highly predictable. (Idioms and other
phenomena make some exceptions, course. The bonds for figures have been discussed
in this chapter; those for groups and words will be discussed in the next chapter.)
Goldberg and other construction grammarians have not understood the nature and
function of grammatical meaning, or the role of syntactic forms in carrying it.
Adherents of construction grammar may well reply, of course, that my account
of grammatical meanings and so on is tantamount to describing constructions. That
is true for “constructions” in its very old general sense. If we understand semantics
correctly, we do not need any new theory here, or any special use of “construction”.
The explanation through grammatical meaning applies to neologistic construc-

tions, as well. For example, Thompson and Hopper (2001, p. 48) give the attested
utterance, “We can’t sample you” (meaning ‘We cannot give you medication samples
as treatment’). The meaning flows from the meaning of sample as ‘medication sam-
ple’ (given by the context), converted from Participant meaning to Process meaning
by being a Predicator, and given the transitive meaning of affecting another Partici-
pant by having a Complement. Each syntactic element signals the hearer how to re-
late the content meanings; establishing new relations for a word follows the same
process as establishing conventional ones. In Construction Grammar, such neolo-
gisms would be incomprehensible, since the unpredictable meaning has not yet been
conventionalised.
There is an important qualification to be made, however. The meaning of con-
structions that is thus predictable is the linguistic meaning; some utterances require
the second phase of interpretation, producing a cognitive meaning, which may be
considerably different or even (by irony) contradictory. “We can’t sample you” relied
on the hearer’s knowledge of medical practice. Goldberg’s example, “She kissed him
unconscious”, relies on our ability to extend our knowledge of passionate kissing; her
transitive example, “Pat cubed the meat,” relies on knowledge of cookery.
7.8.3 Ergativity
Several writers have discussed “ergativity” in English, contrasting it with “transitivity”

as a property of figures or clauses. See, for example, Lemmens (1998), Peeters (2000),
and Halliday (2014). Ergative clauses are exemplified by “Cloud formed over the hills,”
and “Waves undulated across the lake.” Transitive clauses are exemplified by “He ate
three eggs for breakfast” – and by “He ate hungrily,” since transitivity here is a se-
mantic matter, not the syntactic one of the presence of a Complement. The difference
lies in the conceptualisation of what happens. In the “transitive”, the event is con-
ceptualised as doing, as someone carrying out an action. The actor is essential to the
conceptualisation, but anything he acts with or to or on is not essential; causation fits
naturally here, and the actor is frequently a causer – an Agent, as a knid of Actor. In
the “ergative”, the event is conceptualised as a happening, as an event being actualised
in a medium – the cloud or the lake water in the examples above; the medium is the
only essential participant. Causation does not fit easily with this conceptualisation.
It can be represented, through an outside participant, a trigger rather than a cause
(“Cloud formed over the hills, set off by the drop in temperature”); but characteristi-
cally the event is represented as self-engendering.
This issue lies behind much discussion of “the middle”, the “medio-passive”, and
so on, as in “His book is selling well” and “This bread cuts easily.” Those sentences
portray the event as self-engendered, there being no cause and no agent. They fit the
definition of semantic ergativity. In the examples of transitive ‘eating’ and ergative
‘forming’ and ‘undulating’, it is hard to conceptualise the event the other way; but in
many real-world situations, we have a choice. That is quite clear with selling and cut-
ting, just above. Again, we can say, “The bull chased him” (transitive), or we can say
“He fled (from the bull)” (ergative), where his running is construed as an occurrence
rather than something caused.
We can now see that the distinction being made is one of how we interpret the
events we perceive, not one of how words carry meaning; it is cognitive rather than
linguistic. Fled has the same meaning in the linguistic sense, whether we say “He
fled from the bull” ergatively, or “He hesitated, then fled” transitively (the action now
being deliberate, and caused). The point is clear also in the account given by Halliday
(2014, §5.7.2), where it is asserted that cited pairs of sentences present alternative
ergative and transitive “interpretations”. The examples include “The boat sailed” and
“Mary sailed the boat,” and “The cloth tore” and “The nail tore the cloth.” The differ-
ence lies in our taking the cloth, for example, as the medium in which tearing hap-
pened, triggered by the nail (ergative), and taking it as an inert object affected by the
tearing (transitive). There is nothing in the form of the sentences which corresponds
to the “ergative” / “transitive” distinction, thereby expressing it. Tellingly, Halliday
refers to the alternatives as “interpretations”, not as meanings or statements. I con-
clude that the issue is one of how we understand the world, not of how we present our
understanding in language.
I am aware of one morphosyntactic form, however, which appears to be a true
expression of ergativity in English. The “get” passive uses the implicit meaning that
the event is self-engendered to suggest that the Subject is responsible for the Event,
at least in part. For example, “He was arrested by the police” assigns all responsibility
to the police; but “He got arrested by the police,” especially in the context of previous
wild behaviour, assigns some responsibility to the man himself. (“He got himself ar-
rested…” is transitive.) The use is indirect and not clear-cut, but I believe that it does
entail linguistic ergativity; however, it appears to be an isolated instance, not part of a
general pattern in the grammar of English.
The relation of this ergativity to the syntactic and morphological ergativity studied
by Dixon (1994), for example, is instructive. In “His book is selling well” (Halliday’s
“ergative”), the medium of the event, i.e. the book, is expressed as the Subject of the
clause; in the corresponding “transitive”, “The bookshops are selling his book quickly/
frequently,” it is expressed as the Complement or “Object”. The “Object” of the transi-
tive corresponds to the Subject of the ergative: that is exactly characteristic of syntactic
ergativity, which is reflected in ergative morphology. I suggest, then, that ergativity
as usually discussed has its basis in semantics (in the loose sense of understanding as
represented in language). The view of Du Bois (1987) that it is based in discourse is
inexact; transitive and ergative semantics have similar outcomes in morphosyntax and
discourse structure.
7.8.4 Compositionality in figures
Any account of the semantic structure of English must show the extent to which it is
compositional. We considered its application to sense structure in Chapter 5 (§5.2.5),
consider it here, and will do so again later. A variant of the definition implicit in §7.8.1
(which gives some discussion of this topic) is that a complex expression is fully com-
positional if its meaning is completely predictable by general rules from the meanings
of its constituents (Cruse 2011, p. 65). The general rules are taken to be linguistic.
For example: in the context of the USA, “The President went home” means that the
President went to the White House; but that element of the meaning is not composi-
tional, since it does not arise from the words and the grammar, but depends on world
knowledge. On the understanding of semantics followed here, the application of the
general rules consists of following the procedures specified by grammatical meanings.
The meaning of figures, then, is fairly fully compositional. We have seen that the
meaning of figures is specified not only by the words individually, but by the Pro-
cess type, by the number, position and semantic class of the Participants, and by the
position of any Circumstance. Thus, the difference between “He slowly opened the
door” and “He opened the door slowly” is specified by the position of slowly: “…
slowly opened” has slowly as premodifier of the Predicator, but “opened… slowly” has
it as a Circumstance. The difference between “His story convinced no one” and “His
story was unconvincing” is specified by the difference between a material Process
(convinced) combined with a Participant (no one), and a relational one (was) com-
bined with a Property word (unconvincing).
On the other hand, the meaning of Circumstances is often not fully composi-
tional. “Sometimes I’d see someone I knew on the bus and we’d wave” (COCA) can
be interpreted to represent one or other or both people being on the bus, according
to what we know of the situation (e.g. that one of them was on the footpath). The
interpretation depends on information that is not in the utterance itself. Further, the
overall relation which a figure carries is not compositional; the transitive or possessive
relation, for example, is carried by the figure as a gestalt, and cannot be built up by
combining the Predicator and its Participants.
The grammatical meanings that direct the composition are carried by various
sorts of sign. Some are overt in lexis, as with copulas (§7.2.1), or in morphology, as
with auxiliaries and inflections (§7.2.4, §7.3.3). Some are covert, in syntactic struc-
ture, as with clauses signifying figures (§7.1.2), or in still more abstract status, as with
having the status of Process (§7.3.1) or of Subject or Complement (§7.4.3). Phonolog-
ical signs have not been illustrated, but are frequent, as with falling tone demarcating
clauses and therefore figures.
7.9 Conclusion: Figures
Summary
This chapter has described figures as semantic structures built from groups by the
operation of grammatical meanings. It has concentrated on three structures, those of
material Process, mental Process and relational Process, passing over the three others,
for brevity, and because the characteristics of those others overlap those considered
here. Although they have a developmental basis in cognition, those types of figure
structure are linguistic, in having some correlation with morphosyntactic features,
but chiefly in having semantic reality as alternative expressions of the same underly-
ing meaning. Figures consist of Participants, Processes and Circumstances, presented
as complementary constituents. Figure structure is closely parallel to syntactic struc-
ture – Participant + Process + Circumstance parallels Subject + Predicator + Adjunct
for example – but is distinct, as the structure which the syntax realises.
Figures are the highest distinct level of the semantic hierarchy, carrying the
most general grammatical meaning, and directing hearers’ response as acceptance of
statement, and so on. The semantic structure embodied in the figures has often been
equated with propositional structure regarded as a logical structure, or with argument
structure regarded as inherent in all cognitive understanding of events in the world.
In this chapter, those concerns have been put aside as being outside linguistics; the
structure of figures has been presented as the structure we construe for the meaning
we wish to utter.
Looking forward. In discussing figures, this chapter has discussed only part of the
semantic hierarchy; the remainder – the group and sense levels – will be discussed in
the next chapter.
Chapter 8
Hierarchies (2)
Groups and senses
8.1 Introduction
This chapter continues the analysis of hierarchic semantic structure in English, which
was begun in Chapter 7 – structure contrasting with the network discussed in Chap-
ter 5, and built by the processes outlined in Chapter 6, from the elements outlined in
Chapter 4. Chapter 7 dealt with the top level, that of figures; this chapter deals with
the lower levels, those of groups and senses. It will continue the use of the contrast
between a sense’s internal structure and its syntagmatic structure. The syntagmat-
ic structure will perhaps increasingly seem to intrude on syntactic structure; that is
natural because syntactic structure realises the syntagmatic structure of meaning; the
two are distinct, because we are dealing with senses and structures of senses, where
syntax is a matter of words, as morphological forms, and their arrangement.
Two apparent exclusions should be noted: “argument structure” is not discussed,
although traditional linguistics generally includes it as part of the structure of “verbs”
i.e. Event groups, since the issues have been dealt with in §5.4.5, and in §7.4.5; rank-
shifting or “embedding” is not discussed, since we are concerned with structure, rath-
er than function, and the structure of the rankshifted unit and the larger one is not
affected. Since the illustrations are mostly single phrases, extracted from the full utter-
ances, they will generally be set within the text. Again because they are short, they will
usually be given without context or stated source; nevertheless, they are all attested,
unless otherwise stated.
The chapter deals first with the types of group, in turn: §8.2 Entity groups, §8.3
Event groups, and §8.4 Property groups. Then, in §8.5, it deals with the senses of
words and morphemes (which belong on a lower rank, but do not seem to need a
chapter of their own.) Discussion and conclusions follow (§8.6, §8.7).
8.1.2 Introduction to groups
The groups being dealt with in this chapter are semantic groups; I am not using dif-
ferent words for the syntactic and semantic units, as I used “figure” and “clause” for
the corresponding distinction in Chapter 7. They parallel syntactic groups closely, in
ranging in size from one word/sense to many, and in allowing rankshifted units with-
in them. As explained previously, “group” is close in meaning to “phrase”, but in this
book “phrase” is used only for “prepositional phrases”, which consist of a preposition
linked to an Entity group; it seems better to avoid a new term.
As explained in Chapter 7, groups as figure constituents belong to the semantic
classes of Participant, Process and Circumstance. The semantic classes of a group’s
constituents are distinguishable from those three classes, but are closely related to
them: in “Scottish nationalists are preparing the ground for a new referendum on
independence” the group “Scottish nationalists” is a Participant by semantic class; its
head, ‘nationalists’, is an Entity by semantic class, being distinct in lacking the Partici-
pant’s relation to the Process ‘are preparing’. Similarly, the head of that Process group,
‘preparing’, is an Event, lacking the Process’s relation to ‘Scottish nationalists’ and ‘the
ground for…’.
The groups being discussed are defined by structure, not function. An Entity
group, for example, is one that has an Entity as its head, not one that functions as
Entities usually do, e.g. as Subject; it is thus not the same as a “nominal group ”or
“noun phrase”. That general distinction will become important when applied to Event
and Property groups in later sections. The three types of group just referred to make
up the range of semantic groups, corresponding to the three basic semantic classes
outlined in Chapter 4, §4.6. The groups usually called “adjectival phrases” and “adver-
bial phrases” are here prepositional phrases consisting of a preposition and an Entity
group (e.g. “… a new referendum on independence”) or Property groups (e.g. “the
supposedly empty cities” and “…generate power much more cheaply”).
8.2 Entity groups
8.2.1 Introduction
8.2.1.1 Zones
In line with the substantial degree of nominalisation in English, Entity groups are
more highly developed than other groups, having a more complex structure. As noted
by Quirk et al. (1985), the semantic and syntactic structure of Entity groups should be
understood as consisting of zones. Zones are like slots which take fillers e.g. Subject
and Predicator slots; but they differ in that speakers may leave a zone empty, or put
two fillers in it. Entity groups have two general zones, in addition to the head slot:
(1) a premodifying zone, which usually has a determiner and may have premodifiers,
but need not have any element occupying it; and (2) a postmodifying zone, which
may have one or more groups, but which need not have any. Those general zones are
shown in Table 1, with invented examples.
The premodifying zone (determiners and premodifiers) consists of more specific
zones: a determiner zone, and four zones of premodification, described by Quirk et al.
(1985), as shown in Table 2 (based on Quirk et al. 1985, p. 340).
Chapter 8. Hierarchies (2): Groups and senses 179
Table 1. General zones of premodification

Premodifying zone Head zone Post modifying zone
Tigers
Six huge tigers
Tigers bred in captivity
Table 2. Specific zones of premodification, after Quirk et al.

Determiners Premodifiers Head
Zone I: Zone II: Zone III: Zone IV:
precentral central postcentral prehead
our numerous splendid African tourist attractions
all this costly social security
a certain church tower
these crumbling[,] grey Gothic church towers
some intricate old[,] interlocking Chinese designs
all the small carved Chinese jade idols
both the major Danish political parties
I will use descriptive terms instead of numbers to name the zones, as follows.
Zone IV senses will be “Classifiers”, following Halliday (2014), and indicating one
common function of those senses. Zone III senses will be “Descriptors”: they are
factually descriptive senses. Zone II senses will be “Epithets”, which are more highly
descriptive than Descriptors. Zone I senses will be “Reinforcers”, like mere and sheer,
since they reinforce the sense of the head; (Quirk’s examples do not represent them
well). The important characteristics of zones are that: they may be empty; they may
contain several senses; the senses in each zone have a characteristic semantic struc-
ture; the order of the zones is grammatically fixed, but the order of senses within a
zone is not.
To be “a determiner” is to have a certain function, not to belong to a certain class;
similarly, to be “Classifier” or “Descriptor” is to have a certain function, not to belong
to a certain semantic class. The point will become clearer in the following sections.
8.2.1.2 Definitions
An Entity group is a group with an Entity sense as its head. As noted previously, the
Entities that head these groups are developed from the basic class, entities, the sub-
lexical forms developed from what we perceive as happenings. (See Chapter 4, §4.6.2.)
The Entities extend the sublexical forms further out into the mental network, includ-
ing potential links to lexical items and syntactic structures – their bonds; they are thus
truly linguistic, where the sublexical forms are sublinguistic; they function at group
level, most often as heads. In becoming linguistic, they are reconstrued for linguistic
functions, especially for being modified. In Chapter 7, we referred to Participants,

which are slightly larger areas of the semantic network, including Entities and their
links to wider functions – at figure level, that is – such as those of Actor/Subject and
Undergoer/Complement in relation to Processes. A consequence of the development
from sublexical entities to linguistic Entities is that they often denote “things” that are
not real, such as ‘existence’, ‘absence’ and ‘hole’, or “things” that are based on sublexical
events (by nominalisation), such as ‘destruction’.
8.2.1.3 Coverage
This section will not discuss premodifiers that consist of groups (submodified words,
or two or more co-ordinated words – although they provide some picturesque expres-
sions, as underlined in example (1).
(1) “[The former prime minister’s famous one-liners] appeared to be impromptu

but were, in fact, assiduously crafted by his cigar-smoking / three-piece-suit
attired / viciously tongued speechwriter, Ken Hancock, may he rest in peace.”
(Punctuated as in the original; Michelle Boag,
New Zealand Herald, September 1, 2011, p. A13)
This section deals with standard groups – those that are structured as modifier +
head, and that serve on the group rank, that is, as constituents of a figure. Rankshifted
groups are not dealt with separately, since they have the structure set out here, and
their shift in rank does not concern us.
8.2.2 Classifiers
As shown by the examples in Table 1, typical Classifiers are “nouns”, although some
are “adjectives”; but both identify Entities: African means ‘of/in Africa’. The table also
illustrates groups with more than one Classifier; such series of Classifiers have a struc-
ture of their own, constituting constructions. (By “construction”, I mean an arrange-
ment of meanings; it is not a technical term, as it is in Construction Grammar.)
Section 2.2.2 explains the commonest construction; the following section will
use that as a basis to explain the other constructions briefly. For more detail on all the
constructions, see Feist (2012a, Chapter 5).
8.2.2.2 Entity-head construction, with qualia

Introduction. The commonest Classifier construction has an Entity as its head – and
will be called the “Entity-head construction”. It is illustrated in example (2).
(2) “Detz .44 magnum [alloy] bolt-action hunting rifle” (Square brackets indicate
words that have been added to an attested example from an advertisement.)
Table 3. Relations between modifiers and head, in the Entity-head construction
Entity modifier Relation Head
1 Detz is the source or origin of
2 .44 magnum is the size of
3 [alloy] is the material of rifle
4 bolt action is the type of
5 hunting is the function of
The example can be paraphrased as follows, with the elements represented directly
in the example underlined: ‘a rifle of Detz [the manufacturer] origin, .44 magnum
calibre or size, made of alloy, of the bolt-action type, and for the purpose of hunting”.
The paraphrase indicates that the Classifiers rely on relationships with the head which
are only implicit. There are five such relations, as in Table 3.
I will call those five implicit relations “qualia” (and “quale” in the singular), fol-
lowing Pustejovsky (1995; 2012, p. 380), and refer to this construction more simply as
“the qualia construction”.
(I understand qualia a little differently from Pustejovsky. As just shown, there are
five qualia not just four; my Origin quale is wider than Pustejovsky’s Agentive quale;
and I see the qualia as relations rather than as constituents of the head, although they
are directly related to attributes in the head. More importantly, perhaps, Pustejovsky
sees nothing in the text that represents them – no linguistic correlates, such as word
order; they are postulates, derived philosophically. Here, the qualia are linguistically
real, being meanings conveyed by the Classifier construction, and observed empiri-
cally. Nevertheless, there are important elements of agreement between Pustejovsky’s
“generative lexicon” and this work.)
The concept is supported in another form by Barsalou (2012, §3.4), who notes
that “some lake trout” needs the concept LOCATION to integrate the word meanings.
Table 4 sums up this construction. For each position, it gives an example, a
statement of the relation which the quale consists of, and the name by which I will
refer to it.
Table 5 gives attested examples of phrases invoking the various qualia, to illustrate
the range of contexts in which they occur.
Table 4. Entity-head construction, with the qualia

Position Classifier 1 Classifier 2 Classifier 3 Classifier 4 Classifier 5 Head
Example Detz .44 magnum [alloy] bolt-action hunting rifle
Relation origin or size or what type function or –
expressed source dimension constitutes it purpose
Name Origin Dimension Constituency Type Function –
Table 5. Examples of the qualia

Modifier: Modifier: Modifier: Modifier: Modifier: Head:
Origin Dimension Constituency Type Function Participant
quale quale quale quale quale
Fisher & Paykel 7.5 kg Excellence washing machine
Smeg 690 L stainless automatic dish washer
254 mm electric mitre saw
30-pound-test single strand stainless wire
Iranian 16th century brass boat shaped vessel
The positions and relations in this construction are grammatical realities, not
matters of knowledge or logic, just as those of Subject, Predicator and Complement
are. That can be seen in several ways. First, it can be seen from the disambiguation
of ambiguous phrases by the presence of other premodifiers. In “brick kiln”, brick
is ambiguous between the kiln’s being made of brick, and having the function of
baking bricks (the Constituency and Function qualia). But in “reinforced-concrete
brick kiln” it is not ambiguous, because the presence of a preceding modifier (“rein-
forced-concrete”) filling the Constituency position makes it clear that brick is in the
Function position. (That position-dependent ambiguity is like that of digits in num-
bers: in “222”, the meaning of each “2” depends on its position; empty subzones are
place-holders, like the zero in “202”.) Second, the reality of the subzones can be seen
from the fact that some of these Classifiers can be repeated, with successive Classifi-
ers invoking different qualia: “a brick brick kiln” and “an English English teacher” (a
teacher of English from England) would both make sense. Third, the order is fixed: we
cannot say “electronic oil-pressure British gauge” or “British oil-pressure electronic
gauge”, and so on. Finally, the range of examples in Table 5 above, and the freedom
with which people now use such phrases, show that this construction is well accepted.
Internal structure. The internal structure of Classifiers is very simple. They have de-
scriptive meaning only; they are typically high on the precision dimension; and,
because they are virtually always used referentially, they invoke only the necessary
meaning, without associative possible meaning elements. Those characteristics show
in the contrast between the Classifier use of an Entity sense and its use as a head. In
“Roman winter fertility festival”, each of the three Classifiers simply designates a ref-
erent, doing no more than enabling us to identify the class of festivals being referred
to; but in the following phrases, the same words invoke qualities that are part of the
word’s possible meaning: “He brings out the most extraordinary echoes of the past,
what it must have been like to be a Roman”; “Edinburgh only gets seven hours of day-
light in winter”; “… soils of reputedly high fertility”. The characteristics also show in
the contrast between the Classifier, Descriptor and Epithet uses of a sense. The Epithet
and Descriptor uses of high, in Table 6, include descriptive elements which have been
bleached from the Classifier use.
Table 6. Epithet, Descriptor and Classifier uses of high

Determ. Epithet Descriptor Classifier Head
her high humming nasal sound
mild high blood pressure
a typical suburban high street
Classifiers are used in their minimum morphological form. Entity words lose
their plural and genitive inflections: “doctor parking”, and “child language acquisi-
tion”. Classifiers derived from Event senses lose their marking for tense and aspect:
“skim milk”, “barb wire”, “a carry bag”, “a drink driver”. Classifiers derived from Prop-
erty senses lose their adjectival inflection: “Japanese ports” becomes “Japan ports”
in contemporary English. All part-of-speech forms thus converge on an uninflected
form – “parent rights” for “parental rights” and “parents’ rights”, “burn time” for “time
for which it burns” and “time of the burn”. The morphological reduction is a sign
of semantic reduction. In “doctor parking”, we do not conceptualise any doctors as
unique people, or even as countable individuals, but only an abstract type or concept,
conceived without reference to existence in time or space. In “a carry bag”, we do
not conceptualise any event occurring at a particular time. “Our doctors’ parking”
would represent the doctors as specific unique individuals; “parking for the doctors”
would present them as less “individuated” (Timberlake 1975), but as specific people;
in “doctor parking”, they are still less individuated. The result is that the words carry
senses which are reduced to the minimum, and are unspecified semantically as well as
morphosyntactically; the word activates the central referential node, but not normally
the descriptive elements (see Chapter 5, §5.2.3.1). The motivations for the change
in individuation are those of simplicity and speed in processing, and of information
structure (Chapter 10).
Syntagmatic structure and semantic class. As Entity words, the Classifiers have sin-
gle nodes, with no descriptive meaning in referential use, and a minimal conceptual
meaning in descriptive use. As Entities not Properties, they have nothing in their
linguistic semantics that can link them directly to the Entity head. Instead, they use
the quale to make an indirect link. They are like prepositional phrases, which convert
a nominal group into a unit of a different kind, such as as a postmodifying “adjectival”
group, by means of an added relating word (the preposition); similarly, the Classifier
construction converts an Entity into a different kind of unit, a Property, by means of
the relating quale – to modify the head. For example, in “blood pressure”, the Entity
‘blood’ and the Type quale combine into a Property – which could also be expressed
in a postmodifying Property group, “of the blood”. In “fertility festival”, the Entity
‘fertility’ and the function quale combine into ‘for celebrating fertility’. (‘Celebrating’
is contributed by the hearer’s world knowledge.)
The qualia thus link a Classifier sense to an attribute in the head sense, as its val-
ue. (See Chapter 7, §7.3.2, for attributes.) For example, in “brick kiln”, meaning ‘kiln
Constituency
KILN Origin quale BRICKS
Constituency: made of.......
Attributes Function
Type
Dimensions
Diagram 1. “Brick kiln” with Constituency quale
KILN Origin Function

Constituency quale
Attributes Function: to bake .......... BRICKS
Type
Dimensions
Diagram 2. “Brick kiln” with Function quale
made of brick’, the Constituency quale links brick to the constituency attribute of KILN
as its value. See Diagram 1: senses are represented by boxes; concepts are printed in
small capitals; meaning elements are printed in lowercase letters. The qualia are repre-
sented by arrows; they have no boxes because they are not senses, but sense relations.
But in “brick kiln” when it means ‘kiln for baking brick’, it is the Function quale
that is the link, and a different attribute whose value is supplied; see Diagram 2.
Classifiers in adjectival form invoke Entities, not Properties as adjectival forms
usually do. For example, the adjective English denotes a descriptive property in “a very
English old gentleman”, and is an Epithet; but in “the present English team”, it means
‘of/from England’ invoking an Entity to indicate origin. That is the reason why current
usage is increasingly preferring expressions like “the England team” and “Japan ports”
to “the English team” and “Japanese ports”.
Like other modifiers, Classifiers operate with the general grammatical meaning,
“Add this content to the meaning of the head.” That meaning is made more specific for
each Classifier by the construction in which it occurs, which contributes the quale as
a concept to be used in relating the Classifier to its head, just as inflections for person,
number and tense add a conceptual element to their grammatical meanings. So for
“… electronic pressure gauge”, the specific grammatical meanings are “Add ‘electron-
ic’ to ‘gauge’, as its type”, and “Add ‘pressure’ to ‘gauge’, to indicate its function”. That
grammatical meaning is carried by their status as Classifiers, which is signalled by
having an Entity sense and by word order. In the qualia construction, the first posi-
tion signals origin, the second signals dimension, and so on. In practice, readers will
commonly use common-sense knowledge as well, of course: ‘2-inch’ is necessarily
dimensional, so its position need not be considered. Being a word form that usually
has an Entity sense (a “noun”) is a useful marker, but not a reliable one, as we have
seen with English.
8.2.2.3 Other Classifier constructions

Introduction. The other Classifier constructions can be understood best through
comparison with the figures that are alternative expressions for them. I emphasise
that the figures are given here only as an explanatory parallel, not as the source they
derive from.
The first can be illustrated as follows: “The government farm buy-up” parallels the
corresponding figure, “The government is buying up farms”; “consumer fuel purchas-
es” parallels the corresponding figure, “Consumers purchase fuel”. Table 7 shows the
group structure, and Table 8 shows the figure structure.
Whereas that construction has the Process as its head, the next has a Circum-
stance as its head. It is illustrated by “Sound travel time changes with altitude”, as in
Table 9. The parallel figure is “Sound travels for a certain time”, as shown in Table 10.
There are two other constructions, with attributes as the head, but I will not ana-
lyse them here. (See Feist 2012a, §5.5, §5.6.)
Table 7. Structure of the phrase, “consumer fuel purchases”

Classifier Classifier Head
(Actor) (Undergoer) (Process)
consumer fuel purchases
Table 8. Structure of the figure equivalent to “consumer fuel purchases”

Actor Process Undergoer
consumers purchase fuel
Table 9. Structure of “sound travel time”

Classifier Classifier Head
(Actor) (Process) (Circumstance)
sound travel time
Table 10. Structure of the figure equivalent to “sound travel time”

Actor Process Circumstance
sound travels for a certain time
Internal structure. Like Classifiers in the Entity-head construction, the Classifiers in

these constructions are simple in having a minimal descriptive meaning, and in being
reduced in individuation, as in “government farm buy-up”, where farm is less individ-
uated than in “The government is buying up farms”.
Syntagmatic structure and semantic class. Just as the Classifiers in the qualia construc-
tion were members of one semantic class but acted in the role of another class, so the
Classifiers and head here have a double role. In “consumer fuel purchases”, ‘purchas-
es’ acts as an Entity, to head the Entity phrase and be modified. But as a Process in
the corresponding figure structure with ‘consumers’ and ‘fuel’ as the Participants, it
has an Event sense; that Process-Participant relation applies also in the Entity-group
structure. Similarly, ‘consumer’ and ‘fuel’ act as Properties in their role as modifiers
in the Entity phrase, but are Entities in their role as Participants in the underlying
conceptual structure.
Diagram 3 illustrates the figure structure, with CONSUMERS as the agents, PUR-
CHASE as the Event, and FUEL as the Undergoer; the left-to-right order is semantic,
that of transitivity, represented by the arrows; (it is also the syntactic order, but that
is incidental).
Diagram 4 illustrates the structure of the phrase, as an Entity group; the left-to-
right order is now syntactic, to aid clarity, and the modifiers are on a lower level to
represent being subordinate. The words are printed in italics above the boxes, which
represent the senses. The arrows still represent transitivity – not dependence / modi-
fication, which would need a second set of lines.
CONSUMERS PURCHASE FUEL

Agent Event Undergoer
Diagram 3. ‘Consumer fuel purchases’ as figure
purchases
PURCHASE
Event
consumer fuel
Agent Undergoer
CONSUMERS FUEL
Diagram 4. ‘Consumer fuel purchases’ as Entity group

The grammatical meanings which build these constructions resemble the mean-
ings of the qualia construction. For “government farm buy-up”, the specific grammat-
ical meanings are “Add ‘government’ to ‘buy-up’, as Actor for the Event ‘to buy’,” and
“Add ‘farm’ to ‘buy-up’, as Undergoer….”
8.2.2.4 Constructionless Classifiers

Classifiers also occur singly, with semantic relations to the head that do not fit into
any of the constructions just listed. Examples are “child actor” (‘…who is a child’) and
“child psychiatrist ” (‘…who treats children’); “peanut oil”, (‘…made from peanuts’)
and “baby oil” (‘…for putting on babies’). As the glosses show, the Classifiers here have
no conventionalised relationship with the head, and do not fit the pattern of any of the
established constructions. I will call them “constructionless uses”, accordingly. They
do not occur in combination with a constructional Classifier, simply because they
cannot enter that Classifier’s construction. (However, they do occur recursively, as in
“Hero pilot Captain Sully Sullenberger,” which has three appositional construction-
less Classifiers.) Note that the relation we understand to exist between head and mod-
ifier is not specified by the modifier, or by any construction; the hearer is to infer it,
using world knowledge and the context; it is cognitive meaning, not linguistic mean-
ing. Some such expressions have become idioms, of course, with a unitary meaning,
not one constructed from modifier and head; “horse sense” and “horse latitudes” for
instance, and perhaps “glory days”. Others have become single words, such as a grave-
yard, horsehair and horseman.
Since there is no grammatical relationship between a constructionless Classifier
and its head beyond the bare fact that the Classifier is a modifier, the link being a mere
unspecified association, the grammatical meaning is quite non-specific, amounting to
“Add this content to the head in the best way you can think of from the context.” This
is another instance of the “association operator”, which Gil (2005, §2.3) discusses as a
characteristic of the simplest possible form of language; cf. Chapter 4 (§4.5.2).
For further discussion, see Downing (1977), Warren (1978, 1984), and Feist
2012a, §5.7).
8.2.2.5 Discussion of Classifiers

Submodification. Classifiers are frequently submodified, by other Classifiers. In exam-
ple (3), the Type quale Classifier analysis is modified by other Type quale Classifiers
recursively.
(3) “The prepositional [Sub-subtype] networking [Subtype] approach [Type]

analysis” (Sandra and Rice 1995, p. 98)
In example (4), there are six levels of submodification. ‘Trade’ uses the Type quale,
and ‘deceptive’ uses Subtype, to modify ‘practices’; ‘trade practices’ is a construction-
less modifier of ‘consumer protection’; ‘consumer’ modifies ‘protection’ as Undergoer
of the Event ‘protect’, in a Process-head construction. ‘Texas’ uses the Type quale, and
the whole of ‘deceptive… protection’ uses the Function quale, to modify the head,
‘act’. (That analysis is intended to show that my account of Classifiers will cope with
even the most complex of Classifier phrases.) The submodification can be continued
indefinitely.
(4) “Texas [[Deceptive [Trade Practices]] [Consumer Protection]] Act”

(Title of a Texas law)
The Classifier constructions in all their variations are extremely versatile, potentially
very powerful – and sometimes quite obscure, as in example (5), which is in fact
structured, and as in example (6), which is not – and sometimes quite long, as in
example (7).
(5) “Greater Tamaki Consolidated Receiving Environment Stormwater Network

Discharge Consent Prioritisation Consultation Results.”
(Title of a city council report; New Zealand Herald,
December 13th 2014, A23)
(6) “Endangered Species Smooth Dark Chocolate Baby Marine Iguana Bar”
(name of a chocolate bar; COCA)
(7) “You’ll have a chance to take home some classical pieces of memorabilia at
the Houston Symphony Partners’ Mad as a March Hare Symphony Surplus
Salvage Spring Cleaning Clearance Sale.” (Publicity material; COCA)
Compounds. Quirk et al. (1972, §I44–I48) treat as compounds such expressions as

“flower pot,” “reading material” and “air brake,” and many other writers have done the
same. It is clear, however, that they are straightforward examples of the Participant-
head construction, with “flower pot” and “air brake” using the Type quale, “reading
material” using the Function quale. Moreover, there is now an increasing literature
which agrees that they are not compounds, and which in some cases gives a detailed
argument for treating the first word in each of those pairs as a regular premodifier; for
example, Gagné et al. (2005), Bundegaard et al. (2008), Feist (2011) and Feist (2012a,
§5.8.1). Accordingly, I will here simply outline what I believe has gone wrong in the
discussion of “noun + noun compounds,” “compound nominals” and so on. “Com-
pound” began as an etymological term, and as such remains useful. It gradually ac-
quired applications in various areas of synchronic linguistics, e.g. morphology, syntax
and phonology, gaining a criterion in each one. The result is that Quirk et al. (1985),
for example, give syntactic, semantic, graphological and prosodic criteria, as well as
the etymological one; Miller (2014) gives six criteria. The different criteria give differ-
ent results, and each of them has exceptions; but that is thought acceptable, because
‘compound’ is deemed to be a “prototype” concept. Because of the inconsistencies, the
term as currently understood is useless for explaining present-day English. Worse,
keeping the term and the concept has concealed from linguists’ view the varying con-
structions, set out in the previous sections, that have developed in the last century or
two; and has concealed the fact that the constructions are formed according to regular
English principles, without “prototypes”.
Genitives. Some Classifiers carry genitive meanings, as can be seen by comparing al-
ternative forms for the expression: “a tonne weight,” compared with “a tonne’s weight”
and “a weight of one tonne;” “the Japanese Prime Minister” and nowadays “the Japan
prime minister,” compared with “the Prime Minister of Japan” and “Japan’s Prime
Minister.” The point will be discussed further in Chapter 9, §9.5.3, which will con-
clude that these Classifiers form a third formal genitive, in addition to the of- and -’s
genitives.
8.2.2.6 Conclusion: Classifiers

Whereas premodifiers in the other zones are related to their heads as Properties
straightforwardly and directly, Classifiers’ constructional meaning makes a composite
Property (Entity + quale or ofter relation), which relates to the head. The construc-
tions form a number of intricate structures – more complex than any other relative-
ly basic semantic structures in English. Though complex, they are not idiomatic,
but rule-bound – provided that we accept the general sub-rule of English that rules
may be broken by marked forms, the marked forms here being the constructionless
Classifiers.
8.2.3 Descriptors
Typical Descriptors denote colours, shapes and other perceptual qualities: red, white,
square. If not single morphemes, they are typically in adjectival form, i.e. with such
suffixes as -ous and -able, or in participial form, that is, with the suffixes -ing and -en/
-ed. Examples, followed by Classifiers, include: “black iron fences,” and “high blood
pressure”. Some words can be used as either Descriptor or Classifier: it is the particu-
lar sense, not the word as such, which is a Descriptor or a Classifier; see Table 11 for
an example.
8.2.3.2 Internal structure

Type of meaning. Descriptors have non-gradable and relatively concrete descriptive
meaning. They never have affective or attitudinal meaning, as Epithet do. (The con-
trast with Classifiers is in semantic class and function, to be dealt with in the next
subsection.)
Table 11. Classifier, Descriptor and Epithet uses of dirty

Determ. Epithet Descriptor Classifier Head
a simple but effective digital dirty bomb
a little dirty comic book
thin, straggly, dirty blonde rats-tail hair
Table 12. Red and running as Classifier and Descriptor

Word Determ. Descriptor Classifier Head
red permanent red alert
a red silken shirt
running some new running shoes
some running cold water
Dimensions. Since they are not gradable, their descriptive meaning does not vary on
the intensity dimension; and they are bounded, being conceptualised as applying
across areas of space or lengths of time. Their senses are relatively simple, having few
possible sense elements that may be invoked; but they are more complex than those
of Classifiers; so the dimensions of salience and expectedness apply to them. Compare
the pairs of uses in Table 12.
In “permanent red alert”, the Classifier sense ‘red’ does not denote a descriptive
quality, and does not apply to an area, but in “red silken shirt”, the Descriptor ‘red’
is a quality, and we apply it to the area of the shirt. In “running shoes”, the Classifier
‘running’ is like ‘athletics’ in designating only a type (not movement or duration), but
in “running cold water” the Descriptor is conceived with movement in time. (In “a red
silken shirt”, silken means ‘made of silk’ (SOED <1>; it means ‘like silk, glossy’ (SOED
<2>) only as a Descriptor, as in the invented phrase “a silken Red Admiral butterfly”.)
8.2.3.3 Syntagmatic relations and semantic class

Descriptors as modifiers. Where Classifiers create a Property sense only indirectly
through a quale or a grammatical relation, Descriptors denote Properties directly. For
basic Descriptors, the concrete cognitive domain to which the sense belongs provides
the bond; it specifies a value for an attribute of the head. In “red silken shirt”, ‘red’
specifies the value of the colour attribute of ‘shirt’. Accordingly, basic Descriptors are
Property senses, modifying an Entity as head. See Diagram 5a.
The participial forms are more complex. They do function as Properties, as in
“disabled Irish writers” – being disabled is a Property of the writers. But participles
are basically Event words, as in “Disease and accidents have disabled many writers”
(invented sentence); and when used as Descriptors they generally retain some Event
RED SHIRT GLITTERING BEADS
Value of color attribute Event Actor
(a) (b)
Diagram 5. (a) Basic Descriptor semantic structure, in ‘red shirt’; (b) Descriptor seman-
tic structure with a participle, in ‘glittering beads’
qualities, such as denoting a happening and its time (past, for ‘disabled’), and relat-
ing to the head as Process to Participant (the writers are in the role of Undergoer of
‘disable’, in the cited sentence). In “glittering crystal beads,” the beads – the Actor –
glitter, in continuous aspect. See Diagram 5b. (That relationship with their Event base
weakens through historical time, as with “a three-tiered wedding cake”, and “a sur-
prising result”; the senses become “adjectival” i.e. they become Property senses not
Event senses.)
A few noun forms also occur as Descriptors, as with “a silver sound” and “her
copper hair”, but they are commonly accepted, as by SOED, as having become “adjec-
tives” historically. Another syntagmatic structure is that Descriptors denoting time
may call on a happening in the denotation of the head. For example, in “an epic eight-
minute atmospheric adventure”, ‘eight-minute’, being an expression of duration, re-
quires ‘adventure’ to be conceived as a happening, with duration in time.
Thus the general grammatical procedure for combining the sense with that of the
head as a Property applies to most Descriptors, but takes some more specific forms.
For participles: “Treat the Event sense as if it were a Property, and combine it with the
head.” For time senses (such as ‘eight minute’), “Treat the Event sense’s relation to the
head as a Property, and combine it with the head.”
Structure with both Classifier and Descriptor is illustrated in Diagram 6. Part (a)
represents the structure as network, with a chain of nodes and links: shirt has attrib-
utes, one of which is Constituency, whose value is ‘silk’; showing the full network
would include elements of shirt other than its attributes, and attributes other than
constituency, and so on. Part (b) represents the words as constituents (as boxes), with
linkage shown as meaning elements in areas of overlap.
SILK
value
constituency SILK
attribute constituency
SHIRT SHIRT
attribute colour
colour
RED
value
RED
(a) (b)
Diagram 6. Structure with Descriptor and Classifier – “red silken shirt”: (a) as network;
(b) as constituents
Table 13. Grading by submodifiers, as changing the class of old and fat
Epithet Descriptor Head
an extremely fat old pig
an extremely old fat pig
an extremely fat, extremely old [black] pig
Submodification of Descriptors. Descriptors may be submodified – may have the role

of operand as well as that of operator – as with “dark red hair”, and “recently arrived
immigrants”. The submodification may be complex and may be repeated, as in exam-
ples (8) and (9); the submodifiers are underlined. (Note that this is semantic submod-
ification; syntactic submodification may perhaps be different.)
(8) “The floppy-haired, guitar-strumming, Harley Davidson-riding Mr Boudou

was considered… to be a lightweight.” (Economist, June 14th 2014, p. 37)
(9) “Particular highlights… include recording the platinum selling, Classical Brit
award nominated album, …” (Daily newspaper report)
Descriptors are relatively seldom submodified, however. That is partly because they
are semantically simple; but there is another reason, which is much more important
for our purpose. For a Property sense to be submodified for degree, it must be grada-
ble, and therefore an Epithet, as shown by the invented phrases in Table 13: when a
Descriptor sense is graded, it becomes an Epithet sense, because making it gradable
changes its semantic subclass. Old and fat are Epithets when graded, but Descriptors
otherwise.
8.2.3.4 Discussion of Descriptors

Categoriality of zones. With the examples in Tables 6 and 11, it is easy to determine
which zone the word belongs in, because all slots are filled. When there is only one
premodifier, as in “a four-year stint” and “a nationwide revamp”, that is often difficult,
and the sense may seem to be “on the borderline” between zones , or to be a “periph-
eral” instance of a “prototype” category. In fact, it is never on a borderline. Every ap-
parently borderline Descriptor or Classifier will co-ordinate with either a Descriptor
or with a Classifier – one or the other; for example, ‘high’ seems borderline in “high
pressure” and “high street”; but the coordination in “high, increased pressure” reveals
a Descriptor, and the coordination in “the high or ‘main’ street” reveals a Classifier.
The characteristics of the senses it could combine with make clear which zone the
sense is in. We conclude that there cannot be indeterminate or strictly “borderline”
instances in use, and that the distinction between Descriptor and Classifier senses is
categorial, not subject to prototype effects. Such potentially confusing “borderline”
instances occur with all adjoining pairs of zones, and the distinction between the
zones is always categorial.
Cognitive and linguistic areas of meaning. Descriptor senses are close to the cogni-
tive end of the cognitive-linguistic scale (Gentner and Boroditsky (2001); Chapter 4
above, §4.2.1). They are perceptual rather than conceptual; they lack the attitudinal,
affective and social meanings, which language adds to factual meanings, as subjective
comments. Grasping the distinctions between related Descriptors generally requires
experience of the world, as with complementaries such as living/dead, and moving/
stationary, and with other incompatible pairs such as grassy/stony/sandy and running/
standing/walking. However, there is an element of language dependence in the dis-
tinctions between, for instance, ‘red’, ‘crimson’ and ‘scarlet’, which are partly language
dependent.
8.2.4 Epithets
Epithets are typically the colourful adjectives that characterise descriptive speech and
writing, as illustrated in Table 14.
Table 14. Sample Epithets

Det. Epithet Descriptor Classifier Head
the archetypal happy fat man
this beautiful young gay man
an earnest bespectacled young man
a frail-looking elderly Chinese man
8.2.4.2 Internal structure and class of Epithets

Epithets are typically more abstract than Descriptors, but the defining difference is
that they are scalar: their descriptive or affective sense is set somewhere on a scale;
consequently, it is usually gradable. They are also distinct in being able to include
affective, attitudinal and social types of meaning in their structure. Like Descriptors,
Epithet senses are Properties.
8.2.4.2.1 Descriptive meaning and its dimensions

Perceptual meaning versus conceptual meaning. Whereas Descriptor senses are rela-
tively concrete and perceptual, Epithet senses are relatively abstract and conceptual.
That can be seen clearly in their contrast with Descriptor uses of the same words.
In Table 15, bloody as Classifier is referential, meaning ‘having blood in it’; as a De-
scriptor, it is descriptive, meaning ‘shedding blood’; as Epithet, it is chiefly emotive,
but has the faint conceptual meaning ‘bad’. (That is the universal pattern for those
three zones, arising from the historical development: Classifier senses develop into
Descriptor senses, which develop into Epithet senses.)
Table 15. Bloody as Classifier, Descriptor and Epithet

[severe] [frequent] bloody diarrhoea
the traditional bloody reprisal killings
all this bloody modern [British] poetry
Intensity dimension. Being relatively abstract and conceptual is characteristic of Epi-

thets, but Epithets’ defining feature of gradability resides in the scalar intensity dimen-
sion. Epithets are scalar – necessarily having an intensity or “degree” dimension – but
Descriptors are not.
The basic sense of words like big is indeterminate in intensity or degree, and may
be graded in use, either morphologically by inflections such as -ish and -esque, or
morphosyntactically by the system of comparison (-er/-est or more/most), or lexically
by submodifiers like extraordinarily. By contrast, synonyms of big such as huge differ
from it in being inherently determinate in intensity, carrying their grading within
them, as it were. Similarly, words like outstanding, superb and excellent carry their own
grading as inherently superlative, so are not gradable morphosyntactically; they are
scalar, though not gradable.
It is important to note that words such as carved, tailored, black, old, and young
are not gradable when used as Descriptors, but when used as Epithets they may be
graded, by -er/-est or submodification (and have a somewhat changed meaning). It is
senses that are gradable or not gradable, rather than words. With Descriptors, we see
Properties as simply present or absent, but with Epithets we see Properties as present
more or less (gradably), with presence presupposed (Croft and Cruse 2004, p. 167).
Illustrative examples are given in Table 16.
There is an instructive parallel in the fact that we can answer “Oh, it’s quite cool,”
in answer to the question, “How hot is it?” Hot in the question denotes a Property that
may be present or absent; it does not assert a degree of heat.
Other dimensions. As shown in the histories of brave and flare in Chapter 5 (§5.2.2.1,
§5.3.4.2), and in bloody just above, Epithet senses are more highly evolved than De-
scriptor and Classifier senses, becoming more varied and more highly differentiat-
ed. For example, SOED gives two Descriptor senses of blue, both related to colour;
but it gives five Epithet senses, ranging over constancy, depression, political loyalty,
Table 16. Non-gradable and gradable uses of dark and tiny

Word Use Det. Epithet Descriptor Head
Dark Gradable, as Epithet very dark tiny rooms
Dark Non-gradable, as Descriptor a very strong dark side
Tiny Gradable, as Epithet a very tiny little place
Tiny Non-gradable, as Descriptor very dark tiny rooms
Table 17. Bony as evocative Epithet and as Descriptor

a bony wizened face
a long bony face
pedantry and indecency. As a result of that differentiation, Epithet senses also vary
greatly in their dimensionality. On the specificity dimension, some senses form a
range of specific meanings, as with ‘obscure’, ‘arcane’, ‘esoteric’ and ‘occult’; but others
are general. They vary on the vagueness dimension as well, from precise ‘serendipi-
tous’ to very vague ‘frightful’.
As a result of developing various elements within their meaning, Epithets often
have a number of sense elements which vary on the dimension of expectedness. For
example, hot <1>, a Descriptor sense, is simply ‘very warm’; but hot <7>, an Epithet
sense, has ‘danger’, ‘stolen’ and ‘affected by a trade union dispute’ among its many
possible elements; (‘difficult or awkward to deal with’ is the necessary element of that
sense). Similarly, using a word in Epithet position makes it evoke more of its possible
sense elements (partly because of its salient position). In the uses in Table 17, bony as
Epithet evokes such elements as ‘hard’ and ‘rather unpleasant’; as a Descriptor, it has
no more than its necessary meaning.
(The context in BNC for the first phrase includes “… his thin beak of a nose. He
had a bony wizened face and an unhealthy pallor…”. The context in BNC for the sec-
ond phrase includes “The high, arched eyebrows in the long bony face rose slightly…”;
bony does little more than explain long.)
A further consequence of Epithets’ having many meaning elements is that the
senses are often complex, with the elements varying on the salience dimension. SOED
illustrates its sense <4a> of loose with the quotations given as examples (10) to (12).
(10) “Scandal I have given by my loose writings.” (Dryden)

(11) “A rash young fool; carries a loose tongue.” (Carlyle)
(12) “Senators, so loose when it comes to their own behaviour…”
(International Herald Tribune)
The definition of <4a> includes ‘indiscretion’, ‘lacking power of restraint’, ‘lax in prin-
ciple’ and other elements. All of those elements are invoked by each of the utterances
just quoted, but ‘indiscretion’ is salient in (10); ‘lacking power of restraint’ is salient in
(11); and ‘lax in principle’ is salient in (12).
8.2.4.2.2 Affective, attitudinal and social meaning

While it is not the defining feature of Epithets, it is very characteristic of them that
they have a variety of meaning types, often combining descriptive, emotive, attitu-
dinal and social meaning. That is most easily understood through their history, as
discussed above in Chapter 4, §4.4.6 and §4.5.1, with gross.
Table 18. Warm as Epithet, with and without emotive meaning

Epithet Descriptor Classifier Head
With emotive meaning warm neutral colour scheme
Without emotive meaning warm cooked white rice
Table 19. Paradigm of social meaning

Sense Formal Standard Informal
‘very big’ gargantuan huge whopping
‘hit’ smite hit biff
Several points which are implicit in the previous section on variety of meaning
types should be made explicit here. First, some Epithets which often have emotive
meaning lack it in some uses, as illustrated for warm, in Table 18.
Second, while the function of liquid as Epithet is to be emotive (through being
metaphorical), other Epithets are not emotive but have a synonym available to give
an emotive version of the same descriptive meaning. For example, overweight has
fat (in certain of its uses), obese, well-padded, curvaceous, and voluptuous. Next, the
descriptive meaning of Epithets often becomes gradually very vague and weak, and
sometimes disappears altogether, as with nice, lovely, horrible, terrible and appalling.
The absence of descriptive meaning is a distinctive feature of such Epithets.
Finally, social meaning occurs in Epithets in the same ways as attitudinal and
emotive meanings do. We have colloquial comfy and standard comfortable; slang easy-
peasy and standard easy; literary bestarred and standard starry; dialect wee and stand-
ard small. A second pattern is in the paradigmatic double contrast between standard
and formal use, and standard and informal use, illustrated in Table 19. (The social
value of words varies greatly with place as well as time; the ratings in the table are for
New Zealand usage at the time of writing, from my own experience of it.)
Chapter 5, §5.3, on sense relations, expands some of the points made in the last
three paragraphs.
8.2.4.3 Syntagmatic structure of Epithets

8.2.4.3.1 Introduction
To say merely that premodifiers modify their head is a simplification; strictly, each
premodifier operates on the rest of the group, as a unit. However, because of their
abstract and rather subjective semantic nature, Epithets sometimes also operate on
other modifiers individually, creating the networks described in Chapter 5, §5.2, and
sometimes operate on other heads. That flexibility also makes other elements of their
syntactic nature complex.
8.2.4.3.2 Epithet as operator

Modifying the head’s referent
(a) Modifying an Entity in the head
In their developmentally basic form, Epithets are Property senses by semantic class,
expressed by non-derived words (“adjectives”): bold, rash, loose and so on. Their syn-
tagmatic structure is therefore like that of typical Descriptors; they relate to the head
as value to an attribute. For example, in “The great carved stone dragon”, ‘great’ is the
value of the ‘size’ attribute in ‘dragon’. “Round pink face” has ‘round’ as value of the
‘shape’ attribute of ‘face’.
Sometimes however, the Epithet has a non-Property element. The reason for that
is that they vary in their semantic derivation, as indicated by their morphological
form. The adjectival forms are often derived from nouns or verbs; they may now be
understood with a purely Property sense, as with illiterate and dirty, but they some-
times carry an element of their Entity base sense, as with “My infantile bad behav-
iour”, where the Epithet carries the sense ‘of an infant’, just as the Classifier infantile
does in “the dreaded infantile paralysis”. Similarly, participial forms sometimes retain
something of the underlying Event, as in example (13).
(13) “A four-foot long spinner shark went flying through the air behind a very
surprised surfer.” (COCA)
The shark had surprised the surfer, so surprised carries some Event sense.
The nominal forms relate to their heads somewhat as Classifier nominal forms
do. For example, in “a Mickey Mouse outfit” – when it means ‘an inferior organisa-
tion’, not ‘clothing for a certain cartoon character’ – the link is not through adding the
meaning of the modifier to that of the head, as with most Epithets, but through the
same vague associational link used with constructionless Classifiers: “of or pertaining
to Mickey Mouse”, with the connection supplied cognitively, as ‘unsuitable’, and in this
case ‘inferior’. Kafkaesque – ‘associated with Kafka’ – is similar. (It may not be clear
why I have identified those senses as Epithets. The reason is that they have conceptual
senses that are gradable. Being followed by a Descriptor sense is a useful but not reli-
able indicator of being an Epithet.)
The semantic class of the Epithets just discussed is structurally complex. As not-
ed, such senses have an Event or Entity element, retained from their origin; but Entity
premodifiers must have Property senses, so the class is reconstrued as Property to fit
the syntactic context, and the Event or Entity element is subordinated to it.
(b) Modifying an Event in the head

In a further variation, the modified element of the head is sometimes an Event, not
an Entity. The Event element may be apparent morphologically, as in utterances like
“He is a big eater”, “They are hard workers”, and the much-discussed “She is a beau-
tiful dancer”. The Event element may be reasonably clear intuitively, as in nominali-
sations such as “his frank admission”; but it may be obscure, as in “an indiscriminate
massacre” (BNC). In each case, however, that Event element becomes apparent in a
ATTRACTIVE ATTRACTIVE
value value
appearance appearance
attribute attribute
WOMAN WOMAN
attribute attribute
activity: she dances activity: she dances
value value
GRACEFULLY GRACEFULLY
(a) ‘attractive woman who is dancing’ (b) ‘woman who dances gracefully’
Diagram 7. The two meanings of “beautiful dancer”
paraphrase of the group, which needs an Event head word with a manner Proper-
ty modifying it, in place of the nominal construction: “He eats largely,” “She dances
beautifully,” “He admitted it frankly,” and “[They] massacred [people] indiscrimi-
nately.” Diagram 7 illustrates the semantic ambiguity of “a beautiful dancer”. Both (a)
and (b) represent the full ambiguity; diagram (a) represents the meaning ‘an attrac-
tive woman (who dances)’, by having the relevant elements in bold face; diagram (b)
represents the meaning ‘woman who dances gracefully’, with those elements in bold.
(Strictly, (b) elides some of the structure, which includes manner.)
The preceding paragraphs have described the syntagmatic structure of Epithets’
descriptive meaning. The structure for nondescriptive meanings is quite different.
Since they are not conceptual, their bond to the head is not of the sort discussed
so far, such as being a value for an attribute in the head. The grammatical mean-
ing which controls their being built into the compositional meaning of the group is
different also.
The structure of affective and attitudinal meanings can be illustrated from exam-
ple (14).
(14) “We’d all fuss over the unfortunate young woman.” (COCA)
The affective meaning of unfortunate is a feeling – a feeling about the woman, just as
one piece of information can be about another, and as a Comment is about a Topic.
The grammatical meaning for these nondescriptive meanings is thus quite different
from that of descriptive meanings. It can be formulated as, “Understand the speaker
as having a feeling [pity, in this instance] for the referent.” In many contexts, including
the context of the sentence just cited, there is a secondary meaning, “Respond to the
speaker by sharing the feeling”.
For social meaning, the structure is looser still, as in example (15).
(15) “Blown tae smithereens… his gorgeous, bonny head, no there.”

(COCA; fiction)
The modifier bonny, like the preposition tae, marks the speaker as Scots. But there is
not even a speaker intention to create a syntagmatic structure; there is only the arbi-
trary association between bonny and Scots speakers.
(c) Modifying a referent other than that of the head

We have seen that Epithets need not have descriptive meaning , and need not add
their meaning to the head, as other premodifiers must. They can also modify some-
thing other than the head. Sometimes, they modify a referent denoted by the deter-
miner, as in “his own stupid fault” where stupid applies cognitively to the referent
of his, just as the postdeterminer own does. “His older brother” and “his favourite
music” have a comparable structure, as shown by the paraphrases, “He was younger
than his brother,” and “He favoured that music.”
The referent may be more distant syntactically. It may be the head of a different
Entity group, as in example (16).
(16) “I remember laying [sic] there in the lonely hospital room thinking, ‘Here we
go again’.” (New Zealand Herald, August 9th 2014, A17)
Linguistically, ‘lonely’ modifies ‘hospital room’, but cognitively it modifies the speaker.
The referent may even be represented in a subordinate prepositional phrase: “a naked
photo of the mayor” (Huddleston and Pullum 2002, p. 558), or be merely implicit: “an
indiscriminate massacre of commuters” (British National Corpus).
Discussion of Epithet as operator. These variations in Epithets’ syntagmatic relations,

and their power to evoke varied and vivid meanings, are made possible by the seman-
tic elements discussed in Chapter 4. Their range of meaning types, being much great-
er than that of Descriptors and Classifiers, increases their wealth of meaning. Their
variation on the intensity, generality and specificity dimensions adds to the subtlety
and complexity they add to Entity groups. Finally, the variation in their grammatical
meanings, and their potential to be both very general and quite specific in the opera-
tion to be applied, and their vagueness as to what it is to be applied to, add still more
to the range of meaning they allow.
8.2.4.3.3 Epithet as operand

Epithets have varied relations as operands, also – that is, when they are submodified.
Because they are often high on the generality dimension, they are often specified by
a submodifier, as in “very politically powerful people”. (There, the modifying group –
“politically powerful” – is itself submodified, with “very” intensifying it.) Because the
intensity dimension is important for Epithets, they are frequently specified for de-
gree, as in “full-length black leather coat”. Because they are often vague, the reference
is often made precise, is in “that potentially dangerous mix of DNA”. Their affective
meaning is also often modified, just as their descriptive meaning is – mostly on the
intensity dimension – as in “a [sic] absolutely horrible numbing experience”. If the
Epithet has an Event sense (which is unusual), it can be submodified for manner, as in
“a heavy, badly carved dark-stained wardrobe”.
8.2.4.4 Discussion of Epithets

8.2.4.4.1 Comparison with Classifiers and Descriptors
Table 20 gives examples of blind used as Epithet, Descriptor and Classifier, to illustrate
the semantic nature of the zones, and to emphasise again that a single sense can be
used in different zones, with changes in its internal structure. As Epithet, ‘blind’ is
gradable and intended to evoke a range of possible elements – highly descriptive. As
Descriptor, it is still descriptive, but simply factual, being non-gradable. As Classifier,
it is not descriptive, being intended merely to classify the referent. Note that it is not
just the same word, blind, that is varying, but the same sense i.e. <1> “Without the
sense of sight”.
Table 20. Blind as Epithet, Descriptor and Classifier

an old and nearly blind retired political philosophy teacher
an elderly blind black man
a handsome young blind man
8.2.4.4.2 Cognitive and linguistic areas of meaning

Whereas Descriptors represent cognitive meaning primarily, Epithets represent lin-
guistic meaning, with some exceptions such as uses like “nearly blind”, where the
grading turns a Descriptor into an Epithet. Affective and social meanings are obvious-
ly not cognitive, and combining them with conceptual meanings is wholly linguistic.
So are the patterns of conceptual elements that distinguish synonyms. The Epithet
word elderly, for example, and its synonyms ancient, antiquated and aged, all represent
the single cognitive concept OLD – expressed in the Descriptor old. We learn to dis-
tinguish them from experience of English, not from experience of the world.
8.2.5 Reinforcers
Sample Reinforcer uses are “a perfect stranger” and “pure fabrication”. They reinforce
the meaning of the head, rather than add meaning to it; they are thus similar to in-
tensifying submodifiers like very. Further examples, from BNC, are given in Table 21.
Table 21. Sample Reinforcers

Det. Reinforcer Epithet Descriptor Classifier Head
utter deep[,] black conviction
mere modern Christian magic
sheer bloody-minded macho rage
a mere little nine-year-old girl
8.2.5.2 Internal structure of Reinforcers

In an expression like “That’s utter nonsense!”, utter could be replaced by absolute,
complete, perfect, pure, sheer or total, without a significant change of meaning. In this
use, those words are thus synonymous; but in other uses they have distinct mean-
ings: compare “complete translation” and “perfect translation”. The explanation of the
anomaly is that Reinforcer senses develop from Epithet uses, losing their conceptual
meaning. In fact, they are empty of descriptive meaning, differing radically from Ep-
ithets and Descriptors in that respect. (Development from Classifier to Descriptor to
Epithet is at each step an increase in abstraction, with increasingly abstract content;
this step abstracts from content meaning altogether, into grammatical meaning; see
further §8.2.8.5.)
Mere has negative affective meaning, and the others may have some affective
meaning in context also. For instance, in example (17) utter is neutral, but in example
(18) it is emotive.
(17) “With all due respect, that statistic is utter nonsense.”

(COCA; spontaneous; thoughtful discussion)
(18) “But what he said is just utter nonsense.”
(COCA; spontaneous; political debate. Just marks utter as sarcastic)
Their dominant, and usually exclusive, meaning is grammatical; they are “grammat-
ical items” (and consequently do not belong to a semantic class). The meaning is of
the general type, “Adjust the meaning of the head.” Different Reinforcers have differ-
ent specific meanings. Words like complete mean “Adjust the meaning of the head by
maximising it.” Mere specifies “… by minimising it”, as in “mere decoration”. Sheer and
absolute sometimes signify “… to mean neither ‘more nor less than…’ …”, as in “sheer
luck”. (See Quirk et al. 1985, p. 1338, for the three types.)
8.2.5.3 Syntagmatic relations

Whereas other premodifiers are mostly related to the head through the bond rela-
tionship of value to attribute, Reinforcers are not, since they have no content. Their
grammatical-function relationship varies according to the three specific meanings
described just above. Diagrams 8 and 9 illustrate the maximising type, by contrasting
the sentences “That’s nonsense”, in Diagram 8, and “That’s utter nonsense” in Diagram
9; both use constituency format. Conceptual meanings are shown in capitals as usual,
That is nonsense
[ Find previous referent ] [ Relate Subject ABSURD WORDS
and Complement
as Carrier and Attribute ]
Diagram 8. Syntagmatic relations in “That’s nonsense”
That is utter nonsense

[ Find previous referent ] [ Relate Subject [ Intensify
ABSURD WORDS
and Complement the head]
as Carrier and Attribute ]
Diagram 9. Syntagmatic relations in “That’s utter nonsense”
based on nonsense <1>: “Absurd…. words or ideas”; grammatical meanings are shown
in lower case, formulated as instructions to the hearer. The effect of the maximising
intensification is represented by the use of boldface for ABSURD WORDS, in Diagram 9.
(The opportunity is taken to represent the grammatical meaning of pronouns and
copulas as well; only one of the four words in the diagram has content meaning. That
is intended to highlight how far meaning in English often is from being a succession
of concepts.)
In Diagram 8, the meaning of each word (in italics) is given below it – either a
grammatical meaning in square brackets, or a descriptive meaning in capitals. The
entry for is means: “Relate the subject, indicated by the arrow to that, and the Com-
plement, indicated by the arrow to nonsense, in the relation of Carrier to Attribute”.
Their relations with other premodifiers vary also. Since Reinforcers, like other
premodifiers, modify the rest of the group, most Reinforcers do not readily occur with
other premodifiers, because the latter seldom have a meaning suitable for emphasis.
(Epithets constitute an exception, since they naturally accept emphasis, as in “a com-
plete bloody idiot”.) Mere, however, occurs with other premodifiers frequently, since
its deprecating effect is readily matched by negative Epithets and Descriptors, as in “a
mere useless gibbering stop-the-war-at-any-price pacifist,” and “a mere 390,000 live
TV audience on Sky” (BNC). (Those two phrases are rare instances of Entity groups
with all four zones filled.)
Reinforcers cannot be submodified; they do not act as operands. That is a natural
consequence of having only grammatical meaning. (When “grammatical items” such
as prepositions are submodified, as in “right to the end”, it is their content meaning,
not their grammatical meaning, which is modified.)
8.2.6 Determiners
As noted in §8.2.1, “determiner” denotes a function not a semantic class. The zone
where those functions are realised – the determiner zone – has three subzones, il-
lustrated in “all the other three boys.” Quirk et al. (1985, §5.11–5.14)say that the first
subzone has “predeterminers”, including all, both and half. “Central determiners” in-
clude a(n), this/that and other demonstratives, and genitive words and phrases such
as John’s and my uncle Bill’s. Other is a “postdeterminer”, along with the cardinal and
ordinal numerals.
This discussion will be limited to the basic functions of determiners, excluding
such things as the use of the with proper nouns and parts of the body, and the occur-
rence of superlative adjectives as post determiners.
8.2.6.2 Syntagmatic relations

Like the premodifiers, determiners relate to the whole of the rest of the Entity group,
as in “all [the [other [three boys]]].”
Determiners are commonly thought of as grammatical items, and as having a
syntactic function. Accordingly, their semantic nature and their semantic relation to
the rest of the nominal group is usually not made clear in the literature. Quirk et al.
(1985) say that determiners indicate the kind of reference that the group makes. That
is not very clear, however, since “reference” is vague, being wide enough to include
both whether the meaning is specific or generic (which is a conceptual and semantic
matter), and whether the group is definite or indefinite (which is a discourse matter,
guiding the hearer on whether the referent is already known).
I suggest that the relevant function is (for once) indicated fairly well by the stand-
ard grammatical term itself. Determiners set the boundaries (“terms” in an oldish
sense) of what the group denotes. Where restrictive premodifiers help limit what is
denoted descriptively, determiners help limit it non-descriptively. Central determin-
ers (the articles, demonstratives and genitives) do that deictically (in a broad sense
of the term), pointing the hearer to something in the context which will limit the de-
notation. Predeterminers and postdeterminers limit it quantitatively, predeterminers
operating through strict quantification (all, both, half), and postdeterminers working
through quantification in a looser sense, by count (cardinal numerals), by position
in numerical sequence (ordinal numerals), and so on. The structure of determiners’
semantic relations to the rest of the group is thus as follows: quantification [deixis
[quantification …head]].
That explains determiners through the all-purpose concept of function. More
precisely, and in the terms used in this book, determiners have a grammatical mean-
ing such as “Limit the denotation of the group according to the context” (as in “the
moon”), or “Limit the denotation according to the postmodifier” (as in “the books on
the shelf ”).
For non-restrictive Entity groups, there is often no need at all for determiners, as
in “I like dark chocolate.” When they are used, they are simply descriptive, rather than
determinative: in “He saw some big brown Hereford cows in the paddock,” some does
not identify the referent, but gives extra information, as the premodifiers do – some
there means “…and there were several of them”.

The elegant schema just given (other determination consisting of quantification +
deixis + quantification) is oversimplified; the continued expansion of determiner
functions over the centuries has distorted it. Central determiners, which at first were
abstract and purely grammatical (that, the) now include referential meanings and de-
scriptive content (in genitives such as “the present Queen of England’s crown”; even
predeterminers can contribute conceptual information, as in “He saw some…” quoted
just above. Postdeterminers are partly conceptual in relying on numerical concepts
(sixth). Their basic grammatical meaning is in effect combined with other types of
meaning.
8.2.6.4 Discussion of determiners

Cognitive-linguistic scale. Determiners lie at the linguistic end of the cognitive-
linguistic scale (Chapter 4, §4.2.1).
Determiners in relation to other zones. The determiner sense of single illustrates the
point, made previously but worth repeating, that the sense of a word with a basic
Property meaning changes its semantic structure with changes in zone, while retain-
ing some element of the basic sense. Single occurs in all five zones, as in examples (19)
to (23), all from BNC; all the uses use the element ‘unaccompanied’ from sense <1>.
(19) “A single emphatic conventional gesture”: postdeterminer, emphasising the

singular meaning of “a”; SOED <4>.
(20) “The single largest capital investment”: Reinforcer; emphasising “largest…
investment”; SOED <6>.
(21) “A simple, single toroidal winding”: Epithet; “unique, singular”; SOED <7>.
(22) “An indivisible single self ”: Descriptor; “undivided, united”; SOED <5>.
(23) “A young Nigerian single parent”: Classifier; denoting a (sociological) type of
parent; SOED <8>.
Table 22 shows that pattern visually, in the diagonal “movement” of single from top
left to bottom right.
Genitives. Genitive determiners such as “the present Queen of England’s crown” obvi-
ously have a structure quite different from those of the single-word determiners dealt
with so far. They are rankshifted groups with their own structure of premodifier +
head + postmodifying group.
Table 22. Single in all five zones

Det. Reinforcer Epithet Descriptor Classifier Head
a single emphatic conventional gesture
the single largest capital investment
a simple, single toroidal winding
an indivisible single self
a young Nigerian single parent
8.2.7 Postmodifiers
Postmodifying items are usually phrases or clauses. See Quirk et al. (1985, §17.55) for
the rationale for atypical one-word forms, such as “the road back”, “the room upstairs”
and “a house nearby”. A few of those uses are interesting for the difference in meaning
between premodifier and postmodifier use. In “his journey homeward”, journey is
a nominalised Event; homeward as postmodifier denotes its direction descriptively
(‘he travelled towards his home’). But in “his homeward journey”, journey denotes a
countable Entity, and homeward as premodifier is restrictive, identifying the journey,
not describing it (‘that trip’).

Postmodifying items are like premodifying ones in the types of meaning they express;
they differ chiefly in giving extra weight to the Property represented, which is an issue
of information structure, to be dealt with in Chapter 10. Single words are Property
senses, like typical premodifiers. Prepositional phrases, which semantically consist of
a grammatical meaning (in the preposition) combined with an entity (in the Entity
group), are reconstrued as Properties when rankshifted. Adjectival clauses are simi-
larly reconstrued into functioning as Properties.
8.2.7.3 Syntagmatic structure

Relations in the structure. The syntagmatic relations between postmodifiers and their
head are the same as for premodifiers, in that they gain their bond by instantiating an
attribute of the head meaning. They differ, however, in commonly making that attrib-
ute explicit. Compare attested example (24) and a paraphrased version of it in (25).
(24) “The adult worms, bright red in colour and up to 2.5 cm in length, are easily
recognised…” (BNC)
(25) “The bright red [Epithet] 2.5 cm [Descriptor] adult [Classifier] worms are
easily recognised…”
The explicit phrases in (24), which show that the attribute concepts are implicit in
(25), confirm the analysis, given in earlier sections, that the attribute-value relation
constitutes the bond of most premodifiers. It also confirms the qualia analysis of Clas-
sifiers, since ‘2.5 cm’, which is part of a postmodifier in (23), and a Descriptor in (24),
can readily be a Classifier, as in (26).
(26) “Fossilised [Descriptor] tertiary-era [Classifier] 2.5 cm [Classifier] worms”

(invented, but based on (25)).
Significance of the structural order. Postmodifiers are ordered on quite different princi-
ples from premodifiers, the order being grammatically free, but having certain stylis-
tic constraints (Quirk et al. 1985, §17.61–17.64). There is no semantic significance in
the order, as there is with premodifiers: reversing postmodifiers makes no difference
to the meaning. “An American smart bomb” could be rendered with postmodifiers as
“a bomb of the ‘smart’ type, made by America” or as “a bomb made by America, of the
‘smart’ type”. Reversing premodifiers does make a difference: “a Nigerian single par-
ent” is not synonymous with “a single Nigerian parent” (Those are invented phrases
based on the example in Table 22.) Postmodifiers’ variation in order thus concerns
information semantics, rather than structural semantics, and will be dealt with in
Chapter 10.
8.2.8 Heads
The identity of the head of Entity groups, for our purpose, is not straightforward.
Heads are usually defined morphosyntactically, as the word that determines concord
and determines “distribution” of the phrase and its function in the sentence, and so
on. However, we must distinguish between the syntactic head and the semantic head
of an Entity group, for two reasons. The first is that there are several situations in
which syntax and semantics diverge. Ellipsis may eliminate the syntactic head, as in
“one blue eye and one brown”– ‘brown eye,’ that is – but the ellipsed head is present
semantically. With adjectival heads, as in “We should give to the poor,” we again cog-
nitively understand a semantic head, such as ‘people’. Rankshifted phrases and clauses
acting as Entity groups – as Subject or Complement – have no syntactic head either,
and have no syntactic modifier-head structure. Second, even when there is a syntactic
head which has full semantic value, the syntactic and semantic functions in the group
may centre on different words. Quirk et al. (1985, §5.23) instance expressions like “a
lot of time” and “lot of money, ” where lot is the head syntactically but not seman-
tically. Halliday (2014, §6.2.6) says that, in “a cup of tea,” cup is the syntactic head,
but ‘tea’ the semantic head, normally. Similarly, in “He chose a big one,” and, with
marked stress, in “He chose the big jacket,” the semantic head is the modifier, ‘big’. In
some groups, there is no semantic head although there is a syntactic head, because
the whole group is semantically empty. That happens in presentative sentences with
dummy Subjects, like “There is a man at the door,” and “It is raining.” The semantic
head, then, is the semantic unit in the group to which other groups or figures relate.
The concept of Entity-group head developed here is much the same as that
of “Thing” in Systemic Functional Grammar (Fawcett 2000; Halliday 2014); but
Fawcett’s account lacks detail of the internal structure and relationship with modifi-
ers, and Halliday’s is very sketchy.

Descriptive meaning. We have seen in the preceding sections that Entity-group heads
generally act as Entities, providing an attribute for which a modifier provides the val-
ue. Heads relate directly in that way to Epithets like ingenious and extensive, and De-
scriptors like purple and square. They relate indirectly to Classifiers like “American
gun laws”, where the Entity denoted by the Classifier combines with a grammatical
relation to form the Property which constitutes the value.
A second structure occurs with Reinforcers, where the head provides a gradable
quality which the modifier reinforces. That quality may be the whole of the head’s
sense, as with “utter desperation”, and it may be part of the sense, as with “utter fool”
(‘foolish’ + ‘person’). A third syntagmatic structure, still with an Entity sense in the
head, occurs with Event Descriptors, as in “running water” and “broken glass”, where
the head is Actor in the Event (“running water”), or Undergoer (“broken glass”); the
aspect of the Event is specified by the -ing or -en suffix of the participle.
We have seen, however, that the head sometimes provides an Event, for which a
modifier can provide the Actor or Undergoer or Circumstance. In “hurricane proper-
ty damage”, ‘damage’ acts as an Entity for the structure of the whole figure, and could
be modified as an Entity by words such as extensive; but it also acts as an Event, with
‘hurricanes’ as Actor, and ‘property’ as Undergoer: ‘hurricanes damage(d) property’.
Similarly, “We offer overnight on-site explosives storage” (‘ We will store your explo-
sives overnight on-site’) has ‘overnight’ and ‘on-site’ as Circumstances of the Event
‘store’. Finally, the head may represent a Property, with the modifier representing the
Entity which carries it: in “public risk aversion”, ‘aversion’, denoted by the head, is a
Property of ‘public’, denoted by a modifier.
With phrases like “a big mouse” and “a small elephant” (where big and small are
“relative modifiers”), there is some reciprocity in the syntagmatic structure of modifi-
er and head. In “small elephant”, ‘elephant’ partially determines the meaning of ‘small’,
just as ‘small’ partially determines the meaning of ‘elephant’.
Nondescriptive meaning. The syntagmatic relationship between the head and nonde-
scriptive meaning is much looser, as noted in §8.2.4.3, on Epithets: affective and atti-
tudinal meanings are simply about the head’s referent.

In terms of information, modifiers add to the head; but in terms of semantic struc-
ture, modifiers change the head. They specify dimensions that are not fully specified
in the meaning that the word brings to the utterance from the lexicon. For example,
“extreme heat” specifies the intensity dimension of ‘heat’; “mathematical circle” and
“rough circle” make vague ‘circle’ specific, in different ways; “civic buildings” shifts
‘buildings’ on the generality dimension. “On your right” and “her left profile” specify
the point of view dimension. Determiners specify whether the head’s reference is defi-
nite, specific, generic and so on.
In part, however, the Entity as modified does not change its lexicon sense, but
selects from its network of sense elements, or emphasises them. For example, giggle
has both favourable and unfavourable affective meanings as possible elements; ‘un-
favourable’ is selected in “silly, little-girl giggle” (COCA), but ‘favourable’ is selected
in “merry little giggle” (COCA). Similar affective meanings in a modifier and the
head reinforce each other, as with “ugly old brute” (COCA). Social meanings may
be confirmed or reinforced similarly, as with “real cool dude” (COCA), where “real”
and “cool” have various possible interpretations, which are controlled by “dude”. The
head’s structure on the expectedness dimension is also affected; ‘horse’ will have the
likely elements ‘for riding’ and ‘for pulling’ differently selected in different contexts;
when a film was made about “Ed, the talking horse”, a very improbable element for
‘horse’ was activated.
In principle, the head’s internal structure controls the nature of its modification:
time-duration modifiers are possible only if the head include an event element, for
example. But we have great freedom in construing the semantic nature of words, so
the principle is not very powerful; “a ramble” clearly has an Entity – but we can say
“a three-hour ramble”, where the time-duration modifier construes an event element
within the Entity.
English allows speakers to choose whether collective senses like ‘committee’ will
be singular or plural morphosyntactically – “the committee is…”, or “the committee
are …”; organisational unity and membership plurality are both possible elements of
the sense, and the speaker may choose which will be salient, marking the salience by
the form of the Predicator.
8.2.8.4 Semantic class

In Chapter 4, §4.6, Entity, Event and Property were introduced as alternative semantic
classes, but in this chapter the same word has been assigned to two of those classes at
once, in nominalisations of Event and Properties. The resolution of the contradiction
lies in the hierarchic internal structure of heads of Entity groups. In “The estimated
waste generation is significantly higher ….” (COCA), for example, generation is an
Entity at the top level, that of group syntax, which needs an Entity for the determiner
(the) and the modifier estimated. At a lower level, generation has an Event element as a
constituent, namely ‘generate’, which has ‘waste’ as Undergoer. The linguistic reality of
the subordinate Event is evident in two ways. First, it controls the syntax of Classifiers;
we would say “the estimated municipal waste generation”, not “the estimated waste
municipal generation”. Second, it controls the participial inflection of Descriptors, as
noted previously.
Entity, as a semantic class, is the word’s role in the Entity group, just as Participant
as semantic class is the group’s role in the figure. The difference is precisely that differ-
ence in role, with the different relationship and linking elements involved. The Events
we are discussing differ from the basic class, event, in the same way: they have the role
of group head, and relationships with determiner, premodifiers and so on; events are
unspecified in those ways.
I have so far simplified the discussion by restricting it to common nouns as head.
What I have said applies to Entities represented by pronouns and proper nouns, and
by Entity groups and clauses – all can act as semantic heads of Entity groups. The
range of forms demonstrates that being a semantic head is a matter of (semantic)
function, rather than of cognitive content.
Entities are specified enough to have subclasses, which are familiar to us in their
realised form of proper nouns, common nouns, mass nouns, and abstract nouns.
Wierzbicka (1996, Chapter 13, §3) would add “dual nouns”, to allow for nouns de-
noting pluralia tantum such as trousers and scissors. The nature of the subclasses lies
in their position on the boundedness dimension – their degree of individuation (i.e.
degree of boundedness) – boundedness in space, whereas Events are bounded in
time. Abstract Entities are not bounded in any way. Quantifiable Entities, such as
‘water’ ‘petrol’ and ‘air’, are bounded in being spatial, and having mass – mass nouns.
Countable Entities are bounded by being spatial, like quantifiable Entities, and by
representing an outline, which makes them countable – count nouns. Unique Entities
are still more bounded – proper nouns, generally. The class of bounded Entities has
the subclasses animate and inanimate, or “conscious” and “non-conscious” as some
linguists prefer to call them. The main relevance of the distinction is in its control of
choice between he, she and its. Gender is thus included in the concept of animacy.
Uniqueness, outline and mass could be thought of as sub-dimensions of the spa-
tial boundedness dimension; but the concept is rather clumsy. Rijkhoff (2002) calls
having an outline “Shape”, and calls having mass but not outline “Heterogeneity”. They
constitute “Seinsarten”; the concept is useful, to match Aktionsarten. (Rijkhoff ’s ac-
count does not extend to abstractions, unfortunately, or to sublexical meanings.)
The boundedness of the subclasses controls what modifying elements they can
take. Countable and quantifiable Entities take different determiners and types of pre-
modifier because the difference in the nature of their boundedness controls the attrib-
utes they have. That corresponds with the explanation given by Schreuder and Flores
d’Arcais (1989): their “perceptual” nodes are those with space and time bounds; their
functional nodes are free of those bounds; unique Entities activate only the central
node of the meaning, going straight on to activate purely cognitive nodes.
8.2.8.5 Discussion: Entity-group heads

Effects of abstraction. The process of cognitive abstraction, discussed above, can go no
further than to producing abstract content; but linguistic abstraction does go further.
In the process of abstraction referred to as “grammaticalisation”, concrete and then
abstract elements of content are extracted, and then content itself is abstracted away,
leaving, in one form of grammaticalisation, a semantically empty shell word that acts
as a “dummy Subject”, as in “It is raining”, and leaving, in a different form, “grammat-
ical items”, as for example with the noun spite in the compound preposition, “in spite
of ”, and perhaps with Topic as a semantically empty discourse Entity.
Heads with two syntagmatic relations. We have seen that heads can have two syntag-
matic relations, as in “hurricane property damage” (§8.2.8.2). That structure is com-
mon: all “nominalisations” have it, and so do a number of other senses. That analysis
resolves several puzzles in the literature.
Quirk et al. (1985, §17.54), like other works, puzzles over gradience between
“nouns,” “gerunds,” “participles” and suchlike, acting as nominal heads. The puzzle
arises partly from reliance on the traditional concepts of parts of speech, which con-
flate morphology, syntax and semantics, but also from failure to understand this dual-
ity in many heads. Pustejovsky (1995) identifies “dot objects” as a distinct class, which
Cruse (2011, §6.5.1) refers to as meanings having “facets”. ‘Book’ is an example, in that
it can be treated simultaneously as a physical object and as the content of the book.
Another needless class is that of “non-inherent adjectives” (Quirk et al. 1985, §7.39),
as in “an old friend” and “…elected the wrong candidate”. Again, the head is simply
working with two syntagmatic relations at once. These phenomena should not be seen
as indicating classes, but as properties of uses or senses, or as relations between them.
Another instance is that of phrases like the much discussed “a beautiful dancer”.
To Huddleston and Pullum (2002, p. 557), ‘dancer’ there is a “process-oriented attrib-
utive”; Warren (1984, p. 239) analyses “hard worker” as “[hard work]-[er]”; and Quirk
et al. (1985, §7.73) compare them to a verb-adverb construction. Such Entity heads
seem to be constructed semantically as they are constructed morphologically – ‘a per-
son (who)…’, signified by -er, + a qualifying element, ‘(who) dances’, ‘(who) works’,
and so on. Another example is “successful historical novelist”, where ‘historical’ mod-
ifies ‘novel’, and ‘successful’ modifies ‘-ist’, as it were. The distinction between modi-
fication of the “referent” and of the “reference” is similar (Bolinger, 1967; Functional
Discourse Grammar as in van de Velde 2007, p. 206): in “a poor doctor” for example,
the “referent” is the person, and the “reference” is the qualifying element, skill as a
doctor. Still more unsatisfactory distinctions include the distinction between “indi-
vidual level” and “stage level” modifiers (Maienborn 2011, for example), and between
subsective and standard or “intersective” modification (Partee 2010; Murphy 1988,
p. 531ff., for example).
The principle of dual syntagmatic relations, finally, explains “active zones” in
Cognitive Grammar, which is intended to account for the varying way in which modi-
fiers relate to Entity heads. In the account given by Taylor (2002), the concept is vague,
needing identification of the various “active zones”, and an explanation of their nature,
which I believe to be properties of the real-world referents. I suggest that my account
is to be preferred. (See also Langacker 1999, p. 62.)
8.2.8.6 Conclusion: Entity-group heads

The parallel between Entities as a semantic class and Participants as a semantic class
leads to an important general point about semantic classes. They can usefully be rep-
resented as forming a hierarchy. For what in traditional terms would be “nominal
classes”, the highest level is that of Participant (discussed in Chapter 7). It consists of
an Entity, from the level below, and its bond – the element of its structure (such as
AGENCY) which is the term of its relationship to the rest of the figure (such as transi-
tivity). That relationship constitutes its role in the figure (such as Actor or Undergoer).
An Entity in turn consists of a sublexical entity, and its bond – an element (such as
an attribute) that can relate to an element (such as a value) in other senses (acting as
modifiers, for instance).
That explanation is given in terms of classes, as constituents, since those terms are
used so often in linguistics. An explanation in terms of relations is preferable in this
account where meaning – and the mind in general – is regarded as a network. That
explanation goes as follows. The parallel between the Entity role and the Participant
role leads to an important point about their relationship: it is hierarchic. The Partic-
ipant role (discussed in Chapter 7) transcends the Entity role by adding a role in a
figure, which consists of a relationship to a Process structure (such as the transitive
material-Process-Undergoer structure); the relationship holds between such elements
as AGENCY in an Actor, and CHANGE in the process.
Those remarks apply to the relations of Event classes, to be discussed in §8.3, and
the relations of Properties classes, to be discussed in Section §8.5.
8.2.9 Entity group as a unit
The semantic structure of Entity groups, as we have seen it in the preceding sections,
is even more complex than the syntactic structure is, since single words often have a
complex semantic structure, and the semantic relations between words are sometimes
discrepant with their syntactic relations. This section examines how such a complex
assemblage functions as a semantic unit. The way it functions varies with whether its
purpose is descriptive – giving information for its own sake – or identificational –
identifying a referent, with any information given being secondary to that purpose,
and eliminated from the hearer linguistic meaning if not needed.
8.2.9.2 Structure of Entity groups used for identification of a referent

General structure. As noted in Chapter 7, in the discussion of Participants, groups
intended to identify a referent must be sufficiently determinate, that is sufficiently
specific to achieve unambiguous identification. For a generic statement, a bare noun
could be adequate, such as “composers.” For other statements, modifying elements
could be needed, as in “American composers” and “successful American composers”;
and a determiner may be needed, either in combination with modifiers or alone – “all
successful American composers” or “the composers”. The overall semantic structure

is then reflected in the syntactic structure: “all [successful [American composers]]”.
However, the semantic result of composing the group in accordance with that
bracketing is not simply an accumulation of the words’ meanings: the speaker’s in-
tended result is that the hearer identifies the referent, not that the speaker absorbs sev-
eral descriptive concepts. I begin the explanation of that with example (27), in which
a woman describing her travel in Provence sets the scene, and begins the narrative.
(27) “I arrive on a blistering Saturday afternoon on the eve of Bastille Day.”

(New Zealand Herald, October 8 2013, travel section, page 3)
The Entity group “I arrive…” is purely referential; it is purely functional, with no con-
tent of its own.
Acronyms work similarly, in part. Example (28) is from a newspaper report on
treating hangovers.
(28) “Our livers release an enzyme called alcohol dehydrogenase (ADH) […] an
enzyme called aldehyde dehydrogenase (ALDH) […] Some drinks, including
a herbal tea made from hemp seed, increased the length of the ADH process,
and inhibited the ALDH process.” (cited in a New Zealand Herald
news report, from Telegraph Group Ltd of unspecified date)
The acronyms enabled the writer to identify chemicals in the last sentence without the
technical concepts given in the first sentences; like the pronoun “I”, they are purely
referential, simply directing the reader to the reference, without content, which the
acronym suppresses. The group, “the eve of Bastille Day,” from example (27), similarly
acts referentially; it has eve as the focus of reference; and a determiner and a postmod-
ifier to narrow down the reference by calling on the reader’s cognitive knowledge of
French holidays; but those modifying elements are wholly subordinate.
The meaning of a restrictive Entity group as a whole is ultimately simply the iden-
tity of its referent, which relates the whole figure to the “mental model” which the pas-
sage is building up in the reader’s mind. There may be no content, as in “I arrive…”;
there may be content which the reader must bring to mind, as in “the eve of Bastille
Day”; and there may be conceptual meaning available if the reader knows it and wants
to recall it, as with the personality and background of “I”, events in 1789 for “Bastille
Day” and the constituents of ADH. But when the whole group is composed into the
figure, any content meaning falls away, and the significance of that identificational
group will be reduced to its referential value, the identity of the referent.
Words such as Saturday, as in example (27), are names, like proper names such as
Bastille Day, again in (27); they bring the referent to mind directly, without conceptual
meaning being needed. In that sense, groups such as “the ADH process” in (28) are
names. In their role in the figure, they have identificational meaning, in contrast to
their meaning in the mental dictionary; any conceptual meaning the reader may use
in reaching it is not the final meaning, as just explained.
There is support for this account in the linguistic literature. Mel’cuk (2001, §7)
sees the construal of something as unitary or “articulated” as a fundamental issue
in language, applying to lexicalisation, syntactic constructions and so on. There is
also support from psycholinguistics. Schreuder and Flores D’Arcais (1989), as not-
ed previously in Chapter 5 (§5.2.3.1), describe meanings as having a central node,
which is independent of the perceptual and functional descriptive elements; that is,
it is without content. When we hear or read language, that node is activated first, by
the word with which it is associated; when we speak, it is activated after we formu-
late the meaning, and leads to the morphological form that represents it. As Flores
D’Arcais and Schreuder (1987) emphasise, in discussing alternative routes through
the conceptual network, the central node can be activated without the outer content
nodes; since it is presumably linked to experience of the object represented, it would
function as the identificational meaning being proposed here. McClelland, Rumelhart
and Hinton (1986) describe featureless “instance units” denoting an entity; the units
include content (i.e. features) that requires linkage to neighbouring nodes, but that
can be inhibited.
Restrictive Entity groups, then, function as units by narrowing down their mean-
ing to a minimum identificational meaning – the traditional “act of reference”. They
work on what may be called the principle of minimal meaning: the hearer or reader
should retain from the text just enough to achieve the purpose of the moment.
Like many other semantic structures in English, this structure can be altered and
even reversed in marked use. Proper nouns, like Melissa, are usually purely referential,
but when a new office worker identified herself to a client as “I’m the new Melissa,”
the marked use invoked descriptive elements like ‘person who answers the phone and
keeps records’.
The determinateness of an Entity group is thus complex, involving the semantic
subclass of the head (e.g. countable Entity, mass Entity), premodification, determina-
tion, and the function of the group.
Order. The order of Classifiers (§8.2.2) is a set cognitively or “logically”, as a classifying

order. For example, in “Detz .44 magnum hunting rifle”, the most general concept is
immediately next to the head, with progressively more specific concepts progressively
further away from it, to narrow down the reference. That does not apply to the or-
der of the zones, however. In “the global [Descriptor] financial [Classifier] crisis”, the
crisis is identified by the presence of the two properties (‘financial’ and ‘global’), irre-
spective of the order in which they are mentioned: the reference would be the same in
“the financial global crisis”. In the referential use, then, the set order of premodifiers is
syntactic rather than semantic.
8.2.9.3 Structure of Entity groups used descriptively

General structure. Entity groups used descriptively contrast with those used for iden-
tification of a referent. That can be seen from example (29), which ran straight on
from example (27) in its context.
(29) “It is the busiest weekend of the year, and the village’s narrow Roman streets
heave with visitors and merchants selling everything from nougat and olive
oil to fur coats, vinyl and Virgin Mary statues.”
(New Zealand Herald, October 8, 2013, travel section, page 3)
The Entity group “everything from nougat…” serves the descriptive function char-
acteristic of travel writing; but even the definite groups, which appear to be identi-
ficational, serve that function. In “the busiest weekend of the year,” we know from
the previous sentence that the day is Saturday; and, for “the village’s narrow Roman
streets,” it is a given that villages have streets; weekend and streets are semantic heads
structurally, but valueless informationally. In those groups, the syntactic head is func-
tionally insignificant, and it is the modifying items which the reader is to retain, as if
they were heads; the groups are turned inside out, as it were.
As a further contrast, words are intended to evoke many of their possible mean-
ing elements, not only their necessary ones. “Olive oil” and “Virgin Mary” evoke the
lifestyle of southern France, with associations such as being traditional, rural, simple
and Roman Catholic; “nougat,” “fur coats” and “vinyl” evoke a contrasting lifestyle;
“narrow” and “heave” serve to evoke mental imagery.
Descriptive groups thus do not conform to the principle of minimal meaning,
but to its converse principle of maximal meaning: the hearer or reader should develop
from the text all likely meaning, and retain as much information as possible.
The overall structure of descriptive groups is syntactically the same as for identi-
fying groups: “the… [narrow [Roman streets]”, for example.
Order. The order of modifiers in descriptive use is set quite differently from the order
in referential use. It will be useful to consider the specific issues and the general issues
in turn.
The specific issues have been touched on already, in Chapter 5 (§5.5.2.1), where
it was shown that, in “brave young man” and “brave old man”, brave is linked to ele-
ments within young and old, as well as to man. Similarly, in “a good thick slice”, good
expresses approval of the thickness, and intensifies it. Consideration of example (30)
extends the illustration to all three descriptive zones.
(30) “Lush [Epithet] golden [Descriptor] presidential [Classifier] draperies”

(COCA; the draperies were for President Clinton)
The phrase is intended to be evocative, with possible meanings invoked fully; the
senses will interact, like those of brave and good just cited. The draperies are thus not
only ‘lush’ and ‘golden’, but ‘lushly golden’; preceding the Descriptor golden enables
the Epithet lush to modify it. Golden not only modifies draperies, but links with pres-
idential, suggesting the status and power of being president; that is possible because
the Descriptor golden precedes the Classifier presidential.
Generalising from those specifics, we can say that Epithets come first among these
three zones, so that as abstract, general, emotive and attitudinal senses they can modi-
fy more concrete (but still descriptive) Descriptors. Descriptors precede Classifiers, so
that as descriptive senses they can modify referential Classifiers. Reinforcers precede
all three, so that, on the relatively rare occasions when they occur with descriptive
modifiers, they can reinforce them all. For instance, in “a mere 390,000 live TV au-
dience on Sky” (BNC), cited above (§8.2.5.3), the mocking sense of ‘mere’ applies to
‘390,000’, ‘live’ and ‘tv’ (and, strikingly, not to its syntactic head, audience). The order
of premodifiers in descriptive use is set semantically, by their meaning-type structure.
8.2.9.4 How grammatical meanings build the structure of Entity groups

In building the syntagmatic structure of the phrase, Classifiers operate by requiring
the general grammatical procedure of adding the content to the meaning of the head.
That procedure is made more specific for each Classifier by the construction in which
it occurs (§8.2.2 above), which contributes the quale as a concept to be used in relating
the Classifier to its head. So for “… electronic pressure gauge”, the specific grammati-
cal meanings are “Add ‘electronic’ to ‘gauge’, as specifying its type”, and “Add ‘pressure’
to ‘gauge’, as specifying its function”. That grammatical meaning is signalled here by
the fact that the modifiers are Entities by semantic class, and by word order – in the
qualia construction, the first Classifier signals origin, the second signals dimension,
and so on.
So far, Descriptors have been presented as simply operating on the head: gram-
matically, “Add this word’s content to that of the head.” That is a slight simplification.
Strictly, it is “Add this to the following parts of the phrase.” That is shown in the stand-
ard syntactic analysis (compare §8.2.9.2): determiner [Descriptor [Classifier head]].
The Descriptor may also interact with the Classifier; when we read “black iron fences,”
we are likely to conceptualise the iron as black, the fence being made of black iron.
(Compare the section on Entity group on networks, Chapter 5, §5.5.2.1.) Such inter-
action is very limited with Descriptors, however.
Epithets’ grammatical meaning usually links the content to the complex of con-
cepts built up by the rest of the group, as with Descriptors; but as we saw (again in
Chapter 5, §5.5.2.1), it may develop wider links, when affective meaning applies to
the situation rather than the head, as with “I just can’t sell the bloody things,” and
hearers overrule the grammatical meaning in favour of a cognitive understanding of
the relationships.
Reinforcers’ grammatical meaning does not add to the head’s meaning, but ad-
justs its position on the intensity dimension, as in “by sheer guts”. Determiners’ gram-
matical meaning specifies the type of reference the group is to have.
8.2.9.5 Discussion: Entity groups as units

Unity. Restrictive groups are very tightly unified. All elements serve the identifica-
tional meaning, and the senses other than that of the head do not become part of the
meaning of the whole text. Descriptive groups, by contrast, are not unified seman-
tically at all, although they are unified syntactically. “Thanks to my rusty schoolgirl
French” could equally well be stated as “thanks to my French, which is rusty, and of
the schoolgirl variety,” breaking up one group into three; being one group is a result
Table 23. Relationship of Classifiers in “your reading, writing and number skills”
Predicator Complement
Determiner Classifier Head
reading
Improve your writing skills
number
of stylistic choice, not of semantic appropriateness. Such groups are units in the infor-
mation structure, in denoting Participants, but they are not units semantically.
Co-ordination and linearity. The nature of coordination as breaching linearity by re-

quiring the hearer to double back to the previous semantic and syntactic position
was set out in Chapter 7 (§7.8); it was illustrated with “Alistair Brown, the CEO of the
company, has cut 150 jobs”, where the appositional phrase duplicates the role of the
Actor and Subject, “Alistair Brown”.
Coordinate pre-modifiers have the same effect. For example, in “Improve your
reading, writing and number skills,” the three modifiers occupy the same slot, as Clas-
sifiers. The structure is clear when the sentence is set out as in Table 23, which illus-
trates the fact that the three words modify skills equally and simultaneously, not in
sequence.
The function of the comma and of and here (and pauses in speech) is to overcome
the misleading nature of that sequential flow, warning the hearer that the next item in
the flow is not the next structural item i.e. the head.
Compositionality. Entity groups consisting of two or more senses are compositional

in that those senses are, loosely speaking, constituents, and are combined (to an im-
portant extent) by well-defined processes. Moreover, the nature of the processes is
defined by the grammatical meanings, and from the nature of the zones in which they
occur. For example, in “the village’s narrow Roman streets”, Roman as a quale Classi-
fier signifies that ‘Roman’ specifies the origin of the streets; and narrow as Descriptor
signifies that ‘narrow’ is to be built in to the total meaning as a nongradable Property
which is the value of an attribute of the head. Indeed, the grammatical meanings and
the significance of the zones, as I have described them in previous sections, provide
for a complexity, precision and subtlety of compositionality that goes beyond any-
thing I have seen described in the compositionality literature.
On the other hand, there are several important ways in which Entity groups are
not fully compositional. In the apparently straightforward group, “the village’s narrow
Roman streets”, the village’s signifies the location of the streets, but the standard signif-
icance of genitive determiners is possession; this idiomatic use requires reconstrual of
the compositional meaning, from possession to location. We have seen in this chapter
and Chapter 5, on networks, a number of ways in which senses do not combine by a
regular process. Two clear examples in Entity groups are use of expletives like bloody,
which relate to a general situation and not to the word they modify syntactically, and
phrases like “Our happy feet”, cited in Chapter 2 (§2.4.2), where ‘happy’ modifies the
referent of our, and “vacant lots baked hard in the valley sun”, where ‘valley’ has no
definite semantic head. More fundamentally: in restrictive use, the meaning that is
used in the rest of the figure is the identificational meaning, which is a product of
the groups’ sense elements, not a composition of them. For example, with “Some
drinks… increased the length of the ADH process”, ADH helps the to identify the
process; it does not give new information which is combined into the final meaning.
Compare Fretheim (2011, p. 133): “When the description has done its job as indicator
of how… to identify the semantic value of the referring expression, then it is replaced
by a referent which does not contribute to truth-conditional content.”
When Entity groups are used descriptively, the modifiers are not combined into
the meaning of the text straightforwardly, either. As noted above, they do not enter
the text as compositional units of the group, but as separate items of information.
“The village’s… streets” identifies the streets; narrow and Roman assert new informa-
tion; they are equivalent to “the village’s streets are narrow; they date from Roman
times”; there is no unified compositional structure carried into the text.
8.2.9.6 Conclusion: Entity groups as units

The broad semantic structure of Entity groups is as follows: determination [premod-
ification [reference postmodification]]. The structure of determination is: quantifi-
cation [deixis [quantification…]]. Put technically, the structure of premodification
is: Reinforcers [Epithets [Descriptors [Classifiers [head]]]]. Put descriptively, it is:
intensification [abstract, gradable description (and affective modification) [concrete,
non-gradable description [indirect description [head]]]].
There is also a structure on the linguistic-cognitive scale. Classifiers are wholly
cognitive, being a rendering of a cognitive concept. Descriptors are strongly cognitive.
Epithets are partly cognitive and partly linguistic, being conceptual but with a combi-
nation of meaning types unique to language. Reinforcers and determiners are wholly
linguistic, having only grammatical meaning.
8.2.10 Discussion of Entity groups
8.2.10.1 Construal of semantic class

The senses of the words in Entity groups function as specific semantic classes – Prop-
erty, Event or Entity. That class membership is construed, rather than inherent, and
construed according to the semantic function the word is to serve. The sublexical
sense, SPEED, may be represented in an Entity group as an Entity, in “its great speed”,
or as a Property in “a speedy retreat”, or as an Event subordinated within a Property
(see §8.2.8.4) in “a speeding car”.
8.2.10.2 Locus of grammatical meaning

The analysis of Entity groups has shown that we need to expand a little our under-
standing of what can carry grammatical meaning in English. Grammatical items carry
it, obviously; but we see in Entity groups that some content words carry a grammatical
meaning by default. For example, senses such as ‘dangerous’ are inherently gradable,
so default to being Epithets, carrying the grammatical meaning, “Add this to the head
as a gradable Property”; senses that include an affective element also default to being
Epithets, since we project our feelings onto things – “an interesting book” is one we
feel interested in – making the feeling a property of the thing.
That default grammatical meaning of content words can be seen also when there
is insufficient context to specify the grammatical meaning, as in interjections and
texts which consist of jottings. For example, if lecture notes read “Plateau – high –
dry…”, the words would default to their commonest use – plateau as head, and high
and dry as modifiers; cognitive meaning and the context would confirm or falsify
that interpretation, and show whether high and dry should be taken as Epithets or
Descriptors.
We know that syntactic position on its own carries grammatical meaning, but
that it sometimes needs combination with other markers, as when the significance
of those is carried by the combination of its deictic sense and its position (“Those
are…” versus “Those books are”). The zones in Entity groups take that principle a
stage further; each zone carrying a grammatical meaning through its position. In “old
forbidden books”, old is in the Epithet zone, so the meaning is, “Add this as a gradable
Property”; in “curious old writers” it is in the Descriptor zone, so the meaning is, “Add
this as a nongradable Property”. The zones rather than the individual words, carry
the grammatical meaning; any default grammatical meaning the word may have is
overridden or confirmed by the construction. The same point applies to the Classifier
subzones, and the subzones of the determiner zone – predeterminer, central deter-
miner and postdeterminer (§8.2.6.1 above; Quirk et al. 1985, §5.11–5.14). They could
be described as constructions, consisting of semantic class and position, interacting.
8.2.10.3 Scope relationship of premodification and postmodification

I have not dealt so far with the relation between premodification and postmodication,
when both occur in one group – that is, the issue of which one applies its meaning
to the other. That is because there is no set pattern, and because there is no separate
principle at stake. However, I will illustrate the relationship here. In restrictive Entity
groups, premodification may take scope over postmodification, as in example (31),
from the same article on travel in Provence quoted from earlier.
(31) “[I googled] French language courses in Provence…”

The author has already indicated that she wanted a course to improve her French,
and is now indicating that she wanted it to be in Provence, so “in Provence” restricts
“French language courses.” The structure is “[French language courses] in Provence,”

with post modification taking scope over premodification.
The relationship is reversed in example (32), from the same article.
(32) “[I had] a chance to ditch my embarrassing mix of pigeon and sign language.”
[“Pigeon” = “pidgin”]
Embarrassing is given as a Property modifying the mix of pidgin and sign language;
combining pidgin and sign language is not a Property of the embarrassing mix. The
structure is “my [embarrassing [mix of pigeon and sign language]],” with premodifi-
cation taking scope over postmodification.
8.2.10.4 Modals
Some words in premodifier position, such as alleged and fake, are not subject to the
meaning, “Add this content to the meaning of the head word,” as most premodifiers
are; they do not modify the sense of the head, but modify the word’s use. They are thus
modal words; they have been discussed in the literature under various terms, such
as “negational descriptors” (Cruse 2011, §15.6.3), and “privative adjectives” (Partee
2010). They have not been discussed so far, because they do not fit the zone pattern.
They can occur anywhere in the premodifier order, being placed immediately before
what they modify; they are not affected by zone – they do not change meaning with
a change in position, and do not have the semantic characteristics of the zone they
appear to be in. Table 24 illustrates the variation in position, with former apparently
in Descriptor and Classifier positions.
The modals given so far, and others such as probable, suspected, former, replica,
and then (as in “the then president”) are simple in their internal structure. They do not
carry affective, attitudinal or social meaning; and their descriptive meaning does not
provide possible elements that may be invoked or inhibited.
Their semantic relation with other words in the Entity group also differs from that
of other premodifiers. First, the nature of the modification is different. In “his British
former wife,” former does not give a Property of the woman referred to – as British
does – but a Property of ascribing the term wife to her. Second, the scope of modifi-
cation can be quite different. In “apparent high and low cost prescribers” (referring to
doctors prescribing medicine), the scope of apparent does not include the head: the
Table 24. Position of former as modal

Determiner Epithet Descriptor Classifier position Head
position position position
the American former world champion
the former American president
doctors concerned are real prescribers, not apparent ones; the scope is only ‘high and
low cost’.
There are several types. With bogus, imitation, reproduction, and fake, the pre-
modifier signifies that the head should be taken referentially, with its descriptive
meaning disallowed; thus with “He bought some fake whisky”, “whisky” is to refer to
some real liquid, but its descriptive elements do not apply – they are like epistemic
auxiliary verbs. With suspected, alleged, reported, self-styled and so on, the modifier is
an “evidential”, signalling that the speaker is not confident of the validity of the head’s
descriptive meaning; the speaker of “an alleged fool” is not sure whether the referent
is foolish in fact. With former, present, then and ex (as prefix or separate word), the
modifier signals that the descriptive meaning of the head applies at the specified time.
We conclude that modal premodifiers are grammatical items, carrying not only a
grammatical meaning but also a conceptual element, such as that of evidential value
or time relevance. They thus resemble most uses of prepositions.
8.2.10.5 Free order within a zone

As noted in the preceding sections, the order of premodifiers in different zones is set;
but the order of premodifiers within one zone is not. In “He is a rude, horrible, selfish
man,” for example, the three Epithets may be put in any order. The same applies to
words within the same Classifier subzone. In “brick and tile three-bedroom, single
level townhouse,” three-bedroom and single level, which both occupy the Type sub-
zone (§8.2.2.2), maybe reordered to make “…single level, three-bedroom townhouse.”
There are, however, some stylistic constraints on the grammatically free order. A con-
ventional order may develop, as with “ brick and tile” – not “ tile and brick”, and the
speaker may feel that the communicative intention requires a climactic order, as with
the Epithets in example (33).
(33) [A fictional character yearned…] “to appear a strong silent capable man,
unmoved by emotion”. (BNC)
In unmarked use, the repeated Epithets would need commas, and would apply to
“man” independently and equally. Here, the absence of commas and the relative
strength of the senses make ‘capable’ a climactic culmination of the modification, tak-
ing up ‘strong’ and ‘silent’ into its own extended meaning.
For more detail, see Feist (2012a, Chapter 6).
8.2.10.6 Marked order across zones

The order of group elements is like the order of clause elements in being grammat-
ically set; but in both cases the order is an unmarked order, which can be varied as
a marked order, with certain consequences – “He is ready” can be varied to “Is he
ready?” with a syntactic and communicative consequence, and “The rain came down”
can be varied to “Down came the rain,” with a semantic or stylistic consequence. Sim-
ilarly, the position of premodifiers set out above is the unmarked order, and may be
varied for a similar purpose.
Consider example (34).
(34) [John Major, a British prime minister, was advised…] “in his hour of crisis to
maintain a typically British stiff upper lip.” (BNC)
Modification by typically shows that British is intended to have an Epithet sense, al-
though SOED gives it only a Classifier sense, “Of or pertaining to Great Britain or its
inhabitants”. The sense must therefore be taken as an abstract, gradable quality, such
as ‘reserved’, here carrying also an equivocal attitudinal meaning. The word has been
given a marked position, forcing the hearer to give it a new sense. (SOED gives French
an Epithet sense, which was evidently established through a similar marked use.)
Example (35) gives a less familiar type of marked use.
(35) [The lithium-air battery is] “picked as one of the most technical advances in
2010.” (Gliding Kiwi, March 2011, page 53)
Technical, there, evidently means something like ‘highly skilled’, which is an Epithet
sense; but the nearest established sense of technical is “pertaining to… the applied arts
and sciences” (SOED <2>), a Classifier sense.
The marked position in those examples is marked by the submodifier; in exam-
ple (36), the marked use of recidivist is marked by its coordination with senses that
default to Epithet use. The example describes a candidate for drug rehabilitation; re-
cidivist keeps its dictionary meaning, but seems to add the colloquial, disapproving
sense, ‘hopeless’.
(36) “A prolific, recidivist, versatile offender”

(New Zealand Herald, 12th March 2014, A12)
The film, Noah, a $150 million “blockbuster”, was described by its director as in exam-
ple (37); the first biblical is an Epithet, and the second a Classifier.
(37) “The least biblical biblical film ever made.”

(New Zealand Herald, 13th March 2014, A22)
The examples so far have had a Classifier moving into Epithet position. When an Epi-
thet or Descriptor is moved into Classifier position, it loses its descriptive meaning. In
an example used in §8.2.2.5, “Texas Deceptive Trade Practices Consumer Protection
Act”, deceptive no longer has its broad evocative meaning, but has a narrow technical
meaning (as defined in the act).
The marked use may entail changing the semantic class of the sense. ‘Dismaying’
usually has an Event sense; it is converted to a non-gradable Property sense in ex-
pressions like “with startling, dismaying speed” (COCA; Descriptor use), and gains a
gradable Property sense in “The most dismaying aspect of al-Qaeda’s revival…” (Epi-
thet; The Economist, September 28th 2013, p. 12). (The event element remains, subor-
dinated, in these uses, as it does in nominalisations.)
Like marked order of clause elements and marked stress, marked order of pre-
modifiers is a conventionally established procedure for conveying meaning; it breaks
a “rule,” but is itself a kind of rule. It demonstrates clearly the reality of the zones. It
has been a major, though unrecognised, source of new word meanings in English,
like the shift of words from modifier position to head position, in “a Caesarean”, “my
mobile”, “a television”, and so on. (For more detail on marked order, see Feist 2012a,
Chapter 7.)
8.2.10.7 Content relationships among the zones

Classifiers in the qualia construction (§8.2.2.2) can be characterised by the domain to
which their content belongs – origin in place or time, dimension, constituency, and
so on; yet many Descriptors and Epithets are also words of origin, dimension, constit-
uency and so on. For example, the Classifier 17th century, the Descriptor old and the
Epithet ancient all express time of origin. It is worth considering the explanation for
that apparent repetition and confusion, when Entity groups are so tightly structured.
The explanation is in part simply historical: words that were used in one zone
“moved” into a neighbouring zone as they acquired new senses. There is a deeper
explanation, however. The senses in a new zone have a different communicative func-
tion. Those further “forward” in the group are more descriptive, more expressive, or
more personal, being more abstract or having affective or attitudinal meaning (Re-
inforcers excepted). (Some would say those further forward are more “subjective” or
“intersubjective”.) They are more apt for the interpersonal metafunction. The relative-
ly few that moved “back” to the Classifier zone are more precise and referential, better
serving the ideational metafunction.
8.3 Process groups
8.3.1 Introduction
Process groups are semantic groups that have a Process as head; their chief use is as
Predicators. As noted in Chapter 7, they typically have Events as heads; they seem to
have begun historically as Event groups. Chapter 7 also showed that now the Process
groups in some Predicators have heads that are not Events – relational or stative sens-
es: the function of Predicators is to predicate, not to represent Events.
Process groups, then, can usefully be divided to three subclasses: Eventive, rela-
tional and stative Process groups. Groups with a copula could, from one point of view,
be regarded as a fourth subclass, since they function as Predicators; but semantically
they are not Processes, since they belong to no semantic class, having no content.
Some Process groups do not function as Predicators, but as rankshifted modifiers
of other groups, often being nonfinite. But this section will deal primarily with the
finite groups that serve as Predicators; the structure of nonfinite ones will be set out
incidentally.
As with Entity groups, Process groups may be expressed as one word or as many,
and they are structured syntactically as a head with optional dependants. For our
semantic purpose, the head is a content sense (is “lexical”), not an auxiliary’s sense. It
often fails the test of being able to stand on its own, as in “[They] are watching [televi-
sion]”; but that is a syntactic criterion, not a semantic one.
Where Entity groups have one type of grammatical modifying item (determiner),
and several types of content items modifying the head, Process groups have several
types of grammatical modifier, and only one type of content modifier. Where Enti-
ty groups frequently have subordinate groups as postmodifiers, Process groups have
only relatively infrequent particles and single-word modifiers as postmodifiers.
Trousdale (2010, Chapter 3) provides a useful three-fold classification of ap-
proaches to Processes (“verbs” to Trousdale). “Localist” approaches treat motion in
space as the central concept; other uses of verbs are metaphorical extensions. “Aspec-
tual” approaches treat the simplest type of verb (states) as basic; other types, such as
activities and achievements, are more complex extensions of them. “Causal / force
dynamic” approaches take ‘individuals act on individuals’ verbs as basic. I have used
elements of all three, but will in this chapter use the aspectual approach chiefly.
8.3.2 The Finite
Introduction. We saw in Chapter 7 on figure structure that the Entity group’s being
finite is crucial to its function as Predicator. That finiteness may be carried by the
inflection of the Process word or an auxiliary or the do operator; it is an abstract and
a semantic element, which (following Halliday 2014) will be called “the Finite”, as a
constituent of the (semantic) group.
Syntagmatic structure and internal structure. The meaning of the Finite is wholly
grammatical, and is expressed grammatically by the tense morpheme. It directs the
hearer to adjust the meaning of the head, by specifying its tense and mood. Specifical-
ly, for “Oil prices were dipping yesterday”, the meaning of were may be formulated as
“Specify the tense of ‘dip’ as past, and its mood as indicative.” (The significance of the
inflection for person and number is a separate issue.)
Internally, the Finite is quite simple, being a grammatical operator. Syntagmati-
cally, it has two relationships. If the head has the attribute of time of occurrence (as
relational Predicators do not – Chapter 7, §7.2.1), it relates to that attribute. For all
heads, it relates to the whole group as Predicator: finiteness signals that the semantic
unit it belongs to is a figure, being uttered as an assertion or question and so on.
8.3.3 Grammatical auxiliaries
I will follow convention in distinguishing modal auxiliaries such as must and may
from the others, which I will call “grammatical” auxiliaries, and in referring to do used
as an auxiliary as “operator”.
Grammatical auxiliaries are usually treated as expressing aspect rather than time;
Quirk et al. (1985, §4.17) describe aspect as a “grammatical category to do with the
way the verb action is regarded or experienced with respect to time”. The action is
traditionally regarded as “complete” or “incomplete”; various terms are used, such as
“progressive” and “imperfective”.
There are several problems with that view. First, it provides an overlap between
tense and aspect, and gives no coherence to the sequence or other pattern of the auxil-
iaries. For example: Quirk et al. (1985, §4.17) give a mix of syntax, meaning and mor-
phology in explaining the sequence. Second, “aspect” in this view has no consistent
meaning, since its nature depends on the Event type of the action – the “continuous”
or “imperfective” is different for eventive and non-eventive heads, for example; it may
mean “act in progress, actual, present, ongoing, steady state or (dependent) proposi-
tion” (Halliday 2014, §7.4.4). I will accordingly treat those meanings as part of indi-
vidual senses, not as part of the auxiliaries’ semantic system, and not treat them here.
As always, our concern is with semantics; the tense system is described below
semantically, morphology and syntax being relevant only as the way the semantics is
realised.

Introduction. Following Halliday (2014, §6.3), who refines what Quirk et al. call “the
way the verb action is regarded” (cited above), I will treat the grammatical auxiliaries
as part of a system of “primary” and “secondary” tense. Primary tense is illustrated in
“He eats/ate” and in “He has / had a new car”; it denotes past, present or future relative
to the speech event (being sometimes called accordingly “deictic tense”). Secondary
tense is illustrated in “He has eaten” and “He has been eating”; it denotes past, present
or future relative to the time selected in the primary tense (Halliday 2014, §6.3.2).
(Has and had are now auxiliaries, not main verbs.) “The cork is floating” presents the
Event as seen at the time it is happening. “The cork has floated” denotes the past as
seen from the present time; in Halliday’s terms, it expresses “the past in the present”.
“It has been floating” portrays a past Event as currently happening; it expresses “the
present in the past in the present.” (Halliday 2014, §6.3.2.)
Halliday gives a dialogue to illustrate how that “serial tense” system is built up
(2014, §6.3.4); it is shown with the tense designations in Table 25.
Scope of the auxiliaries. Table 25 shows that the tense form (be + [work]-ing) is pro-
gressively modified by preceding auxiliaries. Expressed in bracketed form, then, it is
“[was [going to [have been [working]]]]” – without allowing for the discontinuity of
Table 25. How serial tense is built up

Text Tense form
Does the machine work? Present
It’s not working now. Present in present
But it’ll be working when you next need it. Present in future
Is it going to be working by tomorrow? Present in future in present
It was going to have been working already Present in past in future in past
before you came in; but….
the constituents. The structure parallels exactly the structure of premodifying ele-
ments in Entity groups.
Subsidiary systems of tense. Halliday shows (2014, §6.3.3) that the system of primary
and secondary tense outlined just above has two subsidiary systems. “System II” is
used for projecting a narrative into the past, as in reported speech; it reduces the
options for past forms; for example, the three present tense forms in “She arrived,”
“She has arrived” and “She had arrived” all become “They said that she had arrived”.
System III occurs in nonfinite Process groups and in modalised ones; it is still more
reduced.

As will be clear to the reader, these auxiliaries are wholly grammatical in meaning.
They require the hearer to specify the secondary tense established by the rest of the
group as ‘past’ (for have -en), or ‘present’ (for be -ing), or ‘future’ (for be going to).
Thus for “Is it going to be working by tomorrow?” in the dialogue above: the Finite
(in “is”) specifies that the Event ‘work’ is set in the present (the primary tense); going
to specifies the future (in the present); be …-ing specifies the present (in the future in
the present).
8.3.3.4 Conclusion: Grammatical auxiliaries

Time can be presented lexically – through content semantics – as with adverbials of
time such as “the day after that”; but in this system of primary and secondary tense ex-
pressed in the Finite and the grammatical auxiliaries, it is presented entirely through
grammatical semantics. It is a system that has many forms, but has a simple principle.
8.3.4 Modal auxiliaries
Modal auxiliaries differ from the other auxiliaries in qualifying the likelihood or ob-
ligatoriness (and so on) of some other element in the group. They also differ from oth-
ers in having an element of content. Those qualifying likelihood (epistemic modality)
contain a concept such as LIKELY or PROBABLE; and those qualifying obligatoriness
(deontic modality) contain a concept such as DUTY. (I will not pursue the controver-
sial issue of exactly what varieties of modal meaning there are.)
That small element of content meaning combines with the grammatical meaning,
“Adjust… in the manner specified (i.e. specified by the content of the modal).” For “He
must go,” the hearer must ascribe DUTY to ‘go’ – roughly, ‘For him, going is a duty.” It
is commonly the meaning of the head that is modified, as in example (38), where may
modifies go: the young man’s going is doubtful – and the speaker is confident of what
she is asserting. But in example (39), it is the assertion that is doubtful; the speaker is
not confident of it.
(38) “ ‘He’s a young man searching for his spiritual path,’ she said. ‘He may go
through other iterations, like many of us do’.” (COCA)
(39) “He has – he’s the President. He has to be the leader. Now he may go down
again. But he has got to say, ‘Here’s my plan’.” (COCA)
The semantic modification structure is subject to alteration by phonological structure,

through contrastive stress, as in “He may have been going, but he certainly isn’t now”,
where may modifies been alone, not been going.
8.3.5 Negative as modifier
As we saw in the section on Entity groups, modifiers normally modify the follow-
ing unit, and that is true of the negative modifier, never. But not modifies the unit
within which it is placed (with a couple of exceptions) – the whole group or utter-
ance. Negating “He works” with not requires inserting the do operator – “He does not
work” – presumably to provide the two-part unit to be modified. Thus, “He will not
have worked” negates the future meaning of “will have”; “He will have not worked”
negates the past meaning of “have… -ed”. (Not as submodifier, as in “a not unexpected
result”, works like normal modifiers, and like the negative prefix, un-.)
The negative particles are thus rather like modal Entity modifiers such as fake and
alleged, in being variable in position and semantic scope.
8.3.6 Premodifiers
The “premodifiers” to be dealt with in this section are “adverbs” such as properly,
closely, and regularly. They are distinguished from grammatical items with a modify-
ing function, as with determiners in Entity groups. It is assumed here that the general
rule in English for one-word modifying items is that they precede the word modified.
That is quite clear for submodifiers (as in “very good”), and for focusing modifiers
(as in “He only cracked the vase,” and “He cracked the only vase”), and for Entity
group modifiers (with a few idiomatic exceptions). It appears to be the default rule for
Process group modifiers, so that any modifying word preceding the Process head is
taken to be a modifier in the Process group, not a Circumstance, unless there is a spe-
cific reason otherwise. That is illustrated in examples (40) and (41). (“Mike” refers to
a Customs officer dealing with a suspicious arrival at an airport.)
(40) “It’s going to take Mike a while to properly search his [i.e. the traveller’s] bag.”
(Commentary in a television documentary, on customs procedures; NZTV1)
(41) “It’s going to take Mike a while to search his bag properly.”
(Altered version of 40)
In (40), properly describes the action of searching, and means <3> “Fittingly, suita-
bly” – fittingly for a Customs officer searching, regardless of the object searched. In
(41), placed after the verbal head (and necessarily after the Complement (‘his bag’)
as well, properly means, <5> “Thoroughly,” which takes into consideration the object
searched. It now modifies the predication (‘search his bag’), and is a Circumstance
(Adjunct, syntactically), not a modifier in the Process group. Those principles under-
lie the analyses in this section.

Premodifiers in Process groups are Property words, by semantic class, as premodifiers
in Entity groups are. The two sorts of modifier are naturally alike in their meaning
types and their dimensions. “They have regularly met their manager” and “They had
regular meetings” are alike. They are also alike in their default grammatical mean-
ing; hearers must add the modifier’s content to that of the head, or adjust the head’s
meaning.

Regular structure. Again as with Entity modifiers, Process modifiers usually operate
on one of the dimensions of descriptive meaning in the head. They may specify a
frequency value for the head’s cognitive time dimension as attribute, e.g. “…regularly
met…”, or specify the degree on the intensity dimension, e.g. “Every form should be
closely examined.”
“Closely examine” is represented in Diagram 10, which is based on SOED <1>,
“Investigate… (something) by close inspection…” For the sake of the diagram, that
has been paraphrased as “Investigate (something) by inspecting to a high degree.” That
sense is analysed as having INVESTIGATE, with its manner attribute specified by BY IN-
SPECTING within the sense, and its attributes of time and place unspecified. INSPECT-
ING has its intensity dimension specified by closely as HIGH DEGREE. The diagram
represents all that, as the meaning interpreted by the hearer, not the system meaning
of the words, “closely examine”. The networking lines link the semantic elements; the
underscored lines mark “gaps” where features are unspecified. “Something” in the
diagram is specified in the sentence above by “every form”, in the transitivity structure
INVESTIGATE (something)
Attributes intensity of inspecting: HIGH
manner: BY INSPECTING: dimensions generality:
time: speecifity:
place:
etc
Diagram 10. Attribute and value structure of “closely examine”, in network format
CLOSELY INVESTIGATE
High degree of intensity attribute
Diagram 11. Structure of “closely examine”, in constituency format
discussed in Chapter 7, §7.2.2 and §7.3.2. (Since the diagram represents senses not
words, it would also stand for the synonymous word, scrutinise.)
The structure of “closely examine” is shown in constituency form in Diagram 11:
the dotted box represents the bond, with the ‘high degree’ element of closely specify-
ing the intensity attribute of ‘investigate’.
Process modifiers can also act as a modal: “ Mr Sharif ’s willingness apparently to
allow the Taliban to….” resembles the modal Entity premodifier in “Mr Sharif ’s ap-
parent willingness”, and the epistemic modal auxiliary in the approximate paraphrase,
“Mr Sharif may allow the Taleban…” Such modal premodifiers operate lexically, with
varying shades of content meaning; they parallel the modal auxiliaries, which work
grammatically, in a small paradigm of set meanings.
When there are several auxiliaries, modifiers are placed precisely to modify the
rest of the group. In example (42), the two premodifiers are placed in separate posi-
tions accordingly. ‘Insufficiently’ modifies ‘recognise’, but ‘commonly’ modifies ‘in-
sufficiently recognised in the past’ semantically – syntactically, “commonly” modifies
“been insufficiently recognised”.
(42) “Four elements of language have commonly been insufficiently recognised.”

(Linguistics text)
Process groups are therefore closely parallel to Entity groups; but there is no morpho-
syntactic structure corresponding to the zones in Entity groups. Another difference
from Entity groups is that there is here no recursive modifying structure like that of
Classifiers. A final difference is that some Process premodifiers change the semantic
structure of the head; in example (43), the activity ‘guide’ becomes an accomplish-
ment instead of an activity (§8.3.8.2), since ‘successfully’ establishes an end-point
when the guiding is completed.
Agent event Undergoer Circumstance

He shot Trayvon Martin fatally
Diagram 12. Structure of “He shot Trayvon Martin fatally”
Agent Event Undergoer Circumstance

He shot Trayvon Martin
fatally
Diagram 13. Structure of “He fatally shot Trayvon Martin”
(43) “Sewell seeks to become the first coach to successfully guide a team to victory.”
(BNC)
Structure of incorporation. Increasingly often, I believe, Property senses like ‘success-

fully’ are used as premodifiers of an Entity in situations where they have traditionally
been placed later, and often as Adjuncts. Example (44) illustrates the use; the context
is consideration of criminal responsibility.
(44) “That[,] Zimmerman says[,] is when he fatally shot Trayvon Martin.”

(COCA)
“Fatally” could have been placed at the end of the clause, grammatically and stylis-
tically, and without harm to the information structure; the writer’s choice to place it
earlier implies that a different meaning was intended; there is no contrast with shoot-
ing Martin non-fatally; ‘fatally’ and ‘shot’ must be merged in the interpretation: “fa-
tally shot” = killed.
“He shot Trayvon Martin fatally” would have the group, “fatally”, as Circum-
stance, the word fatally being its head, as illustrated in Diagram 12.
By contrast, “He fatally shot Trayvon Martin” has fatally as modifier in the Pro-
cess group, “fatally shot”. Fatally is thus reduced syntactically from a whole group to a
word, and from a head to a modifier. That is illustrated in Diagram 13, where linked
rectangles are used to highlight the alternative positions of fatally, and fatally’s posi-
tion on a lower line highlights its syntactic and semantic subordination.
In example (45), the senses ‘sexually’ and ‘assaulting’ are unified, just as ‘sexual-
ly’ + ‘violating’ are unified in ‘raping’. The structure is essentially the same in exam-
ple (46), although the unity is less apparent.
(45) “A polygamist leader goes on trial, accused of sexually assaulting children.”

(COCA)
(46) “Whereas these workers have successfully repeated their own experiments,
others in Yorkshire have been unable to do so.” (BNC)
The structure of modifier and head seems to be breaking down in these uses; a struc-
ture of modifier + head is becoming an unstructured lexical unit.
Those are instances of the increasingly common phenomenon of incorporation
in English, with an adverbial being incorporated into the verb here, where formerly
it acted as a modifier. They constitute incorporation because the fusion of senses to
achieve social acceptance of a new concept is the semantic characteristic of incor-
poration. (We will meet an information-structure characteristic in Chapter 9; these
instances have that characteristic, as well.)
Incorporation of adverbials does not seem to be accepted in the literature on in-
corporation. However, it is exactly like the incorporation of nouns, which is now well-
established in English. That is illustrated in examples (47) to (50).
(47) “He was awakened by police later in the morning when they door-knocked
the street as part of their investigation.” (Daily newspaper report)
(48) “Let’s brain drain her, then get rid of her.” (Meaning: ‘Let’s find out all she
knows, then dismiss her;’ spontaneous conversation, 2013)
(49) “The fact is, they’re very easily mind controlled.” (COCA)
(50) “The Super fund trust fund raised $1.5 million a year.” (COCA)
The examples demonstrate the standard characteristics of noun corporation (see

Mithun 1984 and Gerdts 1998): the incorporated noun is reduced semantically,
morphologically, and in syntactic status. For example, door-knocked in (47) has the
phrase “on doors” reduced syntactically to the single word “door”; “doors” is reduced
morphologically to “door”; and ‘door’ is semantically reduced accordingly in individ-
uation, from a countable Entity with its own image schema to a vague Entity not spec-
ified for subclass. Door-knocking is construed as a specific police procedure, which is
conceptually different from knocking on doors. The incorporation is motivated also
by information structure (Chapter 10), as Mithun notes (1984) – reducing the sa-
lience of the doors, in this instance. Incorporated modifiers are written in various
forms: the modifier and head form one morphological word, or are hyphenated, or
are two words – but that does not affect its nature as incorporation. Nor does its con-
sisting of inserting the noun in a group, not inserting it into a single complex word
as in the noun incorporation which is usually studied in the literature; polysynthetic
languages such as Mohawk and analytic languages such as English incorporate nouns
differently. (See Feist, 2013, for further detail.)
Words other than nouns are now being incorporated in Process groups, as in
example (51), which has an incorporated adverb, and (52), which has an incorporated
adjective.
(51) “Fast-forward six days.” (Used as an imperative, in the subheading of a

newspaper article; the meaning is, ‘Skip forward six days in the narrative’.)
(52) “He respiratory arrested on me.” (Written by a paramedic; newsletter of a
paramedic organisation; 2014)
That wider use of incorporation seems to occur more freely in informal English: com-
pare “to front foot”, “to wet comb”, “to cold start” and “my dry erase pen” (adjective
incorporated into premodifying verb). Nouns are also sometimes incorporated into
adjectives, in Entity groups, as in, “…may take off faster than in… risk-averse Europe.”
8.3.7 Postposed particles as modifiers
This section deals with expressions like “run in”, “run out”, “run up”, and “run down”,
as used in utterances like “He ran it in,” and “He ran in his new car” – “phrasal verbs”,
“particled verbs”, and so on.

Semantically, these expressions are not single lexical items, as they may seem to be,
since each word contributes meaning: compare “run in” and “run out” with “take
in” and “take out”. However, there is a gradation from expressions like “He ran off ”
(where “ran” is the Predicator, and “off ” is an Adjunct), through expressions like “He
ran off sixteen copies”, to lexical items like “He measured the run-off.” None of them
are fully compositional, since they are all partly idiomatic, and fully idiomatic phrases
are not compositional at all, of course.
Both the Event word and the particle are simplified in their own structure, being
bleached of some of their meaning through generalisation across contexts. For exam-
ple, “call on”, “call off ” and “call in” do not usually evoke call’s basic element of loud
speech; “called on his resources” and “call off the match” do not invoke the basic posi-
tional meaning of on and off. That resembles the common bleaching of prepositions,
in such expressions as “at sea”, “at risk” and “at stake”.

The syntagmatic structure of particled phrases is variable and complex. In “He looked
over the results quickly”, both looked and over are linked to results, since over retains
some of its prepositional force. There is also, of course, a strong link between the two
elements of the phrase, which survives their being separated in expressions like “He
looked them over quickly”.
The relation between the two words also affects the aspectual class of the phrase;
not only does the activity ‘run’ become an accomplishment (§8.3.8.2) in “run up”, but
the sense of “run down” varies between activity and accomplishment in the ten senses
SOED records for it.
8.3.7.4 Conclusion
That semantic discussion does not deal with what is perhaps the main function of
Process words with postposed particles, namely their various information structure
functions, such as allowing control of focus – compare “He told the driver off,” and
“He told off the driver.” (See Chapter 10 for information structure.)
8.3.8 Heads of Process groups
As with Entity groups, the head is the section of the group which must carry the se-
mantic potential for the group to relate to other groups; for Process groups, the head
must be the focus for the role of Process, relating to Participants. Within the group,
it must be capable of being modified, by both content and grammatical modifiers;
it must also be capable of bearing tense when it is the only word in the group. Ac-
cordingly, in “He has gone”, “gone” is the semantic head, although “has” may be the
syntactic head.

Types and dimensions of meaning. We have distinguished between the head as a con-
tent meaning, and the Finite, a grammatical meaning. The distinction is clear in “He
must go,” where the two are separated in expression (must + go), and in goes (go- +
-es); but it exists also in went.
Process heads bear most of the content meaning of the group, even when it is
complex. Accordingly, they often have descriptive and affective meaning, as in “I
loathe him!” and “I adore it!” They may have attitudinal meaning, as in “I respect it
very much.” They may carry social meaning, as in the bureaucratic jargon of “We de-
liver services,” the legal jargon of “The police allege that the men they have charged…”,
and the colloquialism of “They knock off at 4.30.” Relational heads such as ‘resemble’
and ‘equal’ are much more limited in their range of meaning types, commonly con-
sisting of a minimal concept; that is indicated by the equivalence of ‘resemble’ to the
prepositional meaning ‘like’, and of ‘equal’ to the mathematical “equals” sign.
Process heads, being rich in descriptive meaning, are often defined on the di-
mensions of intensity, vagueness and generality, as well as quality. Meanings such as
‘run’ and ‘walk’ are more fundamental on the basicness dimension than meanings
such as ‘jog,’ ‘lope’ and ‘sprint’, and ‘stride,’ ‘stroll’ and ‘amble.’ Some are defined on the
viewpoint dimension: ‘come’ and ‘go’ are defined from the viewpoint of the speech
situation, and ‘depart’ and ‘arrive’ from a contextually defined viewpoint.
Heads such as criticise, entice and love have many sense elements, which vary
on the expectedness dimension, and vary from context to context in their salience.
Love, for example, has expected elements given by SOED <2> as ‘…attachment’, ‘take
pleasure in’ and ‘admire’. Its likely elements include ‘sexual’ and ‘be in love with’ (sense
<1b>). Clearly, their salience varies with whether love is used in a marital, maternal,
brotherly, or other context.
Aspectual structure. In basic, perceptual senses, the boundedness dimension for Events
is a matter of time, whereas for Entities it is basically a matter of more or less concrete
space. Most Events have duration, and are specified for boundedness: they may have
an initial bound – be specified as having an identifiable moment when the happening
begins – and may have a final bound – be “telic”. Some Events have no duration, as
with ‘arrive’, ‘forget’, and ‘graduate’; they are not specified for boundedness, and may
be called “non-bounded” Events. The pattern set by those elements constitutes the
Event’s “event structure” or “aspectual features” or “Aktionsarten”. (They could also be
thought of as sub-dimensions of the temporal boundedness dimension; cf. §8.2.8.4.)
That structure affects such things as whether Circumstances of duration (e.g. “for two
hours”) or of an instant in time (e.g. “at 6 o’clock”) may be predicated of them.
To an important extent, aspectual structure is a matter of construal, not directly
dependent on the nature of the world as we experience it, giving us semantic choices
in representing our knowledge. That can be illustrated in examples (53) to (55).
(53) “I built sandcastles with the other children”. (BNC): initial and final bounds
(54) “They have built a lovely new leisure centre”. (BNC): final bound only
(55) “… an undercover police station being built in the palace grounds.”
(BNC): neither bound
Aspectual structure as a linguistic phenomenon should thus be distinguished from

our conceptual understanding of building, in examples (53) to (55). Real-world build-
ing necessarily has both beginning and end; linguistic and real-world phenomena
should not be conflated, as they are even by Cruse (2011, §15.4.2), who describes
aspect as “inherent”. The semantic aspectual structure is signalled by morphological
aspect, and by contextual clues.
I have not rated change as an aspectual feature, because, as a defining feature
of Events, it cannot be absent. It is commonly rated as a part of aspect (e.g. Cruse
2011, §15.4.2.1), because of the confusion in which “verbs” are thought of as denoting
change, yet as sometimes not having change as a feature. Nor do I include discussion
of causation as part of the linguistic structure, as Croft (2012) does; causation is lin-
guistically relevant to causative inflections, but here it is cognitive (concerning how
we understand the world), and a lexical matter accordingly (“He caused…”). Habitual
and repeated occurrence are similarly cognitive, subject to lexical expression (“used
to”, “kept on”, “repeatedly”), and without grammatical correlates, not a matter of as-
pectual structure in this sense.
Aspectual structure thus forms a paradigm, which may be set out in Table 26.
Each row gives an example from just above, a suggested descriptive name for the type
of aspect, and a definition of the type through the presence (Yes) or absence (No) of
the feature.
Non-specified Events have traditionally been called “achievements”; non-
bounded Events have been called “activities”; and both doubly bounded Events and
final-bounded Events have been called “accomplishments”. The distinction between
Table 26. Aspectual structure and Event types

Example Name for type of aspect Initial bound? Final bound? Durative?
“I built sandcastles” Doubly bound Yes Yes Yes
“… have built…” Final-bounded No Yes Yes
“… being built..” Non-bounded No No Yes
“… arrive…” Non-specified No No No
initial and final bounds exists in the literature, as in Cruse (2011, §15.4.2), but is not
regularly built into the classification.
Individuation in Process heads. In §8.2, Entities were described as being more individ-
uated if they are more bounded in space. Processes are similarly more individuated if
they are more bounded in time. (See Jackendoff 2011, p. 697 for the parallel between
Entities and Processes in individuation.) Activities, being without a start or end point,
are not individuated at all. Accomplishments are more individuated, since they have
one bound, an end point. Processes with “terminative aspect” (e.g. ‘to expire’, ‘to ex-
haust’) are similar in having that single bound, at the end; and those with “inchoative
aspect” (e.g. ‘to be born’, ‘to set up something’) also have a single bound, although
at the beginning. Since those two aspects (for which see Cruse 2011, §15.4.2.2) are
equivalent to accomplishments in their bounds and individuation, they should be rat-
ed with final-bounded Events (“accomplishments”) in aspectual class. Non-specified
Processes (“achievements”), such as ‘to arrive’, having an instantaneous transition be-
tween two states of affairs, are the most fully individuated.

As implied in §8.3.6, heads of Process groups act as operand through various ele-
ments of their descriptive meaning. Usually, the bond for a content modifier is that
the modifier specifies the value of one of the head’s attributes; for example, the head’s
frequency attribute links to premodifiers such as always and often. The time attribute,
a necessary element in Event senses, acts as a bond for many postmodifying groups,
while specifying aspectual quality. For example, duration provides the bond for “for
six months”; the initial bound provides the bond for “after five minutes”; the final
bound provides the bond for “until 6 p.m.”; punctual aspect provides the bond for “at
6 p.m.” (Clausal Adjuncts are often linked in the same way.) Sometimes, the bond is
more conceptual, with the modifier specifying a dimension of the head content, as
with “closely examined”, discussed in §8.3.6.
Words gain new Process senses in transparent morphological ways, by adding -en
to signify ‘become’, -ise to signify causation, and so on. But they also change sense in
less obvious ways, which are possible only because of the internal structure of their
senses. We have seen that Process meanings relate to modifiers through such attrib-
utes as TIME, PLACE and MANNER, and relate to Participants through such linking
concepts as CAUSE, EFFECT and BECOME. When such concepts are elements of the
Process, they can be inhibited, and dropped from the meaning; even when they are
not part of the head, they are close to it in the mental network, so that the links to
them are easily activated. Thus we can easily understand a reviewer’s very elliptical
description of a film as “a rewarding watch” (= ‘the film will reward you if you watch
it’), since the Event ‘watch’ has ‘Undergoer Participant’ (such as ‘this film’) close to it
in the network, and the Event ‘reward’ likewise. A reporter could say that a prize had
“jackpotted,” (‘become a jackpot’) because BECOME is easily invoked. A mechanical
process was said to “perpetuate” (‘continue indefinitely’) – using the transitive verb
intransitively – because the CAUSE-BECOME link could be treated as possible meaning
instead of necessary meaning, and omitted accordingly.
8.3.8.4 Semantic class

The section on aspectual structure (§8.3.8.2), showed that the aspectual classes of ac-
tivity, achievement and accomplishment are sometimes at least alternative construals
of our experience, providing semantic choices for expression. They thus constitute
subclasses of the class, Process.
We have seen (§8.2.8.4) that nominalised Entity senses contain a subordinated
Event sense; so it is tempting to assert that the Process sense hammered includes the
Entity sense ‘hammer’, since it means ‘drove with a hammer’, and that reddened in-
cludes the Property sense ‘red’. That is not so, however; RED, as a putative element
of ‘redden’, has no morphosyntactic correlate or effect, whereas the Event element in
nominalisations has a correlate in -ing and -ed inflections, and arguments expressed
as Classifiers (e.g. “Saudi oil sales”). Accordingly, RED in ‘redden’ is a conceptual sub-
linguistic sense element, and not a member of a linguistic (semantic) class subordi-
nated within another class.
The relations between Entity class and Process, on the rank above, and entity, on
the rank below, were sketched in §8.2.8.4 above.
8.3.8.5 Conclusion: Process heads

This section on the head of Process groups has dealt with the meaning of words usu-
ally called “verbs,” but the account is quite different from the usual account of verbs.
The view taken here may be summed up as follows. The large class of words that
head “verbal groups” or “verb phrases” have no common characteristic in their sens-
es, since they denote variously events in time, timeless relationships, and static states.
The only thing they have in common is that when used in figures they signify asser-
tion and predication; that is a role which the words have in that situation, rather than
a word class; if we insist on using class in conceptualising these issues, then the class,
being specific to figures (not groups or words), is that of Predicators syntactically, or
Processes semantically. Either way, there is no such semantic class as verbs, and no
such word class partly defined semantically.
8.3.9 Process groups as units

We have seen that the syntagmatic semantic relations are basically straightforward,
and like those of Entity groups: [modifier [modifier [modifier head]]]. That is var-
ied, however, when the group is negated (which creates various structures, according
to stress, for example), and when there is a postmodifying particle. (That summary
is much simplified, passing over the division of secondary-tense modifiers into two
morphemes, as in have -en for the past, for example.)
8.3.9.2 Determinateness
We saw in the section on Entity groups that their Entity core cannot serve as a group
until it becomes sufficiently determinate to take on its role as Participant in the figure.
Similarly, the core of an Process group must be determinate if it is to serve as Process
in the figure.
As Entities must be rendered sufficiently individuated for the predication in
which they are involved, so Entities must be made sufficiently finite. That is, they must
have sufficient time reference to denote a specific event in the mental world being
modelled by the text – to denote a real Process, not a potential one. That is provided
by the tense specification of the Finite. (The time reference may have little cognitive
significance, as when timeless truths or actions repeated in the past and future are
stated in the present tense, with apparent restriction to the present time. That is lin-
guistic time meaning, not cognitive time.) To be fully determinate, Processes must
be assigned secondary tense. Finally, they need boundedness in aspectual structure.
Finiteness + secondary tense + boundedness makes the group fully determinate.
To Halliday (2014, §4.7), determinateness serves to make the proposition expressed
“arguable.” To other linguists, the function is to make it “grounded.”
The discussion given above applies to independent Process groups serving as
Predicators. Subordinate groups are less determinate, according to their status; for
example, Process groups serving as postmodifier of an Entity are not finite, as in ex-
ample (56).
(56) “He’d turn blue and black, the bubbles rising slowly, as he fell deeper…”
(COCA)
8.3.9.3 Complexity
Process groups are simple in not having a zone system, and in lacking potential for
multiple, rankshifted postmodifying groups. They also have far less content mod-
ification than Entity groups do – partly because there are no zones established to
accommodate it – and we do not find strings of Process modifiers, resembling the
strings of Entity modifiers noted in §8.2. That simplicity is compensated for partially
by Processes’ propensity for complex descriptive meaning: some sense elements that
could be expressed adverbially are expressed within the head itself. For example, “He
Table 27. Alternative construals of relative weights

(a) Sublexical construal OBJECT WEIGHT GREATER THAN TONNE
(b) Lexical construal (i) It weighs more than a tonne
(Entity) (Process) (relation) (Entity)
(c) Lexical construal (ii) It’s weight exceeds a tonne.
(Entity) (Entity) (Process) (Entity)
chortled” means ‘He laughed gleefully and loudly’; “He chuckled” means ‘He laughed
gleefully to himself.’
8.3.9.4 Processes as construals

Situations or “state of affairs” as we experience them can be construed into language
in various ways. An obvious instance is the choice between ‘it became red’ with a
Property, and ‘it reddened’ with a Event ,with the element of change included. A sec-
ond instance is that when we see light from a distant lighthouse, we can construe the
perception as an Entity (a ‘flash’ or a ‘light’) or as an Event (something ‘flashed’).
Less obviously, some object having ‘weight’ (Entity), can be reconstrued (for use
as a Predicator) as (i) the object weighing a certain amount, or (ii) weighing more or
less than another object; it is reconstrued as a relational Process. Those close parallels
between the conceptual representation and the two linguistic ones is shown in Ta-
ble 27. Line (a) represents a sublexical construal of the situation (stated in concepts).
Lines (b) and (c) represent the alternative lexical construals (stated in words).
8.3.10 Discussion: Process groups
8.3.10.1 Compositionality of Process groups

Process groups are compositional in some respects, and non-compositional in others.
The syntagmatic structure set out above provides for fairly full compositionality. The
group consists of distinct constituents syntactically, and the semantic constituents are
also distinct, although not identical with the syntactic ones. The meaning of the whole
group is composed by the hearer according to the procedures specified by the gram-
matical meanings, which are well defined. Moreover, Process groups rely on cognitive
interpretation over-ruling linguistic interpretation much less often than Entity groups
do; there is no equivalent here of constructionless Classifiers, or of Epithets’ ability to
relate cognitively to referents outside the phrase itself.
A couple of the features we have noted restrict compositionality substantially,
however. The effect of time specification, both in tense and in content modifiers,
depends on the event structure of the real-world event denoted; hearers must rely
on cognitive interpretation from outside language, making the interpretation non-
compositional. Second, the overriding significance of Process groups, as we have
seen, is that they predicate (in their ideational function) and make an assertion (in
their interpersonal function). Neither significance can be composed from constitu-

ents; they enter the utterance meaning from the speaker’s intention – from “above” –
not from “below,” compositionally. That last point needs emphasis; the literature on
compositionality treats our use of English as if it were no more than the transmission
of aggregated data; it is also – and primarily – the fulfilment of intention.
8.3.10.2 Non-eventive Processes as “verbs”

Section 7.3.5 above noted that some Process meanings involve change and occurrence
in time, but do not involve time passing e.g. ‘believe’, ‘desire’, ‘know’, and that they are
commonly called “stative verbs”. Stativity as a general linguistic concept is unsatis-
factory, since it does not characterise any coherent linguistic class. It cuts across the
semantic classes, and even the traditional word-class categories, since we have stative
adjectives as well as stative verbs (Quirk et al. 1972, §5.38); and we could equally well
have stative adverbs and stative nouns. The situations where it does seem to apply are
better explained, as above, through the concepts of Property, of relation and of being
not eventive (as in this section, i.e. §8.4.4, and Chapter 7, §7.3.5).
Chapter 7 distinguished relational Processes, as well as stative Processes, from
Eventive Processes; it argued, further, that they should not be called “verbs” at all, but
“Processes”. There may have been a time in the history of English when the class of
morphological verbs coincided with the class of Event senses; but over the centuries
those “verbs” have been assigned the ideational function of predicating and the inter-
personal function of asserting, with relational and stative senses increasingly used for
those functions; those “verbs” are no longer simply Event words.
8.3.10.3 Copula groups

The discussion so far has left one type of Process group not discussed directly, namely
the copulas – chiefly be, but also some other words in idiomatic uses, as in “It stands
to reason”, perhaps. Since the meaning of copulas is purely grammatical, they do not
belong to any semantic class – as noted previously, semantic class is being assigned
only to content meanings. Thus they are not Event groups, and do not have the inter-
nal structure or syntagmatic structure discussed throughout §8.
Quirk et al. (1985, §12.32) lists a number of other words as copulas, such as ap-
pear (as in “appear happy”) and become (as in “become older”), noting that they “have
more intrinsic meaning than be”. They do indeed have “intrinsic meaning”, i.e. con-
tent, and should therefore not be rated as copulas, in the uses given. The authors have
presumably noted the heavily grammatical function of those words, but been unable
to resolve the issues, being stuck with the assumption that all such problems are a
matter of classes, not of function.
8.3.11 Conclusion: Process groups
In internal structure, Process groups are Eventive, or relational, or stative, or are copu-
las. In syntagmatic structure, many Process groups have a structure of meaning types
and semantic function. In the example, “must have been often puzzled”, the first el-
ement (the Finite, in “must”) is deictic, “relating the process to the ‘speaker-now’,”
(Halliday 2014, §6.3.1), by tense; then come other grammatical meanings (auxilia-
ries, “have been”); then comes content meaning serving as modification (“often”);
last comes content meaning serving as head (“puzzled”). The syntagmatic structure is
that each unit modifies the rest of the group: …[…[…[…[…]]]]. In all that, Process
groups parallel Entity groups.
Superficially, that structure resembles the zone structure of Entity groups, also.
However, it differs in the crucial respect that zones can take several items of the same
semantic type, but the positions in a Process group cannot.
The general structure just described for “must have been often puzzled” has sev-
eral important exceptions, however. Content modifiers and negatives such as not can
come in various positions; and the group can end with a postmodifying particle.
8.4 Property groups
8.4.1 Introduction
This section deals with Property groups, which are groups belonging to the semantic
Property class – members of the Property class dealt with so far have been expressed
as words, not groups. The structure and relationships of Property groups are more
complex than those of Entity groups or Process groups.
In describing them, I am assuming that the cognitive basis for a Property is a
quality, that perception produces such concrete qualities as RED and HEAVY, and that
more conceptual cognition produces more abstract qualities such as SUCCESSFUL,
OFTEN and NEW. Those qualities are treated as inherent in objects and happenings,
although they are derived by comparison between the objects or happenings. The
qualities are used in the corresponding sublexical senses, and construed into lexi-
cal senses as Properties. The groups they head are used as modifiers of Entities and
Events, which is appropriate for the linguistic realisation of qualities which inhere
in things and happenings. For example, in the headline, “Xero seeks US boost with
brand new platform,” the Entity, ‘platform’, is modified by a group with a Property as
head, “brand new”. In “Xero’s plans have quite often failed”, the Event ‘failed’ is simi-
larly modified by a Property-headed group. In “Xero’s buildings are quite often brand
new,” a Property group is modified by another Property group. Thus English uses
these modifying Property groups in three different ways.
A more serious complication is that we often want to make the possession of

the Property an assertion, instead of simply ascribing the Property to the Entity in
passing. The Property group is then used in a quite different structure: it becomes a
Complement – “Xero’s platform is brand new” – or an Adjunct – “Xero’s plans have
failed quite often.” The Property is separated syntactically from what it modifies.
There is a further complication. The Property groups dealt with so far have had
basic Property senses as their head – and function in a way we are familiar with, as
“adjectival phrases” or “adverbial phrases” – but there are two structures denoting
Properties that do not have basic Properties as heads. In “Xero’s plans have failed at an
uncomfortably high rate,” the group expressing the Property is built up from an Entity
head, ‘rate’. (Note that the prepositional phrase has a preposition and an Entity group
in complementary relation, with the result that there is no head.)
There are, then, two types of Property group: (a) groups like “brand new” and
“quite often” which have as head a sense derived from a cognitive property (“Property
heads” for short), and which are traditionally thought of as “adjectival phrases” and
“adverbial phrases” – see the next section, §8.4.2; (b) groups like “at an uncomfortably
high rate”, which, being “prepositional phrases”, have no head of their own (“headless
groups” for short) – see §8.4.3. The variations in semantic class and structure are re-
flected in variations in their morphosyntactic expression.
8.4.2 Property groups with Property heads
This section deals with Property groups with Property heads, taking “adjectival phras-
es” and “adverb phrases” together, as groups. The equivalence of adjectives and ad-
verbs was shown in the earlier discussion of semantic classes, and it can be seen here
clearly in the close match between the two types of group. Compare the remark by
Xero’s CEO that “expansion had been slower than expected” (attested) with its alter-
native, that Xero “had expanded more slowly than expected”. The underlying cog-
nitive perceptions are of the firm’s size at two different times; the concept INCREASE
IN SIZE is a construal of that situation; and the Entity word expansion and the Event
word expanded are alternative linguistic construals of that. So the “adverbial phrase”
with slowly modifying expanded, and the “adjectival phrase” with slow modifying ex-
pansion, are semantically the same; the difference is morphological, and the choice
between them is motivated by style or information structure.
These groups are used both as premodifiers, e.g. “It was a very successful fete,”
and as postmodifiers e.g. “… his face relaxed and happy”. They are sometimes dis-
continuous, as in “It was a much more successful fete than usual”, where the group
modifies the Entity, ‘fete’.
The distinctive semantic nature of these groups is reflected in their morpho-
syntax: the head may be inflected for comparison, as in “slower than expected” and
“much more successful…”. They are gradable if they are abstract.
The equivalence of “adjectives” and “adverbs” is supported by Lyons 1966,

p. 219ff.), and by Radford (1988, p. 138ff.), who rates adverbs as a special class of
adjectives, noting that the difference between them is simply distributional, which is
complementary; distribution is a syntactic matter, not a semantic one.

Premodifiers. Premodifiers in Property groups commonly consist entirely of gram-
matical meaning, as with very and much more in the examples given just above. There
is sometimes some descriptive meaning: “His first novel was brilliantly successful”
differs from “…extremely successful.” There can also be affective meaning, as in “The
bread was horribly stale,” and social meaning, as in the journalistic description of
produce as “mouthwateringly fresh” (in a passage cited above in §8.2.9). By semantic
class, the content premodifiers are Properties, like the heads.
Postmodifiers. As with Entity and Event groups, postmodifying items are usually par-
ticipial phrases or rankshifted clauses. There are some idiomatic exceptions – enough,
as in “Oh yes, it’s plentiful enough.” They relate to a comparative element, which is
most often a premodifier in the group: in “It all happened much more easily than he
expected,” “than he expected” relates to “more.” But the element may be postposed
(e.g. enough as in “good enough to eat”), and may be represented by the comparative
inflection in the head (e.g. “She ran faster than all the rest of them”). The relationship
of postmodifier and comparative element is that of complementation, rather than
modification.
Head. The heads, as Property senses such as ‘good’ and ‘easy’, have the same internal
structure as the Property senses discussed as premodifiers in Entity groups (e.g. “a
good book”, “an easy ride”). They have descriptive, affective, attitudinal and social
meaning, and they vary on the dimensions of intensity, generality, vagueness and so
on, as in successful, fresh, easily, good, and absolutely smashing.
Their nature on the boundedness dimension is different from that of other con-
tent words, as a consequence of how basic sublexical properties are derived. The men-
tal processes which abstract them from the experience of objects and happenings
abstract them from the spatial dimensions that objects have, and from the temporal
dimension of happenings, making them more abstract than either Entities or Events.
That makes them eligible for mental comparison, which is why Properties are the only
class that has gradable items: Entities and Events may be classified, but have no degree
of intensity on a scale and consequently cannot be graded.
Variation on the generality dimension produces a scale of senses closely analo-
gous to the scale of individuation in Entity senses. The sense of Shakespearean, for ex-
ample, grades from the specific and “individuated” sense ‘pertaining to Shakespeare’
through ‘characteristic of Shakespeare’s work’ to ‘resembling Shakespeare’ (although
the last sense is not recorded in SOED).
The structure of their descriptive elements varies significantly also. As noted in

the section on Descriptors (§8.2.3), modifiers in participial form generally carry an
Event sense. In “recently discovered artefacts”, for example, the Property ‘discovered’
has an Event as a subordinate element.

When Property groups act as premodifiers, their syntagmatic structure is the same
as for a one-word modifier: they operate descriptively on a dimension of the head, or
supply the value for an attribute in the head, as Descriptors or Epithets (§8.2.3, §8.2.4);
or else they add affective meaning, as Epithets. When they act as clause complements
e.g. “That was wise”, or as clause Adjuncts e.g. “She arrived late yesterday”, they are at
a different rank, that of the figure not the group; so they relate to neighbouring units
as groups do, as discussed in Chapter 7. They then sometimes take their own Comple-
ments, “He is quite able to do it,” and “He is capable of getting there.”
8.4.3 Headless Property groups: Prepositional phrases
Properties groups with no head have a range of uses. They are used as postmodifiers
of Entities (e.g. “a star performer among the listed retailers”), and as postmodifiers of
Events (e.g. “by driving at high speed”). They also function at a higher rank, as Cir-
cumstances, expressed syntactically as Complements (e.g. “The meeting is on Tues-
day”), and as Adjuncts (e.g. “He sold his car on Tuesday”).

Structure of the group. I have asserted above that the structure of prepositional phras-
es, as preposition + Entity group, is semantically one of complementation, whatever
the syntactic structure is; but that issue does not seem to affect the structures being
considered here.
Internal structure of the preposition. Prepositions, more than any other words, may
have a fairly even blend of grammatical and descriptive meaning – hence the difficulty
they have caused for linguists who want to divide words between lexical and gram-
matical word classes. Their grammatical meaning is of the type, “Relate X and Y as
co-ordinates”. The preceding item is a Event group or the head of a Entity group; the
following item is a Entity group.
The descriptive meaning is a spatial, temporal or other relationship, as discussed
in the voluminous literature on prepositions. It is rare for prepositions to have any
nondescriptive meaning. One example is anent, which carries archaic and Scots social
meaning. Another example is around in the meaning ‘concerning,’ (as in “I have prob-
lems around that”), which at the time of writing is a modern vogue sense, too recent
to have reached the SOED.
Prepositions have a complex dimensional structure. As their meaning has been

extended from spatial to temporal reference (from “at home” to “at the weekend”),
and then to a still more abstract sense (“at risk”), they have gained a greater range on
the generality dimension, and on the vagueness dimension. They have a clear struc-
ture on the basicness dimension, with spatial reference as basic. They are specified on
the point of view dimension more often than any other group of words: near, beside,
to the right of, in front of and so on depend on the speaker’s location and distance.
There has been a good deal of other discussion of their semantic structure, but it deals
primarily with cognitive rather than linguistic semantics.
Prepositions have a significant connection with Events. The basic sense of many
prepositions is direction – to, towards, from, over and under in some uses, and so on –
which is closely related to movement, and is often included in an Event sense, as with
‘enter’ and ‘pierce’, equivalent to ‘go into’. In many phrasal Predicators, the preposition
shares that movement sense with the “light verb” which acts as Event head; the phras-
al Predicator is then equivalent to a single semantically full head: compare, run across
and meet, and make up and invent, for example. The alternative forms can be seen as
alternative realisations of the situation or “state of affairs” – different ways of differen-
tiating it. In these instances, the semantic distinction between the Entity group and
the prepositional phrase has blurred, although the syntactic distinction remains. They
resemble Events also in taking Complements. Halliday (2014, §5.6) calls prepositional
phrases “minor Processes” accordingly.

When functioning as Complements or Adjuncts, prepositional Property groups have
the syntagmatic structure of those units of the figure; see Chapter 7. When acting as
postmodifiers in Entity groups, they have the structure set out in §8.2.7 above.
Prepositions with descriptive content act as a link for the following Entity group
by sharing an attribute with it. In “He booked a ticket to London”, for example, to
signifies direction in space, ticket implies places travelled to and from, and London
names a position in space. Diagram 14 illustrates that, showing the concepts for “a
ticket to London”, and representing the spatial attribute by the areas of overlap.
In Diagram 14, the structure is represented in constituent units; represented in
network form, as nodes and links, it would appear as in Diagram 15.
To emphasise the point made previously that networks consist of concepts, which
may be represented as node or link according to what is being explained, Diagram 16
presents “a ticket to London” with the link in Diagram 15, ‘direction’, shown as a node.
TICKET DIRECTION LONDON
Diagram 14. ‘Ticket to London’, with to as ‘direction’, in constituency format

direction
ticket London
Diagram 15. ‘Ticket to London’, with to as ‘direction’ (in network format)
in of
ticket direction London
Diagram 16. ‘Ticket to London’ with to (as ‘direction’) as node not link
The grammatical meaning of to, which in its general form is “Link the two Entity
groups”, is specified by its content, into “Link the two Entity groups in a relation of
spatial direction”. In some uses, the preposition is semantically empty, as with the gen-
itive of, and with “different from” and “different to”, where the prepositions would be
antithetical if they had any content. In those instances, the grammatical meaning re-
mains general, the hearer being left to supply the nature of the link, by applying world
knowledge to the context. In some other uses, the conceptual content is attenuated
and vague, as with “in doubt”, “in time” and “in case”; the hearer must combine world
knowledge with what little content meaning is supplied by the preposition.
Most groups in English sentences are linked by relations which are only implic-
it, not stated by a word: in “He booked a ticket to Melbourne”, for example, there is
no word which lexicalises the relation between the Subject “he” and the Predicator
“booked”, or between Predicator and Complement. In contrast, the relation of “Mel-
bourne” to “ticket” is lexicalised. That is unique in the semantic structure of figures.
8.4.4 Conclusion: Property groups
Property groups can form quite complex semantic structures. We can have “not [so
[very much [more easily]]]” (Halliday 2014, §6.4.1), with very submodifying the
modifier much, and more specifying the degree of comparison; and there can be prep-
ositional phrases as postmodifiers. Those expansions of the group make them more
determinate, just as modification makes Entity and Event groups more determinate.
Property groups can be very expressive, and they sometimes carry the weight
of the message (by coming in focus position); but their semantic and syntactic roles
are complex. They thus often seem to obscure the semantic structure of the figure,
which is eminently clear in intransitive figures with Entity + Event, and with transi-
tive figures with Entity + Event + Entity. But they can provide an even simpler and
more basic structure, with Entity + Property, as with “All men are mortal”, and other
profound truths.
8.5 Words and morphemes
8.5.1 Internal structure
8.5.1.1 Internal structure of words

This section deals with words which are complex morphologically, and which there-
fore seem likely to be complex semantically, with the meaning of the whole word built
from the meanings of constituent root morphemes and bound morphemes. I will not
say a great deal on the matter, however, since the disagreements on the issue cannot
be resolved briefly. We have already noted (§2.2.5.2) the disputes about “compounds”.
Hagit and Borer (1998) said that morphology “has always been a battleground”;
Giegerich (1999) said that even within the generative tradition there was only limited
consensus, and Miller (2014) reaffirmed that more recently; Marantz (2013) says that
even the reality of morphology is controversial. Furthermore, I am not qualified to
resolve the disputes.
Writers in some traditions, however, are confident that words have a semantic
structure paralleling the morphological structure. Toman (2001), for example, asserts
that the morphemes in a word have semantic relations that are sometimes those em-
bodied in syntax. That has certainly been one traditional view; SOED, for example,
asserts that the suffix -ish has four distinct senses which it contributes to words. The
frequency of neologism in present-day English confirms that speakers and writers
attribute meaning to roots and affixes: “a world run by datacrats”; “the embiggened
‘Q’ key has been downsized….; “a Los Angeles-scape of motorways”; a tax department
letter states that citizens’ online tax forms will “pre-populate” into their return. The
writer of “embiggened”, in an online blog, assumed that readers would understand
em-, big and -en as constituents, as well as -ed. Some roots and affixes have become
independent words, such as mini, mega and max.
I accept the psycholinguistic resolution given by Aitchison (2012, p. 13): prefixes
and derivational suffixes are integral with the word, which is stored and processed as
a unit, but people decompose words if they need to. We have three parallel procedures
for understanding complex words: taking the word as a unit, which is the default pro-
cedure; analysing it morphologically, which is a backup when the default procedure
fails; and a “lexical toolkit” (Aitchison 2012, p. 13) for making new words – which
is used for neologisms such as those noted in the previous paragraph. Further, we
change our treatment of particular words steadily, as we learn new words and uses.
The assertion that we do not regularly break down complex words is supported by
Schreuder et al. (1985, p. 108), who found that complex words do not take longer
to process.
8.5.1.2 Internal structure of morphemic meanings: Senses

Summary of earlier parts of the book. Morphemic meanings are what has been called, a
little loosely, “word senses”. Their internal structure has been dealt with from various
points of view, in several chapters so far. Chapter 4, on the elements of meaning, gave
a fairly full outline, because morpheme senses – the most versatile of the elements of
language that carry meaning – include most of the elements described in that chapter.
Later chapters have added to that, according to how senses enter into network and
hierarchic structures.
The main points are as follows. Content senses are made up of various types of
meaning: descriptive, emotive, attitudinal and social. (Grammatical meanings will
be dealt with in a later section.) Descriptive meanings, and to a very limited extent
the other content meanings, are structured dimensionally, as physical objects are
structured by the three spatial dimensions. Those dimensions control many of the
relationships between senses, providing the bonds which link them to other content
meanings, and on which grammatical meanings operate.
This chapter has added points arising from how senses are used in context. The
main points are as follows. When senses are used descriptively, they invoke their var-
ious potential meaning elements extensively; but when they are used restrictively or
referentially, the descriptive elements are invoked much more selectively or not at
all, and those that are invoked are used to identify the referent but not as part of the
information which the utterance is intended to convey. Those descriptive elements,
described in Chapter 4 as ranging from necessary through expected and possible to
unlikely, have been seen here as graded in salience, with the necessary elements most
salient and most readily accessible, ranging downwards in a hierarchy to the least
readily accessible.
Semantic primes. Complex meanings are made up of simpler meanings, which are
made up of still simpler ones; many linguists have held accordingly, and very nat-
urally, that there are minimal semantic units, which have no constituents. They
have sometimes been called “primes”, and sometimes “primitives”. I shall call them
“primes”, to match the use in “prime numbers”; it is a structural term; “primitives” will
be a developmental term.
Aitchison (2012) provides a useful critique and simple outline of the issues, fo-
cusing on the work of Miller and Johnson-Laird, and of Wierzbicka (covering Natural
Semantic Metalanguage). Jackendoff (2002, §11.2) gives a more thorough discussion,
including treatment of work by Pinker, Grimshaw, and Pustejovsky. I accept their
adverse criticism of past work, and will not repeat it.
As seen from the perspective adopted in this book, past work on primes is unsat-
isfactory because of its assumption that meanings can be adequately treated as inde-
pendent units. This book has worked with the hypothesis, which I trust has now been
confirmed in the reader’s mind, that meaning forms a mental network. Conceptual
meanings appear to be units, but are in fact miniature networks, extending in various
directions, without a top or bottom where primes could reside, and with boundaries
set only by convention (as for dictionary-making, for example), and not fully distinct.
I follow authors such as Barsalou (e.g. 1999) in understanding linguistic concepts (in
descriptive meaning) to be based on cognitive concepts, which are in turn ultimately
based, directly or indirectly, on multiple perceptual experiences, which may or may

not be deemed to be cognitive, but which are certainly outside language. Nondescrip-
tive meanings are based in non-cognitive modules of the mind, such as the module
for affect. Meanings are learned from other people’s use of language, and are some-
times derived from other elements of the language; but ultimately they must, in an
empirical view of language, be regarded as evolving from perception. According to
Prinz (2002, p. 180), that has been the most popular view, historically. There may be
semantic units which are not related to simpler semantic units, but they are always
based on simpler units from outside language. (In the view that conceptual elements
are innate, their network nature still rules out linguistic primes.)
That is supported by neurolinguists such as Fortescue (2009, 2010), who show
meanings as arising from the “affordances” of perceptions. Consequently, the simplest
linguistic elements are not primes in the sense of having no relation to smaller units,
since they are based on more particular units of perception. That is supported for
concepts by Zwaan and Madden (2005, p. 224) and Hsu et al. (2015), who note that
traces of the original experience remain in the brain.
The view that meaning is based on primes is further weakened by its inability
to deal with grammatical meanings, and by its assumption that meaning is purely
conceptual, the function of language being simply the representation of knowledge.
We have seen, I believe, that that “representational” function is in most utterances a
subordinate one, serving one of the interpersonal functions. Those functions, and
the means by which they are achieved, cannot be analysed into primes. Nor can the
traditional approach deal with holophrases, or with meanings such as assertion and
questioning, which are patterns that cannot be reduced to semantic units, although
they have morphosyntactic units.
The theory of primes was based on a valid intuition, but took it too far; accord-
ing to Aitchison (2012, p. 97), it has been “mostly based on descriptive convenience
and wishful thinking”. Reliance on words is an example of that: word-like primes fit
morphosyntax very conveniently; but, as Jackendoff notes (2002, §11.2), it is not cred-
ible that the fundamental elements underlying the word meanings can themselves
be expressed as words. To Jackendoff, that is because those fundamentals are ab-
stract features. In the approach followed here, it is because they must be or resemble
perceptions.
8.5.2 Syntagmatic structure
The syntagmatic structure of word and morpheme senses has been discussed in each
section of the chapter. The discussion may be summarised as follows. In general, En-
tities relate to Events simply by being Entities; that is, their default link to Events is
what defines them: the physical things on which they are based are capable of action
or are liable to change, for example. Similarly Events default by nature to relating two
Entities; they are linguistic, and that is their linguistic nature. The specific syntagmatic
structure depends on the subclass of the senses, and on the syntax. For example, in
“The committee planned the coming year,” the relation of ‘committee’ to ‘planned’ is
that of Senser to the mental Process, because ‘committee’ is an animate Entity act-
ing as Participant in the mental Process, ‘planned’. The unmarked syntagmatic struc-
ture of Properties and Entities is generally that of a value bonding to an attribute in
the Entity.
There are several marked structures, however. The most important one is that of
nominalisations. Their syntagmatic structure is primarily that of an Entity, relating an
Entity as Participant to its Process, as in “Child protection is…”; but the sense has a
subordinate element of the Event class, relating it to its modifiers (as in “Child protec-
tion is…”, where the modifier ‘child’ is also Undergoer of the Event ‘protect’).
The paragraphs above have dealt with descriptive meaning: nondescriptive mean-
ing is very different. The link between the two senses is very loose, and without basis
in the semantic structure of the two senses; it is simply associational. For emotive and
attitudinal senses, the association is natural through the values of the speech com-
munity that uses the sense; for example, fat commonly carries disapproving meaning
because many speakers of English value slim physique. Social meanings are common-
ly quite arbitrary, however. For example, tassie – ‘cup’ in Scots English – has dialect
value, but saucer, in Scots use, does not.
As to the syntagmatic structure of grammatical morphemes: I have generally not
applied the concept to grammatical meanings, because it applies only where there are
units to be related to each other; in their system aspect, they are units i.e. relations,
with terms implied; they do not have syntagmatic structure, but constitute it.
8.5.3 Semantic classes: Content meaning
Content senses have been classified into semantic classes, which have been identified
(Chapter 4, §4.6) as those of Properties, Events and Entities. Semantic class is a sense’s
propensity to be linked with a member of a certain other class, and thereby serve a
certain function: for example, the nature of an Event is its potential link to an Entity,
to serve as Process. Psycholinguistically, that link is an entrenched link from the sub-
lexical concept which is the core of the sense, which can be readily activated to link it
to nodes of the appropriate class. That link, the bond, consists of a relation such as that
of attribute to value. Put crudely, basic semantic class = sublexical meaning + bond.
The chapter has noted subclasses of those semantic classes, and made the follow-
ing points. Property senses have two subclasses, the gradable and non-gradable, as
in Epithets and Descriptors (§5.2.3, §5.2.4). Entities have several subclasses: count-
able, quantifiable, abstract and so on, corresponding to the morphosyntactic classes
of count nouns, mass nouns and so on (§5.2.8.4). Events have the aspectual types
as subclasses: activities, achievements and accomplishments (§5.3.8.2). That outline
passes over such familiar classes as collective nouns, and issues which Wierzbicka has
discussed (1996, Chapter 13), such as “pluralia tantum” e.g. ‘scissors’ and ‘trousers’
which, she has suggested, should perhaps be rated as dual nouns.
There is a hierarchy of semantic class. At its base are sublexical meanings. They
have no syntactic or phonological reality – it is only lexical items which have that.
They have semantic content, but are barely distinguishable from cognitive happen-
ings, things and qualities, and do not have semantic relations such as antonymy, as
morphemes’ senses do. They thus do not belong to semantic classes, but have been
referred to as events, entities and properties or “qualities”. They belong to the classes
of Properties, Events and Entities when their potential links to other meanings are re-
alised in an utterance. They then have potential for a role and corresponding class at a
higher level in the semantic hierarchy, as Processes, Participants, and Circumstances,
as discussed above.
8.5.4 Grammatical meaning in words and morphemes
Grammatical meaning has been a major topic of this chapter, as of the previous one,
although it has seldom appeared in the headings; both chapters have dealt with hier-
archic structure, and it is grammatical meaning that builds the hierarchy. The definite
and systematic structure of the hierarchy, and the fact that English has carrying con-
tent and carrying grammatical meaning as complementary functions of morphemes,
might suggest that the grammatical meaning is also systematic, in the being carried
simply by morphemes.
However, the previous chapters have shown that that is not so. In the structure of
figures, the linking relations are represented in words (prepositions), but the transi-
tivity relations are not. Most modification is signalled syntactically (by position before
the head), not by words; but modification as changing degree of intensity is often
signalled by Reinforcers and submodifiers, such as very. Modal meaning is usually
carried by words (e.g. possibly), but is grammatical on some occasions (modal auxil-
iaries in Event groups, and the morphological subjunctive).
8.5.5 Discussion: Compositionality of word senses
This section deals with the compositionality of morphemes into a complex word – the
hierarchic structuring of elements into a unit on a higher rank; the compositionality
of sense elements into the networked sense of a morpheme was dealt with in Chap-
ter 5, §5.2.5.
As in that section, some degree of compositionality is intuitively obvious, as in in-
flectional forms such as book/books, derived forms such as rightful, and “compounds”
such as radiosensitive and radiotherapy. How far that extends, however, is not at all
obvious, and is subject to much controversy. For instance, ‘president’ no longer means
‘sitting down first’, and ‘lord’ no longer means ‘bread guardian’ (OE “hlaf weard”).
Conversely, the fact that mini and max have become words shows that, as prefix and
root, they had meaning which was composed into the whole word. I will not pursue
such compositionality further.
The chapter has discussed two situations in which senses are composed in a quite
different way from the standard one. The first is when words are used restrictively. As
we have seen, the final meaning of a word in that use is not so much composed from
its potential elements (i.e. descriptive elements), as that those elements are stripped
away, once they have served the function of enabling the hearer to identify the refer-
ent. The second situation is that of figurative use. Quite different operations are then
applied in the composition; the operations are moderately well defined, in accordance
with the nature of the relevant figure of speech; and the degree of composition is fairly
sharply limited by the dependence of figurative interpretation on context.
8.6 Discussion: Hierarchic structure in groups and senses
8.6.1 Language constraints
We saw in Chapter 4 that English, like other languages, is strongly constrained by the
fact that, as realised, it is one-dimensional, existing only in the “horizontal” dimen-
sion. Consequently, it is highly discretised (Givon 1988, p. 278) into morphemes, each
carrying a relatively very small part of the total meaning. The constraint is countered
in two main ways we have met previously. First, the morphemes, while unitary mor-
phosyntactically, are far from being unitary semantically, commonly having complex
multidimensional meanings. The second is the semantic hierarchy we have studied in
this chapter and the previous one.
Two other useful devices have been studied, but not identified as countering line-
arity. The first is grammatical meaning, which guides the hearer in building hierarchy
out of the flow of words: overcoming that constraint is the reason for its existence.
The second is coordination, which breaks the flow by placing several items in the
same structural slot; see §8.2.9.5 above for pre-modifiers, and Chapter 7, §7.8 for En-
tity groups in apposition. The point also applies to coordinated heads in a group, to
coordinated groups in a figure, and to coordinated figures in a figure complex. The
apparently relentless forward flow of English is there temporarily reversed.
The theme of this section has been how English overcomes the linearity con-
straint; but we have seen in the morphology of groups that English has adopted
another constraint, of its own invention: that certain meanings must be expressed –
although the function of the language is to express what we choose to express. Those
required meanings are the conceptual content of the required grammatical inflec-
tions, such as PLURALITY, and PAST TIME.
8.6.2 Semantic change
The chapter has some noteworthy implications for the study of semantic change.
Some familiar types of change can be best understood through the analytical ap-
proach followed here, particularly through the concepts of types and dimensions of
meaning, as illustrated in §8.2.4.2. For example, “amelioration” and “pejoration” (see
Ullman 1959, p. 101, for example) are less independent and arbitrary when seen as
one stage in the addition of different types of meaning, or as a change in the value of
the affective type. Vague “widening” (Ullman 1959, p. 204) is seen more clearly and
precisely as having three forms: (a) addition of possible meaning elements on the
expectedness dimension; (b) change from abstract to concrete on the descriptive di-
mension; (c) change in the perceptual base and relevant image schemata for the word.
An example of form (c) is from perfect; its original perceptual base was the signing
of a legal document (sense <1>; it has gained new concrete bases in its botanical and
printing senses (<11> and <14>), and new more abstract bases in its mathematics,
music and grammar senses (<7>, <8> and <9>).
A final example of clearer understanding is in grammaticalisation. Grammati-
calisation has usually been treated as a process that occurs in individual words. The
analysis here shows that the recent trend to see those changes as part of larger changes
is correct, since they are related to greater complexity in the structure, refinements
of compositionality, the increased backgrounding of meaning, and the changing use
of signs in the semiotics of English. Further, this analysis confirms that the view of
grammaticalisation as the bleaching of meaning is inadequate; as a word grammat-
icalises, it gains meaning i.e. grammatical meaning, while losing descriptive mean-
ing. That grammatical meaning is sometimes already expressed elsewhere, as when
prepositions acquire the linking meaning expressed previously by syntactic order or
juxtaposition; and sometimes it is a new meaning, as when very gained its significance
of intensification while losing its significance of ‘true’. A major reason for the error has
been the old fallacy, mentioned previously, that being a “functional” or “grammatical”
item, and being a “content item” are alternatives – and, beneath that, the predilection
for explanation through word classes.
It can be illustrated from the grammaticalisation of very, as adjective. The origi-
nal sense, in Middle English, was “Truly entitled to the name, by having the essential
qualities.” (That elides <1> and <4a>.) It lost ‘truly’, then ‘entitled to the name’, being
by then reduced to ‘having the essential qualities’ (<4b>). Note the loss of conceptual
elements in stages – which is a precise description of vague “bleaching”. In the mean-
time, it began to be used emphatically, and gained the function of emphasising. Final-
ly, it lost the remaining conceptual elements, becoming purely functional: <7> “As an
intensive, emphasising….” Diagram 17 shows the changes in very, as a bar graph, as
in earlier diagrams of word history.
This semantic analysis also illuminates one less familiar form of change, internal-
isation/externalisation of meaning, as in the word lousy. In §4.4.1, it was said to have
begun as <1> “Infested with lice”. The point to be noted now is that the Entities which
[Emphasis] [Emphasis] [Emphasis]
Truly
entitled to Entitled to
the name the name
by having by having Having
the qualities the qualities the qualities
<1> <4> (a) <4> (b) <7>
Diagram 17. Grammaticalisation of very
do the infesting, i.e. lice, were included in the meaning. However, when it came to
mean <3> “Teeming”, as in “Shipping magnates lousy with the shekels” (SOED), the
Entities which do the “infesting” (shekels, in the citation) are stated separately, now
being external to the meaning. Part of the meaning, then, has been externalised. The
converse process can be illustrated from lousy, also. The original sense (“Infested with
lice”) evidently did not carry affective meaning; but by Late Middle English, sense <2>
had developed: “Vile, contemptible”, which is purely affective. Between the two, there
must have been (first) a use in which the established meaning, “infested with lice”,
occurred with dislike felt by the speaker but not included in the meaning; and (sec-
ond) a use in which the affective meaning of dislike was regarded as part of meaning,
combined with the descriptive sense as established meaning. The affective meaning
was internalised into the word.
This section and earlier ones on semantic change have been included in what is
otherwise a synchronic study, for two purposes. First, they are intended to help the
synchronic analysis, by giving another view of it. Second, they are intended to vali-
date the analysis further, by demonstrating that it illumines obscurity in traditional
semantic history.
8.7 Conclusion: Hierarchic structure in groups and senses
Summary. As shown in Chapter 7 and in this chapter, the semantic hierarchy of Eng-
lish has two main features. It consists of a series of layers, those of the sense, the
semantic group, and the figure; they are realised in the corresponding syntactic lay-
ers of the word, the syntactic group, and the clause. They have well-defined relations
between the layers, and among the units in each layer. The semantic hierarchy has a
clear and well integrated structure.
The second feature is that that simple basic structure is made much more complex
in several ways. There is a partially distinct layer above, that of the figure complex;
the units of paragraph/paratone, chapter and so on can reasonably be regarded as
realisations of inchoate higher levels still; and a still lower level, that of the meaning-
ful individual sound (Chapter 2, §2.5.6, and elsewhere). Further complexity arises
from the reasonably familiar phenomenon of rankshifting or “embedding”. The final

element of complexity is the differentiation of units, especially the constituent units
of groups, into more and more specialised types, different in both semantic type and
semantic function. That specialisation is greatest in Entity groups, and lowest in Prop-
erty groups.
Conclusions drawn. Grammatical meaning has been set out as a system in Chapter 6,
and has been applied broadly in Chapter 7 and in more detail in this chapter. In Chap-
ter 7, it set up a hierarchic structure with simple horizontal relations between the
groups, and a vertical relation to the figure as the abstract entity the groups constitute.
In this chapter, its structure has been more complex; that can be seen in the contrast
between example (57), which gives a sentence used previously, in standard form, and
example (58), which gives it with bracketing to indicate the grammatical relations.
(57) “The surrounding street environs comprise a variety of restaurants, conveni-

ence outlets and cafes.” (New Zealand Herald, August 10th 2013, page C1)
(58) [[The [surrounding [street environs]]] [comprise] [a [variety [of [restau-
rants,] [convenience outlets] and [cafes]]]].]
The grammatical structure of the whole sentence, “The surrounding street environs
comprise…”, can be seen more fully in Diagram 18. To reverse the usual emphasis on
content meaning, the grammatical meaning is represented in capitals. Thus the top
three lines represent the grammatical meaning, “Relate the groups in the sentence as
Carrier (i.e. ‘environs’), relational Process (i.e. ‘comprise’) and Attribute (i.e. ‘variety).”
The left side of the diagram shows that the hearer must make ‘environs’ definite, and
RELATE GROUPS AS -
CARRIER RELATIONAL PROCESS ATTRIBUTE
environs comprise variety

MAKE MAKE
DEFINITE DEFINITE
RELATE RELATE
RELATE RELATE
restaurants
surrounding RELATE
outlets
street
cafes
RELATE
convenience
RELATE
Diagram 18. Outline grammatical meaning of a whole sentence; (“The surrounding

street environs…”)
Identfiable
surrounding surrounding
street street
street environs
environs environs
environs
Diagram 19. Effect of grammatical meaning in an Entity group (“The surrounding

street environs”, in set-theory view)
must relate ‘surrounding’ to ‘environs’. The bottom right of the diagram represents the
listing construction, signalled by the comma and and in “restaurants, convenience
outlets and cafes”.
Diagram 19 represents the effect of grammatical meaning in “The surrounding
street environs” as an entity group in terms of set theory: it narrows down the refer-
ence from the wide scope of “environs” on its own, to the narrow scope of the whole
phrase. “Identifiable”, in the innermost circle, represents the effect of the definite arti-
cle. (Set theory fits reference reasonably well; but it ignores non-descriptive meaning,
and even with descriptive meaning, it is inapplicable to descriptive use.)
The grammatical meanings of modification given in this chapter and previous
ones are the equivalent of “dependency” in some theories of linguistics, such as Word
Grammar e.g. Hudson (1984). I suggest that they explain the phenomena being
described as “dependency” rather better. They provide useful distinctions between
different types and levels of dependence. They provide a rationale to avoid the arbi-
trariness by which it is decided that “verbs” do or do not “depend” on “subjects”, and
for the direction of the dependency, on which authors disagree.
Chapter 9
Hierarchic structure (3)

Information structure
9.1 Introduction
This chapter continues the study of hierarchic semantic structure, moving on from
the structure of morphosyntactically defined content units (in Chapters 7 and 8) to
the structure of content as such, which is information structure (Chapter 2, §4.1.3).
The chapter deals with the three forms of information structure in turn: first rel-
evance structure (§2), then orientation structure (§3), then salience structure (§4).
Discussion and conclusions follow, as usual.
The chapter does not deal with “dynamic semantics” (e.g. Stalnaker 1978), the
study of how the hearer’s mental model of the information being conveyed changes
progressively, sentence by sentence. (That is one of the exclusions noted in Chapter 1.)
There is also no discussion of syntagmatic structure, as in Chapters 7 and 8. The rea-
son for that is a consequence of the fact that the structure here is one of content – of
cognitive semantics; the syntagmatic connections are those of cognition – the logical
connections of thought – not the linguistic connections of sense and group structure.
9.1.2 Introduction to information structure
Information structure, as understood traditionally. “Information structure” (Gundel

2012) is much the same as “information packaging” (e.g. McNally 1988; Collins 2009)
and “information flow” (Du Bois 1987; Chafe 2001). (Even “information” has varied
considerably; see Harder 1996, pp. 49–52). All three terms stand for structure which
is usually thought to be outside syntax, although no well-defined area is identified for
it; Gundel (2012) notes that some writers place it somewhere in grammar, but that
some place outside grammar. However, Büring (2007) includes it in syntax, following
Jackendoff (1972).
The structures regularly include Topic and Comment, and may include others.
Topic and Comment may be illustrated from example (1).
(1) “£100m is a useful sum of money.” (The Economist)

“£100m” expresses the Topic, and “is a useful sum of money” expresses the Comment.
Most writers deal with the distinction as a matter of what is “given” (i.e. stated in the
text, or familiar to the hearer) and what is “new”. Beyond that, no other forms of
information structure are accepted by consensus; even the meaning of terms varies
considerably.
A good deal of the disagreement about information structure has arisen, I be-
lieve, because it has become well developed in English only in the last century or
two. Its being so recent has had the important effect that information structure has not
become fully grammaticised (realised in grammatical form), in that it lacks consistent
and unequivocal marking. It evolved by exploiting structures that developed for oth-
er reasons; for example, Topic is a reconstrual of morphosyntactic subject, and it is
signalled largely by signs which are not visible or audible, because they are not words.
Consequently, the nature of information structure is rather obscure. A further effect
of its being historically recent is that linguists have often passed it by, in favour of
syntax and morphology.
Information structure as understood here. Here, information is taken to be content of

all kinds: descriptive content such as images, concepts and facts, and affective content
such as feelings and attitudes; it can be represented by morphemes, groups, clauses
and so on; it is not to be equated with propositions (as in Lambrecht 1994, p. 5), which
are matters of knowledge or logic. As content, it excludes grammatical meaning. The
content being structured is that of cognitive semantics, rather than of linguistic se-
mantics; as a consequence, these structures do not relate consistently to those set out
so far in the book.
I will first define information structure ostensively, as the group of three struc-
tures introduced in this section; a conceptual definition will be given in §9.6. The first
structure is that of Topic and Comment (illustrated in example (1)), which may be de-
fined loosely as what a figure or similar chunk of English is about (Topic), and what is
said about the Topic (Comment). What unites those two correlated terms is that they
concern relevance; I will accordingly refer to the structure as “relevance structure”.
The next structure is based on Halliday’s distinction (2014) between Theme and
Rheme. The Theme orients the reader or hearer to what is to follow, so that structure
will be called “orientation structure”. In example (2), the Theme is underlined.
(2) “In most parts of the British economy, £100m is a useful sum of money –
enough to build half a dozen schools or even buy a couple of decent strikers
in the English Premier League.” (Original form of example (1))
The last structure is that of Halliday’s Rheme, which consists of several items of in-
formation arranged as flow of reader interest, with a culminating item. It consists of
everything after the Theme. In example (2), ‘£100 million’ is developed by the in-
formation about a useful sum, by the information about schools, and finally by the
culminating piece of wry humour about the importance of football in Britain. The
nature of that culmination varies greatly, with genre, speaker’s intention and so on,
Chapter 9. Hierarchic structure (3): Information structure 257
but can be generalised as a rise in salience; the structure will accordingly be called
“salience structure”.
I have divided Halliday’s Theme-Rheme structure into two structures, for the
following reasons. They are united only by the fact that, in Halliday’s view, the two
together make the whole clause; but that is a morphosyntactic issue not a semantic
one. More important, they have different functions, as my names for them indicate.
“Theme” and “Rheme” are clumsy terms, being not self-explanatory, and subject to
much misunderstanding; but it seems better to keep to terms which are fairly well
established, through Systemic Functional Grammar.
Information units. All three of these are structures within an “information unit”. That
is an amount of information which the speaker presents as a unit, by grouping the
information together in an intonation contour, or an equivalent unit of writing such
as a clause. See example (3); it has five information units, underlined.
(3) “We deployed the panels and sank the spikes and sat it out, but we took a real
battering and there were telephone poles flying down on us.”
(Spoken by a man who observed a passing tornado from inside his vehicle.
New Zealand Herald, 15 April 2013)
There are usually several items of information within an information unit, and with-
in the Rheme; but in informal English, there may be only one item; by default, an
information item is a semantic group, with the semantic head as a focus. It is use-
ful to speak of “items”, but information cannot be reduced to items; as Lambrecht
notes (1994, p. 51), “information arises by RELATING something new to something
that can already be taken for granted” (Lambrecht’s emphasis); items in isolation are
not informative.
The scope of information units varies considerably. In speech, they are often very
short, and bear little relationship to clauses. Writers often use clauses as units to mark
off information units; but they often treat an introductory group as an information
unit (separated by a comma), which then acts as a rankshifted unit within the infor-
mation unit established by the rest of the clause. Similarly, an introductory subordi-
nate clause (with its own information structure) may be treated as an introductory
part of a larger information structure. Both speakers and especially writers sometimes
set up extended information structures, at the length of a paragraph, or even more.
Status in grammar. The status of information structure in the grammar of English

is, in my observation, very seldom specified. It is a problem – which writers seem to
avoid – because it is fairly clearly not a matter of syntax, nor of phonology, nor of lexis;
it does not fit into any accepted understanding of semantics; so it seems to be neither
linguistic content nor its expression. On the other hand, it does seem to be part of
language, not of knowledge or the structure of knowledge.
The status assigned to it here is that it is a structure of content, as content, and
as construed and expressed by the speaker. Specifying that it is a structure of content
as content distinguishes it from the structures presented in Chapters 7 and 8, which
are structures of content as embodied in morphosyntactic units. Specifying that in-

formation structure as the speaker’s construal and expression is important. It is not
inherent in the content itself, as logical structures (e.g. propositions) or psychological
structures (cf. Dubois 1987, p. 815). Further, speakers may choose not to use it, so not
all utterances have information structure. To identify it, we must read the signals the
speaker has put in the utterance, not work from the nature of the information.
A second element of the status of information structure is that it is built by means
of its own form of meaning – specifically, a form of grammatical meaning. The nature
of part of an utterance being marked as Topic, for example, is that it serves to guide
the hearer in interpreting and applying the utterance. It functions like a chapter head-
ing or a lecture title; it is implicit guidance from the speaker: “This is the area of your
knowledge which I am going to give you information about.” Being guidance makes it
grammatical meaning. Meaning must have a sign that carries it; because information
structure has developed recently, as noted above, those signs are not systematic, and
in some places are absent. In that, it contrasts markedly with syntax, which has fully
differentiated units, well-defined relationships, and fairly clear systematic symbolic
representation.
The signs are linguistic, but the relations set up are either cognitive, e.g. the about-
ness relation of Topic and Comment, or evaluative, as in Rhematic structure. The
linguistic relations of bonding, discussed in Chapters 7 and 8, are not affected, and
not replicated. (There will be no sections on syntagmatic structure in this chapter.)
That nature of its status provides further reasons why information structure has
not been discussed by linguists until recent decades, and has been not well under-
stood: it is a structure of content, not of words, groups or figures, and it is consequent-
ly hard to see – hard to distinguish, that is.
9.2 Relevance structure
9.2.1 Introduction
9.2.1.1 Definition of Topic and Comment

In general use, “topic” means what something is about, and “comment” means what
is said about it; I will sometimes use the words in that general sense, with lowercase
initial letters; but I will distinguish a technical sense of the terms. “Topic” (with an
uppercase initial) is content marked morphologically or syntactically to show that
it is to be treated as the topic of a sentence or other stretch of language. “Comment”
(again with an uppercase initial) is content “expressing information which is relevant
to… the addressee’s knowledge of ” a Topic. (The quotation is from Lambrecht (1994,
p. 131); but to him, the relevance is to a referent, not a Topic.)
Topic and Comment are terms in a relationship – a relationship of content; they
are not independent entities; especially, they are not morphosyntactic entities; al-
though the Subject very frequently states the Topic, other parts of a clause can state it.
The Comment is thus a correlate of the Topic; it is content other than the Topic, and
has no marking of its own: because Topic and Comment are complementary, hearers
can assume that everything not marked as Topic is Comment. (Strictly, Comment is
not the whole of the remaining content, since material in the Theme, to be discussed
below, does not belong to Comment.)
I emphasise that a Topic, as a linguistic entity, exists only when marked as such.
Every informative utterance is about something, so has a topic (with lowercase t), or
“logical subject”. That logical topic can be identified by direct statement, as in “The
next topic to be considered is…”, and “That’s an interesting question” as the beginning
of a turn in conversation. In neither situation is the topic marked linguistically, so no
linguistic structure is created: there is no Topic-Comment structure.
Marking something as the Topic, then, tells hearers that that is what the utterance
is about – what the utterance is relevant to – guiding them in structuring the informa-
tion in their minds, by identifying the point in their mental network of knowledge to
which the information should be attached. As noted in the introduction to the chapter
(§9.1), it amounts to an instruction, “Treat this as the point to which the information
being given is relevant.” The Comment gives the corresponding instruction, “Treat
this information as relevant to the Topic just specified, and fit it into your knowledge
of the subject accordingly.”
Speakers commonly use the same item as morphosyntactic Subject and
information-structure Topic, but the two need not coincide, as in example (4).
(4) “The people who were watching, we may criticise them for a lapse of decency.”
(New Zealand Herald, February 4th 2015, p. A2)
“The people who were watching” states the Topic; it is marked as Topic by the re-
sumptive pronoun them. The Subject of the clause is we, which, in the information
structure, is part of the Comment.
So far, Topic-Comment structure has been applied to statements. It applies also
to questions, as in “Is £100m a useful sum of money?”; the Comment here is informa-
tion asked for, not information asserted. It applies also to exclamations, as in “What a
useful sum of money £100m is!”; again, ‘£100m’ is the Topic; the Comment is its high
degree of usefulness. I take it that commands are not structured in this way; this is a
structure of information, applying when the speaker’s intention is to supply informa-
tion; commands are concerned instead with gaining action; Topic and Comment are
irrelevant, although a command may have a logical topic.
9.2.1.2 Utterances without Topic and Comment

Not all sentences or other utterances have a Topic; speakers do not structure all of
their utterances that way. Lambrecht (1994, p. §4.1.1) notes two kinds of utterance
that do not take Topic-Comment structure. The first kind occurs in main figures
where everything is of equal importance, and in focus – “event-reporting” sentences,
such as “The children went to school”, in answer to the question “What happened?” I
will deal with that reporting structure in §9.2.2.
The second of Lambrecht’s kinds occurs in many subordinate figures, since such
figures usually give the background to the Topic-Comment in the main figure. An
example occurs in “After the children went to school, he had to clean the house and
go shopping.” (“He” = John, already established as the Topic.) The utterance is about
John, not the children or their going to school; and the first figure has nothing to
mark a Topic, so there is no Topic – and therefore no Comment. There are also figures
without Topic-Comment structure of their own, but contributing to that structure
within a larger scope. For example, some figures introduce a Topic, without making
a Comment, like “It’s raining,” and “There’s a man at the door,” where the existence of
rain and of a man at the door are presented as Topics for the hearer to comment on,
or for the speaker to develop in the next utterance. These are “presentational” figures.
I will deal with all of those types in §9.2.3, as “loose structures”.
These types form a scale. At one end is the simple reporting type, in which there
is no differentiation among information items according to the relevance. At the other
end is the Topic-Comment structure, in which the items are fully differentiated and
the distinction is made consistently. The rest of this section is arranged according to
that scale.
9.2.1.3 Scope
The literature on Topic-Comment structure does not specify its scope, but the scope
is implicitly taken to be the figure. That is natural, because the structure is fully devel-
oped as a linguistic matter only at that level. However, the scope is seen to be much
wider if we allow that the structure may be even less systematic (regularly used and
paradigmatic) than it is in figures. In particular, paratones (Chapter 2, §2.5.2) are of-
ten structured with Topic and Comment. Perhaps we should also allow that book and
chapter titles state Topics – they certainly state topics. In units below the figure, Entity
groups are sometimes given a Topic-Comment structure, as will be shown below.
9.2.2 Reporting structure
Reporting structure is almost antithetical to Topic-Comment structure. Instead of

making a distinction between what is relevant and what it is relevant to, the text makes
everything equally relevant. This is the “event-reporting” structure of Lambrecht
(1994), where everything is “in focus”. It is illustrated in example (5).
(5) “A man who was forced to drive for more than 1 km after a gunman hopped
into his vehicle has told of the terrifying experience. Kieran Colmer was
driving to his parents’ house after work when a man carrying a gun opened
his passenger door and got in, instructing him to drive. The 22-year-old said
he could concentrate on only one thought during the ordeal on Sunday after-
noon. ‘I just drove and tried not to get shot,’ said Mr Colmer.”
(New Zealand Herald, February 4th 2014, p. A4)
In that example, almost every syntactic group gives new information. Where a group
refers to the same event or person as a previous one, it is given different expression to
give at least minor addition to the information. For example, the main participant in
the story is first “a man”, then “Kieran Colmer”, and then “the 22-year-old”. (Similarly,
“hopped into his vehicle” becomes “opened his passenger door and got in.”) In Topic-
Comment structure, that participant would be made a Topic, by use of a pronoun or
repetition of his name. Here, where we expect that sort of continuity, we are forced to
identify the same person in different expressions. Those variations are indexical signs
of the writer’s intention of reporting as much detail as possible, and of filling in all
parts of a two-dimensional picture as it were, rather than of creating a linear structure
of exposition. Although not conventionalised as part of the grammar of English, or
used consciously for this purpose, the signs instruct the reader, “Use this detail to fill
out the overall picture.”
9.2.3 Loose structures
One situation identified by Lambrecht (1994) fits neither of the structures discussed
so far. In it, the focus for speaker and hearer is on the Subject, not on the last informa-
tion in the figure, as in “The children went to school”, in answer to the question “Who
went to school?”. Such utterances have “argument focus” to Lambrecht; the focus,
being limited to one part of the information, is “narrow”. (Lambrecht includes an-
other situation, that of figures which change the topic, as in “No, my car broke down”
in answer to the question “Did your van breakdown?”. I rate that as having Topic-
Comment structure, with “my car” as the Topic, marked by the contrastive stress that
it would be given.)
In another form of loose structure, occurring in informal speech, all parts of the
information are articulated with logical topics and comments on them, but no con-
sistent linguistic Topic-Comment structure is established. Example (6) illustrates this
structure. (Key: // = tone group and information group boundary; ^ = silent beat.)
(6) Manageress: “// ^ in this job Anne we’re // working with silver // now silver
needs to have love // {Anne: // yeah //} you know the // ^ people that buy
silver // love it//”
Anne: “//yeah // guess they would //”
Manageress: “// yeah // mm // ^ well naturally I mean to say that it’s // got a
lovely gleam about it you know // and if they come in they’re // usually people
who love beautiful things // ^ so you have to be beautiful with it you know //
~ and you sell it with beauty /” (Halliday 2014, §3.5)
The overall information structure of the passage does not have the linear orderliness
of strict Topic-Comment structure, or the reporting structure’s pattern of filling the
canvas, but it has a real pattern nevertheless. It has two focal concepts, SILVER and
LOVE; they are linked in the central assertion, “Now silver needs to have love,” and
the other concepts attach to one of them; the structure works by centres of relevance.
This difference in structure results from a difference in the balance of speaker
intentions. The speaker here gave low priority to the ideational function of developing
topics logically, and higher priority to the interpersonal functions of engaging with
the hearer, and making certain emphases.
9.2.4 Topic-Comment structure
This section will concentrate on how Topic-Comment structure is marked – how
speakers signal that a certain part of the content is to be treated as Topic, and an-
other part as Comment. That is a little complex; the comparable matter of how mor-
phosyntactic Subject is signalled is simpler, since it has been conventionalised into a
paradigm.
9.2.4.2 Topic-Comment structure in figures

There are several types of marked syntax which are used to signal Topic status. Some
are illustrated in example (7).
(7) “I bet she had a nervous breakdown. That’s not a good thing. Gallstones, you
have them out and they’re out. But a nervous breakdown, it’s very bad.”
(Birner and Ward 1998, p. 5)
In the third sentence, the marked use of gallstones, preposed and without formal syn-
tactic structure, signals a change of Topic; so does the similar use of nervous break-
down in the last sentence. In the third sentence, the repetition of the sense ‘gallstones’
in them and they creates Topic continuity; in the second sentence, the element of rep-
etition in that also marks Topic status, while linking the two sentences. The example
also illustrates an expression of the Topic which is not in the Subject “you have them
out”. (Marked syntactic constructions are sometimes called “topicalisers”, even when
they are being used to mark something else, such as stressed Theme.) In example (3)
above, the ellipsis of the Subject marks it as Topic and creates continuity (“We de-
ployed the panels and […] sank the spikes and […] sat it out”).
In unmarked syntax, the marking of Topic is much less obvious. The main mark-
er is the use of pronouns, as in “… and they’re out,” in example (3). I emphasise that
pronouns are markers of Topic, not symbols of it; there is nothing in the nature of
pronouns to symbolise it, and there is no consensus in English that pronouns should
always have that significance. Their operating as markers relies on the assumption
that the referent is fairly salient in the hearer’s mind, and on the association of salient
referents with Topics. In presentational figures, the use of the dummy subjects it and
there marks their function of introducing a Topic. In speech, other devices are used
also: in example (8), pause was used to mark the African Union as the Topic, by giving
it phonological weight.
(8) “The African Union [pause] will increase its force in Somalia.”
(BBC World Programme news)
I believe that misrelation is gradually becoming accepted syntactically, with marking

Topic as one of its functions. Consider example (9). The context is that migrants are
travelling up the Nile to Sudan.
(9) “When Sudan approaches, after a cold night under the stars on deck, the men
roll up blankets and their wives braid the hair of young daughters. Squinting
into jewelled sunlight bouncing off the water, more and more of the shore
comes into view on both sides.” (The Economist, February 16th 2013)
There is no nominal group that can act syntactically as Subject for the Predicator
squinting. Fifty years ago that construction was quite unacceptable; “misrelation” was
a gross solecism, in formal English; now, it seems quite acceptable – clearly so to both
writer and editor, in example (9), and presumably so even to educated readers. I sug-
gest that the misrelation here marks ‘men, wives, daughters’ as the Topic of the second
sentence; omitting mention of them is like using they as marker: the writer assumes
that readers have those referents sufficiently in mind to relate ‘squinting’ to them, and
having them in mind constitutes their being the Topic. ‘Squinting into the light’ is the
Comment on it.
9.2.4.3 Topic-Comment structure in larger and smaller units

Topic-Comment structure in paragraphs and paratones
In writing, nonlinguistic graphological devices such as headings, typeface and font of-
ten signal information structure; so does paragraph indentation. It is not immediately
clear whether those devices are linguistic; but they should be regarded so, because
they are not only correlates of information-structure meanings, but signs symbo-
lising them.
It is well-established, I believe, that the tones which constitute the end of an into-
nation group represent the end of an information unit (Halliday and Greaves 2008),
and clause-final punctuation represents them in writing. In larger units than the
figure, phonology often works the same way: it marks a whole intonation group as
the Topic of a number of following intonation groups, which constitute Comment
(Halliday and Greaves 2008; and see Chapter 2, §2.5.2, on paratones). Change of
Topic is marked by phonological signs such as change of speed and change of pitch
(Halliday and Greaves 2008). Syntax also has a role: one of the functions of pronouns
is to keep a Topic running from sentence to sentence.
Example (10) illustrates a paragraph constructed as an information unit, with the
first figure as Topic, and the remainder as Comment.
(10) “Tourists are crazy – we see some old city walls leading up to a fortress at the
top of an incredibly high clifftop so, naturally, we have to climb them. The
word masochist must have been coined by a shepherd on his way up St John’s
Hill to the fortress at the top, 260 m above sea level.”
(New Zealand Herald, January 27th 2015, travel section p. 3)
‘Tourists are crazy’ is marked as Topic informally, by its relation to the remaining
figures: they amplify it, like figures with a resumptive pronoun.
Topic-Comment structure in groups

The Topic-Comment structure is also, but less commonly, seen within a nominal
group, as in example (11).
(11) “A few optimists believe they could finish the job by 2017. The especially
cheerful Mr Bhattarai predicts it will come much sooner.”
(The Economist, June 22nd 2013, p. 27)
‘The especially cheerful’ is not needed to identify the referent (Mr Bhattarai); it is
there as extra information; it is Comment about Mr Bhattarai as Topic, just as it would
be if put in a subordinate figure: “Mr Bhattarai, who is especially cheerful, predicts….”
Proper nouns do not normally take an article; so the, in “The especially cheerful…”
marks the anomaly, and signals the Topic-Comment structure. The signal is in fact
quite precise – indeed, it is a symbol: it is possible to use the indefinite article in such
constructions, as in example (12), an invented variation on example (11), with the
context changed.
(12) “A few optimists believed they could finish the job by 2017. An especially
cheerful Mr Bhattarai predicted it would come much sooner.”
The remark about Mr Bhattarai is no longer generalising, giving a characteristic of the

man, which is symbolised by the definite article in (11); it is commenting on his state
of mind on that occasion, which is symbolised by the indefinite article.
Occasionally, premodifiers within a group can contribute to information struc-
ture in a less regular and almost idiosyncratic way, as in example (13).
(13) “Although another 5p in the pound may not drive many Britons to move their
families from the Home Counties to the Gulf ’s arid shores, it will encourage
foreigners to go elsewhere instead.” (The Economist, February 1st 2014)
“Arid” is redundant, to the extent that it is unnecessary for identifying the referent; we
therefore seek some other significance for it. That significance is: Britons are unwilling
to move their families to the Gulf just because it, like its shores, is arid. Its redundancy
marks it as a signal; its meaning, to be interpreted cognitively from context, is an oth-
erwise unstated step in the argument.
9.2.4.4 Discussion: Topic-Comment structure

In the literature, the relation of Topic to Comment is often defined as the relation of
the “given” to the “new”. The great variation in the meaning of those terms is evidence
of their being unsatisfactory, and in fact the definition simply does not fit the evidence.
Comments are frequently “given” material, not “new”, because in exposition writers
frequently make the Comment an explication of foregoing material – that is what a
good deal of expository writing consists of; and in general use speakers frequently
repeat a point. Topics are frequently new, in all varieties of English, because speakers
and writers quite often establish something new as Topic, as they introduce it.
An example of new information as Topic occurs in example (9) above, in the
last sentence: ‘The word masochist…’ Example (14) illustrates figures in which all the
Comment is given. It comes from a discussion of clause structure, which has already
included secondary clauses, parataxis, hypotaxis, expansion, cohesion and so on. (The
Comments are underlined.)
(14) “Conjunctions may be used to mark the secondary clause in both parataxis
and hypotaxis; but different classes of conjunction are used. With parataxis,
linkers are used, but only when the logico-semantic relation is one of expan-
sion (e.g. and, more, but). Linkers may also serve a cohesive function.”
(Halliday 2014, §7.3)
Halliday (2014, p. §3.5) formulates the perception underlying “given” and “new”
much better as “newsworthiness”. But even that formulation is not precise; the issue
is what the speaker wishes the hearer to attend to most closely, whether it is news-
worthy or not, and whether or not the hearer even wishes to attend to it. Lambrecht
(1994, p. 206) is close to supporting that, in saying that the issue is what the speaker
is asserting; but “asserting” suggests that the issue does not apply in questions. There
is nothing in the content of Topics or Comments which characterises them; it is the
speaker’s signals to the hearer which characterises them.
Some other writers see the creation of Topic-Comment structure as the speaker’s
attempt to adapt the utterance to whether the information is active in the hearer’s
mind: Vallduví (1992, as cited Ward and Birner 2011, Chapter 73); Du Bois (1987,
p. 815 – “concept activation state”). That makes the matter psychological, not linguis-
tic; like other versions of the given and new concepts, it does not fit instances such
as example (13), where the writer knows that all concepts are active, but makes one
Topic and others Comment.
9.2.5 Conclusion: Relevance structure
Topic and Comment give structure to the utterance by constructing a relationship

between two complementary parts of the content. To emphasise: the two parts are not
independent – there is no Topic without Comment, and vice versa, just as, with the
relationship of fatherhood, there is no father without a son. The units are not syntactic
ones, such as Subject or Adjunct, but parts of the content.
The existence of logical topic and comment on it is so nearly universal in inform-
ative English, and the existence of linguistic Topic and Comment is now so familiar
to speakers and hearers that the structure needs minimal marking. Specific signs sig-
nalling the Topic are common, nevertheless, in a variety of syntactic and phonological
forms, although not in the form of lexical markers, as in some languages. The exist-
ence of Comment is signalled only indirectly or negatively, through the presence of
Topic as only part of the utterance. Although the Topic-Comment structure is created
by a sign or signs, the structure is itself a sign, since it has significance; namely, the
way in which the hearer is to relate the content to the knowledge which the cumula-
tive meaning of the text is building up; that is a grammatical meaning. To elaborate
that point, made in the introduction: it is an instruction from the speaker, and a cor-
responding procedure for the hearer to carry out. It can be paraphrased as: “Treat the
content marked as Topic as denoting the area of knowledge to which the utterance is
relevant, and treat the remaining content as the information that is being asserted as
relevant to it.”
This relevance structure has a formal linguistic reality within figures realised as
syntactic clauses. There is often a similar structure in paragraphs and whole texts, but
at those levels there are no unequivocal symbols, but linguistic markers, and often
marginally linguistic signs such as layout and section numbering.
9.3 Orientation structure: “Theme”
9.3.1 Introduction
Speakers begin utterances with an introductory section that orients the hearer to
the information to come – the Theme (Systemic Functional Grammar e.g. Halliday
(2014); MacWhinney (1999); see Hasan and Fries (1995) for useful discussion. (The
term is not very helpful, since its sense here has little to do with its usual sense in
everyday usage, and it has a various senses in the linguistic literature.) For example,
guides on a bus tour orient their passengers by saying things like “On your right, …”.
In a narrative, “Later, …” will warn the hearer that the story is about to jump forward
in time. In description, “In the background, …” is like an instruction on where to
place the coming details in your mental “picture”. The Theme is complemented by the
Rheme – the rest of the utterance – in which the speaker traces a “route” through the
information from the Theme as starting point.
Just as Topics are a formalisation into grammatical structure of the natural prac-
tice of starting an utterance by telling the hearer what it is about, so Themes gram-
maticise our natural practice of ensuring that hearers are mentally ready for what we
say, as we begin.
Themes are not marked by any construction or marking word; they are signalled
simply by their initial position, except that the independence of their function is
sometimes marked by a comma or by a corresponding intonation unit and pause in
speech (Halliday 2014, §3.5). In unmarked use, the Theme extends to the end of the
Topic or Subject; for marked use, see Halliday (2014, Chapter 3).
9.3.2 Classes of Theme
Introductory note. Themes may be divided into three general classes, to illustrate the
range of what they do. The classes are not categories (mutually exclusive classes with
abstract definitions), and do not have distinctive markings that differentiate them.
The exposition here is based on Halliday (2014, §3.4).
9.3.2.1 Textual Theme

Textual Themes orient the hearer or reader to the context, either textual or situational
context. Often, they orient the reader figuratively by orienting the action literally, as
in example (15), from a newspaper report of the effects of a flood.
(15) “Fourteen kilometres north, the funeral procession… was unable to pick up
mourners…. Six and a half kilometres east of Chertsey, the fire brigade was
around the corner.” (New Zealand Herald, February 12th 2014, page A22)
In that example, the first Theme orients hearers by establishing a relationship between
the forthcoming information and the previous information; in the second Theme, the
relation is with something outside the text i.e. Chertsey. (This is the “textual function”
of Halliday (2014).) Modal Adjuncts also have this function, as in example (16). In its
context, it followed data suggesting that California had not made great progress under
the governorship of Jerry Brown.
(16) “In fairness, Mr Brown has not been idle. He has adjusted school finance
formulas.” (The Economist, November 30th 2013, p. 36)
The Theme here orients readers to the countervailing evidence that is to come.
9.3.2.2 Interpersonal Theme

Some themes orient the hearer in a quite different way: they establish or confirm
the relationship between a speaker and hearer; they include greetings, farewells and
vocatives such as “John, …”. Fairly obviously, that applies in writing when they are in-
cluded in the sentence; in speech, it applies when they are included in the tone group
(cf. Halliday and Greaves 2008, §5.2.2.6); otherwise, they constitute an utterance on
their own.
It is very striking that the “speech acts” of exclaiming, commanding and so on are
signalled in the very first morpheme of the utterance. Exclamations begin with a wh-
word (“What a …!”, or “How …!”). In commands, the finite Event word comes first,
which is otherwise impossible in English. In questions, the first morpheme is either

a wh- word (what, where, when etc.) or the finite auxiliary (e.g. “Are you coming to-
night?”). Those morphemes act as Themes, orienting hearers to the response expected
of them, before they hear any of the content.1 Statements have the default orientation:
if none of the three previous conditions are met, the hearer must expect a statement.
Some “discourse markers” act as Themes, such as well, look and oh, in some uses.
Their intonation sometimes indicates which of their meanings is intended, but of-
ten provides a thematic effect independent of the lexical meaning, by expressing the
speaker’s affect or attitude, and thus guiding hearers in how they should respond to
the coming content.
9.3.2.3 Topical Theme

Topics, since they state what utterances are about, naturally constitute a good means
of orienting hearers; they serve the orientation function through their relevance func-
tion. Topics have been illustrated already, but I give example (17), for a particular
purpose.
(17) “Of the African states surveyed by the International Budget Partnership (IBP),
a pressure group, 24 turned out already to produce 58 of the budgetary doc-
uments needed.” (The Economist, November 2nd 2013, page 60)
Semantically, the phrase underlined modifies ‘24’, the Topic of the sentence – “24 of
the African states…”. It has been removed from its syntactic position and placed first,
to be Theme. (Compare Quirk et al. 1985, §17.50.) Readers may be sceptical of the
reality of Themes, especially since they have no distinct marking such as inflection;
but example (17) shows that Themes have considerable power, since they produce
syntactic distortion.
Since figures regularly begin with a Topic, or at least with a Subject as logical
topic, Topics constitute a default Theme, so that there is a Theme in most utterances,
even if the speaker provides no textual or interpersonal one.
9.3.2.4 Discussion: Types of Theme

The three Theme types occur together, in various combinations. Halliday gives exam-
ple (18) as an extreme.
(18) “Well, but then surely, Jean, wouldn’t the best idea be to join in?”
(Halliday 2014, §3.4)
1. One value of do as a prop word is that it provides a distinctively interrogative morpheme to

come first in questions, as in “Do you know him?” The older alternative was to invert the finite
Event word and pronoun, as in “Know you him?”; but that left a finite Event word as the first
word, leaving the hearer momentarily in doubt as to the proper orientation – to a command or
to a question.
‘Jean’ is a vocative interpersonal Theme; ‘surely’ is a modal interpersonal Theme; ‘the

best idea’ is a topical Theme; ‘but then’ operates as textual Theme; ‘well’ is an inter-
personal Theme.
Themes are sometimes chosen quite deliberately, but are often chosen intuitively,
showing the speaker’s perspective or preoccupation, as in a speaker’s choice between
“I’ve got a headache”, which indexically signals preoccupation with the self, and “My
head hurts”, which signals the speaker’s preoccupation with his head (Halliday 1989).
The Topic frequently provides a Theme, even when the speaker has no conscious in-
tention of including a Theme.
Choice of Theme is also related to genre: narrative typically has temporal themes
like “Then, …” or “Later, …”. Description typically has spatial or locative themes like
“In the background, …” and “To the right, …”. Exposition typically has abstract and
logical themes like “Therefore, …” and “As a result, …”.
The Theme’s role as orientation to the coming content has a parallel in the role
of sentence, clause and phrase beginnings as orientation to the syntactic structure,
signalling to readers how they will probably have to parse the structure they are em-
barking on. (See Hawkins 2004.)
9.3.3 Themes occurring at ranks other than the figure
9.3.3.1 Ranks above the figure

As with topics, Themes are often provided for passages of text greater than the figure,
right up to whole books, where epigraphs (briefly) and some prefaces (more lengthi-
ly) orient readers to the text explicitly. Some headings above a text or within it serve
similarly.
Grammatically expressed themes can apply to a whole article and to paragraphs.
A report in a news magazine (The Economist, July 6th 2013, p. 28) had “Bhutan at
the polls” and “Debt and discontent are growing” as subject headlines, and “Happy
and you know it?” as a Thematic one. The text began, “Arrows thud into a wooden
target.” That acted as the Thematic sentence for the first paragraph, about Gyeltshen,
an 89-year-old who was watching archery. That paragraph and the next, both about
Gyeltshen and his attitudes to change, acted as Theme for the article about growing
debt and unhappiness in Bhutan. See Fries (1983) for supporting discussion.
It was noted in the previous section that many “discourse markers” such as well,
look and oh often act as Themes for a figure. They also often serve that function for
turns in conversation (as well as claiming the turn), letting the hearer orient himself
or herself to the approach which the speaker will take.
Conjunctions state the conceptual relations between figures in a figure complex,
but they also serve as themes – textual themes. And, but and or signal that readers
must co-ordinate the following information with previous information; because,
when, if and so on signal that the coming information must be subordinated. Relative
pronouns are similar: which and who as relatives prepare readers for descriptive
(subordinate) figures, just as they prepare readers for a question when they are
interrogatives.
9.3.3.2 Ranks below the figure: Semantic group

As Halliday notes (2014, §6.3.1), groups often have Thematic structure also. Just as
topical Themes, by being referential, ground the figure in the hearer’s experience, so
Entity groups are given Thematic grounding by a determiner, through its deictic ef-
fect. Consider the beginning of an example just above: “Of the African states surveyed
by the… IBP, a pressure group, …”. The underlined the warns readers, as it were, that
in the coming group they will have to search their minds for a referent; the underlined
a lets them know that they will not have to.
Event groups have themes in a similar way, in the Finite. Being grounded in time,
it correlates with the nominal group’s grounding, which is in literal or conceptual
space. Example (19), from a news report of an accident, illustrates the point.
(19) “It looks like they were going across some waterfalls and the mum fell with
the baby from 10 m high.” (New Zealand Herald, June 2nd 2015, p. A1)
The past tense form, were, sets the rest of the Predicator, and the rest of the report, in
past time; it shifts the orientation away from the present relevance set by the present
tense of the main Predicator, looks.
9.3.3.3 Conclusion
Initial position is Thematic at all levels of analysis, from text to group. It accords with
the familiar discourse principle that what is urgent comes first.
9.3.4 Discussion: Orientation structure
9.3.4.1 Other views

Theme has been given little attention by that name (though Lambrecht 1994 uses it),
but its presence and nature has been noted often enough. Thompson (1978) and Fries
(1983) record it as “orientation”; Chafe (2001) describes it, as part of “discourse flow”.
9.3.4.2 Speech participants as themes

Misrelation, illustrated in §9.2.4.2 with the function of setting a Topic, also serves to
set Theme. Consider examples (20), (21) and (22).
(20) “For the avoidance of doubt, the subsidy available for long-term residential
care is only payable for up to 90 days.”
(From the admission agreement for an old people’s home)
(21) “Repeating an observation made in other chapters, what is typologically
interesting …, is that …”. (Dingemanse 2011, §362)
Those examples contrast with the older style, in example (23).
(22 ) “Returning now to the question of fatherless children, we find among the
Trobrianders a trend …” (Malinowski 1929, p. 169)
As the example (22) illustrates, with “we”, it is the author, or the author and the reader,
to whom the prepositional or participial phrases we are discussing must be related.
Thus in example (20), the avoidance of doubt is being carried out by the author for the
sake of the reader; in example (21), it is Dingemanse who is repeating the observation.
The practice here, then, is for writers to leave implicit a relation to participants in the
speech situation, assuming that readers are conscious enough of those participants to
relate them easily to prepositional and participial phrases. It has for long been axio-
matic that speakers can assume that the hearer is aware of the speaker-hearer relation
in spoken dialogue; but now writers assume that readers, also, have that degree of
awareness.
That is part of a wider trend over the last 500 years, noted by Halliday (e.g. 2000,
pp. 228–229). It is a “new alignment of grammatical forces” (2000, p. 228), leading
to organising discourse as a flow of information (as opposed to expression of knowl-
edge), and towards regarding writing as personal communication, as speech is, not
just as a treatment of subject matter, displaced from context. English has developed
more resources for information structure, and has given greater prominence to it.
9.3.5 Conclusion: Orientation structure
A Theme is a signal, guiding hearers to the best orientation for comprehending or ap-
plying the content of what follows. The orientation may be textual, towards previous
text; or it may be interpersonal, between hearer and speaker; and it may be ideational,
towards the knowledge being conveyed (through Topic). As instruction to the hearer,
it is fundamentally a grammatical meaning.
Semiotically, Themes as grammatical meanings are represented in an unusual
way. First, they are represented, not by a word or other physical sign, or by syntactic
structure, but by position; initial position in the content is a symbol of being Theme,
just as following a Topic is a symbol of being Comment. Second, whereas the inter-
rogative and some other elements of syntax are signalled by the position of words or
groups – syntactic units – the signal here is the position of a content unit.
Theme, as the realisation of orientation structure, is moderately widely recog-
nised in the literature as existing in figures, but it is not so recognised in groups, where
the article, auxiliary Event word or preposition which introduces the group orients
the hearer in a similar way. As with relevance structure, there is often such a struc-
ture in paragraphs and whole texts, sometimes created in the same way, with readers
taking their orientation from the initial content, but sometimes created cognitively,
through layout or bibliographical means, not linguistic ones.
There are many variations in the thematic structure of English, such as “It was
Jane that started it,” where the Theme is “predicated”. (That is one example of so-called
“topicalisation” where the function is to adjust Theme, not Topic.) For detail, see
Halliday (2014, Chapter 3), for example.
9.4 Salience structure: Rheme
9.4.1 Introduction
Definition and description. “Salience structure” is the pattern of relative importance of

the items of information, which the speaker creates in an utterance to guide hearers
in evaluating the information, and in assigning to it the importance which the speaker
intends. The pattern is carried by the “Rheme”, which is the section of content follow-
ing the Theme, and normally including the Process, any subsequent Participants, and
any Circumstances. Example (23) illustrates the normal structure. It is from a news
report discussing prospects for farming exports; Rolleston is a farmers’ leader. The
Rheme is underlined, and divided into items of information.
(23) “On the positive side, Rolleston | noted | the dollar | had become | more
exporter-friendly |.” (New Zealand Herald, January 29, 2015, p. B20)
“On the positive side” provides textual Theme, and “Rolleston” provides topical
Theme. Each item of information consists of a group, and is marked phonologically
by phrasal stress. ‘Noted’ had been repeated several times already, and ‘the dollar’ is a
given in any discussion of exporting; so both items had little communicative impor-
tance. ‘Had become’ denotes a change, so was important. ‘More exporter-friendly’ af-
fects farmers’ incomes; it was the most important item – the “information focus”. That
focus is “what the speaker is drawing particular attention to” (Halliday and Greaves
2008, p. 103). It is expressed – symbolised – phonologically as the main stress in the
figure. Its final position is a marker of being focus – a default indicator, not wholly
reliable, but generally the only one in writing.
Example (24) illustrates an exceptional structure, where the Process (‘were’) is
not part of the Rheme; as a copula, it carries no information so is outside the salience
structure.
(24) “The prices were | strong | but not fantastic |.”

(From the same report in New Zealand Herald as example (23))
Place in information structure. Salience structure complements the orientation struc-

ture: the Theme orients hearers to the information to come; the Rheme is that infor-
mation, and guides them in structuring the salience they are to assign to its various
items. The salience structure complements the relevance structure as well. The Topic
sets the centre of relevance, with the Comment setting out what is relevant to it. Just as
the Theme includes the Topic (the hearer is to be oriented to the centre of relevance),
so the Rheme gives a structure of salience to all that is relevant (in the Comment).
The nature of the importance depends on the speaker’s intention, and generally
accords with the genre: it may be logical importance in the argument, or dramatic
value in the narrative, or strength of feeling in expressive speech, for example.
Because the Theme generally includes the Topic, and the Rheme generally coin-
cides roughly with the Comment, it is worth emphasising another difference between
the two structures. Using a Topic-Comment structure is optional for the speaker, but
Theme-Rheme, being a default structure, is almost unavoidable and virtually universal.
Establishment of salience as grammatical meaning. Because its significance is not con-

tent meaning but guidance to the hearer in using the content, salience structure is
established by grammatical meaning. A verbal formulation would be, “Give the items
of information the degree of importance indicated by their position in the order, with
the stressed/last item being the most important.”
Terms used. Like “Theme”, the term “Rheme” is not helpfully self-explanatory; it has
remained current since its use by the Prague School, with changing meanings; but
it seems better to use an established term than to invent another one. Other terms
sometimes used for heightening and lowering salience are “foregrounding” and
“backgrounding”. I use “salience” in preference, since it allows more naturally for gra-
dation, and for speakers to make quick or temporary changes in an item’s importance.
The dominance of the terms of “given” and “new” in the literature is one reason
why previous writers have given little attention to salience structure: writers have un-
derstood the phenomena of salience in those terms, consequently subsuming them
under Topic and Comment. (Also, it has been ignored because it has been associated
with the Prague School.)
Structure of this section. The section deals with blocks of content of increasing size.
Salience of information within an information item is dealt with in §9.4.2; within an
information unit, in §9.4.3; and within a paragraph or whole text, in §9.4.4. §9.4.5
gives discussion, and §9.4.6 gives conclusions.
9.4.2 Salience within an information item
As noted above, each item of information typically consists of a syntactic group (which
may be realised as a single word), marked by phrasal stress; it usually therefore con-
sists of a Participant, an Event or a Circumstance. There are some exceptions, where
speakers use marked stress to divide or combine what would normally be individual
items of information. Since the marking is by stress, and groups have a set pattern of
stress, there is little scope for speakers to structure the item; but the increasing com-
plexity of Entity groups in modern English creates a need for it, and their well-defined
structure (see Chapter 8, §8.2) provides the means for it. This section deals with how
it is done.
We saw in Chapter 8 that nominal groups often refer to several Entities, not only
the central Entity denoted by the head. In the sentence “She researched children’s
language learning,” she and learning denote Entities which participate in the discourse
as Actor and Goal. However, the phrase “children’s language learning” refers not only
to learning , but also to language and to children, both of which participate in the
situation, and could become Participants in the discourse (if used as Subjects, for
example). I will accordingly distinguish between “discourse Participants” (with an
upper-case P, for a technical term), and “situation participants” (with a lower-case p,
for a general term). This section shows how speakers manage the relative salience of
the various situation participants they refer to.
9.4.2.2 How salience within an item is managed

Meaning type. Since grammatical words have little or no content, the content value of
a phrase rises and falls as content and grammatical words alternate. Rhematic value
“undulates” within the group accordingly. In “noted | the dollar | had become | more
exporter-friendly”, cited above, the content senses (in the underlined words) forming
little crests of salience.
Placing a Participant as a determiner, a premodifier, or postmodifier. In Chapter 8,

§8.2.2.2, we saw that writing “doctor parking” instead of “parking for the doctors”
reduces the degree to which we conceptualise the doctors as unique individuals – re-
duces their individuation. Reducing the salience of Participants within an informa-
tion item is the main use of that device.
Examples (25) to (27) illustrate the alternatives, with ‘parent’ used in a determin-
er, a postmodifier and a premodifier.
(25) “He would have to swear an oath that denies his parents’ rights.” (COCA)
There, the parent is the subject Participant, and is not only a specific person, but also
central in the discourse. Accordingly, the parent is kept salient in a determiner group,
and so referential and fully individuated.
(26) “It is simply wrong that the video game industry can be allowed to put their
profit margins over the rights of parents and the well-being of children.”
(COCA)
There, there are no specific parents, so the concept of parents is reduced in syntactic
status by being in a modifier position, and reduced in semantic specificity by being
indefinite; but a contrast is needed between parents and children (mentioned in the
following phrase), so the participants retain some individuation, by being expressed
in the plural form, and in a whole, though rankshifted, group.
Table 1. Alternatives for ‘parents’, with progressively lower salience

Degree of salience Determiner Premodifier Head Postmodifier
Highest the parents’ / their rights
Medium the rights of parents
Lowest parent rights
(27) “In a survey, issues of confidentiality and parent rights were found to be
among the most common ethical dilemmas.” (COCA)
In that example, parents are in the background, not conceived as individuals, and not
contrasted with any other group; accordingly, they are given minimal salience, the
concept now being generic and abstract, and expressed with full morphological and
semantic reduction, as “parent” – singular in form, although plural in reference. (The
adjectival form, parental, would suit the function equally.)
Those variations are tabulated in Table 1.
Controlling salience of Properties. Choice of position in the Entity group is used to

control the salience of Properties, as well as of Entities. In example (28), the Property
of being able to be carried has low salience. A standard expression of it is given in (b),
as a prepositional phrase. In (a), the phrase is reduced to a single word; and the Event
word carrying is reduced to carry , losing its transitivity and aspect features; it has
become abstract and generic, with minimal salience.
(28) a. “It has a comfortable carry handle.” (COCA)

b. “… a comfortable handle for carrying it.” (Invented alternative)
Salience of participants in postmodifying group. In example (29), the postmodifying

section of the complex Entity group is given a Rhematic structure. (The three post-
modifying groups are underlined.)
(29) “But the fatal shooting on May 9th of a 65-year-old Taiwanese fisherman by
the Philippine coastguard has shifted Taiwan to centre stage.”
(The Economist, May 18th 2013, p. 29)
The example comes from a discussion of international relations in east Asia, so the
Philippines coastguard has been placed in focal position within the postmodifiers to
set it against Taiwan, later in the figure. The “of…” group has been separated from
“shooting”, rather unidiomatically, to give “…Taiwanese…” more salience, and “on
May 9th” less salience.
9.4.3 Salience of items within an information unit: Rhematic structure
9.4.3.1 Marked and unmarked Rhematic structure

Unmarked use. As noted above, figures have by default a structure of increasing sa-
lience – a “Rhematic structure” – with information items as the units of structure.
The most salient item is the information focus, realised and marked in speech by the
phonological focus, which is created by the strongest stress in the information unit. In
unmarked use, the strongest stress is the last phrasal stress. That Rhematic structure
is the default in English as the “principle of end weight”; it is difficult to avoid using it;
it is part of English grammar.
The “weight” of the content is important, as well as the position; any item of in-
formation gains in salience for hearers as it gains in detail, descriptive force, or extra
layers of meaning of different types. Speakers give an item of information that detail
and force for the very reason that it is salient to them; that is why length has become
a marker of weight, and particularly of information focus. (I take this to be what is
usually meant by “weight” in “end weight”, although some writers seem to include
syntactic complexity.)
In writing, the structure is the same, although there is no distinct signal that
matches stress in speech. I believe that it is in fact marked below full consciousness,
however, by subvocalisation – the mental process in which the mental patterns for
vocalisation are activated, without the motor nerves and vocal organs being activated,
during both hearing and reading; see Feist (2013, §3) and the references there.
Marked use. There are several sorts of marked Rhematic structure. Example (3)
above had as one Rheme; “There were telephone poles flying down on us”. The usu-
al structure would have only “Telephone poles were flying down on us” as Rheme;
the marked use of “there were” (normally a presentational construction) functions to
make a whole figure constitute the Rheme, to give the telephone poles more salience.
A more familiar marked use is the use of stress to move the focus forward, as in the
invented expression, “No, I gave Jack the ticket,” where the stressed word Jack is the
whole of the Theme. Another marked use is predicating the Theme, as in “It was Jane
that started it” (Halliday 2014, §3.7).
In example (30), the speaker was emphasising that you could reduce your taxes by
claiming deduction for even minor things.
(30) “You can tax deduct your tea and coffee.” (Spontaneous conversation)
The speaker’s alternative expression would have been “You can deduct your tea and
coffee [from your income] for your taxes.” Her actual expression reduced a Circum-
stance group syntactically to a single word on the lower rank of modifier; it is morpho-
logically reduced, to singular form, and semantically greatly reduced in individuation
(see Chapter 8, §8.2.2.2). As noted previously (Chapter 8, §8.3.6.3), those changes
constitute “noun incorporation”. The incorporation here looks very different from the
incorporation studied traditionally, which occurs in polysynthetic languages; but the
difference occurs simply because English is largely analytic, not polysynthetic. More-
over, traditional study, as in Mithun (1984), asserts that control of salience is one of
the two main functions of incorporation; Riehl and Kilian-Hatz (2005) state that it
is the main function in the African and European languages they studied. The use of
incorporation in many forms has been increasing rapidly in recent years, especially in
informal English. (See Feist 2013 for further discussion.)
9.4.3.2 Examples of Rhematic structure

In examples (31) to (33), the Rheme begins after the vertical bar; the phrasal stresses
have single underlining; the focal stress has double underlining.
(31) “Police | are investigating the suspicious death of a man whose body was
found in the Tauranga suburb of Merivale yesterday.”
(New Zealand Herald news report)
The Rhematic structure runs right through the rankshifted figure (“whose body…”),
and is written to take the stresses rather colloquially, not matching the syntax precise-
ly. Readers of the report might think that the fact of finding a dead body was more
salient than the fact that it happened “yesterday”; but the writer clearly did not intend
to make it so, since “yesterday” could have been placed earlier in the sentence. (Jour-
nalism emphasises what is recent.)
Example (32) followed example (31).
(32) “The man, […] | was found by neighbours beside Merivale School about 10
a.m.” (New Zealand Herald report)
“Was found” effectively drops out of the Rhematic structure, having no information
value because it repeats an item from the previous sentence; it would be unstressed
in speech. The information focus is again on the time, building a strong narrative
structure for the text.
The ellipsed part of (32) is given as (33).
(33) “… | believed to be in his early 50s….”
The ellipsed portion, a subordinate figure, has its own Rhematic structure. Its Topic
has been omitted, again because it has no information value, as obvious; similarly, it
has no Theme, for much the same reason.
Examples (34), (35) and (36) illustrate the effects of position by contrasting three
uses of ‘stupidly’.
(34) “Tall, thick branches fortified with sharp thorns snagged my clothing and tore
my flesh. Blinded by darkness, I had stupidly run through Dad’s rose garden.
‘Ow,’ I cried.” (COCA)
‘Stupidly’ has little salience, because the word is syntactically subordinate to “run”, in
the group “had stupidly run”; in speech, “run” would take the phrasal stress, leaving
“stupidly” unstressed.
In contrast, an item of information is Rhematically stronger when it is placed as a
separate group, as in example (35), from romantic fiction. Stupidly is now more sali-
ent, to continue the effect of “speechless” in the previous sentence; but it is not placed
at the end, as is possible, because the dazzling eyes are to be the focus, to lead on to
further erotic description.
(35) “Casimir was speechless. He stared stupidly into her dazzling eyes. Again she
pressed lightly against him, raising herself slowly to tiptoe.” (COCA)
In example (36), ‘stupidly’ is placed to constitute the information focus and be a cli-
max, although it could have been placed earlier. A hung-over student describes a
lecture.
(36) “If words were balloons, these words were floating up to bounce against
the ceiling of the windowless fluorescent-lit lecture hall, colliding with one
another and drifting about, stupidly.” (COCA; fiction by Joyce Carol Oates)
The Rhematic build-up achieved by the syntactic placement of stupidly is aided by the
semantics: stupidly is now primarily emotive, whereas it is primarily descriptive in the
previous examples.
In example (37), the concept of stupidity is taken right out of the Rheme.
(37) “But following the post-Guns-’n’-Roses incident, I had realized how my stu-
pidity had endangered not only my life, but also that of my friends.” (COCA)
The speaker’s stupidity has been backgrounded as assumed information by being

made the Topic of its figure; “my stupidity had endangered not only my life” could
have read “I had not only endangered my life stupidly …”.
9.4.3.3 Discussion: Salience of items within an information unit

The nature of Rhematic structure can usefully be illustrated negatively, as in example
(38), which comes from a notice at the front door of an old people’s home.
(38) “Please reverse wheelchairs down the ramp.”
The sentence was written as if it were speech; the stress was intended to fall on reverse,
the point being that if you push the chair forwards you might lose your grip on the
handles. But as we read the sign, we take the stress to be on on “down the ramp”, and
we wonder, “Where else would you reverse the wheelchair, if not down the ramp?”,
and “Why would you reverse it at all?” The misreading demonstrates the reality of
Rheme in English grammar.
Apart from example (38), the examples discussed have all been statements. Other
types of utterance also have Rhematic structure. For example, a question about the
crime report in example (32) could be either, “Was the body found beside Merivale
School about 10 a.m.?”, or “Was the body found about 10 a.m. beside Merivale
School?”. “No, meet me outside the post office at 6 o’clock,” would be a command
with marked Rhematic structure (where bolding marks contrastive stress).
The account of salience structure given in this section is supported generally but
indirectly by DuBois (1987); his “Preferred Argument Structure” applies the issues to
cognitive structure and its syntactic expression.
9.4.4 Salience of information units in larger units
9.4.4.1 Salience in a figure complex

Example (39) is a simple illustration of salience in a figure complex. Halliday (2014,
§7.6) gives other examples from speech, with analyses.
(39) “So instead of getting seven shillings a week, I got about fifteen shillings a
week.” (Spontaneous speech; cited by Halliday and Greaves 2008, p. 103)
These structures are created primarily by intonation patterns, specifically by repeti-

tion of the same tone in successive groups, and by patterns such as a rising tone fol-
lowed by a falling tone. That is aided by stress; in example (39), “fifteen” would have
had extra stress, to contrast with “seven”. Syntax often provides markers that confirm
the structure. An example is the subordinate figure “instead of getting…” in exam-
ple (39), which is placed ahead of the main figure, which is unusual in spontaneous
speech; the two figures thus make a Rhematic structure, with the second one in focus.
A complex and subtle illustration from writing is provided by example (40),
which is the first sentence of a columnist’s article arguing that a recent obituary of a
noted writer was sexist.
(40) “If you’re a woman who has been made to feel a little slovenly and neglectful
this week by the realisation that you haven’t had your uterus steam-cleaned
(as promoted by Gwyneth Paltrow), you might have been prompted to won-
der whether you will ever stop being told your body is disgusting and finally
find yourself acknowledged for other qualities, such as your character and
achievements.” (Stephanie Merritt, Observer, reprinted in
New Zealand Herald, February 2nd 2015, p. A19)
The long first figure, up to “Gwyneth Paltrow”, has its own Rhematic structure, with
three rankshifted figures built to a sarcastic focus through the references to the uterus
and to Gwyneth Paltrow. The main figure (beginning “you might”) builds similarly,
through rankshifted figures and the emotive use of “ever” and “finally”, to a mock anti-
climax with the understated virtues, “character and achievements”. The long, complex
figures together make a two-part Rhematic structure.
9.4.4.2 Salience in a paragraph and whole text

Just as incorporation is used to reduce salience within a figure (§9.4.3 above), so it is
sometimes used to reduce the salience of a whole figure within a paragraph. That is
done twice in example (41).
(41) “Fully driverless cars, in particular, may take off [in China] quicker than in
litigious America or risk-averse Europe.”
The cognitive meaning of the last phrase is that the change will be slower in Europe
‘because it is averse to risks’. That explanatory subordinate figure has been reduced in
salience by reduction to a pre-modifying group, ‘risk-averse’, within the main figure;
the noun risk has been incorporated into averse, to facilitate that. The process entails
reducing the syntactic status of risk(s) from head of its group to modifier. “Litigious”
is used in the same way, being semantically a reduction of ‘which is litigious’.
As in a number of other areas, phonology marks large-scale salience structure in
speech, where syntax does not. As noted in Chapter 2 (§2.5.2), pitch creates phono-
logical paragraphs in speech, each consisting of several intonation units, and there-
fore several information units (Tench 1996, pp. 23–24). These paratones begin high,
and fall in fairly even steps; those stages in the pitch fall create the Rhematic structure.
In formal speech style, they are further marked by a pause.
Example (42) illustrates this structure of salience. In this passage of spontaneous
speech from an interview, the speaker is a businesswoman discussing the merits of
dried seaweed for making fertilisers and animal feed supplements; intonation units
are numbered.
(42) “[1] You know it’s got almost every known element, [2] but it’s a packet of
potential value you know. [3] It’s packed with amazing nutrients, growth
hormones, vitamins, minerals – [4] if you dry it properly. [5] Yeah. [6] Ah.
[7] So it’s like a – a smorgasbord feast for, you know, for soil, and plants, and
for animals.”
(New Zealand TV1, Country Calendar, “The Good Weed”, 10th May, p. 2014)
Most of unit 1 was highly salient, being abnormally high in pitch, and with strong
stresses on “every”, “known”, and “element”. Units 2 to 5 fell, in even steps; the rumina-
tive unit 6 matched unit 5. Unit 7 formed an unusual kind of focus: it was spoken in a
lower pitch range, and with more relaxed rhythm, highlighting its double function as
summary and final comment.
Since each paratone typically begins higher than the end of the previous one, a
structure through the text is created. Naturally, some paratones begin lower, to mark
backgrounded material, for example. In my own observation, they are also marked by
changes in speed and rhythm.
The graphological marking of paragraphs by indentation, and the structuring of
texts by paragraphing are therefore the direct equivalent of phonological marking;
both are therefore part of English grammar, not matters of “logic”, nor mere typolog-
ical conventions.
9.4.5 Conclusion: Salience structure
Theme and Rheme, as information structures, specify how the content of the utter-
ance is to be structured in the hearer’s mind; they do not add any content of their own.
Their meaning is consequently grammatical. Each word and phrase in the utterance
has its own meaning; none of them mean ‘Theme’ or ‘Rheme’; that meaning is carried,
not by the words, nor by content, but by the content order.
While Theme and Rheme give structure to the utterance by constructing a rela-
tionship between two parts of the content, the Rheme has its own internal structure,
that of increasing salience towards the end of the utterance. The nature of the salience
varies with the speaker’s intention, genre, and so on; it may lie in the content’s value
for proving the argument, or in vividness of concrete description, or in force of affec-
tive meaning, and so on. It thus resembles communicative dynamism, as discussed
by the Prague School. For a survey of the types and forms of salience, see Chiarcos
et al. (2011). Salience structure within figures has often been discussed, but it occurs
also in larger units, and in smaller ones. (There are a number of variations in salience
structure not discussed in this section, such as co-ordinated Rhemes, co-ordinated
foci, and foci in marked position.)
The structure is wavelike, in that it undulates, is repetitive, and usually builds up.
It thus differs from syntactic structure, which is primarily a composition of equivalent
units, and differs from the relevance structure established by Topics and the salience
established by Rhemes, which are field-like to an important for extent. The wave-like
structure is illustrated in example (43), which gives a sentence from a news report,
and example (44), which gives its Rheme, with larger type size representing increased
salience. (Vertical bars divide the information items.)
(43) “Chirpy Elijah Amoah is in remission after twice fighting back from leukae-
mia.” (New Zealand Herald, August 18th 2015, p. A1)
(44) “… | in remission | after twice fighting back | from leukaemia.”
The variation shown is a combination of variation within an information item (§9.4.2)

and variation of items within a unit (§9.4.3); it is realised in speech as rhythm (see
Chapter 11, §11.5.4, and Chapter 2, §2.5.5, for the analysis of rhythm).
Semiotically, the Rheme is like the Theme in consisting of content units, not of
syntactic units, and in being signalled abstractly; but the sign consists of the order of
content, not of its position. The structure, then, is a sign; as with the other forms of
information structure, its significance is in how elements of contents are to be relat-
ed to each other. That grammatical meaning, formulated in its system aspect, is that
that chunk of information is the Rheme; stated in its hearer aspect, it is, “Assign each
chunk of information a degree of salience as signalled” – signalled by its position or

phonology.
9.5 Discussion: Information structure
9.5.1 Relation of information structure to the bases and elements of language
Intentions and function. Information structure arises from a secondary speech inten-
tion, not from the speaker’s primary one (Chapter 3, §2.2): whatever the primary
intention – such as informing, persuading or entertaining – speakers almost inevita-
bly guide their hearers – for various reasons – in structuring and using the concepts,
impressions and feelings that we have discussed as “information”. As a guidance to the
hearer, information structure is an interpersonal function.
Semiosis. The structuring of the content is represented rather differently from the con-
tent itself. Content is almost entirely represented symbolically, by words, and its syn-
tactic structure is indicated by fairly precise abstract symbols such as word order. The
representation of information structure is much more varied, using lexis very rarely,
syntax occasionally, and phonology frequently. It is also often imprecise, as in the use
of pitch. In a state of excitement, we naturally speak at a higher pitch, so that high
pitch is an index of excitement; but it has been conventionalised to act sometimes
as a marker of Theme, and, when combined with volume at the end of an informa-
tion unit, it marks phonological and information focus; changes of pitch are therefore
quite ambivalent.
Cognitive and linguistic areas. The relationship between cognitive and linguistic se-
mantics illumines the nature of salience. Barsalou (2012, p. 240) notes that consistent
selective attention to a component of experience develops conceptual understanding
of it; that is parallel to the linguistic development that constitutes a Rhematic struc-
ture; I suggest that selective attention is the cognitive equivalent of salience. That is
supported by Hsu et al. (2015), who describe salience as the psychologically measur-
able capturing of attention.
Confusion as to the difference between what is linguistic and what is cognitive
has caused a great deal of the disagreement on the nature of information structure.
The concepts of “the given” and “the new” have been central in the discussion. But as
noted by Dahl (1976) as cited by Fries (1983, p. 117), the terms are generally badly
defined; Dahl identifies seven concepts which are confused with one another; and the
two terms are used for both relevance structure and salience structure. The reason
for that confusion is that the terms are not used for features of language – meanings
or structures that have a lexicogrammatical sign.2 They are used for what the hearer
probably knows as background knowledge, or what the hearer has in mind at the mo-
ment of the speaker’s utterance. Moreover, that hearer awareness is what the speaker
calculates or supposes it to be, according to the linguist’s theory. “Given” and “new”
are matters of cognition, and are subjective, without any means of empirical verifica-
tion. That criticism applies still more strongly to specific information statuses such
as “identifiable,” “available,” “recently mentioned,” and so on; see Prince (1983) for a
particularly complex classification. Hinterhölzl and van Kemenade (2012, p. 1) avoid
that trap by asserting that information structure is not linguistic in the narrow sense
at all, but pragmatic; however, it is in fact linguistic, in having specific linguistic forms
as its signs, and in having linguistic meaning in the form of grammatical meaning.
The concepts of Topic and Comment are commonly also cognitive, in the liter-
ature. They are not defined linguistically, and are not distinguished from the logical
and rhetorical concepts of topic and subject. They are like the rhetorical and stylistic
concepts of introduction, development and conclusion, and of listing and summary.
Syntax. The correlation we have seen between information structure and syntactic
structure seems to have a historical explanation. Presumably, Topic has developed
from the subject of thought, and Comment from predicates in thought. Theme has
perhaps developed from the use of introductory words in the relatively free word
order of Old English, and introductory phrases in Middle English. Rheme, including
focus, has evidently developed from the ancient and presumably rhetorical pattern of
end weight in English.
9.5.2 Compositionality
There is very little compositionality in information structure, since it consists almost

entirely of relationships, not assemblies of parts. The exception is in the Rheme, to the
extent that it is a series of information items.
Where there is compositionality, in Rhematic structure, it has a different nature
from the compositionality of the content hierarchies. The sequence of information
items may have syntactic linking, as between Process and Complement for example,
but there is no purely informational linking. Where the Rheme has a series of Circum-
stances, even the syntactic structure cannot create links. However, there is another
kind of composition, which broadly consists of a series of rising steps, and which
more exactly consists of a series of waves. We have noted previously that the group
which constitutes each information item has a head and a phrasal stress, making a
2. “Lexicogrammar”: structurally, lexis and syntax (and phonology) can be distinguished; but
both carry meaning, and they sometimes carry alternative realisations, so they can accordingly
be linked as “lexicogrammar”; but it will not be treated as an analytical category. It will be dis-
cussed in Chapter 14, §3.5.
peak which is at once semantic and phonological. In unmarked use, the Rheme is
therefore a series of rising waves, culminating at the focus.
The pattern varies with genre, and more subtly with the writer’s style. Informal
speech has longer troughs and higher rises to the peak, as in example (45). Two people
are looking at a map and discussing where a house might go; information items are
divided by a vertical bar, and stressed syllables are underlined. It is characteristic of
the information structure of spontaneous conversation that there is commonly only
one item of information in each figure.
(45) “There’s a nice low but firm bit here | which is a house site | where those trees
are | and then the fence is [pause…] half [pause…] or is nearly | to the top of
that low hill.”
(Wellington Corpus of Spoken New Zealand English, TDPC080_th:0280)
In formal speech and in writing that pattern is reversed, since the information is dens-
er, as in example (46); note that the Predicator has two items of information. (Only
the Rheme is analysed.)
(46) “Examinations | set nationally and marked blindly | would give way to | assess-
ment of pupils’ work | by teachers who knew them.”
(New Zealand Herald, February 2nd 2015, A24)
9.5.3 Case
This section deals with the apparently irrelevant subject of case, arguing that, however
it may have been used in the past, case in nouns (distinguished from pronouns) is
now used for information structure.
The lack of consensus on the nature of case in recent work shows that a substan-
tial improvement is needed in our understanding of it; see, for example, Butt (2006)
and Allen (1995). In the history of ideas, improved understanding has often come
from finding a general concept that simplifies and unifies previous views, which then
can be seen as partial. To illustrate that from the present issues: the nature of he (ver-
sus him) and the nature of she (versus her) are not to be found in the words taken
singly, but in a single more general concept, such as that of subject form or of nomi-
native. It will be argued here that, similarly, the nature of case is to be found in a single
concept that is more general and abstract than particular cases with their “structural”
and “lexical” forms.
Jakobson (1936/1990) points the way to a single “invariant” general meaning for
each case.3 The accusative means that “some action to some extent affects, is directed
3. It is not easy to apply Jacobson’s work to modem English: what he wrote was mostly applied
to Russian, although it seems to have been intended to apply to Indo-European languages gen-
erally; he changed his mind on some issues over the years, and often expressed a single idea in
various ways; and I am relying on a translation of his work.
at, or is manifested on the stated entity” (1936/1990, p. 341). The dative means “the
existence of the object independent of the action” (1936/1990, p. 358). The genitive
means “limitation of the signified entity’s participation in the content of the utter-
ance” (1936/1990, p. 377). Jakobson also describes those meanings as “features”, each
of the cases just listed being marked for one of them. The nominative is defined as the
case that is not marked for any feature; it has a function, that of naming, rather than
having a meaning; it is used for the Subject because it has the function of identifying
the referent and often the Topic. In my terms, the cases other than the nominative
indicate to the hearer the nature and extent of the referent’s involvement in the action;
that is, they guide the hearer in interpreting the content, as grammatical meaning
setting up information structure. (Jakobson applied that analysis to Russian and Ger-
man; I believe it also applied to Old English, where it worked with the relatively “free”
word order.)
The application to modern information structure is as follows. The adnominative
genitive has always been used to mark an Entity subordinate to the head Entity, and
therefore less salient than the Participant: in “The king’s son was…”, the son is the Sub-
ject, and the king is less salient; that genitive has always shown “limitation of the sig-
nified entity’s participation in the content of the utterance” (in Jacobson’s phrase). We
saw in §9.4.2.2, that the two genitive forms exemplified by “the committee’s chairman”
(a determiner) and “the chairman of the committee” (a postmodifier), have a third al-
ternative, “the committee chairman” (a premodifier), which, as a member of the same
paradigm, must be rated also as a genitive form. (For further discussion, see Feist
(2012b, and work cited there.) Section 9.4.2.2 also showed that the three forms are
used to signify degree of salience or “participation in the content”: the determiner gen-
itive in -‘s/-s’ signifies greatest salience; and the premodifier genitive, as in “the com-
mittee chairman”, signifies the lowest salience or participation. That general concept
of signifying salience or participation in the content unifies the many and otherwise
quite diverse “meanings” of the genitive, such as the ten given by Stefanowitsch (2003).
The dative, as in the “dative alternation”, detaches the discourse Participant some-
what from the action; that is acknowledged in the conventional terms “oblique” or
“indirect” object. For example, in “He gave the book to Mary”, Mary is more salient
and more “independent of the action” (in Jacobson’s phrase) than in “He gave Mary
the book”. (Here, salience given by position is distinct from the salience of participa-
tion given by case – which is marked by to.)
Those general meanings are particularly clear when the speaker has a choice be-
tween cases – which is often true in Russian, was often true in Old English, and is
sometimes true in modern English – and they apply even when there is no such choice.
We conclude, then, that present-day English has case in Entity words as a system
which in many uses expresses an abstract and general meaning concerning discourse
Participants’ involvement in the situation. (Case in pronouns appears to be different.)
I have identified the abstract meaning as “salience”, suggesting that it is expressed in
various ways – morphologically by inflection (e.g. – committee’s), and by prepositions
(of and to), and by morphosyntactic construction (as in “the committee chairman”).
There is some support in recent literature for that argument. It fits the common
view that case should often be regarded as abstract (e.g. Haspelmath 2009, p. 508).
DeLancey (1981) argues that in some languages case is used to represent the view-
point from which speakers describe the event, and for attention flow. Givon (1979)
says that the dative shift controls topicality. Butt (2006, p. 191) acknowledges that
discourse factors are involved in case, discussing topicality, specificity and bounded-
ness as motivations for choice of case (2006, pp. 190–191); she believes that there is
more to be discovered on this approach (2006, p. 191). An instance of the “meaning”
of the case being specificity is that in some contexts the Russian accusative will make
the noun specific or definite (e.g. “the tram”), while the genitive will make it indefinite
(“a tram”). Kiparsky (1998) argues that the partitive in Finnish signifies incomplete-
ness of the action; Ramchand (1997) makes much the same argument for Gaelic. The
literature on “quirky case” can also be seen as supporting my argument; for example,
in Icelandic, several different cases can be used for Subjects. Wierzbicka (1981, p. 58)
believes that each case has a single core meaning, though she does not seem to see the
meaning as being quite abstract. Mitchell (1985) and Traugott (2008) are consistent
with the view that case functioned in this way in Old English. The historically recent
development of the premodifier genitive fits the argument by Halliday (e.g. 2000)
already mentioned (§9.4.3.2 above), that English has in recent centuries increasingly
developed resources to aid the flow of information.
The argument here applies to pronouns, but other issues affect them as well.
9.5.4 Cohesion
Cohesion is presented by Halliday (e.g. 2014, Chapter 9) as if it were information

structure, or in some other way part of “discourse”; and it is treated similarly by other
writers. But its status in English grammar is not shown precisely in that work, and
the four means of achieving it (lexical cohesion, conjunction, reference, and ellipsis
and substitution) are only loosely related. Halliday says (2014, §9.2) that it creates
meaning, but the nature of that meaning is not stated; all that is shown clearly is that
text can be more or less cohesive. The argument for regarding cohesion as a distinct
structure in English is therefore not strong; Halliday himself says (2014, §3.5) that it
is not part of “structure” i.e. syntagmatic structure, and it fairly clearly does not have
paradigmatic structure.
However, the discussion by Halliday et al. does show two things. The first is that
there is in English texts a structure – weak or strong – of varied and criss-crossing
semantic relations; that constitutes a network (Chapter 5, §5.5.3). The second thing
shown is that cohesiveness is important in English.
9.6 Conclusion: Information structure
9.6.1 Summary
The primary semantic structure of a text in English consists of figures, groups and
senses, set up grammatically – that is, set up by morphosyntax (which we notice most
in writing) and by phonology (which we notice in speech); it is thus a structure of the
content, but of the content as grammatically defined units. Information structure is
a parallel and complementary one, the structure of the content as such. It consists of
three elements: salience structure (Topic-Comment), orientation structure (Theme),
and salience structure (Rheme). It is created by the speaker’s guidance to the hearer
on how to relate parts of the message content to each other, and to the hearer’s existing
mental content. Vallduvi and Vikunen (1998) give a similar definition.
In the history of linguistics, information structure is relatively new, and uses un-
familiar terms and concepts; but it implements three familiar fundamental princi-
ples: what is already established comes first; what is uppermost in the speaker’s mind
is expressed first; what is most important comes last. (The first two are from Croft
2003, p. 66.)
This structure is rather different from syntactic structure. It is strongly linear:
Theme is identified by its initial position, and its function of orienting would fail if
it came last; and Rhematic function is created by sequence. Thus, information struc-
ture exploits the linearity of language, creating value out of a constraint. A second
difference is that apart from the Rheme, its units are not composed into a pattern that
constitutes a different kind of unit, as groups are composed into a figure; the units are
terms in relationships which make up structure, just as relationships such as marriage,
employment and citizenship make up social structure.
The hierarchic nature of information structure is two-fold, as follows. In the
linear structure just discussed, the information unit divides into the Theme and the
Rheme, as subordinate units which constitute it. The Theme consists of up to three
units: textual, interpersonal and topical Themes. The Rheme has a focus item, and an
indefinite number of other items of information. In a non-linear second structure,
the information unit divides into Topic and Comment. All of that is the maximum
structure; as noted previously, speakers often do not develop utterances into a full
information structure. Diagrams 1 and 2 represent the two structures.
Information unit
Theme: orientation Rheme: salience
interpersonal textual topical (other Rheme units) focus
Diagram 1. Linear information structure: Theme and Rheme

Information unit
Topic Comment
Diagram 2. Non-linear information structure: relevance
9.6.2 Conclusions drawn
Types of grammatical meaning. The chapter has extended our understanding of gram-
matical meaning beyond what has been seen in the preceding chapters, since the
grammatical meanings which build information structure, named in §6.5 as the in-
formational type of grammatical meaning, is rather different from those which build
morphosyntactic structure, the syntactic type. This section will set out and develop
those two types.
The function of the syntactic type, in building morphosyntactic structure, is to
aid the representation of experience, in the ideational metafunction; the information-
al type, in building information structure which guides the hearer, serves the inter-
personal metafunction. The syntactic type operates on lexical and morphosyntactic
units; it interprets their place in the morphosyntactic structure in a relatively mechan-
ical way, with operations such as adding, changing, or replacing an element of mean-
ing, and assigning units a status as coordinate or subordinate. The informational type
operates on units of content; it gives an evaluation of the content’s relevance, orienta-
tional value, or importance in the message; it is thus evaluative in nature. Thus “inter-
pretive” and “evaluative” will now be preferred terms for the two types of grammatical
meaning. The signs realising the interpretive meanings are concrete and morphosyn-
tactic, typically being grammatical words and bound morphemes, but also being the
order of words and groups. The signs for evaluative meanings are more abstract, and
typically phonological, so also including order of content.
That distinction between the two types of grammatical meaning can be summa-
rised as in Table 2. The detailed distinctions, and the overall distinction, are intended
as useful guides; the types do not seem to be categorial.
Neither of those two types accounts for “speech acts” – the grammatical mean-
ings which guide hearers as to the response the speaker expects – acceptance of the
statement, a reply to the question, obedience to the command, or a sympathetic re-
sponse to the exclamation. Those meanings work on a higher level, that of intention;
Table 2. Interpretive and evaluative grammatical meaning

Characteristics Interpretive grammatical meaning Evaluative grammatical meaning
Function Ideational Interpersonal
Materials Lexical grammatical units Content units
Form of meaning Interpretive Evaluative
Signs Words and morphemes, typically Phonological features, typically
the interpretive and evaluative meanings are ways of realising that intention. This
“discourse grammatical meaning” – if a name is wanted – is also different in form and
operation, not entailing the development of structure.
Systematicity. As shown above, information structure has grown erratically in English;

it nevertheless constitutes a system that is moderately coherent, in having the single
function of evaluation – it evaluates the relevance and salience of the content. Since
the lexicogrammar presents the content, showing the content’s relevance and salience
complements what the lexicogrammar does. Also, it is systematic in being structured
internally by complementation, as noted in §9.4.1: what the message is to be about
(Topic) complements what is then said about it (Comment); how the message is to be
approached (Theme) complements how it develops from that starting point (Rheme);
Theme and Rheme complement Topic and Comment by providing evaluation.
On the other hand, the regularity of information structure as a system is weak-
ened by the fact that evidentiality, which is a comparable evaluation, operates sepa-
rately through modal auxiliaries; so the system is incomplete. It seems also to overlap
the morphosyntactic system: the grammatical meaning of conjunctions serves to in-
dicate the organisation of content, as Topic-Comment structure does; and modifiers
give information about the head, as Comment gives information about Topic – that
is what makes possible the use of modifiers as predication in uses like “An angry Mr
Blair…” (see §9.2.2.3). Above all, the semiotics of information structure is unsystem-
atic. Most of the signs are markers not symbols, so they are generally ambiguous, and
appear or do not appear as if at random – the use of pronouns and pause to mark
Topic, for example. The marking of Comment is very weak, consisting only of the fact
that it follows a Topic; so is the marking of Rhematic structure apart from focus, being
equally abstract; in each case, the hearer may well interpret the marker as indicating
no more than that the utterance is moving on to the next syntactic unit.
Recent research is showing that information structure in English is older, and
better integrated with the rest of grammar, than has been shown in earlier literature.
There was some development of information structure in Old English; see, for exam-
ple, Traugott (2008) on left-dislocations, Petrova ( 2012) on focus, Taylor and Pintzuk
(2012) on information status. Indeed, that work suggests that the “free” order of Old
English was in fact an information-structure order.4 The profound changes in Middle
English that brought a relatively fixed order changed all that; but the strictness of the
order made any departure from it strongly marked, and therefore clearly meaningful.
That potential for new sorts of meaning and new sorts of structure has been realised
over the centuries in a re-working of information structure; it has proceeded gradual-
ly and piecemeal, exploiting departures from fixed order, and other marked uses – one
4. The research also suggests that the mysterious first position that “verb second” typologi-
cal theory cannot explain is simply the Theme position, accepting adverbials as interpersonal
Themes, nominals as Topical Themes, and so on. Imperatives and questions no longer need be
excepted from theory, since the initial verb also sets Theme, as orientation to expected response.
feature of syntactic order here, and another feature of morphology there. There has
not been time for the information structures to be systematised fully.
Linguistic reality of information structure. The issues of whether information structure

is linguistically real, and of its place in the grammar of English, have been raised in
several sections of the chapter. It never appears in the linguistic literature as a section
of grammar, along with syntax and morphology. Even in outlines of grammar within
Systemic Functional Grammar such as Halliday (2014), it does not appear in the table
of contents, although it is discussed in the text. But the evidence in this chapter leads
to firm conclusions on those issues.
Morphology, syntax and phonology are regarded as branches of grammar partly
because each has a distinct type of sign. Information structure does not – it uses pho-
nological and syntactic signs. But while some specific signs are shared with syntax and
phonology, others are not. The topicalising constructions and the distinction between
premodification and post modification have no other significance than in informa-
tion structure. More significantly, two whole systems of English phonology seem to
exist for information structure alone: the tone system, i.e. the contour of rising and
falling tone over an information unit, and the stress system – they do not have value
for communicating knowledge.
Moreover, sharing of signs as means or “organs” of expression does not prevent
information structure from being a distinct system, just as the human speech, breath-
ing and digestive systems are distinct although they share bodily organs. Distinctness
of a system is a matter of how the organs are used, and particularly of function. The
three types of information structure analysed in the chapter cannot be reduced to
morphosyntactic or phonological structure, and not even to the more familiar seman-
tic structure analysed in the previous chapters. Its function, that of helping hearers
relate the communicated information to the hearer’s existing information, is not re-
ducible to other functions either.
We conclude that information structure is linguistically real, and a substantive
section of English grammar. It is a form of the speaker’s intended meaning, alongside
the content; it is realised in (linguistic) semantics and syntax and phonology, through
informational or “evaluative” grammatical meaning (§6.5), as the content is realised
(also in semantics and syntax and phonology), but through syntactic or “interpretive”
grammatical meaning.
Chapter 10
Other structures
10.1 Introduction
The previous chapters have set out the major semantic structures of English – the net-
works and the hierarchies – which present themselves as made up of compositional
units. This chapter deals with the minor structures, introduced in Chapter 4, §4.9,
which are each structured as a continuum, or are indeterminate in structure.
We will consider first the indeterminate structures, where the meaning is not
well differentiated, having no bonding elements to relate the semantic units to other
units – where language does not provide an argument structure, for example, and
where the intention to speak is too urgent to allow differentiation. Those units with-
out internal structure are discussed in §10.2. We will also consider situations where
children, for example, have various functional meanings to express but do not have
enough differentiation of syntactic or lexical expression to be able to formulate dif-
ferent utterances for the different meanings. Those units with several functions are
discussed in §10.3. Being units without syntagmatic bonds, those forms of meaning
fit into neither the networks nor the hierarchies discussed so far.
The remaining structures are continuums, rather than structures of units: fields,
to be discussed in §10.4, and wave structures, in §10.5. They were introduced in
Chapter 2, §2.5.7.1.
10.2 Semantic units with indeterminate structure
10.2.1 Introduction
The semantic structures studied so far consist of units which are bonded to other
units, necessarily having an internal structure which includes an element which pro-
vides for the bond, such as an attribute which can be specified by a value in another
unit. We now come to study semantic units that have no structure of that kind – al-
though they may have constituents of different meaning types, which may in turn be
dimensional. The familiar instances of such semantic units are “interjections”, which
will be important in this section, but there are other types as well. (“Interjection” has
various meanings, and is primarily a syntactic term, so it will not be used as a strict
term here.)
Identification of these expressions needs care. First, many of the words and
phrases used in this way are also used in other ways. Second, it is not easy to draw a
distinction between semantic units which structure conversation linguistically, and
those which structure conversation socially and accordingly belong in the study of
discourse analysis, rather than linguistics. Third, some items seem simple because the
whole expression is elliptical; they are excluded also.
These unstructured semantic units have been classified here by metafunction,
and within that by more specific function. I have benefited from the classification of
interjections by Ameka (1992), using his “expressive”, “cognitive” and “phatic” types,
but adding others. The section is structured according to the classification.
10.2.2 Ideational function
Few structureless semantic units are ideational: when we express concepts, we reg-
ularly relate them to others, making them a part of a larger semantic unit, and giv-
ing them structure. Among the few ideational examples are expressions for weather,
earthquakes, and fire, such as “It’s raining”, “There was an earthquake”, and “There’s a
fire.” The sense, ‘fire,’ for example, does not differentiate the burning from the thing
that burns; there is no structure of Event + Entity; the sublexical form has its cognitive
base in a whole situation, not narrowly in a thing or in a happening. That is reflected
in the fact that there are usually synonymous alternative Entity and Event forms to
realise such sublexical senses, as in “The fire burned all the wood” (with two entities
and an Event), and and “Rain is falling.” Similarly, representations of bodily experi-
ences such as “It’s hurting” and “I’ve got a pain in my ankle” are undifferentiated units
semantically .
10.2.3 Interpersonal function
10.2.3.1 Social relations

Some interjections establish or maintain a social relationship, without having any
content: they have “phatic” function. (The term comes from Malinowski 1930.) They
include the vocalisations rendered as “mmm” or “mhm”, “uh-huh”, and “yeah” in some
of its uses, but also greetings, and some uses of ah, oh, I see, and so on. Speakers
are not conveying any information of their own, but are intent on assuring the hear-
er that they are attending fully to listening and sharing. Example (1) illustrates the
phatic function; it is part of spontaneous conversation among friends; speaker A is
telling a story.
(1) A: “And he said there were vets and doctors evidently cos it was human eyes
he was working [Speaker laughs] on when he wrote.”
B: “Sounds quite… I see. [All participants laugh.] I don’t think I want to
know too much.” [Speaker laughs]
Chapter 10. Other structures 293
A: “No it doesn’t sound too good but you see he organised all this before he
went.”
(Wellington Corpus of Spoken New Zealand English, T_DPC022_th)
B’s response and all the laughter are phatic (while possibly serving other functions as
well), and so is the first part of A’s rejoinder. (A’s ‘But you see” signals his switch from
phatic response back to narrative.)
It does not seem possible to define a meaning for these expressions, since they
are not symbols that carry definable meaning, or expressions of affective meaning.
They have the indexical quality of indicating something about the speaker, but they
are intentional, not being mere by-products. They are signals – of a state of affairs or
an intention – but many of them are barely conventionalised, consisting of a suitable
contour on any of various consonant and vowel sounds.
Other phatic expressions, however, are fully conventionalised as words, such as
greetings and farewells, and terms of address. They are symbolic in being convention-
alised and in being arbitrary from a synchronic viewpoint. (Expressions like “good
day” and “farewell” and even “goodbye” are fully motivated from the historical view-
point.) But they too are signals, rather than symbols: saying hello is a thing you do,
not a piece of information you pass on to someone – you can say hello by saying “Hi!”,
“How are you?” and so on ; a greeting’s significance lies in the fact that you say it, not
in any content. This is language as action, not language as communication.
10.2.3.2 Interpersonal action

Directive function. Some of these structureless units resemble commands, in having
a directive function – or “conative” function to Ameka (1992) – and are related to
the speaker’s wishes. They direct the hearer to be quiet (Sh!), to explain (Eh?), or to
respond (Hey!). They are like the phatic interjections in the nature of their meaning
and semiosis.
Conversation management. In example (1) just above, “but you see” warned the hear-
ers that the narrative was about to start again. Eh and uh encourage the hearer to
respond when used at the end of an utterance with an interrogative intonation; some
uses of er and um implicitly request the hearer not to respond, or not yet, at least. Such
interjections or “discourse markers” manage the interchange in conversation. Intro-
ductory well often does so also, by acknowledging the previous remark; you know
often acknowledges the other speaker, as participant in the conversation.
Semiotically, this group is quite mixed. Er and um have an indexical basis in
meaningless vocal gestures (Wierzbicka 1992, p. 178), but have gained conventional-
ised meanings. Well and you know began as a regular linguistic symbols, but have lost
their descriptive meaning (i.e. have grammaticalised), to be like er and um in having
only an imprecise social meaning which is highly context-dependent, thereby being
not fully symbolic.
These meanings are like grammatical meanings in providing guidance to the

hearer – particularly anyway in example (2), which warns the hearer that the speaker
is not going to pursue the existing line of thought.
(2) Kotb: “Is she talking about you?”

Gifford: “I think so.”
Kotb: “Yeah. Anyway. She feels like we all go through the ups and
downs….” (COCA)
10.2.4 Expressive function
The expressive function was introduced in Chapter 3, §3.3.2, as the vocalisation of

affect. Its basis is in natural expressive sounds such as grunts and cries; some of those
have been conventionalised as expressions like whew, ugh and bah, which are barely
to be rated as words. Interjections such as heck, struth, and blimey are partially es-
tablished as words; others have become fully established, as with yucky, which has
syntactic status as a modifier.
Expressive utterances, by definition, are not intended for an audience, but serve a
function for the speaker alone, such as release of tension or of an excess of affect; they
are “a flooding up of the previously contained feeling”, says Goffman (1981, p. 99).
Many words with a different origin and basic function are pressed into use as ex-
pressive interjections, especially those with religious or sexual meanings, generally
combining the expressive function with an interpersonal function such as conveying
feeling or causing some effect in the hearer.
The meaning type, obviously, is affective, and the main dimension on which they
vary is that of intensity; but “meaning” here does not indicate something communi-
cated to a hearer – the expressions have function, rather than meaning.
The method by which that function is achieved is not symbolism or iconicity,
since there is no abstract content that can be represented by symbols or icons. The
clue to the method is in the words’ origin: sound. The sounds work as a physical
release of physical tension, thereby releasing the affective tension. Initial plosive con-
sonants release pent-up breath from tense lungs, as in blimey, blow it, bloody, bastard,
and so on. Aspirated consonants work similarly, but more slowly: heck, aha, and so
on. Less obvious qualities are the length of final consonants, as in struth, and the
abruptness of the ending in words like heck, yup and nope. The same effects can be
achieved with those consonants in descriptive words, by intensifying the feature con-
cerned: the slight aspiration given to initial stops in English is intensified for emphasis
in b[h]illion, b[h]rilliant and so on. Consonants are lengthened, as in million. Some of
the interjections discussed in the section on the descriptive function also have this
expressive function; wham and bang are obvious examples; sorry and thanks are ex-
pressive when spoken with extra force and length on the initial consonant.
This mode of significance has been known as “sound symbolism”, defined by
Nuckolls (1999, p. 225) as “linguistic sounds such as phonemes, features, syllables, or
tones that can be meaningful”. The term is misleading, because the effect is not sym-
bolism, and it is not a representation of sound, as the term suggests. As noted above in
Chapter 2, §.2.5.6, on onomatopoeia, the minimal meaningful unit is smaller than the
morpheme; it is often a phoneme, and often a single feature of the phoneme, such as
aspiration or being plosive. Accordingly, a better term for this phenomenon of using
a single sound or phone as meaningful is “phonic use” of language (contrasting with
“phonemic use”), to mark its prime characteristic, that it is the sound itself that carries
meaning, not a combination of phonemes. It should be noted finally that the effect
comes from the physical action of articulating sounds; it is not auditory. (This phonic
use of linguistic sound has been illustrated here with single words; it will be seen in
context in Chapter 11, on realisation.)
This phonic mode of significance is ignored in most linguistic literature, although
it is taken for granted in some: see Nuckolls (1999), and the references there; see also
Firth (1957) and Halliday and Greaves (2008, p. 169) on “phonaesthesia”. For a recent
fairly full account, see Feist (2013).
10.3 Semantic units with multifunctional structure
10.3.1 Introduction
The last section dealt with expressions which are unitary in function; this section
deals with semantic units which have several functions. The functions concerned are
semantic ones: functions of utterances for speaker and hearer, or “speech acts”. In
most utterances, that function is a distinct, fully differentiated one, symbolised clearly,
as command, question and so on. It is not made clear in these utterances, because the
speakers are either children who simply cannot achieve that in their early years, or
adults who are overwhelmed by urgency or who intend a special effect.
I distinguish two types of multifunctional semantic units, holophrases and ideo-
phones, to be discussed in the following sections.
10.3.2 Holophrases
Children’s holophrases. Slobin (1970) observes that children at the “two-word” stage
(when they are limited to utterances with the equivalent of two adult words) have half
a dozen fairly distinct functions which their utterances are to serve, but not the ability
to differentiate their utterances accordingly. For example, “dolly dress” can express a
desire (to be given the dress), or can be a description (of the doll’s clothing), or can
state ownership (what the doll possesses) – or can ask a question. At the earlier stage
of “one-word utterances”, the function may be even less distinct, and barely linguistic.
Karmiloff and Karmiloff-Smith (2001, p. 64) record that “juice” may simply refer to
an image of a carton of juice, or signal that someone in the room is drinking juice, or
draw attention to the situation.
These utterances are “holophrases”. They are utterances which serve, or can serve,
several semantic functions, but whose form does not represent the functions sepa-
rately. “Juice”, in the example just given, can serve as an act of reference, an act of as-
sertion, and as a linguistically informal act of drawing attention. In a further instance,
a small boy was taken to the beach as a treat, but on coming in sight of the sea, he ex-
claimed loudly, “No bath!” The utterance was at once an expression of fear, an urgent
request, and an implicit statement. (Note that in that utterance, “bath” was the best the
child could do to say what he meant; it did not have any of its established meanings;
it was an imitation of the word bath, and not in fact that word as a pairing of form
and established meaning: the common term, “one-word utterances”, is a misnomer).
Holophrases are in part simply responses to a situation, like the calls and cries
from which they develop. The morphosyntactic expression may be fairly well differ-
entiated, but their functions and any meanings they have are not, because children use
the one expression as their response to many situations. From the speaker’s point of
view, the holophrases have several meanings, one of which may be dominant. From
the hearer’s point of view, their meaning is indeterminate, with the result that there are
several possible appropriate responses; the meaning and response must be resolved by
reference to the context. In the system aspect, they are ambivalent among the various
possible meanings. They usually stand alone; if they do fit into an utterance, they are
not well integrated with it.
Semiotically, holophrases cannot be classified simply into one of Peirce’s types.
They are partly indexical, particularly with children in the first few months of talk-
ing, as indicators of mood or need. They have a symbolic element, in being based on
adult words. They come nearest to being signals, however, as their phonological form
matches the situation, and their significance is not fully conventionalised; the child’s
carers must learn to interpret them.
Adults’ holophrases. Holophrases are developmentally primitive, but adults use them
too, exploiting their communicative potential. By combining several intentions, they
are economical in saving time, and are forceful. In example (3), when grandma says
“Whatever,” she is intending to convey the information that she does not care about
calling Cassandra by her correct name, and is intending to put mama in her place, and
to express her impatience at the sidetrack about names.
(3) “In fact,” [Mama] continued, “… Maybe I’ll home-school her, you know?”
Daddy laughed, but Grandma did not look pleased. “Cassandra,” she began.
“My name is Cassie, Anne.” Mama frowned at her mother-in-law.
“Whatever,” Grandma said. (COCA; fiction)
Adults’ holophrases, then, are linguistic forms which have several uses; but there is
nothing in the form to specify which of the possible uses is intended. Hearers are left
to infer it, using the context and the distinctive intonation they would be spoken with.
That is usually easy; for instance, in example (4), so cannot be a conjunction: the usual
significance is contradicted and thereby cancelled. It comes from an argumentative
discussion of American politics.
(4) Beckel: “There hasn’t been a brokered convention since the ’40s.”
Bolling: “I understand that. So?”
Beckel: “So, you’re not going to have a brokered convention here because by
the time – you still have a lot of states to go.” (COCA)
On the other hand, holophrases are often expressions which retain some of their
standard meaning, and are often formed by a reduction of a standard utterance to one
word or group, as in examples (5) and (6). Example (5) was spoken by a man who had
lost control of a truck carrying 4.5 cubic metres of gravel down a steep hill.
(5) “The brake just went to the floor, I pumped it and nothing.”
(New Zealand Herald, news report)
Example (5) is a statement in the narrative, and an expression of feeling, and a dra-
matic evocation of experience, with the semantic “emptiness” of nothing imaging the
speaker’s experience. It is a piece of language where the expression of experience is
direct, not the indirect expression which is mediated by articulated language.
(6) “Clarice moaned again, and this time the sound swelled… to a scream. ‘David,
my leg, my leg! Lord Jesus have mercy, David, I think my leg must be bro-
ken!’ ” (COCA; fiction)
At face value, the instance in (6) is simply an exclamation, but in the context it ex-
presses very strong feeling (as indicated by “scream”), gives information about what is
wrong, and is a call for help.
Other examples of holophrases reduced from full utterances include “Sorry” and
“Thanks”, which retain a minimal conceptual element from such figures as “I’m sorry
about that” and “I thank you for that”, from which they presumably derive. Obviously,
they have a social function as well as making a statement, and they express feeling.
Other holophrases are words like wham and bang, when they stand alone as utter-
ances, but also when integrated into the utterance, as slam is, as adverb in example (7),
and as verb in example (8). They are multifunctional in giving factual information,
expressive effect, and often affect – dislike, in example (7), presumably.
(7) “They’re coming slam into the greenhouse.” (SOED example)

(8) “Charlotte slammed the back door.” (SOED example)
A final type of adult holophrase is in some uses of “discourse markers”. Redeker (2006,
p. 345) lists six functions for you know, and five functions for well; when a use has
several of those functions at once, it is a holophrase.
Longer holophrastic expressions. Many expressions are semantically standard to the

extent that they have an internal structure which builds the utterance composition-
ally, but are like holophrases in having several semantic functions at once. The use of
proverbs and of idioms like “Finders keepers, losers weepers” often works this way;
the proverb has its own meaning or function(s), but speaker has his own purpose in
citing it in a particular situation, as in example (9). The situation is that arguments
have been made to move the Prado Museum in Madrid to a different building, or to
build a new one; but Mr Portearroyo has suggested redeveloping the existing building.
(9) “ ‘A bird in the hand is worth two in the bush,’ counters Mr Portearroyo.”
(COCA; The Economist)
Quoting the proverb makes a statement – the content of the proverb – but makes an
argument beyond that of the content, by implying that proverbial wisdom supports his
view; and it thereby gives a warning – against flouting the authority of that wisdom.
Holophrastic expressions, like true holophrases, gain effect from their brevity and
consequential evocativeness. They are very close to the poetic effect illustrated in ex-
ample (10), where “to do” seems to have a wide range of reference.
(10) “Theirs not to reason why,/ Theirs but to do and die.”

(Tennyson’s poem, “Charge of the Light Brigade”)
Givon (1979, §7.10) supports this analysis, noting further that such expressions char-
acteristically deal with the speech situation only, without any reference displaced to
another time or place. So does Wray (2002); these expressions form the core of the
third of her five lexicons, as types of formulaic utterance.
Conclusions: holophrases. Holophrases are not made up of compositional elements,

so have no internal semantic structure, although (as in the children’s examples) they
may have some inadequately connected morphosyntactic elements. They are low in
descriptive meaning, and may be empty of content. The only meaning, in a narrow
sense of the term, which they can be said to have is a grammatical one – which is more
pragmatic than semantic: “Find a meaning for this for yourself, from phonology and
the context, by implicature.” When grandma says “Whatever” in example (14), she
avoids doing what is asked, gives no socially acceptable response, and presumably
speaks with sour intonation; her hearers will also detect an acerbic tone; they draw
appropriate inferences.
Holophrases are highly dependent on context. That is particularly clear with chil-
dren’s holophrases, but applies also to adult utterances like “Whatever!”, since, in a
different context, that may be a cheerful and cooperative expression.
10.3.3 Ideophones
Typical examples of ideophones are pell-mell, hoity-toity, riff-raff, knick-knack and

doolally. The obvious feature of those is their reduplicative structure with the varia-
tion of the initial consonant or the vowel; but that is subordinate to the effect of that
structure, which is to draw attention to their phonic use of English sounds – that is,
using sounds as sounds, not as phonemes.
Some ideophones have regular syntactic function; flim-flam, for example, is used
as a noun, an adjective and a verb; they can the submodified as in “Our politicians are
too wishy-washy to take a stand on this issue.” (SOED). Others are more limited, be-
ing only an interjection (hoity-toity) or Complement of a verb (“throw a wobbly” and
“go doolally”). The last two examples highlight further characteristics of ideophones:
words with ordinary uses, such as wobbly, can also be used as ideophones, with their
phonetic features emphasised (such as the initial /w/); and they verge on falling out of
the grammatical system altogether, since doolally and many others have no use apart
from the idiomatic phrase.
Some are quite polysemous; jimjam (singular) means “A fanciful or trivial article,
knickknack”, but jimjams (plural) in modern colloquial use means ‘pyjamas’, and the
jimjams means “A state of nervous agitation, the jitters” (SOED). Others are mor-
phologically productive; the prefix ker- (“imitative of the sound or effect of the fall
of some heavy body” – SOED) forms kerchunk, kerflop, kerplunk, and others, both
nouns and interjections.
The examples already given show that ideophones commonly carry descriptive
meaning, some being basically descriptive and acquiring ideophonic value in context,
e.g. wobbly. Others are primarily emotive and attitudinal in meaning, as with hoity-
toity; that value can vary, as with knick-knacks, which is used favourably in exam-
ple (11) and unfavourably in example (12).
(11) “The furniture, the pictures, the loved knick-knacks, all told their story of a
lifetime spent together.” (BNC)
(12) “The house was always cold, full of stale animal smells and decorated with
grotesque knick-knacks…” (BNC)
Ideophones, then, vary in both internal and syntagmatic structure, ranging from
regular semantics of meaning types and dimensions to idiomatic semantics. Their
distinctiveness is, first, their combination of phonic use – which they share with ono-
matopoeia and expressive interjections – with regular semantics – which they do not
share with those other expressions. Second, they are distinctive in being multifunc-
tional as other words are not (“words” excludes interjections here.)
They have an extra function, however, which arises from a secondary intention, a
function such as being vivid, being imaginative in the use of language, or being play-
ful. Such a function is virtually inevitable, since the content of an ideophone always
has an alternative expression, so preferring the ideophone has significance. However,
since the nature of the function is context dependent, and is often conveyed by pho-
nology as much as by the ideophone as word form, it is a little difficult to illustrate.
Examples (13) and (14) do illustrate it moderately well, however. In example (13), a
boy tosses various solid objects into a deep hole.
(13) “You’d hear a kerrrrrplunk. Fascinating. Worms, beetles… chicken feathers

and/or droppings, wouldn’t do that. No kerplunk. Not fascinating.” (COCA)
(14) “ ‘It’s a sweet sound, isn’t it?’ Jim asks, … listening to the sound of palm fronds
in the breeze punctuated by the bloop! plop! kerplunk! of golf balls upon the
waters.” (COCA)
In (13), the ideophones combine with the informal syntax to reinforce the evocation
of the boyish world. In (14), the variation and neologism in the ideophones depict
experimental curiosity in the writer, and give the writing playfulness; they contribute
to an imaginative dimension in the writing evident elsewhere: “… golf balls upon
the waters”.
10.3.4 Discussion: Semantic units with multifunctional structure
Some utterances are like holophrases in having multiple functions, but are not holo-
phrases in fact, because they have a single dominant and clearly stated function. The
routine announcement, “The committee is now in session” functions as a statement,
but also often functions as a warning (that people must now be quiet), and as an
instruction (to pay attention), and is also a piece of self-assertion (establishing the
speaker as chairperson). “Dinner!” is an elliptical statement that dinner is ready, but
functions also as an invitation. Other such routine statements are “Ladies a plate,”
and “Six for six-thirty.” Some stock rejoinders are similar: “You and whose army?”, “I
wasn’t born yesterday,” “Don’t tell me about it!” and “Silly me.”
Figures of speech also serve a secondary intention parallel to the primary one
of conveying descriptive and affective meaning. As with ideophones, the utterance
comes to be aesthetic or humorous in function. It can be illustrated by example (15),
spoken by a man who reported that at 11 a.m. on Christmas Day he got a message on
his phone.
(15) “My friendly bank had sent me what everyone must want for Chrissie. Not.
My latest credit card statement had arrived, complete with the cost of gifts for
all and sundry. Just the thing to put me in the holiday spirit.”
(New Zealand Herald, December 30th 2013, p. 16)
“Not” is straightforwardly a negation; but it also converts “what everyone must want”
to irony, and creates a joke.
The same effect sometimes results from elliptical colloquial English, as with a sign
in a car window, such as: “$8,000. Ph. 654321”: an offer of sale, a statement of price,
and a request to be telephoned. Discourse markers are not only often polysemous,
having various meanings within the same communicative function, but also some-
times serve different functions. That is illustrated in example (16), which comments
on Mr Brown, governor of California, who has been supporting proposition 30.
(16) “Regrettably, Prop 30 will worsen the state’s dependence on the volatile
income of a small number of wealthy residents. But hey, Mr Brown is a poli-
tician, not a policy wonk.” (The Economist 20th of October 2012, p. 35)
The passage uses “but hey” ironically; instead of expressing light-hearted attitude,
it expresses sarcasm; instead of signalling a change of subject, it signals remorseless
continuity; and it creates a tone of bitter playfulness in the switch from formality to
informality just as the point being made becomes more serious, not less.
10.3.5 Conclusion: Semantic units with multifunctional structure
Multiple meaning in language is naturally thought of as potentially confusing polyse-

my, or counter-productive ambiguity; but in this section we have seen that it can have
its own value. By combining several functions into one expression, it is economical; it
is often vivid or forceful, and evocative.
10.4 Semantic units with field structure
Fields as semantic structures were introduced in Chapter 2, §2.5.7.1, with reference to

Halliday (2014, §6.2.4), following work by Pike, who suggested an analogy between
meaning and energy. Whereas most meaning consists of units, which can be com-
pared with photons or electrons as particles, some meaning is better understood as
waves in a continuum, with a rise and fall of significance, and some meaning is better
understood as a field, like a magnetic or gravitational field which acts on anything
within it.
The fields we are concerned with are set up by affective and social meanings,
forming areas within the meaning of a text where all content is modified. Social mean-
ing regularly acts this way. For example, the colloquialism of “I’ll”, used in preference
to “I will”, sets the relation between speaker and hearer as that of social equality; but
that significance applies throughout the utterance, not only at the moment when “I’ll”
is spoken. A field is created – in time, not space – in which everything is understood
to be uttered in social equality.
In example (17) a woman was giving serious advice in a friendly way.
(17) “You can be like, ‘I’m going to sit here and pity myself ’, or you can be, ‘This
did happen but I can heal myself ’. ”
(New Zealand Herald, July 2nd 2014, A17)
Most of the diction is formal, to give the serious tone, but colloquial be like sets a
friendly relation with the reader which will underlie the seriousness, throughout the
utterance; it is renewed by I’m and the second “be [like]”.
Obscenities and blasphemous language provide other clear instances. In exam-
ple (18), about a football player, bloody refers to goals syntactically, but the speaker is
not angry with goals; semantically bloody creates a field of anger that covers Deane,
his club, and its management.
(18) “Brian Deane has scored his first goal… but really he was bought to SCORE
[sic] the bloody things.” (BNC)
Vandelanotte (2002, §3.2) describes such attitudinal adjectives as taking scope over
a whole situation. Terms of address constitute another straightforward example: be-
ginning an utterance with “Your grace” can make the whole of a short speech deeply
respectful.
Example (19) provides a more complex illustration. It was an attack on Los Ange-
les (“LA”), and particularly on the cult of the perfect figure. Paltrow was a well-known
actress.
(19) “Paltrow may have found a niche peddling physical perfection in slovenly
Britain, and be the only woman in our whole flabby, rudimentary land,
and be able to pronounce ‘quinoa’, but out here [in “LA”] macro-heads are
ten-a-penny. (Paltrow has endearingly admitted to enjoying the odd pint of
Guinness and ‘a cigarette on a Saturday night: it’s just the right amount of
naughty’).
There’s no right amount of naughty in LA. Bodies aren’t just temples – they’re
the bloody Taj Mahal. You have to have successfully eliminated every food
group… in order to be considered a health freak.”
(Columnist for the Telegraph daily newspaper, August 2013)
The passage sets up several fields which continue through the rest of the passage,
being renewed periodically, and applying to everything to which they might be rel-
evant. First comes formality, begun with niche and peddling, reinforced by rudimen-
tary, and renewed later with eliminated; in the meantime, the strength of the tone of
formality leads us to discount the informality of elided there’s, as an undertone. That
informality, however, sets up as another field, picked up by aren’t and they’re; the two
fields continue, in parallel as it were, with the text alternating between them. As well
as those social meanings, there is the emotive meaning of sarcasm. It is begun by
derogatory peddling and confirmed by slovenly; the existence of that field makes us
take the reference to macro-heads and to pronouncing quinoa as sarcastic, even if
we do not understand the references; and it forces us to take “the right amount” as
unfavourable (by irony) although at face value it is favourable. Clearly, bloody spreads
its social and affective meaning widely through the passage, but we see that words like
slovenly and rudimentary, which have a strict, narrow syntactic scope, create fields of
meaning also.
Field structure is simple, in that it is not differentiated into constituents, and the
procedures for constructing it are not well differentiated either. For the hearer, there
are guidelines rather than rules for the limiting the area of the field. Social meanings
relating to the speaker individually create fields extending indefinitely: the speaker
does not stop being American or upper-class or elderly. Social meanings creating a
field of relationship such as informality, deference or authority extend until an in-
compatible meaning occurs. For example, the field of formality created by “… you are
required to conform to all the relevant regulations” reaches a boundary if the speaker
continues “Okay, so let’s talk about ….”
When the basis of the feeling is specified, an emotive meaning will not construct
a field at all; the extent of its application is then defined linguistically. In “This is a
lovely warm room”, the admiration is restricted to the warmth, which is restricted to
the room; the extent is then defined by the syntactic structure; the scope of feeling is
the scope of modification. When the field is set up phonologically, as by heightened
pitch, it will last as long as the phonological feature does. Often, the field will cover
one information unit, as in the example above about scoring goals. At times, however,
hearers must use real-world knowledge of the situation or of the speaker, to delimit
the field.
The concept of field has a broader, less well-defined application in the concept
of context, and can help to make that the concept clearer. We can profitably take the
notion of frame, as part of context, as a field extending over a certain taxonomic do-
main; similarly, a scenario can be taken as a field extending over a certain period of
time or series of events. The meaning of that much discussed word, bachelor, is set by
the field created by the frame in which it occurs. That is shown by the following ex-
amples; they are OED citations (modernised in spelling), and have the word creating
the field underlined. For SOED sense <1>, “A young knight”: “This knight is a worthy
bachelor.” For <2>, “An unmarried man…”: “His wife! I have heard him swear he was
a bachelor.” For <3>, “Junior member… of a trade guild”: “A barge also of bachelors
of the major’s craft.” This view helps explain sense variation as discussed earlier in
the book (Chapter 5, §5.2.2 on brave, and §5.3.4 on flare, for example). It also affects
our understanding of compositionality, since these fields affect the way we combine
meanings compositionally.
Grammatical mood, which necessarily affects the whole of an utterance, and mo-
dality, which sometimes does, also create field structures. The mood is set, as inter-
rogative or declarative for example, by the Subject and Finite together; the hearer
interprets the rest of the figure accordingly; and that field of meaning is sometimes
reinforced by a tag at the end – “is it?”, “isn’t it?” and so on. Similarly, an initial modal
Adjunct colours the whole figure, and is sometimes reinforced by a modal auxiliary.
Reinforcers act over the small field of the pre-modifying string: in “a mere 250,000 live
television audience”, mere downplays all of the modifiers, as well as the head. Finally,
negation is sometimes sustained by “negative polarity” items such as any (cf. some),
neither (cf. either), nor (cf. or) and budge (cf. move); this field has its own name – “the
scope of negation”.1 Double negatives commonly act in the same way, strengthening a
field of negation, not producing a positive, as in example (20).
(20) “Yeah… We won’t be going, we’ll be stopping here. Isn’t no point in going
home.” (BNC)
Field structures, then, are created most clearly by affective and social senses, which
carry the grammatical meaning, “Attach this sense to all the content of the relevant
field;” they are created also by some grammatical meanings. The extent of the field
is often not specified, the hearer being left to determine it, using some linguistic
guidelines, and some general knowledge; with grammatical fields, the extent is seen
as “scope”.
10.5 Semantic units with wave structure
10.5.1 Introduction
Wavelike structures were introduced in Chapter 2, §2.5.7.1, as patterns of rise and fall
in feeling, or in the importance of ideas; they are usually expressed as a rise and fall in
pitch or stress. The significance in the wave is in the height of the crest, the depth of
the trough, and in whether the trend between them is a fall or a rise.
10.5.2 Wave structure in information
Structure in figures. In Chapter 9, Rheme was presented as a series of upward steps.

Considered more closely, those steps are small-scale waves. In general, the head of
each group constitutes the crest, since being the head signifies importance; preposi-
tions and auxiliaries are low points; premodifiers constitute the rise; postmodifiers
constitute a fall; and any pauses constitute a trough. In practice, some whole groups
are semantically too light to form a wave, and in groups like “a big one” the modifier
forms the peak.
Example (21) illustrates those points; it was spoken spontaneously; the wave
crests are underlined.
(21) “If your mummy was going to make a frock, what material would she use?”
(Couper-Kuhlen 1986, Ch. x, §3.2)
If you read the sentence silently, the content words indicate to you which words carry
the most information value; but if you hear it, or read it with a sense of its vocalisation,
1. This field effect is “semantic prosody” to Sinclair (1998, p. 20): “The reason why [a lexical
item] is chosen over and above the semantic preferences that also characterise it.”
then the rhythm carries that significance. That is the general and almost universal
significance of rhythm – what it exists for, its “meaning”. In the form of rhythm, wave
structure is entirely ordinary, as example (15) shows, and as widespread as rhythm,
though little noticed.
Information, then, is structured as waves. The Rheme typically consists of a series
of waves, rising to the focus, as discussed in Chapter 9. (Figures (14) to (19) illustrate
the point.) The Theme or Topic or both form one high point, balancing the focus in
the Comment. Those structures are realised by the tonicity system – the pattern of
stress. The information unit as a whole forms a larger-scale wave, with the focus as its
crest, realised by the tonality system – the rise and fall contour of pitch. The realisa-
tion is primarily phonological, but operates in silent reading through subvocalisation,
which reinforces intellectual perception of importance, and the alternation of gram-
matical and content meaning – along with “Zipf ’s law”, that the shortest words, being
the most frequent, are generally the least important.
Structure in paratones. Paratones (see Chapter 9, §9.4.4.2) are also structured as waves.
Example (22), from spontaneous speech, illustrates the effects in small scale.
(22) “The whole farm was just a sea of ragwort!! It was everywhere!”
(New Zealand TV1, Country Calendar, “Chaos Springs”, 5th April, 2014)
The underlined parts were stressed, each being stressed more than the one before
it. The speaker’s surprise and horror at the spread of the noxious weed rose and fell
accordingly. That undulating structure was reinforced by the information structure:
“The whole farm” identified the Topic; “just a sea” represented the first element of
the Rheme, and “of ragwort” represented the focus of the Rheme. “It”, an unstressed
pronoun, and merely repetition of the Topic, was very low in importance; the stressed
word “everywhere” was the focus of its figure, and the climax of that section of the
text – a high crest following the low trough. Along the way, the trend was a rise, sig-
nalling to the hearer that the text was incomplete, and that the speaker had stronger
feeling yet to come.
Example (23) illustrates the effects in writing, with much longer waves, and with
less rise and fall. It is a review of a new car model.
(23) “The outgoing car’s been hailed for its fine balance for some 16 years, and fans
were concerned the recipe would be lost in the model change.
They needn’t have worried. The iconic sports car may be bigger, but not by
much; and the lightweight, fine-handling, nicely-balanced and affordable
sports recipe appears to have been retained.”
(AA Autofocus, Autumn 2005, page 7)
The long Entity group, “the lightweight, fine-handling, nicely-balanced and affordable
sports recipe”, illustrates two further points. Compared with the rest of the sentence,
the whole string of premodifiers has heightened stress, and forms a crest – a very long
one. However, it has its own internal structure, a rather unusual one: from ‘nicely
balanced’ onward, the details repeat what has already been stated, or are otherwise
unimportant; so the crest of the wave comes very early in the group.
10.5.3 Wave structure in aesthetics and emotion
Emotion. The discussion has so far has dealt with descriptive meaning; emotive mean-
ing also creates wave forms. Example (24) illustrates that; the speaker is describing the
skipper of his ship.
(24) “I tell you, this quiet, bowed, bandy-legged, almost deformed little man was
immense in the singleness of his idea, and in his placid ignorance of our agi-
tation.” (Joseph Conrad, Youth)
As in example (23), a series of premodifiers, signalled by the commas, creates a rising

trend, developing our expectation of a significant peak. Here, it is a peak of feeling – of
dislike – not of concepts in an information structure; that peak in the Topic has a bal-
ancing peak in the Comment (“immense… singleness”) – the feeling of admiration.
Aesthetics. It will be clear to readers that literature creates wave structures, both in
the small and strictly linguistic scale of rhythm, and in the larger scale of climaxes to
scenes and chapters. Those structures, like other literary devices, rouse the imagina-
tion and stir feeling – which are aims of conversationalists, advertisers, politicians and
so on, as well. Part of Lincoln’s Gettysburg speech, example (25), illustrates that. It was
delivered as part of the dedication of a memorial to soldiers who died in the American
Civil War; the quotation here is the well-known ending.
(25) “It is rather for us to be here dedicated to the great task remaining before us –
that from these honoured dead we take increased devotion to that cause for
which they gave the last full measure of devotion – that we here highly resolve
that these dead shall not have died in vain – that this nation, under God, shall
have a new freedom – and that government of the people, by the people, for
the people, shall not perish from the earth.”
(From the Bliss copy of the speech)
The long series of slowly rising and quickening rhythmic waves creates a rising trend
which builds expectation and a forceful climax – with a pattern of marked stress on
the prepositions by and for – and a short falling trend leading to a quiet resolution.
The secondary aesthetic intention reinforces the primary political one.
10.5.4 Conclusion: Wave structure
The phonological indicators of peaks and troughs are not symbols, with convention-
alised and sharply defined meanings. Their basis is the expressive value of forceful
and rhythmic sound, so they have an indexical element; however, rhythm has become
conventionalised in part, so they act as markers. Fundamentally, I believe, these struc-
tures work as music does, below consciousness, having effect rather than “meaning”.
10.6 Discussion: Other structures
Interjections like bah and whew certainly seem to be marginal to language, and the
other phenomena discussed in this chapter, such as holophrases and ideophones, may
seem to be marginal, also; but in some ways they are central.
They are important functionally. The structures with no internal structure studied
in §10.2 – interjections, roughly – are valuable for making expressions brief, punchy
and forceful; they are appropriate to the expressive and interpersonal metafunctions,
as well as to ideation, whereas discursive language suits only ideational well. The mul-
tifunctional structures – holophrases and ideophones – provide imaginative stimula-
tion, by carrying several layers of meaning at once.
The structures in this chapter are also noteworthy for highlighting some impor-
tant points about the semantics of English that are not apparent in the structures dis-
cussed in Chapters 5 to 9. They rely on underspecification: full specification is not
necessary or inherent in English. For all of them, function is more important than
content, and some have no content at all: fundamentally, language is functional; and
where language does have content, conveying content is the function. Just as writing
is secondary to speech, so spoken words as patterns of phonemes are secondary to the
patterns of pitch and stress: phonology is both basic and central to language. Just as full
specification of content, and even of intention, are not essential to language, nor is be-
ing articulated into syntagmatically organised units, and being integrated into a text.
10.7 Conclusion: Other structures
Outline of the structures, and their relations. When we analyse connected text, and
when we read traditional linguistics, the whole structure of language seems to be hier-
archic – of a kind that can be naturally presented in tree diagrams. When we think of
the meaning which underlies text, the whole structure seems to be that of a network,
like that of the knowledge that seems to underlie it. How are the structures studied in
this chapter related to those structures?
The interjections studied in §10.2 are almost notorious for not fitting into the syn-
tagmatic horizontal structure of hierarchy; they do not relate to other words. But they
have a place in the vertical dimension, commonly ranking as figures, like exclama-
tions: “Wow!” ranks with “That’s remarkable!” as a whole utterance; unstressed “oh”
generally ranks with groups, in being an item of information. They are very weakly
related to other elements in the lexical and semantic networks, although they do have
a place in the mental networks from which they rise. The reason for their lack of posi-
tion in the other structures is that those structures are fundamentally descriptive and
conceptual in their semantics, being adapted to the ideational function which needs
bonds between its elements; by contrast, interjections serve other functions. (The de-
scriptive forms in §10.2.2 are exceptions to those generalisations.)
Holophrases, by their nature as utterances, do not fit into the semantic hierarchy
of a text. Ideophones are integrated into the semantic hierarchy in which they occur,
to varying degrees; the phonetic features of words with standard syntactic function
and semantic structure, like crash, scribble and squander, can be given extra weight as
they are uttered, giving them ideophone use; they are then will integrated. In uses like
“It fell kerplunk into the pool”, there is only weak integration – the utterance without
kerplunk has acceptable syntactic and semantic structure. Most holophrases and ide-
ophones use words with some standard meaning, so do fit into the semantic network
of sense relations.
The field structures studied in §10.4 have a natural, if somewhat surprising, place
in the semantic hierarchy. In the “bloody Taj Mahal” quotation, example (19), the
affective and social significance of bloody is to be applied to something – the whole
situation, if not to the Taj Mahal – so it is a modifier, and subordinate in the hierarchy,
although not in the usual way.
Waves relate to the other structures in a quite different way. They relate to the
semantic structure of modifiers and heads, and to the information structure; on the
other hand, those structures can occur without the wave pattern. Wave structures are
semi-independent, therefore.
Conclusions drawn. The semantic structures discussed in this chapter have raised
some of the issues first raised in Chapter 3, §3.4.2, of how English counters some of
the constraints imposed by its own nature and structure. The structures of syntax and
lexical semantics are highly adapted to the expression of information. Interjections,
however, make us see language as almost biological, being affective and a response to
stimulus, and as having value for the speaker alone; phatic expressions show language
as social communion; directive expressions show language as a mechanism for con-
trolling people. Holophrases give a cue for response, not information on which it may
be based; ideophones are often a form of play. English escapes from what may seem to
be its nature by adopting other natures.
Looking forward. The first four chapters of the book presented the basics of seman-
tic structure. Chapters 5 to 10 have set out networks, hierarchies, and other general
structures. The next chapter begins the last section of the book, which shows how the
general structures are realised, focusing on the alternatives which English offers.
Chapter 11
Realisation (1)
Interpersonal functions
11.1 Introduction
The previous six chapters have set out the semantic structures of English, considered
as structures; this chapter and the next set out the ways in which those structures are
realised. In a secondary sense of the word, however, it presents another form of struc-
ture, that of the strata or stages through which meaning is realised. It also entails the
structural relation of the potential to the actual, and the relations among the alterna-
tive forms which are options for realisation.
The two chapters serve several purposes. The primary one is to identify and de-
scribe the ways in which meanings are realised in English, illustrating realisation, not
treating it exhaustively. Secondary purposes are to aid readers’ understanding of the
structures already presented (by incidentally describing them from a fresh perspec-
tive), and to clarify the relations among the strata, as a basis for discussion in a later
chapter.
The chapters also have a persuasive or even polemical purpose. They aim, first, to
convince readers that the realisation of meaning cannot satisfactorily be described as
“coding” or as “mapping” onto speech or writing, since there is always choice among
alternatives, and since many of the realisations are indirect or complex. Second, they
aim to show that realisation is a much more important and substantial issue than it is
usually taken to be. Generative grammar sets out utterances as the “projection” of a
structure which is implicit in a sublinguistic level; much functional grammar is simi-
lar, in positing syntactic properties as part of lexical entries, so that syntactic structure
flows naturally from word choice; in both approaches, there is not much to be said
about realisation, once you understand words and the speaker’s (cognitive) intended
meaning.
The full treatment of realisation has been divided between this chapter and the
next. This chapter deals with the structure for interpersonal functions; the next deals
with the structure for the ideational function. The presentation of the interpersonal
functions flows from the intention “down” to the act of speaking; but the ideational
function is presented from the “bottom” upward. The distinction between the two
forms of realisation is one of expository perspective rather than of linguistic reality,
since in each function the two forms of realisation combine, and are partially inter-
dependent. However, there is important validity in varying the perspective, because
realisation of the interpersonal functions is often not fully worked out at the lower
levels, whereas in the ideational function lower levels are fully worked out, and hear-
ers are not so aware of the speaker’s intention.
There are two related issues which the chapters do not consider. They do not deal
with processing issues, although processing does affect realisation. Following the con-
clusions of Hawkins (2004) and Bencini (2013), I am presuming that the important
relevant effects of processing issues have been built into the grammar of English, so
need not be accounted for here separately. Second, in keeping with the approach and
presentation followed elsewhere in the book, I do not attempt the formalism given by
Halliday and Matthiessen (1999, §19.6) and Fawcett (2000), for example, with formu-
laic realisation statements and operators.
The sections of this chapter each deal with one of the interpersonal functions –
the Expressive function, and realising emotion, for example.
11.1.2 Introduction to realisation
Realisation as process. Realisation is the process of converting a speaker’s meaning or

intention into actuality. It is indicative of linguists’ confusion about realisation that
that definition, based on SOED’s basic sense of realise, is more useful than SOED’s lin-
guistic sense: “manifest (a linguistic feature) in a particular phonetic, graphic, or [my
emphasis] syntactic form.” That is confusing, because the graphic and syntactic forms
are not alternatives, since the syntactic form is itself “manifested” in a phonetic form.
The nature of realisation can be shown more concretely from a typical piece of
news reporting: “Then he saw a terrified little boy peering out a window and heard
his desperate screams for help.” The reporter had to conceptualise the events, choose
words for them (terrified or scared? peering or looking?), construct clauses (will the ac-
tion of peering make a separate finite clause?), and type them out with correct spelling
and punctuation. In all that, what was potential became actual; abstract knowledge
was rendered into abstract linguistic form, and then into physical form. The process is
much more complex than is suggested by the common terms “mapping” and “coding”.
I am also distinguishing this sense from “realisation” of language in the brain, as
discussed by Lamb (1999) for example, and the “realisation” of the meaning by the
hearer. Readers familiar with Systemic Functional Grammar should also note that this
sense is distinct from the sense used in early work of Halliday – the realisation of gram-
matical systems, not of meaning, which is referred to as “instantiation” in Halliday and
Matthiessen (1999) and Fawcett (2000).
Realisation as meaning-form relation. That explanation has been in terms of a pro-

cess, but “realisation” also means “the relation between a set of meanings and a set of
forms” (Fawcett 2000, p. 34). That relationship between the potential and the actual
in language is the primary concern for understanding semantic structure; analysis of
Chapter 11. Realisation (1): Interpersonal functions 311
the speaker’s mental processes is an aid to that. Moreover, description of realisation

as process is a little inaccurate and unsatisfactory, because we do not have a good psy-
cholinguistic account of it, and the stages often overlap. These two chapters will view
realisation both ways, but the process view is subordinate. It will be used sometimes as
evidence; chiefly, it will be used as an expository metaphor, using the stages in the pro-
cess to stand for the layers in the structure. The study is linguistic, not psycholinguistic.
Other issues. In the reporter’s mental process outlined above are several important
issues that must be resolved. Phrases such as “a terrified little boy” and “peering out
a window” put the words one after the other, but in the real world there was no se-
quence of terror and boy, or peering and window – pictorially, each needs a snapshot,
not a motion picture: realisation entails making a linear and directional structure
from the non-linear, multidimensional network of meaning (cf. Chapter 3, §3.4.2, on
linguistic constraints.) Second, it is clear that syntax makes a “stratum” distinct from
meaning, but it is not clear whether morphology is distinct from syntax in English.
Third, the meaning being expressed seems to make a seamless whole, but the expres-
sion consists of discrete units i.e. words; does the process of realisation include indi-
cating how the hearer is to reconstruct the meaning from the items compositionally?
Finally, we should consider the role of the semiotic strategies (naming, describing
and pointing), and the sign types (indexes, icons and symbols). Those issues apply to
the realisation of conceptual information, in the ideational function, where the stages
run through the set sequence of strata from semantics through lexis and syntax to
graphology or phonology.
When we consider the realisation of other intentions, in the interpersonal func-
tions, the issues are rather different. For example, to realise the intention of asking the
hearer a question, the speaker must choose between using the rising tone (tone 2) in
the phonological stratum (Chapter 2, §2.5.3.2), the interrogative construction in the
syntactic stratum, or the question tag in the lexical stratum. Moreover, realisation of
conceptual meaning may not be called into action at all, since emotive and attitudinal
meaning may be expressed by phonology alone. Again, those meanings and social
meaning are often realised by the dimension of word meaning, but that dimension
does not need further realisation. Finally, realising interpersonal intentions such as
greeting people and expressing a feeling use the ideational function of expressing
conceptual meaning as a means. In all those ways, the realisation of interpersonal
functions is different from that of the ideational function. It starts from the most gen-
eral and abstract level, the “top”, and works “down”; it offers choices among forms of
realisation; it proceeds by differentiating a general intention into secondary intentions
and into the nature and content of what might be said. The ideational function starts
from the concrete and specific, the “bottom”, and works “up”; it does not offer as much
choice among forms of realisation; and it works by combining many elements, not by
differentiating one element into many. As the “up” and “down” metaphors suggest,
the two meet “in the middle”; the nature and location of that meeting will need to be
examined also.
Other perspectives. The top-down view is illumined usefully by considering it from the
developmental perspective. Children in their first year of life have a drive to commu-
nicate with carers; that general intention becomes differentiated into the functions of
expressing needs, getting practical help, sharing feelings and so on, and only later gains
support from the ability to formulate a thought in language. Psychological and psy-
cholinguistic understanding of the interrelations among language ability, perception,
emotion and so on enables us to understand how language intentions are realised.
Although the top-down understanding is accepted regularly in developmental
studies, it is not often adopted in linguistics. Among the work that does accept it
are Functional Discourse Grammar (e.g. Hengeveld and Mackenzie 2008), Harder
(2007), and Fortescue (2007).
11.2 Realising the Expressive function
11.2.1 Introduction
As explained in Chapter 3, §3.3.1, the Expressive function is the vocalisation of affect,

serving to release tension – releasing the physical and affective energy which is not
discharged in goal-directed action or gesture or paralinguistic features of language.
When it is the sole motivation of an utterance, there is no descriptive meaning, since
it does not involve the mental faculties of judgement and intellect; some vocalisations
of pure affect have become conventionalised, as in whew and bah. Expression is effec-
tive in releasing tension because of the physical effort required – the effort to breathe
out strongly, to tighten the vocal cords, and to restrict the airflow in the articulatory
organs – and the greater the affect expressed, the greater the volume, and the higher
the pitch.
Semiotically, then, Expression is indexical. Hearers take significance from it in
the same way as they take the significance ‘fire’ from seeing smoke. Speakers do not
intend it to have any significance; it does not have meaning in the usual linguistic
sense of the term. Often, however, speakers combine the Expressive intention with a
second intention, that of showing those present how they are feeling, with the result
that there is an Expressive element, which is in this circumstance intentional and
therefore part of linguistic meaning, combined with other elements.
The chapter examines the different forms of realisation, in turn.
11.2.2 Holistic realisation of Expression
Example (1) comes from a newspaper report of a woman who saved her husband
from drowning, by jumping into the water and supporting him until help came.
(1) “Senior Sergeant Martin Paget … said … ‘She held him at the rear of the boat
and started screaming for help.’
A woman in a boat berthed nearby said the wife’s screams shocked everyone
into action. “The lady was screaming, ‘Help, help’. ”
(New Zealand Herald, A7, April 30th 2014)
The utterance, “Help, help”, was in part deliberate, a piece of regular communicative
English using the imperative. But it was uttered as a scream, so it was in part also lan-
guage in the Expressive function, evidently realising the rescuer’s panicky state, in a
holophrastic utterance. (It presumably used the phonological and phonetic methods
discussed below, as well as a lexical item.)
11.2.3 Phonological realisation of Expression
Example (2) comes from a video recording of a plane landing at San Francisco airport,
which struck the beginning of the runway heavily, and skidded along belching smoke,
before exploding. The words were presumably spoken unintentionally.
(2) “Oh my God. Oh my God. [Repeated half a dozen times] Lord have mercy.”
(Television news, July 7, 2013. Written down immediately from memory,
but not recorded electronically)
The phrase, “Oh my God”, was spoken fairly quietly, with the pitch dropping a little,
and without noticeable stress or rhythm. “Lord have mercy” was spoken with little
of its usual stress pattern, and falling further in pitch. The speaker’s voice was tense
throughout.
As spoken, the repeated “oh my God” had none of the normal features of English
phonology: neither stress, neither of the falling and final tones (1 and 5), nor rhythm,
and the tone group did not signify a unit of information. It worked Expressively by
that abnormality – by being marked, in all those ways. (It also drew on the paralin-
guistic feature of voice quality, and the emotive value of calling on God.) The utter-
ance served the Expressive function purely, being a release of the speaker’s horror, and
not addressed to any audience.
More often, the phonological features used for the expressive function are marked
stress, high pitch, and the rise-fall tone (tone 5), as in example (3). It was spoken by a
man almost overwhelmed at having been present at his child’s birth, and at having cut
the child’s umbilical cord himself.
(3) “I cut the cord! … [Pause for about 5 seconds] … I cut the cord! … [Pause for
about 5 seconds] … I cut the cord!” (UK documentary television
programme, broadcast in New Zealand on July 2nd 2014)
The repeated statements are quite neutral lexically, but had great Expressive power
in the marked phonology – as did the extraordinarily long pauses, as a paralinguistic
feature.
11.2.4 Phonetic realisation of Expression
The utterance “Help!” in example (1) was screamed. It relied for its expressiveness on
what screaming is: loud and high-pitched vocalisations, uttered forcefully. Hearers
take the pitch and volume as indexes of the forcefulness, and thence of the strength
of the affect they express. That utterance of help was thus an instance of the phonic
use (“sound symbolism”) discussed in Chapter 10, §10.2.4. This section amplifies that
account, applying it in the context of the interpersonal function, and using another
perspective.
Gussenhoven (2004, §5.7) makes forcefulness one of the three “biological codes”
that language uses – the “effort code”. Animals respond to forcefulness as a signal of
the affective state of other animals, and use it as a signal of their own state. The second
code is that of “frequency” (that is, pitch), which signals the size of the animal uttering
it, bigger animals’ sounds being of lower pitch. The third code is that of “production
phase”: in the first phase of being produced, a call rises in pitch or volume, and it falls
in the final phase; the call thus signals incompleteness or completeness. Gussenhoven
cites experimental evidence to show that the codes apply in human speech, and illus-
trates its relevance cross-linguistically.
Screamed help used those codes, especially the effort code: the volume, the high
pitch and the aspiration of /h/ all needed effort, signalling urgency. The production
phase code would have had some significance, as the utterance would have risen in
pitch and volume, signalling incompleteness, and need for help thereby. Thus the
phonetic realisation reinforced the sort of phonological realisation discussed in the
last section.
Phonetic realisation applied also in example (2). The phrases were low in pitch,
without energy or effort, signalling the speaker’s low, almost depressed, state. Each
phrase – and the whole series – tailed off, signalling finality and hopelessness.
Finally, a great deal of swearing is Expressive in function, gaining its effective-
ness from its phonetic value, in the ways discussed in Chapter 10, §10.2.4. The words
typically begin with heavy aspiration, as in “Shit!”, or plosives as in “Bugger!”; They
typically end with a complementary effect, as in the sharp ending of “Shit!” and the
tailing off in “Damn it all!”
Discussion. The treatment here of the phonic value of language fits well with other
elements of language with a biological basis. Besides the obvious general reliance on
breathing, there is the evident dependence of tone groups on our out-breath, the de-
pendence of tonicity on extra force of the lungs, and dependence of syllables on rhyth-
mic mouth movements as in chewing (Davis and McNeilage 2002). More specifically,
Gussenhoven’s codes seem to constitute a biological basis for the particular meanings
of intonation. The effort code is a natural basis for realising strong feeling by high
pitch, and for realising semantic focus by raised pitch and volume. The production
code creates signals of incompleteness and of finality, so is a natural basis for realising
breaks within the sentence and sentence endings respectively.
11.2.5 Lexical realisation of Expression
“Words” such as hmmm and ugh which were formed from Expressive sounds hardly
count as lexical realisation of that function, and other words do not suit the function
naturally, since they have descriptive meaning, or have grammatical meaning whose
nature is still more unsuitable for Expression. However, there are a couple of ways in
which other words come to be Expressive.
Some words have been bleached of their descriptive meaning over the last centu-
ry or so, leaving their phones available for phonic instead of phonemic use. Examples
include bloody and bastard; fucking has lost it in some varieties of English. Initial /b/
sounds, and aspirates such as /h/ are readily used in the effort code; initial stop con-
sonants are easily given extra aspiration. Such words, then, carry Expressive meaning
indirectly through the phonetics of their pronunciation.
A second way in which words are used for Expression is through obscenity and
blasphemy. The use of such words in many social situations is conventionally taken to
be shocking; speakers can accordingly use them semi-deliberately to express shock.
There was perhaps an element of that in example (2), “Oh my God”, although for
many speakers the expression is fairly neutral socially. Example (4) shows it inter-
polated into a statement of information; the remark refers to a submerged car from
which the speaker rescued a man about to drown, at great risk to himself.
(4) “It was on a shit on an angle, really.”

(New Zealand Herald, May 24th 2014, A5)
A third link between lexis and expression is as follows. Since affect becomes integrat-
ed with other mental faculties – domesticated, as it were – as emotion and attitude,
affective use of words grades imperceptibly into emotive and attitudinal use; “Hell!”
becomes “helluva”, for example.
11.2.6 Conclusion: Realisation of Expression
The Expressive function is realised in several ways: directly, by the phonological and
phonetic value of what is said, and holistically; and indirectly, through lexical mean-
ing. It is sometimes realised on its own, in interjections, but it is often realised in
combination with other functions, especially that of conveying emotion.
11.3 Realising emotion and attitude
11.3.1 Introduction
The intention to convey feeling, and the linguistic means which fulfil it, develop nat-
urally from the spontaneous and not fully deliberate expression of affect studied in
the last section. Infants bewail their hunger, pain and fear, in the Expressive function;
then, gradually realising that the carer responds accordingly, they begin to convey
emotion deliberately, emotion being a development from affect. The ability to convey
attitude develops somewhat later, as the ability to evaluate situations develops; atti-
tude and emotion are differentiated from generalised affect. Linguistically, they form
emotive meaning and attitudinal meaning, as introduced in Chapter 4.
All of the linguistic strata are used to convey emotion; the sections to follow deal
with them in turn – phonology in §11.3.2, and lexis and syntax in §11.3.3. The expres-
sion of emotion and of attitude are often integrated, since our attitude to a thing often
controls the emotion we feel; but attitude will be dealt with separately, in §11.3.4.
11.3.2 Conveying emotion by phonology and phonetics
Introduction. The passage in example (5) is taken from a documentary television pro-
gramme about a couple who developed a poorly managed farm. It was spoken by a
neighbour, who set out to compare the farm as it was when the couple took it over
with the way it was at the time of speaking; in fact, however, the passage is much less
description than expression of emotion through phonology and phonetics. (Ragwort
is a noxious weed, very difficult to get rid of, and illegal to have on your land.)
(5) “The whole farm was just a sea of ragwort!! It was everywhere! You just
couldn’t believe how much ragwort could grow in one place! And, I dunno
how he’s done it but I think it’s just sprays – put a lot of time and effort and
work into it. Um. But he’s never gone round with a knapsack or some herbi-
cide or anything like that. Look at the place! It’s got no ragwort! And – um –
yeah – once again it’s pretty mind-blowing stuff actually – so – unbelievable!”
(New Zealand TV1, Country Calendar, “Chaos Springs”, 5th April 2014)
Phonology. The phonology acts as direct realisation of the affective meaning. It uses
stress, particularly: example (6) is the first sentence, with single underlining for words
spoken with a normal degree of stress, and double underlining for words spoken with
a high degree.
(6) “The whole farm – was just a sea of ragwort // it was everywhere you just
couldn’t believe how much ragwort could grow in one place”.
The effect comes not just from the quantitative amount of stress, but also from the fact
that subordinate words like whole, just and could were given stress normally reserved
for head words.
Rhythm is important also: the first sentence was heavy and slow, as the underlin-
ing suggests. In the middle of the passage, stresses were lighter and far more widely
spaced, making the rhythm quick and almost urgent; the syntax aided that, since its
longer structures provided longer intonation groups, which let the rhythm flow more
freely. At the end, the pattern of marked stress used at the beginning was used again.
Tone was important also, with the exclamatory tone, tone 5 (Chapter 2, §2.5.2), being
used frequently.
Phonetics. The phonetic value given to the words aided the strength of feeling. In the
first two sentences, the stressed vowels were lengthened: “The whole farm was just a
sea of ragwort! It was everywhere.” The first three sentences and the end of the pas-
sage (from “once again”) were quite high in the speaker’s range. That lengthening and
the high pitch, like the speed and increased stress, also expressed strong feeling here,
as they often do. They derive from the effort code, since more elaborate articulation
needs greater effort in speaking. With the stressed words at the end – mind-blowing
and unbelievable – the speaker resumed the same style of heightening the phonetic
value of the sounds.
Discussion. Words that in themselves have emotive meaning seem to be the obvious
way of expressing emotion, so it is striking that in the passage’s extensive expression of
emotion there are none, until the end: in most of the passage, the feeling is conveyed
wholly through sound, using an iconic strategy in which phonological rise and fall
images the rise and fall of feeling and information value. The descriptive meaning
is merely support for that, the speaker using it to imply the emotion. The speaker
changed his strategy to descriptive and symbolic semiotics at the end, however, and
summed up with the strongest emotive words he had available – mind-blowing and
unbelievable. Unbelievable acted as a holophrase, to sum up the passage – content and
emotion – in one word.
11.3.3 Conveying emotion by lexis and syntax
Emotion is often conveyed through lexis and syntax, not through phonology, espe-
cially in relatively formal English. That is illustrated in example (7) – cited previously,
in Chapter 9 – where the speaker is a businesswoman discussing the merits of dried
seaweed for making fertilisers and animal feed supplements.
(7) “You know, it’s got almost every known element, but it’s a packet of potential
value, you know. It’s packed with amazing nutrients, growth hormones, vita-
mins, minerals – if you dry it properly. Yeah. Ah. So it’s like a – a smorgasbord
feast for, you know, for soil, and plants, and for animals.”
(New Zealand TV1, Country Calendar, “The Good Weed”, 10th May 2014)
This speaker used lexis, chiefly, to convey her admiration for seaweed. She used words
with emotive meaning (amazing), and favourable attitudinal meaning (value, nutri-
ents). She also used words that in themselves have no emotive meaning, but she em-
ployed them in figurative use (Chapter 4, §4.7.4): almost every known element was
hyperbole, and packet, packed, and smorgasbord were metaphors.
The syntactic structure aids that in an interesting way. There are two complex
structures of coordination: the clause beginning “It’s packed with” has four co-
ordinated Complements; the last clause has for linking three co-ordinated nominal
groups. Moreover, both co-ordinating structures are in marked form. The unmarked
pattern in English is to list most items with a pause or comma, and the last item with
and; but the first structure here has no and, implying that the list is unfinished, and
giving the suggestion that it could go on indefinitely. The second structure uses both
and and a pause for each item, again making the listing emphatic, and strengthening
the feeling.
The passage was spoken with subdued intonation contours and reduced stress,
which were out of keeping with the emotive intention, giving hearers the impression
that she was restraining her feelings. They burst out in the phonology occasionally,
however, in the stress and long vowel given to amazing, and especially in the weighting
given to the climactic phrase, “smorgasbord feast”, by the anticipatory pause, higher
pitch and extra stress, along with an upward movement of the eyes and jerk of the head.
This speaker’s favouring of content, and realisation of emotion through diction
and syntax rather than through phonology, makes a strong contrast with the strate-
gy of the speaker in example (5), who diluted the content and favoured phonology.
However, they were alike in using figures of speech. Example (5) has metaphor (“a sea
of ragwort”), hyperbole (“mind-blowing”, “unbelievable” and “You just couldn’t be-
lieve”), and “rhetorical command” (an expression to match “rhetorical question”),
in “Look at the place!” Figurative language is a major characteristic of some styles
of spontaneous and spoken English, not only of literature. In much recent linguistic
literature, particularly in work following the tradition of Lakoff and Johnson (2003),
figurative language is treated as a means of conveying information or understanding
the world; so it is worth noting that here it is an everyday means of fulfilling the
interpersonal functions of language, without information value, being redundant in
content or simply empty.
11.3.4 Conveying attitude
Attitude is often implicit in expressions of feeling, since we necessarily have a favour-

able attitude to things we admire, and an unfavourable attitude to things we fear.
To convey attitude alone, the only significant means is to use words with attitudinal
meaning. Passage (8) illustrates it. The speaker is a farmer discussing his stock, and
crediting his son for the animals’ high-quality.
(8) “[They’ve] done very well – all the cattle and calves – and the calves are quite
exceptional. So it’s a credit to him – how he’s farmed them. He’s definitely
carrying on, ah, what I did.”
(New Zealand TV1, Country Calendar, “A Dream Run”, 17th May 2014)
The content is remarkably vague, for remarks that are intended to praise. The speaker
is relying, not on descriptive meaning, but on attitudinal meaning. That is overt in
well, exceptional and credit, and in its intensification by the modifiers very and quite.
But there is covert approval in farmed and carrying on. In this context, farmed uses its
possible meaning elements of productivity and respect for the environment; carrying
on evokes the meaning elements of responsibility, tradition and loyalty. The effect is
heightened by the syntax, which places the attitudinal words at the focus of their in-
formation group. (“Carrying on” is followed by a spoken pause – “ah” – which divides
the clause and the figure into two information groups.)
11.3.5 Conclusion: Conveying emotion and attitude
This section has shown that emotion and attitude are sometimes realised almost en-
tirely through phonology and phonetics, and sometimes largely through lexis. (There
are whole lexical classes, such as terms of address, devoted to conveying attitude.) Fig-
urative use of language is also important. Even syntax, which is usually portrayed as a
mere abstract structure, devoid of even descriptive meaning, has a role in conveying
affective meanings.
11.4 Establishing personal and social relations
11.4.1 Establishing personal relations
Passage (9) illustrates strengthening personal relations. The first speaker and her hus-
band had planted 700 kauri trees in an area of gorse, at considerable cost and with no
possibility of financial return. The speaker and her grown-up daughter are at a high
point, looking out over the area.
(9) Wife: “Ah. The kauris are starting to come through the gorse a bit more now.”
Daughter: [Pause] “Mmmm.”
(New Zealand TV1 documentary on farming life, September 2014)
At “starting…”, the wife turned to her daughter, and gazed at her meaningfully during
the rest of what she said. The daughter kept on looking out over the scrub and the
trees as she replied.
The woman’s deep commitment to nature and the family farm, shown earlier in
the programme, and her facial expression as she speaks, indicate that the interjection
ah is direct expression of strong feelings – relief and pleasure. Syntax does nothing
here to aid sharing feeling, although it could have, through exclamatory structure.
Phonologically, both “Ah” and the following sentence were spoken quietly and evenly;
but there was a slight rise at the end, inviting a response and thereby indicating the
wish to share the feeling. The daughter’s reply was equally quiet and even. The lack of
overt expressiveness, apart from the drawn out “mmmm”, in fact increased intimacy,
because it implied “I don’t need to tell you…”
In the passage, personal relations were strengthened by the sharing of feelings,
which were conveyed by interjections and markedly flat intonation, and the implicit
sharing of values and attitudes. It was achieved primarily by the socially intimate act
of stating feelings, and by the pragmatic inferences to be drawn from the growth of
the trees. Its effectiveness was thus largely not linguistic, but the nature of the seman-
tics was crucial to it.
11.4.2 Establishing social relations and social status
There are several specific functions that are included in the general function of es-
tablishing social relations and social status. Summarised from Chapter 4, §4.4.5, they
are: establishing lower or higher status in the speaker-hearer relationship, by means
of “style”, in the sense of Cruse (2011, §3.4.2), i.e. by means of the degree of formality
of utterances; establishing social role, through language characteristic of the “field”
or area of discourse; establishing social identity through dialect – socioeconomic, re-
gional, and occupational; and establishing socio-economic class status. They all use
word choice, syntax and phonology in ways that have been illustrated sufficiently
already.
Some utterances used in establishing social relations are empty of descriptive
meaning, as with hello and goodbye, and politeness forms such as please. Other utter-
ances that do have descriptive meaning are so conventionalised for the situation that
the content is often negligible, as with talk about the weather in some social groups.
All those utterances are like a wave of the hand in greeting or farewell, and a formal
smile; it is not the content of what you say or do that counts, but the fact that you have
done it. The semiotics are quite different, because the sign is the fact of making the
utterance, not the linguistic form. Here, language is gesture.
11.4.3 Discussion: Personal and social relations
Semantic classes. Ah and mmm in example (9), sir as a title and as a form of address,
hi, thanks, boo! and hurrah! are meaningful – are in that sense semantic – but do not
belong to any of the semantic classes set out previously. Nor are they grammatical
items, which were excluded from the classification. We see that such a classification is
of limited value, since it purports to be a classification of word senses in general (or
perhaps of lexical items), but does not cover them all. We are reminded that words,
like language in general, are fundamentally functional rather than bearers of meaning
as a representation of our knowledge of the world. The semantic classes are classes of
only descriptive meaning; and classification is of limited value in linguistics, especial-
ly if it does not rest on strict definitions.
Social meanings as field. The words listed in the previous paragraph do not work as
“particles”, contributing another item of meaning to the accumulating total. Rather,
they work in sustaining a field (Chapter 10, §10.4). That fits, of course, with the nature
of all personal and social relationships, which carry on indefinitely but need periodic
renewal.
11.4.4 Conclusion: Personal and social relations
This section has illustrated the fact that an utterance’s significance, as something
broader than “meaning”, is often created rather than stated. Speakers use indirect
means, relying on hearers’ being aware of the range of options that speakers have, and
accordingly taking significance from the choice made. That social importance of how
you speak parallels the importance of what you talk about and what you say.
There is a wide range of means for realising the intention of establishing social re-
lations and social status; phonology, syntax and lexis all have forms associated with a
range of social relations. Fulfilling that intention is a major function of English, along
with conveying information.
11.5 Guiding hearers’ use of meaning
11.5.1 Introduction: Realisation of grammatical meaning
Previous sections dealt with realising content (affectual rather than conceptual); this
section deals with realising the speaker’s guidance of how hearers should use that con-
tent in constructing the total message. Put in terms of meaning types, those sections
dealt with affective meaning (§11.3) and social meaning (§11.4); this section deals
with realising grammatical meaning.
Grammatical meaning has been described as working at two levels. At the general
level, it guides hearers to the overt response speakers want: acceptance (of a state-
ment), reply (to a question), action (in response to a command or request), or the
sharing of emotion (in an exclamation). It is dealt with first, in §11.5.2. At the spe-
cific level of building semantic structures, interpretive grammatical meaning (§9.6.2)
relates syntactically defined units of content, building the hierarchy of figure, group
and word sense; that is discussed in §11.5.3. Also at the specific level, evaluative gram-
matical meaning creates information structure; that is discussed in §11.5.4. Guiding
hearers’ attitude to the content (“modality” or “irrealis”) is dealt with in §11.5.5.
11.5.2 Guiding hearers’ overt response
The most familiar realisation of guiding hearers’ overt response is syntactic, through
the structure of Subject and Predicator, as in “He does it”, “Does he do it?”, and “Do
it!” – grammatical mood. Exclamation can also be realised by the grammatical items
what and how. However, the basic form for guiding hearer response is phonological,
by the rise and fall of tone (Halliday and Greaves 2008); that is tone’s basic func-
tion. If the desired response is signalled by syntax, no phonological realisation of the
meaning is needed, but intonation is usually used anyway, with mood and tone being
matched: declarative mood and falling tone; interrogative mood and rising tone; ex-
clamatory mood with high rise and fall tone; and imperative mood with level tone.
However, tone frequently replaces syntactic mood, not just reinforces it, as in the
third sentence in example (10) – “There’s a bloody lot…”.
In writing, punctuation provides a graphological substitute for some of the pho-
nology, with the full stop, exclamation mark and question mark standing in for the
relevant tones. If we take the tags is he and isn’t he as lexical items, then lexis is another
stratum in which this guidance can be realised.
Actual usage is not as simple as the syntax-phonology match above may suggest.
Speakers provoke the desired response in various ways, often combining them. That is
illustrated in example (10). It was spoken by a man quoting what a woman nurse said
to him, while he was in hospital with leukaemia.
(10) “Jesus Christ, Joe! Snap out of it! There’s a bloody lot of worse people in here
than you are! Get your act together, boy!”
(NZTV1, Country Calendar, May 25th 2014)
Syntactically, the first sentence consists of two interjections, forming an exclamation;

of the others, two are imperative, and one is declarative. But phonologically, they were
all exclamatory: throughout, the nurse was using tone to guide Joe into responding
emotionally. In two sentences, she was using imperative syntax to goad him into ac-
tion, as well as emotion. In the statement, “There’s’ a bloody lot…”, she was guiding
him into accepting fact, to lead to emotion and action. (Other things were going on at
the same time, of course: she was expressing her own affective state – see §11.2 – and
establishing social superiority, as a nurse and as someone experienced, through tone
and the use of boy – see §11.4.2.)
“Jesus Christ, Joe!” and “Get your act together, boy!” were each spoken in a sin-
gle intonation group: the units for guiding hearers’ overt response are phonological,
not syntactic – or graphological, as our punctuation sometimes makes us think. That
apparently minor fact has an important consequence for the long-running and unre-
solved discussion of “speech acts”. Tonality is a single phenomenon, and it is linguistic;
the relevant tones make a paradigm. Therefore, we are dealing with a fairly simple
linguistic category, which is much better understood as guidance of hearer responses
than as “speech acts”. The category is also clear-cut syntactically, although the para-
digm of forms is not so symmetrical.
Phonological realisation is basic. It is used in utterances that are incomplete sen-

tences; “out” can be statement, question, command or exclamation according to the
tone used. If the signals given by phonology and syntax conflict, phonology domi-
nates: hearers respond according to the phonology, and usually take the syntax as a
signal of attitude. In example (10) for instance, the hearer took the syntactic state-
ments (“There’s a bloody lot…”) and exclamations (“Jesus Christ!”) as signals of sar-
casm. Phonology is mastered by children before the relevant syntax. To scholars who
take an evolutionary approach to language origin, phonological realisation must have
preceded syntactic realisation.
For the sake of understanding the relation between meaning and its signs, it is
worth noting that the category of assertion is usually covert, being clear only through
contrast. As Dimroth et al. (2003) note, “He is Minister for War” as a neutral state-
ment contrasts both with “He was Minister for War” (past tense) and “He is not Min-
ister for War” (denial). The Predicator in “He is Minister…” serves to signal present
tense, and to signal the fact that the speaker is making an assertion. The finiteness of
the Predicator signals assertion, but only covertly.
11.5.3 Guiding hearers to syntagmatic structure
General. This section deals with interpretive grammatical meaning, which consists
of guiding hearers in interpreting content by relating its items – the morphosyntac-
tic units discussed in Chapters 7 and 8. The section illustrates the variety of ways in
which that grammatical meaning is realised.
All strata have a share in the task. Phonology uses stress to signal items of infor-
mation; the tonic stress marks the “head” item of the figure, the information focus;
phrase stress marks the heads of semantic and syntactic groups. That has a basis in
the effort code, since higher pitch requires greater effort (Gussenhoven 2004, p. 85).
In lexis, content items outweigh grammatical items, and Entity and Event senses out-
weigh Property senses.
Syntax also has a vital role, via the position of words in the group, and the posi-
tion of groups in the clause. Semantics itself – the meaning of words, that is – also has
a role, since the significance of word position often depends on the meaning of the
word. Even when we know that a word is in determiner position, we need to know
its meaning before we can interpret it as predeterminer, central determiner or post-
determiner. Likewise, we often need to consult both the meaning of the word and the
meaning of the context before we can decide whether to interpret a nominal premod-
ifier as Epithet, Descriptor or Classifier; compared “blue murder”, blue coat” and “blue
shift” (see Chapter 8, §8.2.2).
As we have seen earlier, the phonetic value of words, distinct from their phone-
mic value, can also signal the interpretation of content. The role of aspiration of initial
consonants is now well established as a signal of intensification, paralleling the role of
intensifying words like very, as when the ch- “That’s childish!” is aspirated, and when
the t- in “How tiresome!” is given aspiration beyond the normal.
Phonology is not usually treated as having a grammatical role. But it should be

considered more basic than lexis for grammatical meaning, since grammatical items
signify how to relate subordinate words to head words which are already identified by
tonicity, just as phonology can carry meaning even where there are no lexical words,
as with drawn-out “Ooooh” and “Aaaah”, given appropriate tone and stress pattern.
Relation of grammatical meanings to the strata and layers. Primarily, subordination

and its opposite, “superordination”, are realised phonologically, by stress, as noted just
above. (They are the meanings, “Treat this item as head of its unit,” and “Treat this item
as subordinate in its unit.”) Coordination (“Link two equal items as having the same
function”) is realised lexically by conjunctions such as and, or and but. Complementa-
tion (“Link two equal items as having different functions”) is realised syntactically, in
the Subject-Predicator-Adjunct construction, for example. Thus, those general gram-
matical meanings (subordination, coordination, and complementation) operate on
complementary strata, just as those meanings are themselves complementary.
The specific grammatical meanings contrast with the general meanings in their
realisation, as in their function. They are realised, with only minor exceptions, in lex-
icogrammar, through the interaction of lexical meaning and syntactic position. For
example, premodifying senses may determine the meaning of the head, as in “a boy”,
or add their meaning to the head, as in “a six-year-old boy”, or adjust the meaning
of the head, as in “a mere boy”, according to their position and their own meaning.
The minor exceptions mentioned are that marked stress and phonetic aspiration may
carry specific grammatical meanings.
It is striking that grammatical meanings are themselves stratified, like the mean-
ings they organise.
11.5.4 Guiding hearers to information structure
The grammatical meanings studied in the last section built morphosyntactically de-
fined units of content into groups and figures. The evaluative grammatical meanings
studied in this section build information structure out of the descriptive meaning as
message or “information”, through evaluating its relevance, orientation, and salience,
as described in Chapter 10. Sections 11.5.4.2 to 11.5.4.4 each deal with one of those
structures within the figure; Section 11.5.4.5 deals with the structures on a larger
scale, in paragraphs or texts.
11.5.4.2 Relevance structure

Introduction. We are concerned here with how speakers realise their intention that
hearers should interpret a particular concept or fact as the Topic of the utterance.
As previously noted, that guidance is not realised very systematically. In particu-
lar, status as Comment is not marked, being inferred by hearers as implied by pres-
ence of a Topic; Topic itself is sometimes unmarked, being taken by default as being
expressed by the Subject, in contexts where Topic-Comment structure has been al-
ready established; and when it is marked, the signs for it are quite varied, English
having no specific Topic markers as some languages do.
Signals of Topic status. Realisation of the speaker’s intention that certain content be
Topic is done in various ways. Use of a pronoun indicates continuity of the item, and
therefore topicality. So does ellipsing the Subject, and repetition, either of the same
word, or of phrasal content in a resumptive pronoun. New topics are often signalled
by various sorts of “dislocation”, or by the presentative construction, “There is …”: that
is its reason for existence. They can also be signalled by an almost holophrastic con-
densation of a situation into a single word or phrase, as in the journalistic headline,
“I formula-fed. So what?”
In speech, phonological signals are used also, such as weakening the stress on the
Subject. According to Gussenhoven (2004, §5.8), the production code is used: a high
beginning signals a new Topic, and a low beginning signals continuation of the Topic.
That seems valid, but only to a certain extent, since high and low beginnings some-
times have other significance, in the signalling of turn-taking, for example. My own
observation shows that in speech pauses after Subjects are markers of being Topic (and
also of being Theme), as when a newsreader begins a new item “China [pause] has…”.
There is much more that could be said, on marked forms, for example (see
Halliday 2014; and Birner and Ward 1998), and on change of Topic, which has its own
signals. However, we have seen enough to conclude that hearers and readers are alert-
ed to Topic status by a wide range of signs, in both syntax and phonology, but not by
a lexical marker. That fits my suggestion that information structure is relatively recent
in English, having yet to establish systematic marking.
11.5.4.3 Orientation structure

Introduction. Orientation is carried out through the Theme, which is initial content
that orients the hearer to the content that follows, as the Rheme.
Theme. Morphosyntactically, Theme status is signalled by position: the hearer takes

as Theme whatever comes first. That may be something which appears to be stated or
placed for that purpose, as many Adjuncts are, but it may be the Topic; both of those
coincide with a unit of clause structure. On the other hand, the Theme may take the
form of a vocative or other expression catching attention, or expressions like a drawn
out well or oh which express speakers’ attitudes to what they are about to say; those
forms stand outside the clause structure.
In groups, determiners and auxiliaries have a Thematic role as well as their con-
tent role. In a prepositional phrase, the preposition signals that a nominal group is
to follow, and may specify the relationship, e.g. as spatial or temporal. (See Chap-
ter 9, §9.3.3.2.) In clause complexes, the form of the Theme is more predictable: the
first clause acts as Theme for the whole complex, as in “While we’re in the mood for
gambling, let’s have a punt,” (the first sentence of a business columnist’s newspaper
article). Relative pronouns and prepositions act as Theme for individual subordinate
clauses, as with “by making a deal with an Asian investor,” (from the same article).
Conjunctions (and, but, or) act as themes for co-ordinate clauses. In all those instanc-
es, themes orient the hearer by specifying the relation of the following content to the
rest of the clause or complex. (See Chapter 9, §9.3.3.1.)
Phonologically, Themes are marked by a general difference in the level. Typically,
the pitch is higher, but it may be lower, and differences in rhythm and degree of stress
may be used. They are sometimes also marked by being made a separate intonation
group. For instance, in example (9), exclamatory “Ah” was such a separate group, act-
ing as Theme, signalling the speaker’s daughter that she was to treat “The kauris are
starting to …” as feeling to be shared, not mere statement, as the syntax indicated.
At the scale of whole texts, Theme is of course often stated explicitly, either within
the text, as in “I want to begin by …”, “The next point…”, and “By the way”, or outside
the text, as a heading.
11.5.4.4 Salience structure

High salience: being in focus. Guidance as to what readers and viewers are to take as fo-
cused is realised regularly by stress in speech. Gussenhoven (2004) seems to be correct
in attributing it to the effort code, though we should presumably include length and vol-
ume of the sound, as well as a pitch, to which Gussenhoven restricts the effort signals.
Syntactic end position is a regular signal of focus, in both speech and writing.
A familiar example is the use of the “dative alternation”: “He gave the book to Joan”
has Joan in focus; “He gave Joan the book” does not. Morphology is used as well: in
“He gave her the book” uses the pronoun (instead of a name) to help keep Joan out of
focus, by eliminating content. In marked use, pro-forms can signal that focus is not at
the end, as expected; that is illustrated in example (11).
(11) “By January most of the Romanians and Bulgarians who wanted to move had
already done so.” (The Economist, May 17th 2014, p. 51)
“Done so” means ‘move’; avoiding move and using an empty pro-form indicates to
the reader that the focus is elsewhere, namely on already. The expression “had moved
already” would also have focused already.
Demonstratives can also signal that information should not be focused, as in ex-
ample (12).
(12) “I visited the dairy and saw those mountains of manure and I just kind of
shook my head and went, ‘Really?”. (New Zealand TV1, Country Calendar,
“Chaos Springs”, 5th April 2014, as for example (5) above)
“Those” marks the ‘mountains of manure’ as familiar to the hearer, and therefore
not important as information. Also, the figure was run on with the next two figures,
making a simple single intonation group ending with “really”, which was the focus;
the hearer was thus guided into treating the previous four clauses as subordinate,
informationally.
Word choice also contributes to signalling focus. Compare the use of warn and
warn off in examples (13) and (14), where the focus is underlined.
(13) “Something in the glitter of the dark eyes warned him.” (COCA)
(14) “David’s belligerent stance warned him off.” (COCA)
In (13), the focus is narrow; the broad focus in (14) is created by the use of the phrasal
verb.
Word choice contributes to focus in a different way in example (15).
(15) “Meanwhile the Tory war machine has been quietly uncovering a minority of
rogues and racists among the UKIP’s candidates.”
The UKIP’s candidates were mentioned just before, so have little information value.
Rogues and racists, by contrast, have their information value heightened by their emo-
tive meaning; indeed, it is their emotive content, not their referential meaning, that
is in focus. Consequently, the location of the focus is clear, although it is not in the
normal place and is not marked directly.
The statement, “We will price match any competitor”, would be expressed in
standard English as “We will match any competitor’s prices”; but it was used as a re-
tailer’s slogan because it put the competitors in focus. The prices have been moved out
of focus in the final phrase by incorporation into the Event group – a regular function
of noun incorporation. Similar examples are “He trail blazed his own way” and “to
acid test an idea”; the salience of the now rather clichéd phrases, “to blaze a trail” and
“the acid test”, is reduced by the incorporation of trail and acid into the Event group
as Predicator.
The meaning that guides viewers and readers in assigning focus is thus realised
in all strata of expression. The various realisations are often used together, reinforcing
each other; but they are independent to the extent that none is obligatory, and that
contrastive stress overrules other signals such as end position.
Gradient changes in salience. The relative status of items of information in the Rheme
other than the focus is indicated by means already mentioned as ways of keeping
information out of focus: less stress and lower pitch in speech, and earlier position.
Another use of pitch is the marking of apposition by an abrupt change of pitch.
That can be illustrated by examples (16) and (17), from example (8) above.
(16) “They’ve done a very well, all the cattle and calves, and the calves…”
(New Zealand TV1, Country Calendar, “A Dream Run”, 17th May 2014)
In (16), the appositional group was spoken with a sharp step down in the fundamental
frequency of the voice, then a step up again. In (17), the first step was upwards.
(17) “So it’s a credit to him, how he’s farmed them.” (As for (16))
A final use of pitch is that speakers sometimes mark the beginning of the new in-
formation by a rise in pitch – an “up jump” (Halliday and Greaves 2008, p. 173).
Example (18) illustrates the use. (Bolding indicates main stress; underlining indicates
higher pitch.)
(18) “I / want to know / how to /… /how to / cope with / stress / really.”

(Halliday and Greaves 2008)
Cope carried an up-jump, marking that information and the remainder as far as the
tonic stress as important information. Speakers do not always mark the unimportant-
important transition phonologically, but when they do, the up-jump is the typical
method.
There are several syntactic methods for controlling salience. A figure will be less
salient if it is realised as a subordinate clause, not a main clause, as with the segment
following “[i]” in example (19). It will be less salient still if rankshifted to group level,
or reduced in structure to a group. In example (19), segment [ii] reduces to a group
the figure, ‘He gave the deals in return for business favours’; segment [iii] similarly
reduces the figure, ‘A court judgement stated that fact.’ Cut-price reduces the figure,
‘He cut the price on the deals,’ to a single modifying word, with minimal salience.
(19) “A wealthy Auckland businessman, [i] whose links to the National party led
to a senior minister’s resignation, gave cut-price real estate deals to a Chinese
politician [ii] in return for business favours, [iii] according to a court judge-
ment.” (New Zealand Herald, June 3rd 2014, news report)
Another syntactic method is to treat the Entity as part of the circumstances, not as
a Participant. The figure underlying the phrase, “exit velocity of the emissions [from
a rubbish incinerator]” (COCA), could have been realised as “the emissions’ exit ve-
locity” or “emission exit velocity” or “emission velocity at exit”; each variant gives a
different degree of salience to ‘emissions’; that control of salience is the main function
the Classifier constructions set out in Chapter 8, §8.2.2.
The points above apply to whole items of information; finer degrees of salience
can be marked for individual senses within an item, in the ways discussed in Chap-
ter 9, §9.4.2 – choosing among “the university’s budget”, “the university budget”, and
“the budget of the university”, for example. Acronyms reduce the salience of de-
scriptive meaning elements, in favour of reference of the whole item: “ATM” lets the
concepts of telling and automation disappear; the expression is referential, without
descriptive function.
Some linguistic discussion uses terms such as “foregrounding”, and makes a two-
way distinction between foreground and background. I suggest that the discussion
above shows that the situation is too complex for a two-way distinction, and that we
need more precise terms.
11.5.4.5 Information structure in paragraphs and texts

Introduction. The literature on information structure limits its discussion to sentenc-
es, although the structure exists in larger units, particularly in speech, as work by
Halliday and Greaves (2008) and Tench(1996) demonstrates; see also Chapter 2, §2.5.
This section will show the ways in which information structure is realised in those
larger units, in speech. Writing uses some of the same methods, and some, such as
headings, which are not linguistic; so it will not be discussed directly here.
Paratones (Chapter 2, §2.5.2) are set up phonologically. They are not so systemic
as to be obligatory or to have a paradigm of forms; but they are readily recognised by
hearers, consisting of distinct patterns of features, in varying combinations.
Methods of realisation. The beginning of a new paratone is realised by a general dif-

ference in one or several of the phonological systems: tonality, tonicity, and rhythm.
Commonly, the pitch is higher, especially at the beginning (Wichmann 2000, p. 25);
sometimes the pitch range is a wider or narrower (Wennerstrom 2001, p. 24); rhythm
and degrees of stress can change also. Specific devices include the marking of Theme
for the paratone, as when a BBC a documentary switched between sections set in
alternating locations with the alternate phrases, “In Antarctica, …” and “In the Falk-
land Islands, …”, spoken lower and more heavily, and followed by a pause. Reference
to a previous paratone can be made by lowering the pitch (Wennerstrom 2001, p. 73).
Short linking passages may be marked by a pitch range which is lower than that of the
substantive material (Wichmann 2000, pp. 26–28).
Unity within a paratone, and thereby difference between successive paratones,
is established by the general lexicogrammatical system of cohesion (Chapter 10,
§10.3.4). A less familiar way of giving information structure to a passage is through
choice among synonymous expressions. A news report on the release of an Ameri-
can prisoner of war in Afghanistan used the following to refer to him: “her soldier
boyfriend”, “Bowe Bergdale”, “Bowe”, “man”, “he”, and “soldier”. (New Zealand Her-
ald, June 3rd 2014, A4.) The choice of expression controlled the participant’s salience
relative to other information, while exploiting the need for repeated reference, and
mitigating its redundancy as information.
Analysis of a passage. Example (20) illustrates the paratones in spontaneous speech.

The passage is from a semiformal interview; it has been marked with “NP” for “new
paratone”.
(20) “Long-term, the seaweed industry has high value for New Zealand. [NP] You
see, the seaweed industry world-wide is huge – billions and billions of dollars.
New Zealand’s got the most amazing seaweeds here, and we have got the
opportunity, for a sunrise industry, that can, um, you know bring money
into our country, give jobs to local people – oh – around these remote coastal
areas, you know. [NP] So, yeah, I suppose it’s a – it’s a – y’know – a bit of a –
big vision.” (NZTV1, Country Calendar, “The Good Weed”, May 2014;
as for an example in §9.4.4.2)
The first paratone was short and phonologically neutral, establishing it as introduc-
tory. The second had an Expressive element, huge being very heavily stressed, and
the /b/ in both occurrences of billions being aspirated; that expressiveness suited the
low information density. It had a strong and emotive three-beat rhythm on “billions
and billions of dollars [pause]”. It was thus different in content and phonological style
from the first paratone.
The third paratone, with a stronger intention to inform, used less stress, and was
more formal in syntax and diction. The rhythm changed to a standard iambic one
(“New Zealand’s got the most amaz-…”), but with stronger stresses than is normal in
conversation; it then developed climactically with a single (marked) stress per rhyth-
mic phrase – sounding almost artificial: “we have got the opportunity, for a sunrise
industry, that can… bring money into our country.” The second paratone is the only
one to have Topic structure within the figures (marked by the repetition of industry).
The last paratone of the passage lost its rhythmic quality as it became reflective;
the fundamental pitch dropped, and the pitch range narrowed; there was much less
stress. The impending change was signalled at the beginning by the discourse marker,
yeah, as well as by the content of the conjunction, so. Its conclusiveness was strength-
ened by a double stress on “big vision”.
There is also an overall structure in the passage. At the beginning, “long-term” is
a Rheme, orienting discussion away from everyday activities; “the seaweed industry”
establishes the Topic. “So, yeah … big vision” rounds off with its lower-level content
and phonology, and with “big vision” as a final, emphatic summary.
Finally, we should note that most of the passage was “displaced” discourse; that
is, it dealt with material away from the deictic centre of the interview; but the last
paragraph comes back to that centre, with I, and yeah addressed to the interviewer.
It is noteworthy that syntax does nothing to create the paratones that match the con-
tent – only phonology does.
I emphasise that the passage was spoken spontaneously, without deliberate plan-
ning, although its information structure uses many features of phonology and lexis.
The corresponding structuring in writing, into paragraphs, sections and so on, seems
simply a graphological – and rather reduced – equivalent, deriving from natural
speech structure.
11.5.5 Guiding the hearer’s attitude to content: Irrealis
As well as building information structure, evaluative grammatical meaning realises

the judgement that what is being uttered does not apply straightforwardly in the real
world; that attitudinal judgement, or “propositional attitude”, will be identified here
as “irrealis”, a broad term, to include “modality”. English offers several quite different
ways of realising that judgement. It can, of course, be done explicitly – that is, lexical-
ly; but it can be done with several grammatical forms, to be illustrated by examples
(21) to (24), from a single news magazine article, “Lagarde for president” (The Econo-
mist, February 8th 2014, pp. 10–11).
In example (21), the irrealis is rendered grammatically by the subjunctive mood.
(21) “If the EU were a company, its board would have been sacked: if it were a
football team, it would have been relegated.”
In examples (22) and (23), it is rendered by a modal premodifier.
(22) “One person – who is not a declared candidate – would be far better: Christine
Lagarde.”
(23) “One supposed mark against Ms Lagarde is that ….”
In (24), it is rendered by nonfinite predication.
(24) “The bigger obstacle to Ms Lagarde becoming commission president lies in

the Lisbon treaty.”
A finite Predicator would shift the predication into reality, as a positive assertion:
“Ms Lagarde becomes commission president” or “Ms Lagarde became commission
president.”
A final grammatical method, in (25), is that of a modal auxiliary.
(25) “Employers seem to think that they can push unions off a cliff.” (News report)
In phonology, some variations of the intonation contour preceding the meaning stress
can express speaker attitudes (Halliday and Greaves 2008, pp. 175–176).
On modality, there is in the literature very little agreement – as to what the cat-
egory includes, and what subcategories exist. As shown by Van Linden (2012), some
writers divide modals into deontic and epistemic types; some add dynamic modals as
a third category; some make a four-way distinction. Van Linden goes on to discuss
categories not clearly distinct from them, such as grammatical mood, rational mo-
dality (i.e. reasonable/unreasonable) and evaluation (i.e. surprising, hilarious, and so
on). There is no consensus on the concept. Similarly, there is no consensus on what
the forms of modality are. “Irrealis” is a more recent term, with a semantic rather than
syntactic use.
What links the uses of the term, and the examples above, is “semantic”, in some
sense. Indeed, on grounds of meaning, we should include in modals, and therefore
in irrealis, the clause, “He’s in his 60’s, I’d say, but you can’t rely on that”: it serves the
same function as the regular modals such as may, and is for practical purposes syn-
onymous with them.
That clause makes clear why people use these expressions: to warn the hearer
that the proposition being stated should not be relied on fully; the expressions guide
the hearer towards taking a cautious attitude to the relevant content. That, then, indi-
cates what I take the term “modal” to denote: a type of grammatical meaning, guiding
evaluation of content. That defines the hearer aspect of modals; in the speaker aspect,
they can be seen as a qualification of the assertion being made – as in the probability
examples above – or of the ascription of a property, as in “the alleged thief ” and “a
fake gold coin”.
The “epistemic” modals discussed so far have the type of grammatical meaning
which may be paraphrased, “Treat the content modified cautiously, in adding it to the
mental model being built up.” Other modals carry a different warning, of course; for
example, “deontic” modals can be paraphrased as, “Treat the content as a statement
of obligation, not of fact.”
This use of words like probable, which carries out the interpersonal function of
guiding hearers, must be distinguished from uses which carry out the ideational func-
tion, which will be dealt with in the next chapter, in §12.6.5.1.
This section has not given a formal statement of how modality is realised, since
its main argument has been that modality does not constitute a satisfactory semantic
category; but it has shown that the significance of modality, as it is usually discussed,
is grammatical meaning, and it has shown incidentally various phenomena by which
“modality” is realised: by explicit statement, by morphological forms, grammatical
items, and content items.
11.5.6 Discussion: Guiding hearers
11.5.6.1 Signs used in realising information structure

The semiotic nature of information structure is unusual. In most linguistics, the semi-
otic sign is taken to be the word; in a more recent school, constructions are taken as
signs, since they have a meaning of their own, additional to the meaning of the words.
Here, however, we see other types of sign. The fact that certain content is to be taken
as Comment is not signalled by any overt sign, but simply by the presence of Topic
(which does have an overt sign). The significance of pronouns as marking topics is
secondary to their basic significance (identifying the referent from the antecedent; in
that role, they too are are covert signs. The use of position and order as signifiers is
familiar to us (for imperative and indicative mood, for example) although not often
included in semiotics; but here it is the position and order of content, not of words
or other syntactic units that are significant. Most of the signs concerned, then, have
their basic and sometimes complete significance from their lexical or syntactic value;
their information-structure significance is additional – indeed, epiphenomenal, hav-
ing evolved later.
The linear and directional nature of language was presented in Chapter 4 as a con-
straint, which limited the expressive power of language. Here, however, its linearity
has been exploited, as noted briefly in §11.9.6: Topic and Comment are linear chunks
of the utterance, rather than syntactic units; and Theme and Rheme are necessari-
ly linear units. Theme and Rheme exploit language’s directionality also: the Theme
would not orient the hearer, and Rheme would not have its build-up, if the text could
be heard or read backwards. In its development, English has not only used existing
signs in a new way: even constraints have become signs.
11.5.6.2 Speech “sub-acts”

“Sub-acts” are units that make up a whole “speech act”. The concept, used in Function-
al Discourse Grammar (Hengeveld and Mackenzie 2008; see also Dik and Hengeveld
1997), is useful in some ways. For example, utterances can be described as made up
of a referential sub-act and a predication sub-act; modifiers generally ascribe prop-
erties to the head Entity, but modal modifiers qualify the act of ascription. Van de
Velde (2007) deals with the old problem of “criminal lawyer” by treating lawyer as
making a “referential” act, and criminal as making a separate “predicating” act, not
an ascriptive act.
However, the concept has the same limitations as that of speech act (see previous
chapters, §3.3.4.3 and §6.2.5.2), and its application is too wide. For example, and em-
bodies the speaker’s act of linking, and a embodies his act of determining the referent.
That is, the concept applies to everything in an utterance; and where everything is an
act, there is little point in labelling it so.
11.5.7 Conclusion: Guiding hearers
The realisation of hearer guidance is very varied. It includes some precise and regular
forms, such as modal auxiliaries; but for identifying Topic the signs may be lexical,
morphological or syntactic, and may be absent. For identifying Theme and Rhematic
status, the guidance consists of order, an abstraction which is not in itself a linguistic
form, and it is the order of content which carries the significance, not the order of
words or groups.
I have suggested that this very unsystematic state of affairs exists because hearer
guidance is a relatively late development in the history of English. It is part of the
broad trend in modern English orienting the structure of the language to hearers
rather than to content (Halliday 2000; see above, §9.4.3.2 and §9.5.3).
11.6 Provoking imaginative responses
11.6.1 Introduction
The range of interpersonal functions discussed in the preceding sections is, at face
value, comprehensive; but there is another function at work, which has not been given
its own section because it does not occur independently in the everyday language to
which the book has in fact been restricted. In all the interpersonal functions, speakers
often seek an imaginative response from their hearers, as well as the primary and
immediate response; they have a secondary intention as well as the primary one. That
can be seen in many of the examples quoted, although the speakers might think of the
effect as “vividness” or “humour”, and so on.
To some writers, this is the “aesthetic” response, or the “poetic function”
(Jakobson 1960), or “language play” (Crystal 1988), or speaking with “style” in the
sense of SOED <13a> “elegance, refinement or excellence of … expression [or] form”.
In the conceptual scheme of this book, it is roughly equivalent to the figurative use
of meaning (Chapter 4, §4.7.4), being applicable to the meaning types studied in the
preceding sections, and the descriptive type whose realisation is to be studied in the
following chapter. Figurative use includes not only lexical “figures of speech” such as
metaphor and metonymy, but also syntactic forms such as balance and climax, and
certain uses of phonological forms such as marked rhythm and alliteration.
11.6.2 How imaginative responses are provoked
Introduction. The responses achieved by the imaginative use of English cannot be di-
vided into categories, but it is worth noting some broad classes, for the sake of making
clear the range of functions which English serves.
Aesthetic response. At one extreme is the aesthetic response. That can be roughly de-
fined as the enjoyment of the form for its own sake – language as a medium, in this
case. Even serious news magazines seek it, as in examples (26) and (27), where the
aesthetics was that of cleverness or incidental wit, not that of high art, although they
share with poetry the devices of rhythm, rhyme and assonance.
(26) “Secrets, lies and America’s spies”

(Leader headline, The Economist, June 15, 2013 p. 11)
(27) “Bond has schmoozed, boozed and bruised his way through women and
villains alike.” (The Economist, October 20th 2012 p. 72,
under the heading “Booze, bonks and bodies”)
In literature, the imaginative response also brings depth of semantic structure and
enhancement of the reader’s response to the whole text. Example (28) is from a novel;
the character, an upper-middle-class London woman, is going to a funeral in a little
rural village.
(28) “The London black suit she wore, though severe and matt, somehow failed
altogether to look like mourning.”
(Elizabeth Bowen, “The Heat of the Day”, p. 67)
In unmarked use, London would be a Classifier (Chapter 8, §8.2). Here it has been put
in Epithet position, making it descriptive and associative, evoking fashionableness
and city manners, for example.
Lüdtke et al. (2014) confirm psycholinguistically that it is features such as those
above (features of “style” and “form” to them) that rouse aesthetic response. Bohm
et al. (2013) found neural correlates of aesthetic response to reading, and found that
it occurs spontaneously – even when not sought for or required.
Humour. Although humour is commonly generated by the content, it is sometimes

purely linguistic, as in puns, and on other occasions it is sharpened by the form in
which it is realised. In example (29), from spontaneous conversation, speaker “JM”
achieves humour by the simple device of dividing the figure ‘I bought a computer for
$20’ into two utterances, with the ‘$20’ thereby taking speaker “FR” completely by
surprise.
(29) JM: “I bought a computer.”

FR: “Did you?”
JM: “For $20.”
FR: “What?” (Wellington Corpus, T_DPC068th.txt, 0975)
Example (30) comes from a group discussion of looking after people with dementia.
The speaker was sharing with a group the frustrations of looking after her mother,
who also needed hearing aids – referred to as “ears”, in the example.
(30) “As soon as she takes her ears out, she can’t hear.” (Personal observation)
The metonymy of ears provides wit, the statement being simultaneously impossible
and obviously true. It relaxed the group’s tension very successfully.
Entertainment. Entertainment often involves the responses discussed above, but is

given separately here to include the element of performance, which is in my opinion
a seriously underrated explanatory factor in linguistics. By “performance”, I mean the
physical actions, gestures, and facial expressions used by speakers to heighten the ef-
fect of what they say. Example (29) will have relied on a deadpan delivery of “I bought
a computer” to make it seem a complete piece of information, and delivery of “For
$20” after a pause, and with a grin. The physical articulatory actions needed to pro-
nounce ideophones, such as skedaddle and hoity-toity, is part of their effect.
The speaker of example (31) had listened to his partner’s scornful account, to an-
other person in the conversation, of how one of the city’s universities took all the best
of the foreign students; the other university was allegedly content to take the “crumbs”
left over; (31) was delivered to cap the partner’s account.
(31) “Out comes the dustpan and sweeps them all up.” (Personal observation)
The remark was a spoken with a strong out emphasising the syntactic inversion; with a
swooping intonation contour from out down to dustpan, and up again through sweep
to up; the rhythm was highlighted, speeding up and slowing again, as the pitch fell and
rose. The partial opening and closing of the lips in /sw/ was emphasised – in phonic
use of the sounds – as an iconic representation of sweeping up, and the stressed long
vowel in the word was lengthened further, making the word a miniature drama of its
own. The balanced clauses provide the substructure for the phonological effects, and
is aptly concise. The utterance was a mini-performance, entertaining, and spontane-

ous – and quite normal in everyday English.
In example (32), a farmer is responding to his wife’s suggestion that they buy the
“mountains” of cow manure she has seen, to use for compost. (It repeats and extends
an example given in §11.5.4.)
(32) “I visited the dairy and saw those mountains of manure and I just kind of
shook my head and went, ‘Really? Why? [pause]would we want to do that?’. ”
(New Zealand TV1, Country Calendar, “Chaos Springs”, 5th April 2014)
The speaker has dramatised the narrative, dividing it into little scenes, one per clause,
and acting it out with a shake of his head and a shrug; went, used where said would
suit, denotes action, telling the reader that performance is to follow. The lexis and
syntax help the effect. Mountains is hyperbole for “big heaps” (though it is used partly
because his wife had just used it), and it brings imagery into the meaning. Each action
is preceded by weak wording, to enhance it: “I just kind of shook my head” and “…
went, ‘Really?’….”. Really, and the exclamatory free-standing why have no descriptive
meaning, and act as holophrases, expressing the speaker’s amazement, briefly and
forcefully. “Why would we want to do that?”, being a rhetorical question and low in
information value, was spoken to the interviewer/camera, as a challenge to the hear-
ers, not simply as narrative. Finally, phonology adds to the drama: the quoted speech
was pitched fairly high, giving the speaker as dramatic personage a different “voice”
from the speaker as narrator. I have given considerable detail here to emphasise the
importance of dramatic performance in everyday English.
Sense of play. I trust that the element of playfulness has been clear in some of the
preceding examples, particularly in the use of the ideophones, but further examples
follow, to illustrate the fact that playfulness is common enough in everyday use of
English, and even in serious use.
Example (33) is from a newspaper report of a cricket match. It plays with alliter-
ation and hyperbolic imagery, which are congruent with poetry but not journalism,
and would certainly not be taken seriously by the newspaper’s readers. (Willow al-
ludes to the bats’ being made of willow.)
(33) “[Williamson and Taylor] posted a 121-run partnership for the third wicket
by cajoling the ball rather than indulging in willow warfare.”
(New Zealand Herald, January 20th 2014, B1)
Example (34), from the same report, plays with metaphor similarly. (“Baize” refers to
the grass playing surface.)
(34) “Other strokes wouldn’t have looked out of place on a snooker table, such was
their adhesion to the McClean Park baize.”
(New Zealand Herald, January 20th 2014, B1)
Example (35) is about the elements in so-called “empty” space other than hydrogen
and helium.
(35) “Though these other elements are a mere soupçon of the interstellar soup,
they do give it a real flavour.” (The Economist, March 16th 2013, p. 77)
As Crystal (1988) shows, play is not only very widespread as a function of English, but
has an important role in children’s development of skill with language, just as young
predators’ play develops their skill in hunting.
11.6.3 Discussion: Provoking imaginative responses
In the section on conveying emotion and attitude, we saw that figurative language is
important in rousing that response. In this section we have seen that it is also impor-
tant for gaining and holding attention, for humour, for drama and for playfulness, and
that it is a frequent and natural part of language in our daily situations. That shows, I
suggest, that it is a mistake to regard “making sense of the world” (Gibbs 2005) – cog-
nition – as the main function of such figurative language as metaphor, as much cog-
nitive linguistics does, particularly conceptual metaphor theory. According to Gibbs
(2005), for example, intimacy is closeness – ‘closeness’ helps us understand intima-
cy – and affection is warmth. The discussion above shows that such a view is wrong
empirically. It is also wrong necessarily: for the metaphor of interstellar space as a soup
(example (35)), we must first understand both space and soup; the metaphor follows
that understanding, so was not a means of achieving it; it thus cannot be a means of
conveying it, as is thought by Lakoff and Johnson (2003) and Gibbs (2005) among
many others. That is supported by Barsalou (1999, §3.4.1), and Barsalou and Wiemer-
Hastings (2005, pp. 133–134). The mistake, I suggest, is in taking as metaphors expres-
sions that have lost their figurative quality, and are by definition no longer metaphors.
Their “conceptual metaphors” are analogies, consisting of literal descriptive meaning.
11.6.4 Conclusion: Provoking imaginative responses
Achieving imaginative responses uses the same means of realisation as other func-
tions do, and it occurs in a wide range of genres – certainly not only in poetry or
other literature. However, it extends the nature of English and the constraints under
which it works beyond what is true for those other functions; and it alters the way the
elements of English are used.
In a famous phrase, using English is “doing things with words”; but those things
are usually taken to be external to language itself; here we have seen doing-things-
with-words as the thing done; using language can be an end, as well as a means. The
semiotic constraints are also extended: poetry especially can carry meaning indexi-
cally, iconically and symbolically all at once. Language is linear; that usually means
that it carries significance on a single line; but in this section we have seen it develop
several lines of meaning at once, in parallel, and operating on two levels at once – of
seriousness, and of humour or playfulness.
11.7 Discussion: Realisation of interpersonal meanings
Strata. As we will see more fully in the next chapter, on realisation of content, the
strata of English usually act successively, as it were, each realising the previous one:
meaning is embodied in words which are embodied in syntax and so on. Here we
have seen that emotion, for example, can be embodied independently in lexis, syntax
and phonology. The different strata may reinforce each other; or the significance of
one stratum may overrule that of another, as when phonology turns a syntactic state-
ment into a question; and the two may combine to realise a meaning that neither has
on its own.
Negation. This book does not set out to account for how particular semantic domains
are realised in English, but this section discusses two of them briefly, since they in-
volve the realisation of intentions.
Negation – negative polarity – has a use in the realisation of information, where
an informative proposition may be negated instead of affirmed. It has other uses in
English, however, one of which is illustrated in examples (36) and (37). They come
from opinions given to a newspaper reporter on the banning of synthetic cannabis.
(36) “I reckon normal weed is way better because you don’t get no side effects.”
(New Zealand Herald, 29th April 2014, A10)
(37) “I can’t go back to weed, it doesn’t do nothing no more. I’ll just give up, stick
to ciggies.” (Source as for (36))
The double negative in example (36) is different in function from the double negative
in “It is not impossible that he will come.” Along with the triple negative in (37), it ex-
presses an attitude – an attitude of disapproval or disappointment. Further, it creates
a field of meaning: the field can be intensified by adding more negative morphemes,
and, because it is a field not an item, those morphemes can be placed more or less an-
ywhere. I suggest that this gives a satisfactory resolution to the puzzles about double
negatives.
Polarity, which here denotes the intention to either affirm or deny, can be realised
in many ways. It can be realised phonologically by compressing the span of fall-rise
tone (tone 4), giving an effect of “gentle negation”; it is used, for example, in one ver-
sion of “He won’t be there” (Gussenhoven 2004, p. 88). It can be realised lexically in
words like opposite and negative, and grammatically of course, by the “operator”, not.
It can be realised morphologically, with negative prefixes such as un- and de-. How-
ever, that simple neatness evaporates when we realise that negative “not unhappy” is
positive “happy”; realise that “He was negatively buoyant enough to walk on the sea
floor” means that he was positively heavy enough; and that words like any are felt
to be “negative polarity items”. The complexities that appear in a thorough exami-
nation of antonymy (as in Cruse 2011, for example) illustrate the point further. We
conclude that positive and negative verbal groups (with and without not) may make
coherent morphosyntactic categories, but that positivity and negativity as semantic
domains do not.
Markedness. Through the various sections of this chapter, markedness has become
increasingly important. It was important in how we convey emotion, since the un-
marked forms will convey moderate emotion, but strong emotion calls on marked
forms, especially in phonology. Markedness is crucial in evoking imaginative re-
sponses, especially in poetry – we could say that poetry is total markedness – but also
in everyday prose. For the realisation of interpersonal functions, which we have seen
to be fundamental in language, markedness is essential. It is not an optional extra, or
even an exception to the rules: it is a rule – to express certain meanings, you “break”
certain “rules”.
11.8 Conclusion: Realisation of interpersonal functions
Summary. The chapter has discussed realisation of the following interpersonal func-
tions: the Expressive function; conveying emotion and attitude; establishing person-
al and social relations; guiding hearers’ use of meaning; and provoking imaginative
responses. The realisation was sometimes in words, but was more often syntactic
or phonological, and sometimes graphological; it was sometimes in two or more
strata at once.
Conclusions drawn. The main conclusion to be drawn about the realisation of inter-
personal meanings in English is that speakers have a wide choice in the means they
use. Speakers can use lexis, syntax, or phonology to carry most sorts of interpersonal
significance; they can choose among a variety of unmarked forms, and choose be-
tween unmarked and marked forms. In some circumstances, the choice can be made
by personal preference, but there are often constraints. The main one is that to rely on
phonology, and to use marked tone and tonicity, is to be informal, while using syntax
where you could use phonology is to be formal. There are less binding conventions
about the social groups and social situations to which the various strata and forms
are appropriate. Although marked forms often create informality, they often will fulfil
a secondary function in formal language, such as being entertaining or eloquent, or
flaunting your wit.
Putting that conclusion more formally, we can say that the relationship between
meanings and the forms that realise them is one-to-many, in each direction. That is,
for each interpersonal meaning there are generally several alternative expressions,
and generally each form can express any of several meanings. The main exception to
that is in lexis: brilliant, for example, can express a vague descriptive meaning, atti-
tudinal meaning, social meaning, and emotive meaning; but for attitudinal meaning
the relation is one-to-one – the attitude can only be favourable. Even there, however,
there is perhaps a one-to-two relation since the attitude can be reversed by ironic
phonology.
The nature of meaning has become uncertain in this chapter. In previous chap-
ters, we could take meaning as basically descriptive or conceptual, allowing emotion
and attitude to tag along with it in particular words, and could treat grammatical
“meaning” as essentially different – perhaps being “meaning” in a metaphorical sense.
Here, however, even meaning in the broad sense of significance has often seemed to
be better described by the terms “function” or “response” or “effect”, or to be better re-
garded as an action of the speaker’s, such as establishing (relationships), guiding (the
hearer), or provoking (responses). That is important for the book’s goal of describing
semantic structure in English.
Another conclusion is that these functions do not regularly use the progres-
sive stages assumed by writers who do discuss realisation, such as Lamb (1999) and
Halliday (2014) – stages in which words realise semantics, syntax then realises words,
and phonology realises syntax. Indeed, the main conclusion we can draw is that in re-
alisation of interpersonal meanings, the strata are alternatives, rather than a sequence
of stages. That needs some qualification, of course: interpersonal functions sometimes
do use statement of information as a means to the goal, and that information must be
realised through the strata (see Chapter 12); and the alternatives can occur together.
A further conclusion – a less obvious one – is that in this sphere phonology is
basic. It is always available, as syntax and lexis are not, to be used in the interper-
sonal functions of English; and for several functions it can be used alone, without
lexicogrammar. Interpersonal syntax such as the exclamatory construction is a late
development, historically; functions like attracting attention cannot be realised syn-
tactically. Also, when there is a syntactic realisation, it sometimes appears to derive
from phonology, as with end position marking focus. Occasionally, the only alterna-
tive to the phonological means is an obvious imitation of it, as with underlining as an
alternative in writing for marked stress. Since we know that speech is basic to writing,
the conclusion is hardly surprising.
Finally, the chapter has shown that the imaginative use of language permeates a
good deal of real-world English; it must be accounted for in description and analysis
of the language. It has presumably escaped much attention in linguistics because it
does not have its own forms, as well as because of the traditional emphasis on the
expression of thought, inherited from philosophical traditions.
Looking forward. The next chapter will complement this one, and complete the treat-
ment of realisation, by considering how the ideational function is realised.
Chapter 12
Realisation (2)
Ideational function
12.1 Introduction
This chapter continues the study of realisation, but moves on from the interpersonal
functions to ideational function, in which we use our understanding of the world
to convey information to others. The perspective of exposition also changes, being
now from the bottom upward. Thus, §12.2 deals with the realisation of meaning into
words, §12.3 with the realisation of words into groups, and so on. As with the previous
chapter, this chapter complements the earlier chapters in presenting a different kind
of semantic structure, in presenting the earlier structures from a different point of
view, which should be illuminating, though it entails a little repetition.
12.1.2 Introduction to ideational realisation
Interpersonal and ideational realisation. Ideational meaning complements the mean-

ing conveyed in the interpersonal functions in two ways. It is based on cognition,
whereas interpersonal meaning is based on affect, imagination, conation or social
attitudes and behaviour; it thus consists of descriptive meaning and excludes emotive,
attitudinal and social meaning. It also complements interpersonal meaning in that,
although sometimes an end in itself, it is generally a means for achieving a further end
(one of the interpersonal functions).
The two types of realisation are also complementary, broadly speaking. The reali-
sation of interpersonal functions differentiates the original broad intention into more
specialised intentions and functions, until relatively discrete items are formulated,
which can then be spoken. The realisation of the ideational function begins from a
well differentiated function, and combines senses which are always discrete into pro-
gressively larger units, which are then realised as a sequence of words. (Later, we will
need to modify that simplification.)
In language, realisation entails a crucial issue which is not apparent in the term,
“realisation”. If an athlete realises his ambition of winning an Olympic gold medal, the
win has the same nature in the reality as it does in the potential; but in the realisation
of ideational meaning, that is not always so, since the meaning is changed in some way
as it is realised. That is an important theme of this chapter.
Outline of the process. Although realisation is primarily the relationship between

meaning and form, it will be treated here as a process, for expository clarity, as in the
last chapter. Realisation, then, begins when the information to be conveyed is formu-
lated into cognitive structures, traditionally described as propositions or information,
and referred to in previous chapters as being in the realm of cognitive semantics. The
concepts are then provided with links to semantic functions; for example, the concept
WEALTH gains a link that will enable it to function as an Entity sense, as ‘wealth’, or a
different link for functioning as a property sense, as ‘wealthy’. That stage will be called
“semanticisation”. The next stage is “lexicalisation”, in which the sense is linked to a
word, such as wealthy or rich or prosperous. Successive stages of “syntacticisation”
follow (given sequential structure, as syntagmatic semantic structure or as syntactic
structure): senses are combined into semantic groups, which are linked to syntactic
group structure; the semantic groups are combined into figures, which are linked to
a syntactic clause structure. Phonological or graphological form is then assigned. The
last stage is articulation – physical realisation as speech or writing. The sections of the
chapter each deal with a stage in the process.
The process presented here is fairly close to the consensus model given by Bencini
(2013, §21.2.1). This model differs from that one in being oriented more to linguistics
and less to psychology, and in linking “relational meaning”, “structural processing”,
“function” and “constituent assembly” very closely as grammatical meaning and its
realisation.
Qualifications. As that outline shows, the account is idealised to a significant extent;

little attention is given, for example, to the realisation of meaning in incomplete sen-
tences. Other qualifications should be made also. Since the focus is linguistic, the
“processes” described are formulated in such a way as to make clear meaning-form
relationships, not mental processes as seen psycholinguistically; but they are intended
to be generally compatible with psycholinguistic understanding.
Ways in which they are not fully realistic include their not taking account of the
facts that processes are sometimes concurrent, and sometimes work in a different
order (Lamb, p. 1999), as when we first choose a general sentence structure, and then
choose words to fit it, and when we use a ready-realised chunk from a previous ut-
terance. Similarly, the range of options available, and the constraints on the choices
made are simplified.
As a description of the abstract relation between meaning and its realisation, the
account here cannot be based primarily on empirical evidence (from psycholinguis-
tics), although the use of psycholinguistic evidence might suggest it should be. The
argument is that the description of structure given must be correct, to explain the
observable phenomena of English.
Chapter 12. Realisation (2): Ideational function 343
Illustration of the process. The need to postulate all these processes, and their nature,
can be illustrated by the contrast between examples (1) and (2). Example (1) comes
from a newspaper table of information about a horse named “Wonder Anvil”, who
was to race in a trotting race that day. Example (2) is a paraphrase of it.
(1) “(36-2-2-3) $13,885 […] Mo(29-2-5) St(7-0-0)”

(New Zealand Herald, May 22nd 2015, On Form p. 7)
(2) “[Wonder Anvil has] in his career raced 36 times, had 2 firsts, 2 seconds and
3 thirds; he has won $13,885 in prize money; … has raced from a mobile start
29 times, gaining 2 firsts and 5 other places, and raced from a standing start
7 times, gaining no places.”
In (1), the whole message had been semanticised, in that it was formulated as meaning
which could be represented symbolically, by either linguistic or mathematical signs.
The information on the type of start was lexicalised to the extent of two items (mobile
and standing) and realised further by the linguistic signs, “Mo” and “St”. The other
information was not lexicalised, but was represented by mathematical signs such as
“36”, and general purpose non-linguistic signs such as “$”, or else omitted.
The message was then syntacticised into units which resemble linguistic groups
but have the force of utterances, e.g. “$13,885” = “He has won $13,885”. The gram-
matical meaning is realised by signs with a quite different significance from usual; in
“Mo(29-2-5)”, the brackets signify a subgroup, and juxtaposition without spaces, as in
“Mo(29-2-5), not “Mo (29-2-5)” signifies post modification. A single space signifies
breaks between utterances, not breaks between words; hyphens connect any minor
details; the overall sequence signifies drop in salience – the reverse of linguistic Rhe-
matic structure.
The example illustrates the nature of the realisation, with its stages of semanti-
cisation, lexicalisation and syntacticisation. It also emphasises that there is nothing
inevitable about using words – they are only one form of realisation – and that the
signs used in realising meaning and intention have a fairly arbitrary and certainly
variable significance.
12.2 From intentions to words
12.2.1 Introduction
The first stage of bottom-up realisation, formulating word senses from the commu-
nicative intentions, is presented here as having three stages: conceptualisation, seman-
ticisation and lexicalisation. Their nature will be explained through analysis of one
passage in detail, with other passages discussed briefly to make points not covered.
12.2.2 Unmarked realisation into words
The processes of formulating word senses for realising information will be illustrated
from example (3). Like several later ones, the passage comes from a television docu-
mentary on farming life. Television was chosen because it would provide a record of
both the text and how it was spoken, and also a record of the gesture, facial expression
and so on with which it was delivered. The particular programme was chosen at ran-
dom. The speaker is Jenny; Steve is her husband; they run a farm, and a business that
makes and sells compost. Double slashes mark the end of intonation groups; dashes
mark pauses. (The explanation of ideational realisation here complements the expla-
nation of interpersonal realisation of parts of this passage in Chapter 11, §11.5.4.4
and §11.6.2.)
(3) “In Utah where we were living there were – very large dairy farmers who kept
their animals indoors the whole time // so they had mountains literally and
physically mountains of manure that had been stacked up for – decades //
and – um – knowing that cow manure’s one of the best manures for gardening
// I decided that – um – perhaps it would be good thing for us to get into the
compost-making business // [laugh] and so I came back all excited to Steve
and he didn’t quite know what to think //”
(From Country Calendar, New Zealand TV1, April 4th 2014)
12.2.2.2 Intentions, and conceptualisation

The speaker’s broad intention to reply to the interviewer differentiated into a main
intention of giving information, and a subordinate intention – presumably less delib-
erate – of expressing feeling. The information was to come from the speaker’s cogni-
tive network of concepts representing the state of affairs of farming in Utah and her
life there, which must have included such concepts as: ALWAYS, DAIRY CATTLE, FARMS,
FARMERS, INDOORS, JENNY AND STEVE, KEEP, LOCATION, OF-GREAT-SIZE, PAST TIME,
POSSESSION, and UTAH. That cognitive state of affairs is suggested sketchily in Dia-
gram 1; it is intended to allow for potential realisations quite different from those in
the passage, and to emphasise that the cognitive relations are quite distinct from the
linguistic relations which use them.
Little theoretical importance is attached to that selection and formulation of con-
cepts; for example, the choice between one-word expression and phrasal expression
of them has no implication about how complex these concepts are. However, it is in-
tended to embody certain points made above. For example, one abstract concept such
as OF-GREAT-SIZE underlies synonymous senses such as ‘large’ and ‘big’. There are no
verbal concepts distinct from nominal concepts; POSSESSION can be realised as ‘to pos-
sess’ or ‘possession’. Similarly, the conceptual stratum does not have the linkage pro-
vided by Predicators, prepositions and so on; that is created during the realisation; the
linkage between Jenny and Utah (“where [we] were living”) represents the cognitive
JENNY AND STEVE
INDOORS PAST TIME
ALWAYS LOCATION UTAH
POSSESSION FARMS
DAIRY CATTLE FARMERS
KEEP OF GREAT SIZE
Diagram 1. Cognitive state of affairs underlying example (3)
network links between JENNY, LOCATION, PAST TIME, and UTAH. (It is assumed here
that in the network itself there is no distinction between node and link; it is structur-
ing a thought that creates the distinction.)
12.2.2.3 Semanticisation
Semanticisation is the production of meaning elements from concepts: ‘dark’, ‘pur-
plish’ and ‘red’ from DARK, PURPLE and RED. The elements combine into senses, such
as ‘maroon’ and ‘plum-coloured’ – or occasionally constitute a whole sense, as with
‘red’. (Givón (2009, §4.2.1) calls that process “co-lexicalisation”, which denotes a useful
concept, but I believe the process is sublexical.) These are the elements that are possi-
ble, expected or necessary in the meaning of words, as discussed in previous chapters.
The concept, OF-GREAT-SIZE, is represented in the sense element ‘large’ (which can be
lexicalised in the next stage as large or big, or as huge if intensified.
As noted previously, the senses elements formed here correspond closely to a
cognitive class. Concrete sense elements are close to perception; event class items, for
example, denote what are perceived as happenings; they entail duration (the basis for
aspectual features), but are too close to perception to have tense or aspect; similarly,
entities realise what are perceived as things. Thus these elements are as yet semantical-
ly indeterminate, and do not (yet) belong one of the semantic classes.
These sense elements have links to each other and to linguistic functions, as well
as to their bases in cognition. They are not entrenched, being available for much con-
ventional linking, and to novel linking; they are the basis for the bonds which form
syntagmatic structure as senses are composed into groups and figures, in the syntac-
ticisation stages. The links are of varying strengths, and many senses have links to
various structural uses. ‘Horse’ has links to ‘cow’, ‘mammal’ and so on, and to Entity
modifier use (“a horse paddock”), to Entity head use (“a horse”), and Event head use
(“He horsed around”). These changes constitute crossing the borderline from cogni-
tive semantics into linguistic semantics.
12.2.2.4 Lexicalisation
Lexicalisation is the step that follows semanticisation, as certain potential links are
selected and activated. In it, the speaker selects previously used patterns in the sense
elements which are associated with a particular word form; occasionally, a speaker se-
lects elements and groups them in a new pattern which he associates with an existing
or new form; either way, the result is a sense, and its combination with a form makes
up a “lexical item”. For example, the concept, CAUSATION, is often semanticised as
‘cause’ and often as ‘result’; the meaning ‘cause’ can be lexicalised in cause, because, or
the reason for, the meaning ‘result’ can be lexicalised in result, consequence, or effect.
‘Placed’ + ‘saddle’ + ‘on’ can be lexicalised as put (the) saddle on or as saddled; that is,
the realisation can be “unitary” or “articulated” (Mel’cuk 2001, Chapter II, §7). Lexi-
calisation is this not a simple matter of one-for-one coding of sense to word.
Examples in the passage include the following. Living does not lexicalise the con-
cept LIFE/LIVE; there is no contrast with DEAD or INANIMATE. Rather, it is part of a
lexical idiom, “to live in (a place)”, which is used partly to denote a person’s place of
residence, and partly to provide the clause Predicator (allowing the place of residence
to come at the end of the clause as focus); it lexicalises LOCATION. ‘Utah’ identifies a
place, but carries no descriptive value, no abstract concept; it does not need semanti-
cisation; its lexicalisation bypasses that stage, linking cognitive knowledge to the word
form. The isomorphic principle of realising one concept by one word is breached in
the use of two words (“very large”) for the concept OF-GREAT-SIZE. Conversely, in-
doors realises a complex of concepts such as LOCATION INSIDE BARNS. The two uses of
manure are different lexicalisations of the same semantic item, ‘manure’; in “moun-
tains of manure”, it is lexicalised as a mass noun; in “one of the best manures”, it is
lexicalised as a count noun.
I believe that social, affective and attitudinal meanings are combined with de-
scriptive meaning at this stage. (That is because there do not seem to be any distinct
senses which consist only of those meanings, singly or together.) For example – in
formulaic expression – ‘cow-dung’ + ‘formal/technical’ = manure; but ‘cow-dung’ +
‘substandard/vulgar’ = cow-shit. This is the outcome of a convergence of the main in-
tention with the inevitable secondary intention of setting an appropriate social level.
It is also the point where sense relations of synonymy, hyponymy and so on impinge
on realisation.
The meanings’ dimensions become determinate as they are associated with par-
ticular word forms. For instance, ‘farmer’ has the potential, and the requirement, for
individuation; by the grammar of English, it must be individuated as a countable
sense; and by the information to the conveyed, it will be farmers; but it is as yet neither
a generic plural nor a specific plural.
At the same time, the potential links, i.e. bonds, to particular structural functions
become actual; for example, the entity sense ‘horse’ becomes the countable Entity
sense, to be realised as a horse. (It can, however, be realised as the Event sense to
horse (around): see §11.2.3 below.) Similarly, events become Events, and properties
become Properties. The senses thus become members of a semantic class, beginning
the process of categorisation and re-categorisation which is one of the characteristics

of ideational realisation. (This narrative, suggesting that realisation follows a strict se-
quence, is misleading here: the realisation of particular structural bonds is usually not
a consequence but a motivation; our speaker realised STACK as an Event (‘to stack’)
not an Entity (‘a stack’) to keep its salience low and put ‘for decades’ in strong focus
(“had been stacked for decades”, not “had been in stacks…”).
We have seen that sometimes concepts are combined (into a single word), or
omitted (left implicit); but sometimes concepts are added. For example, when JENNY
AND STEVE is realised as we, the concept of THE SPEAKER is added, in the grammati-
cal meaning, ‘first person’. It functions interpersonally, establishing personal relations
with the hearers (see Chapter 3, §3.3.5), in accordance with one of the speaker’s sec-
ondary intentions; we note it here because it illustrates the realisation of concepts that
are not part of the message.
12.2.2.5 Discussion: Unmarked realisation into words

In this passage, what has come out of the processes so far is not exactly what went
in. While it does realise some at least of the underlying knowledge, it has undergone
quite extensive and important changes in the processes of being semanticised and
lexicalised. First, some concepts that are lexicalised have been changed in the process,
as discussed in the preceding notes. Second, some essential concepts – POSSESSION,
LOCATION and PAST TIME – have no obvious place in the lexis; and third, some lexical
items – “there were …” – have no source concepts. (The last two points will be dis-
cussed in the next section, on syntax.)
The extent of what is semanticised is also worth noting. The speaker’s use of busi-
ness illustrates an important issue. It seems likely that in using the word she was think-
ing of no more than making compost on a large scale – without the minor elements,
‘for profit’, ‘customers’, ‘making deliveries’ and so on, which “the compost making
business” evokes in this context. She seems to have chosen the first reasonably apt
word that came to mind: the word which memory throws up for lexicalisation is not
always exact. Similarly, some meanings in the passage are quite vague – “a good thing”,
“indoors” and where and how the couple were living, for example. The speaker was
evidently following a Gricean maxim of minimising the content of the utterance. In
contrast, the choice of animals, over the more obvious cows, is noteworthy for what it
includes – not just cows, but the necessarily present heifers, bulls and calves – quite
exact, although general.
Finally, we should note that the discussion so far has dealt exclusively with con-
tent meaning and with content words. Grammatical words and morphemes have not
appeared yet, being introduced in the syntacticisation stage, building the content
words into a syntactic structure.
12.2.3 Marked realisation into words
Nominalisation and related phenomena. In standard, unmarked realisation, a concept

denoting something that is a thing cognitively becomes a semantic Entity, and a cog-
nitive happening becomes a semantic Event. In the familiar phenomenon of nomi-
nalisation, the realisation is incongruent: a cognitive happening becomes a semantic
Entity. A less obvious form of marked realisation is illustrated in the passage with
“very large dairy farmers”. The speaker’s informative intention is to ascribe to the
farmers the semantic Property of being people who run dairy cattle; but the semantic
Property is realised as the Entity ‘dairy’; the Classifier construction (Chapter 8, §8.2.2)
is used to signal the reader that a Property must be derived from it.
The conceptual content of inflections exemplifies another type of marked real-
isation. PLURALITY and FARMER are distinct concepts, which are readily lexicalised
distinctly in many or several, and in farmer; but in “dairy farmers” two concepts are
realised together, in a single word. That applies also to past tense inflections.
Use of phonic realisation. A further marked lexicalisation occurs in phonic realisation.

In onomatopoeia, as in a sentence such as “The cat meowed”, the lexicalising is done
partly by imitative sound, but partly by the standard method of conventionalised
morphemes, since the word now has an abstract dictionary meaning, “to make the
sound of a cat”. In many uses at least – as in examples (4) and (5) – expressions such as
crash and bang use the phonetic features of their consonants to represent features of
the sound. Stops such as /k/ and /b/ suggest a sudden onset; long consonants such as
/sh/, /m/ and /ng/suggest long and fading sounds; voiced consonants suggest resonant
sounds.
(4) “She crashed down the four steps.” (BNC)

(5) “So I banged the window and said get off the fence!” (BNC)
These words denote sounds evocatively by their own sound; so they are not arbitrary
signs, as discussed by Saussure (1915); nor are they symbols in the tradition of Peirce,
nor even indexes, since they do not gain their significance from co-occurrence. They
are sometimes partially iconic, through metaphor, as when /b/ represents abstract
heaviness force, rather than a sound; but they lack iconicity’s characteristic of replicat-
ing the pattern of the thing signified. It is almost true to say that, when used as strict
onomatopoeia, they are the thing signified: the /wh/ sound of wham is the /wh/ sound
of the thing – although strictly, of course, the linguistic sounds are not identical with
the real sounds. In a further breach with traditional semiotics, phonemes here are
meaningful in themselves; the smallest morphosyntactic unit is not the morpheme.
(See Chapter 2, §2.5.6.)
In ideophones (Chapter 10, §10.3.3) such as “It whooshed past me,” the lexical
sign is more indexical than abstractly symbolic. Some ideophones, as in “They fled
pell-mell,” use sound symbolism, but are holophrastic and expressive, rather than be-
ing onomatopoeic realisation of ideational content.
Those uses of sound symbolism complement its use in interpersonal expression

(see Chapter 11, §11.2.4).
Two or more senses lexicalised in one word. Compounds provide for two senses to be
lexicalised in a single word. So does some neologism, as when a journalist lexicalised
‘remove from’ and ‘storage’ in a single word, reporting that his newspaper felt it had
to “de-mothball the words glorious and triumph”, for a royal tour. Noun incorporation
has a similar use. “Babysitting” condenses the senses of a whole figure, ‘people sit
with (look after) babies,’ into a compound. More recent instances include “to test-run”,
“hand-holding” and “queue-jumping”. (See Chapter 9, §9.4.2, for the function of such
uses in information structure.)
“Content words” that do not realise a concept. In “they had mountains, literally and
physically mountains of manure,” literally and physically are content words which
have been bleached of their content, functioning only to emphasise size, already de-
noted hyperbolically in mountains. In “Vidal went into cardiac arrest”, the happening
is lexicalised in arrest; went denotes no movement or other change that would con-
stitute an event, being used simply to make a Predicator for the clause. I believe that
the same is true of “In Utah, where we were living …”; live is usually stative, but the
speaker was not intending to convey the information that she was in the state of being
alive; the word is used as a convenient Predicator, not as a conceptual word. In all
these instances, words that are apparently realising information are in fact realising
an interpersonal function.
Concepts that are intended but not realised. As noted previously, the expression “very
large dairy farmers” does not specify what is ‘large’. It is clear from the reference to the
large quantity of manure that the speaker was thinking of the number of cows needed
to produce it, not of the size of the farm – or of the farmers’ physique. But the concepts
NUMBER and COWS have not been realised.
Conclusion. Realisation of concepts is not a smooth process that works flawlessly; nor
is it automatic or wholly independent: speakers can modify it for the sake of some
intention other than conveying information.
12.2.4 Conclusion: From intentions to words
Realisation of the intention to inform begins with processes that are rather remote
from making language real physically. It begins with conceptualisation, in which indi-
vidual concepts “precipitate” out of the seemingly diffuse network of the knowledge to
be conveyed. Semanticisation forms senses that include the concepts, or are intimate-
ly related to them. Lexicalisation links the senses to word forms.
There are important differences between the information to be conveyed and the
information contained in the words. We have noted several in the last section, on
marked realisation; for example, the conversion of event-related concepts into entity-
related words in nominalisation. There are two more fundamental differences. First,
it appears that conceptual meanings are always structured a little differently from
the underlying cognitive concepts. Second, and more fundamentally, the conceptual
words are content words, and will act as nodes in the meaning network set up by the
utterance: there is (as yet) nothing in the realisation that will convey the links.
It will be fairly obvious to readers that the account just given is idealised. For
example, we often begin to formulate what we say with certain words set in advance,
because we want to use a certain word the previous speaker has used, or because we
must use the technical words belonging to the subject matter. Later sections will sim-
plify the issues in a similar way.
12.3 Syntacticisation (1): From words to groups
12.3.1 Introduction
General. Senses must be not only formulated into words, but also put in order and giv-
en structure, with the order signifying the structure, in the process of syntacticisation.
This section deals with senses being built into semantic groups, which are realised as
syntactic groups. The next section deals similarly with figures and their realisation,
typically in clauses. The two phases of syntacticisation are partly interdependent, as
the phases of realisation into words are.
Syntacticisation has several elements. The senses being realised are changed in
form a little, since senses operating in a semantic or syntactic group are not exactly
the same as they are in the mental lexicon. Second, a semantic relation between them
is constructed. Finally, a syntactic structure must be built, to realise the semantic one.
It would be possible to narrate the semantic and syntactic processes separately, but the
two will be combined, since syntax is not a direct concern.
The analysis will deal only with descriptive meaning. As well as being not directly
relevant to this chapter on ideational meaning, the other meaning types are not fully
involved in syntactic structure. The social meaning given by slang or dialect words,
for example, is realised completely by the word alone. Again, affective and attitudinal
meanings may be syntacticised to the extent of modifying the Subject, as in “The lousy
fellow didn’t pay me back” (the feeling applies to the fellow); but any predication will
apply to the referent of the Subject, not to the emotive meaning: “…didn’t pay me
back” is predicated of the fellow, not of the feeling of resentment. Moreover, non-
descriptive meanings seem to lack the bond structure which provides for the syntactic
relationships which descriptive meanings enter into; see Section 12.3.3.4.
The semantic inadequacy of words. Works on linguistics sometimes give the impres-
sion that the meaning of an utterance is conveyed by the content words alone. If that
were so, then the meaning of example (6) would be adequately conveyed by exam-
ple (7), since the latter lists the content words in alphabetical order, in their base form.
(6) “In Utah, where we were living, there were very large dairy farmers who kept
their animals indoors the whole time.”
(From example (3), with the punctuation normalised)
(7) “Animal dairy farmer indoors keep large living time Utah.”
Linguistic works with an awareness of compositionality imply more is needed; for ex-
ample, “structural (syntactical) arrangements” are relevant, as well as “lexical mean-
ings” (Andreevsky, Rosenthal and Bourcier 1985, p. 380). Assuming that formulations
such as “structural… arrangements” denote grammatical meanings, let us add the
grammatical words and the grammatical inflections to the list in (7), producing ex-
ample (8).
(8) “Animals dairy farmers in indoors kept large live the their there time Utah
very we were were where who whole.”
In making the rather obvious point that such a listing makes nonsense of a perfectly
good English sentence, I am emphasising that content words on their own are very
ineffective – even the addition of grammatical words does little to help – and that
grouping and ordering are crucial. To that extent, meaning is structure.
Goal and coverage. This section and the next set out to illumine that semantic struc-
ture, by showing how semantic and syntactic structure are built as meanings, and how
their underlying intentions are realised.
This section will analyse the first passage studied in the last section, and as before
will discuss other passages more briefly. The body of the section presents unmarked
realisation; a note on marked forms is given in the discussion section.
12.3.2 Grouping the words
Speakers form the senses, and the words that realise them, into groups. That grouping
is part of the realisation of information, which is our primary concern. But the analy-
sis given in the preceding chapters, especially in Chapter 9, on information structure,
shows that grouping is in part the realisation of that interpersonal function, not only
of this ideational function.
That is illustrated in the sentence in the last few examples: in “there were … farm-
ers”, the underlined words both constitute groups which are empty of information,
being fillers to let ‘farmers’ come at the end of the figure, in focus. In example (3),
‘kept ‘animals’, ‘indoors’ and ‘the whole time’ are each presented as a single concept;
animals and being indoors are perceptual realities, but keeping is not. ‘Kept’ is a lin-
guistic concept, not an independent real-world reality, formulated and expressed in
an independent group to make a Predicator, and to let the animals and being indoors
come later in the Rhematic structure. Groups are formed in the top-down realisation
of speaker purpose, as well as in the bottom-up realisation of information. The two
processes combine here.
12.3.3 Structuring the group
The grouped words must now be made determinate. As explained in Chapter 8, being
determinate means being specified or defined enough for the grammatical function
being served, such as being Subject or Predicator. The process includes specifying
what is underspecified in the lexis, and making groups referential where necessary;
those steps in turn include individuating Entities to the necessary degree, and making
Events finite to the necessary degree. For example, events as conceived before they are
lexicalised are not specified for finiteness – are neither past nor present, for example;
‘live’ was lexicalised as semi-finite living; ‘be’ was lexicalised as finite were (in “there
were…”). This section outlines such semantic structuring of the group, which deter-
mines the syntactic structure.
12.3.3.2 Identifying referents

Straightforward expression of information about the world puts some real-world
thing or happening (a “referent”) as head of each group, as in the invented sentence,
“The farmers herded their cattle into barns.” In our passage, that principle is followed
for many referents, such as Utah and the farmers. There and were, however, are also
heads; but they are semantically empty, not lexicalising any information, and not de-
noting referents. The speaker used this marked structure because she wished to treat
the farmers as the Topic, but that Topic needed to be introduced; she accordingly used
the presentative construction, there + copula. Ideational function was overruled by
information-structure function. The concept CATTLE, in “dairy farmers”, could have
been structured as a head, in “They raised dairy cattle”; but the speaker deemed it to
be subordinate, and did not make it referential, as a Participant.
12.3.3.3 Specifying the modifiers’ functional relations

For the content and speaker intention to be realised adequately, the intended modi-
fiers’ functional relations must be specified. Senses not selected as heads are thereby
assigned the status of modifiers; they complete the semantic and syntactic specifica-
tion of the group. In addition to the status, speakers must make clear the functional
relation to the head. That consists of actualising the appropriate links: the semantic
link of modifier to head, and the syntactic link of dependent word to independent
word – the semantic and syntactic bonds studied in Chapter 8.
The modifiers in the passages modify their heads in significantly different ways.
In the nominal groups, their and the determine their heads; large and dairy add infor-
mation to theirs. In the verbal groups, were (in “were living”) acts on its semantic head
to specify the tense as past. (There is sometimes also a functional relation of modifier
to modifier; that has been discussed in Chapter 8, e.g. §8.2.4, and will not be discussed
here, to keep the account relatively simple.) Words like very, which are syntactically
submodifiers, are modifiers semantically; in “very large dairy farmers”, very modifies
large as head; but its function is to intensify the head’s content, not to add information
to it, or specify it.
The modifiers’ relation to each other is apparently not signalled. But when mod-
ifiers are coordinated, as in “literally and physically mountains”, the coordinate rela-
tion is realised by and, or (in a longer series of modifiers) by commas in writing and
the listing tone in speech (Halliday and Greaves 2008, p. 172). (For detail of coordi-
nation, see Chapter 7, §7.8, and Chapter 8, §8.2.9.5.) So the absence of those signs in-
dicates that successive modifiers are in their unmarked subordinate relation; nominal
premodifiers’ relations are realised by their zoning.
The functional structuring has different sorts of motivation. Content modifiers
such as large and dairy realise the informative intention. Were realises the subordinate
intention of providing a setting for the narrative. It is not obvious that the and their re-
alise anything; in fact, they realise the subordinate intention of being clear, necessarily
adopted by anyone who speaks. Put negatively, speakers must not confuse hearers by
failing to be clear, which requires such words. Put positively, speakers must use such
words to make the reference clear, and incidentally to signal that a nominal group is
beginning. (See Hawkins 1994 on early recognition of syntactic structures.)
Their illustrates the realisation of another intention – managing the salience
of information – a sub-process in syntacticisation. It denotes the farmers, and the
phrase could accordingly have been worded: “the farmers’ animals”; but farmers was
the third-to-last word uttered, so the speaker eliminated the content meaning, replac-
ing it with a pronoun to point back to that content word. That aids the information
packaging of the clause, reducing the salience of ‘farmers’ in its phrase to almost zero.
In Chapter 8, §8.2.9, we noted the process of reduction in nominals’ degree of individ-
uation, whereby “the farmers” as a whole nominal group can be reduced to the status
of determiner in “the farmers’ animals”, and could be further reduced to the status of
modifier in “small farmer co-operatives”. Here, the process is taken one stage further
than that, by being reduced to a semantically empty pronoun.
12.3.3.4 Adjusting semantic dimensions

Introduction. As the modifiers’ functional relations are being set, the senses’ semantic
dimensions are also being specified more fully, according to the conceptualisation
that is being realised, and to the speaker’s intention. That involves a change in the
senses’ membership of a semantic class; for example, the sublexical sense ‘farmer’,
consisting of the concept FARMER and potential links to semantic functions, now has
those links actualised and becomes a fully semantic Entity.
Boundedness dimension. Senses for countable entities are realised as count nouns e.g.
farmer, which specify that they are fully bounded, and their number is now specified
by morphological inflection e.g. farmers. If the speaker’s intention for information
packaging reduces the salience of a count noun, it is accordingly reduced morpholog-
ically and syntactically, as described in Chapter 8, §8.2. That reduction is illustrated
here in “cow manure”: the word denotes cows in general, but the plurality and all
other boundedness of the entity has disappeared; the sense is now abstract, and is
effectively a Property, namely ‘bovine’.
In the passage, several other Entities are apparently not specified for bounded-
ness, but for different reasons. For dairy, the reason is that it has an abstract sense,
so cannot lose individuation any further. Utah is a proper noun, so is inherently fully
individuated. For the pronouns we and who, it is the antecedent that would be speci-
fied. There, in this use, is a dummy Subject, does not have an Entity sense, and has no
content to be specified or reduced.
Event senses, as well as Entity senses, may need to have their dimensions adjusted
as they become fully linguistic. The Event ‘live’ is fully specified as bounded by the fi-
nite form it is given, “were living”; were specifies its tense, and -ing specifies its aspect.
The nature of boundedness in Event forms is shown clearly in two other words in the
passage: knowing and excited. In “Knowing that cow manure …”, ‘know’ is not bound-
ed by either starting point or end point, being stative; the -ing denotes extension in
time as nouns like day and year do, but does not denote passage of time. Excited has
an apparently verbal -ed suffix, but in this use (“I came back all excited “), the sense
is in the Property class, not the Event class; it has lost the temporal boundedness that
the Event ‘to excite’ can have – which is illustrated by its synonym, enthusiastic, which
denotes a Property.
Other inherent dimensions. On the specificity dimension, general senses are often
made specific through their modification, as here with ‘dairy farmers’ and ‘farm an-
imals’. On the vagueness dimension, the value of getting into the compost-making
business is left vague, in “a good thing”, not made precise as ‘profitable’ or ‘interesting’.
‘Large’ is left vague as to whether it denotes land area or cattle numbers. The point
of view dimension is affected also, as in “had been stacked”: the main narrative uses
the present (the time of the interview) as viewpoint; but the construction have + -en
(“had been stacked”) realises a different viewpoint, that of the time when the speaker
was living in Utah. (That is another place where one of the speaker’s secondary inten-
tions – guiding the hearer – influences the realisation.)
Contextual dimensions. Management of one of the contextual dimensions, salience,

has been illustrated several times already. The realisation of the other contextual di-
mension, expectedness, is interdependent with it. For example, since the salience of
‘dairy’ has been restricted by its status as modifier, the range of possible meanings it
evokes has been restricted also; compare this use of ‘dairy’ with its use (in a slightly
different sense) in “What to do if you love dairy but hate the after effects?” (COCA),
where it denotes many products and evokes many associations.
12.3.3.5 Achieving reference

Reference is achieved through a group’s being determinate; indeterminate groups
such as “big horse” and “living” will not refer. The nominal groups in our passage
are determinate for different reasons. The groups “we” and “Utah” are determinate
and thus referential through the semiotic process they use: Utah uses naming, and
we uses deixis; names and deictics are inherently referential. Entity groups that use
the descriptive semiotic process must in general be made determinate through some
modifying item, since descriptive heads are not in themselves determinate. In “the
whole time”, the determiner the does that. In “their animals”, the possessive does that,
through reference to the farmers. “Very large farmers” is taken as indefinite because
there is no determiner; the absence of determiner is a marked sign signifying indef-
initeness. Note that farmer in that form is indeterminate, but that making it plural
as farmers renders it sufficiently determinate for generic use. Nominal groups with
content heads, then, are made referential by a combination of content modifiers and
grammatical determiners. The determiners both create the group’s reference and
signal it.
Event groups achieve reference chiefly through grammatical modification, not
content modification. In “were living”, for example, ‘live’ becomes referential in be-
coming finite through were + -ing; the sense ‘live’ is not referential in phrases like
“a living wage” and “living people”, because it is not a head and is not finite. ‘Keep’
becomes referential through its inflection into kept.
Property senses are rarely used referentially; ‘large’ is not, in our passage. How-
ever, ‘indoor’, although often used non-referentially as a modifier as in “ an indoor
plant”, is used referentially in this passage: the farmers “kept | their animals | indoors |
the whole time.” Indoors is head of its phrase, constituting a Circumstance which is
the focus of the figure; as such, it directs the hearer’s mind out of the linguistic world
into the world of reality being described – it is referential. Potential senses which are
properties, in the sublexical semantic network, are very little changed as they are real-
ised in the Property class. The only important difference is that their intensity dimen-
sion must be specified, morphologically with -er/-est for comparative and superlative
degrees, lexically by submodification, or negatively – absence of modification signifies
positive degree.
Conclusion: reference. Reference is a part of the significance that is not compositional:

it has no elements that can be put together. The meaning is emergent, arising from the
whole structure as a gestalt. Reference gives the group unity, since it is the group as a
whole that refers.
12.3.3.6 Discussion: Structuring the group

This section has shown senses as members of fully linguistic semantic classes – de-
fined patterns of activation – emerging from being mere potential in the lower-level
semantic network. That has entailed specifying modifiers’ functional relations, and
adjusting semantic dimensions. The groups now have the potential to identify refer-
ents as the speaker intends, and will direct readers’ minds to them successfully. Spec-
ifying the functional relations further entails creating the bonds studied in Chapter 8,
by realising the links that are only potential in the sublexical senses.
We see in this passage the main function of morphology in English: it carries

grammatical meaning. At group level, it specifies lexical meaning more fully, and
specifies senses’ syntagmatic relations; its function is thus both syntactic and lexical.
12.3.3.7 Conclusion: Structuring the group

The structuring outlined in this section is the first step in giving the meaning of an
utterance a two-dimensional hierarchic structure, as a substitute for the multidimen-
sional structure of meaning in the semantic network.
Constructing a group, as a stage of realisation, entails a considerable amount of
specification, both within the meaning of individual words and in the meaning of
the group as a semantic unit. Steps two and three, achieving full specification and
achieving reference, make the group determinate, and therefore eligible for a role in
the figure and clause.
At this stage, the group that will eventually be realised as “very large dairy farm-
ers” is structured as follows: there is a head, farmers, with dairy dependent on it, and
with large also dependent on it; large has very dependent on it; neither word-forms
nor senses have order, which is assigned in the next stage.
12.3.4 Ordering the group
The word senses, with their dependency relations, could aptly be written in a hier-
archic pattern such as an organisation chart, which would represent those relations
directly; but the conventions of writing prohibit it; and such a pattern is of course
impossible in linear speech. The word forms must be arranged in order; and grouping
is the first phase.
A convention of English require that modifiers be placed next to their heads,
creating syntactic groups, although that is not required in languages such as classical
Latin. The convention creates significance, of course: being next to a head signifies
modification; it is a linguistic sign. Being before the head is also a convention, and
has significance, for information structure (see Chapter 10). (In French, which puts
most of its modifiers after the head but some before it, the significance is semantic.)
For the significance of the relative order of premodifiers, as in “large dairy farmers”,
see Chapter 8.
One other note on ordering may be made here. The phrase later in the sentence,
“the whole time”, illustrates a possible exception to ordering as a separate process,
since it may be that the phrase is “stored” in memory as an ordered unit.
12.3.5 Signalling the group structure
In introducing syntacticisation, §12.3.1 emphasised that an assembly of content words
will not constitute a meaningful utterance; the utterance needs also an indication of
the relations among those words. Yet as we read, we do not see words or anything else
that signify “This is the head” or “This modifies that.” Readers and hearers must follow
covert signs symbolising grammatical meanings. Previous chapters have listed and
illustrated the grammatical meanings that structure English; this section will illustrate
them in operation, showing their contribution to realisation of meaning.
12.3.5.2 Semiotic strategies

The contrasts between the three semiotic strategies of naming, pointing and describ-
ing make some contribution to the signalling. Proper names signal headship, since in
unmarked use they do not function as modifiers. Deictics such as that on its own and
nominative personal pronouns such as he also signal headship. Proper names in the
genitive, such as John’s, and subordinated deictics as in “that book” signal dependence
as determiners. The descriptive process, however, is unhelpful in signalling structure,
since it occurs in heads and modifiers, and even in determiners e.g. numerals and
adjectives such as certain and whole. In Old English and Middle English, Entity senses
were regularly used as heads of Object and Subject groups; Event senses were regular-
ly used as heads of Predicators, Property senses were regularly used as modifiers or as
Complements of the Predicator; that applied so regularly that the semantic class of the
descriptive senses could signal a sense’s place in the semantic structure (aided by mor-
phology). Moreover, exceptions to that rule were well signalled, as with the genitive
signalling an Entity’s use as modifier. In present-day English, however, speakers are so
free to vary the function of semantic classes, and morphological signalling is so much
reduced, that semantic class cannot be used reliably as an indicator of structural role.
The value of semantic class for signalling structure – and of the descriptive strategy as
a whole – is negligible.
12.3.5.3 Need for signals

Previous chapters have shown that grammatical meanings guide hearers in struc-
turing the content; so, to understand the realisation of structure, let us examine the
grammatical meanings in our passage. First, the general meaning I have called “super-
ordination” (§12.3.5.3), (“Treat this item as head of its unit”) sets up the status of head;
the general meaning, “subordination” (“Treat this item as a subordinate in its unit”),
sets the status of modifier. Second, specific meanings set the functional structure. The
meaning “Add this to the head” adds ‘large’ and ‘dairy’ to ‘farmers’, constructing the
full content structure of the group. The meaning “Determine the head …”, carried by
their and the, determines the reference of their heads; their directs the hearer to deter-
mine ‘animals’ by reference to the farmers as owners; and the makes the reference of
time definite (if we take the semi-idiomatic expression compositionally). Very directs
the hearer to intensify the meaning of large. (I am deferring to a later section the
grammatical meanings of in, where and so on, since they concern clause structure.)
That account of the meanings does not explain how hearers know which words to
apply superordination and subordination to, nor even that such procedures do apply
in a particular utterance. Nor does it explain how hearers know which of the specific
meanings (“Add …”, “Adjust …”, “Determine …”) apply to a modifier. In language, as
elsewhere, meaning is carried by signs. What signs carry these meanings?
In the traditional linguistic approach, it would be natural for us to think that
parts of speech are the signals. Being a noun would signal being the head of an Entity
group; being a participle would signal being head of an invented group. Conversely,
the signal of modifier in an Entity group would be an adjective, and so on. But in our
passage, the noun dairy is a modifier and indirectly signals a Property, and the parti-
ciple living is ambiguous, since it is very frequently used as an Entity group modifier.
Gardening, whole, stacked and back can all be used as various parts of speech, and are
used here in their less common ones. Thus, the parts of speech are are of little im-
mediate use. Fundamentally, they are of no use: in the traditional approach, to know
whether dairy is head or modifier, we determine its part of speech (inflection being
unreliable); but to determine its part of speech we must determine whether it is head
or modifier – the argument is circular. Moreover, identification of part of speech is
unreliable, since the parts of speech are indefinable “prototypes”.
We must look elsewhere to find how modifier/head status is signalled. The
matter is somewhat complex; the explanation is divided among the four following
subsections.
12.3.5.4 Signals that a word is a head of a group

A word’s status as head of the group is signalled in various ways. Phonology provides
what seems to be a reliable sign: phrasal stress is the sign that a sense is the semantic
head. (As noted previously, it may not be the syntactic head, as in “a cup of tea” and
“a big one”.)
There are several other indicators, none of them reliable on its own; they are thus
markers of the status, rather than symbols of it (Chapter 3, §3.4.3.3). Syntactically,
position at the end of a group is a marker, but postmodifiers are reasonably common.
Grammatical modifiers, such as articles and auxiliary verbs, seem to be reliable in
identifying the head (as the word they modify), but they are not always present in a
group. Morphology provides other indicators. First, nominative and accusative case
mark heads, while genitive case marks modifiers. Second, to a considerable extent,
it is heads which inflect; in “I decided that perhaps it would be a good thing for us
to get …”, the underlined words are both the inflecting items and the heads, with the
exception that good can inflect for comparison.
12.3.5.5 Signals that a word is subordinate

English grammar also provides signals that a word is subordinate, complementary to
signals of headship. Meaning type is a marker: with the exception of pronouns, words
with purely grammatical meaning are necessarily subordinate, e.g. possessives like
their, determiners like the and whole, and prepositions. Words with dominantly con-
tent meaning are subordinate if they are not final in the group. Compare the following
pairs, where words alternate being head and dependent as they alternate being final
and non-final: “a slate roof ” and “some roof slate”, “a pottery kiln” and “kiln pottery”,
“a swim school” and “a school swim”.
12.3.5.6 Signals that a word is co-ordinate

We have seen in previous chapters that, in interpreting co-ordinating structures, hear-
ers must relate the second of the co-ordinated items to the structural position of the
first. That is illustrated in two ways in our passage. In “literally and physically moun-
tains”, “literally” and “physically” are co-ordinate as premodifiers of “mountains”, oc-
cupying the same structural position. Therefore, in parsing the structural place of
“physically”, we must go back a step, by going back to its coordinate, “literally”. The
same procedure occurs in the larger structure “… had mountains, literally and physi-
cally mountains”, where the underlined phrase is in apposition to the preceding word
“mountains”. The signal in the first of those examples is the obviously co-ordinating
conjunction, and. But in the second example, there is no overt marking in the lexis or
syntax; readers must infer the structure from the adjacency of “literally and physical-
ly mountains” to the Complement group “mountains”, and the fact that they cannot
find any other syntactic function for the group, such as being a second Complement
(because of their co-reference). Those who heard the television broadcast containing
the passage were better off than those who read this transcription, however, as the
speaker used the listing tone (Halliday and Greaves 2008, p. 172), providing a phono-
logical signal where there is no lexicogrammatical one. (This is one instance where a
phonological signal is unique, with no equivalent syntactic or morphological signal.)
12.3.5.7 Signals of a subordinate item’s specific function

In Event groups, the function of a grammatical premodifier is signalled clearly by
whether it is a grammatical or a modal auxiliary. Content modifiers, whether preced-
ing or following the Event head, need no further signalling, because there are no sub-
types of content modification in Event groups. Examples are such utterances as “he
has often done that”, and “she seldom goes out in the evening.” (“Adverb of time/
manner/place” makes cognitive, not linguistic, distinctions.)
In Entity groups, the situation is much more complex, as was shown in Chapter 8,
§8.2. In principle, a premodifier’s function of reinforcing the head’s meaning, or de-
scribing the head, or identifying its referent, is realised by its zone, which is signalled
by its position. In practice, however, its position is often not signalled, there being
no other premodifiers. The hearer must use as clues such markers as the sense of the
word, any submodification, and the presence and sense of any other premodifier. For
example, in “My father left her for a woman in high high heels” (COCA), the first
high must be an Epithet, with the function of adding a gradable Property to the head,
because it can be submodified, as in “her impossibly high high heels” (COCA); the
second high must be a Classifier, with the function of identifying the type of heels,
since the alternative, a Descriptor, would not combine with high as an Epithet. In “She
was wearing high heels,” the precise function of high is not specified.
12.3.6 Discussion: Syntacticisation into groups
We have so far studied utterances where the semantic structure of the expression
matched the structure of the meaning as it was conceptualised – the two were isomor-
phic, or “congruent”. The passage illustrates some realisation that is not congruent. In
“They had mountains, literally and physically mountains of manure that had been…”,
literally and physically are used incongruently. In the intended (speaker) meaning, lit-
erally modifies the speaker’s use of language (‘literally speaking’), but in the expressed
(system) meaning it modifies ‘mountains’. ‘Physically’ belongs in a different domain
and cannot be coordinated congruently with ‘literally’. The realisation is incongruent,
but the meaning is clear, and the effect vivid.
“The compost-making business” is an instance of another type of incongruent
realisation, noun incorporation. We have noted incorporation previously in an in-
formation-structure function; here it has a semantic function. It serves to present a
complex concept as a unit. “The business of making compost” has an Event and an
Entity, two semantic units. Incorporating ‘compost’ – or as some would say, forming a
compound – signals that a single semantic unit is intended, an Event.
12.3.7 Conclusion: Syntacticisation into groups
This section has argued that as lexicalised senses are formed into groups that may be
combined into clauses and expressed in writing, they are grouped, structured within
the group, ordered sequentially, and provided with the necessary signals of the struc-
ture. That has been presented rather as if it were a sequence of phases, but the asser-
tion, put more precisely, is that the relation between intended meaning and realised
meaning has those elements of grouping, structure, sequence, and signalling.
We conclude that the ideational function requires not only realisation of concep-
tual content, but realisation of something that will show the reader or hearer how that
content is to be structured semantically. Since that guidance as to structure is gram-
matical meaning, grammatical meaning as well as content meaning must be realised.
That remedies the gap noted in §12.2.4, namely the absence of links between concepts.
In general, realisation simply converts into a new form what was present at the
previous stage; and syntax realises meaning indirectly, by realising the words that
realise the meaning. In this vital matter of the links between concepts, however, the
realisation is direct from the underlying meaning. Since linkage has been added, the
output of this stage of realisation is significantly different from the input provided by
the previous stage.
We conclude also that semantics and syntax have a complex relation. In content
semantics, the two are interdependent for their existence. They are “autonomous” (i.e.
distinct) in their structure, and approximate isomorphism in their hierarchic struc-
ture; the reasons for departing from isomorphism are various, including giving infor-
mation structure priority over semantics, the intention to gain special effect, and the
speaker’s clumsiness. However, when we include grammatical meaning in semantics,
and remember that it is realised primarily by syntax, we see that semantics and syntax
are not parallel or autonomous, that semantics is logically and temporally prior to
syntax, and that syntax is therefore dependent on it.
12.4 Syntacticisation (2): From groups to clauses
12.4.1 Introduction
As with the first stage of syntacticisation, the second stage – of realising groups into
clauses – involves the double process of realising semantic units into larger structures,
and of realising those structures into syntax. That is, semantic groups are realised into
figures, and the figures are realised into clauses.
A figure is often a structure consisting of items of knowledge, and so are the var-
ious information structures, commonly. It should be remembered here that figures
structure the knowledge for the sake of cognitive understanding, independent of the
hearer’s ability to absorb it, in the ideational function, which is our concern here.
Information structures, by contrast, shape the knowledge according to the hearer’s
ability to absorb and remember it, in the interpersonal function.
12.4.2 Structuring the units
The units to be dealt with here, described in previous chapters, are semantic groups,
functioning as Participants, Processes, or Circumstances. They are realised as syntac-
tic Subjects, Predicators, and so on (Chapter 2).
The structuring relationship being established is quite different from the rela-
tion used in groups. First, it is complementation, not dependence (§6.2). Second, it
runs right through the figure, whether it is a transitive relation (in material or men-
tal Processes) or a non-transitive relation (in relational Processes). In the default, a
Complement complements both Subject and Predicator, and Adjuncts complement
all preceding groups.
That relation is not present in the groups which are the raw material for the pro-
cesses here; it is added here – realisation is “creative” in this respect. The relation can
be seen as emerging “from below”, from the roles they acquire as they are structured
into a figure, or as being imposed “from above” by the speaker’s intentions. It both
relates and sequences the groups.
12.4.2.2 Processes
Introduction. Process, which is the semantic role of the Predicator, has three types that
we have been concerned with – and three others that we have passed by for the sake of
simplicity. (See Chapter 4 and Chapter 7.) The types have been presented as follows.
Material Processes deal with the world of outer experience; they have the features of
change and input of energy. Mental Processes deal with the world of inner experience;
they have the features of having no input of energy, but of involving change. Relational
Processes deal with the world of abstraction from both inner and outer worlds; they
have the features of having no input of energy, and involving no change.
Formulation of process type. The Process type may be formulated relatively congru-
ently, with material Processes realising physical happenings, for example, or relatively
incongruently. Once the speaker of our passage began her narrative, she formulated
information congruently, with a mental happening realised as a mental Process (“I de-
cided that …”), and a physical happening realised as a material Process (“I came back
…”). Before that, however, she wanted to set the scene, for which stative Processes are
appropriate; so although the physical activities of farming congruently form material
Processes, she formulated them with relational Processes – stative (kept, stacked up),
possessive (had), and existential (“There were large dairy farmers”). (Existential pro-
cess is one of the three types mentioned but not discussed in this book.) Those non-
material Processes minimise action and the passage of time, which helps establish the
first forty words or so as background to the narrative.
Semantic class. As a semantic class, Processes are generally developed from Events.
That entails determining the Process type, which in turn entails determining the roles
which the Participants will take. (A material Process takes Actor and Undergoer roles,
for example.) It also entails activating the potential links to those roles. Those changes
constitute the change in semantic class from Event to Process – a necessary step in
the realisation. The copula were, in “there were …”, is one of the exceptional Processes
that are not developed from Events; being purely grammatical, it is not based on any
semantic class.
12.4.2.3 Participants
The formulation of groups into figures entails assigning Participant roles to many
of the Entities. For example, at the beginning of the passage, the group ‘very large
dairy farmers’ is assigned the role of Participant by being linked to the Process ‘were’;
syntactically, it is realised as Complement to the Predicator, were: “There were very
large dairy farmers”. In a subordinate figure, the speaker assigned it the role of Actor
Participant, in “who kept their animals…”, realising it in a pronoun. ‘Utah’ was not to
be salient enough to become a discourse Participant, so was assigned to the Theme,

outside the Rhematic information structure: “In Utah, where we were living, …”. (It
could have been assigned the role of heading a Complement, in focus: “We were living
in Utah.”)
Establishing those roles entails changing the semantic class from Entity to Partic-
ipant, in the same way as Events become Processes; that realises (activates) the bonds
between the Participant and Process – the bonds studied in Chapter 7. More precisely,
it is not simply an Entity sense that becomes a Participant, but the whole determinate
Entity – the Entity as modified by pre- and post-modifiers and determiners. It is ‘very
large dairy farmers’ that is the Participant, not ‘farmers’. The same point applies, of
course, to Processes; ‘were living’ is the Process, not just ‘living’ or ‘live’.
Some Entities remain as Entities in the finished figure. ‘Dairy’, for example, hav-
ing been kept subordinate, remains an Entity, with low individuation, not becoming
a Participant.
I emphasise the intangible bonds as elements of the structure, although they have
no obvious reality, such as expression in an affix. In “I came back all excited to Steve,
and he …”, for example, the conjunction and provides an obvious link between the
clauses, and the preposition to provides an obvious link between ‘Steve’ and the rest
of its clause; but there is no such overt linkage between I and came, between came and
back, and so on. The links are real, because the groups are bound semantically into a
figure, and syntactically into a clause; without them, there is no structure, no asser-
tion – no utterance, in effect; it is those bonds that determine the order. But they are
invisible, being realised syntactically, not lexically.
12.4.2.4 Circumstances
Properties such as time and place are construed cognitively as circumstances of the
happening; they are realised congruently as semantic Circumstances, as with ‘in Utah’,
‘for decades’, ‘indoors’ and ‘the whole time’ in our passage. Note that most of those
have Entities, becoming Properties only when reconstrued into prepositional phrases.
As we saw in the Chapter 7, the syntactic relation of Circumstances to the rest of
the figure – realised as Adjunct to the rest of clause, usually – is variable, is sometimes
ambivalent, and sometimes indeterminate. Here, in “who kept their animals indoors
the whole time”, ‘the whole time’ has a semantic link to ‘indoors’; we can conceptualise
‘indoors the whole time’ as a single complex property. But in denoting duration, it
must have a link to something the animals did, which is presumably STAYED, which is
not lexicalised; by default, it is to be linked to the Event, ‘kept’ (which has the farm-
ers, not the animals, as its Subject). There is no signal to indicate which of those the
speaker intended, and nothing in the morphosyntax to require one of the readings as
the grammatical one: the relation is indeterminate. (There was, after all, no need for
the speaker to specify it; and, structured in this way, ‘the whole time’ forms the infor-
mation focus of the clause.)
The speaker must have had BARNS (cognitively a thing) in her conceptual under-
standing of the situation. However, she did not realised it in barns, as a Participant
Entity, but in a Property, ‘indoors’, as part of a Circumstance: the farmers “kept their
animals indoors.”
12.4.2.5 Discussion: Structuring clause units

Links. As noted incidentally in the previous sections, the semantic links between
groups are established in various ways. The semantic links between Participants and
Process are inherent in their very nature: to be an Actor Participant is to be an Entity
which carries out the Process; to be a material Process is to be an Event which has
an effect on an Undergoer Participant. Those relationships are added to the seman-
tic structure of Entities and Events as they become Participants and Processes. In
“I came back …”, it is that whole structure of Entity plus Actor Participant which is
realised in the syntactic Subject “I”; the Predicator, “came”, realises the structure of
Event plus material Process. That is why English syntax provides no explicit linkage
between Subjects, Predicators and Adjuncts. (Since there seems to be a grammatical
principle that syntax must provide morphosyntactic links between its units, as with
prepositions, conjunctions and relative pronouns, the absence of that explicit linkage
is anomalous, though we never notice the fact.)
The linkage of Circumstance to the rest of the figure is variable, just as the nature
of its relation to the other groups is variable. Often, prepositions create an explicit,
lexical link in the morphosyntax. Semantically, there may be a conceptual link which
the preposition realises, as with “in Utah”, which realises the conceptual meaning LO-
CATION, as well as the grammatical meaning of group relationship.
Often, the preposition makes no conceptual link, as with “of manure”. As is well
known, the genitive of is used for situations with quite varied real-world relationships,
such as possession, origin and being part of. Since hearers must decide for themselves
which relationship is relevant, we conclude that the of does not mean any of them:
the word has only the grammatical meaning, “Relate the preceding and following
nominal groups”, without specifying how to relate them. If we think that the prepo-
sition carries one of those specific meanings, we are reading back into the linguistic
semantics what is given only in the cognitive semantics. In our passage, the linking
for is similarly underspecified: compare “for decades”, “for gardening”, and “for us”.
Where there is no preposition to link a nominal group, as in “kept indoors the whole
time” (which could be rendered as “kept indoors for the whole time”), hearers must
infer it from the existence of a nominal group which is not realising a Participant.
(Indoors functions as a group which also has no preposition; the preposition has been
historically compounded into it – “(with)in+doors”.)
The linkage structure here is noteworthy in that the semantic link exists between
the groups, thereby differing from the preposition in syntax, which is a part of a group.
Semantics and syntax are usually correlated more tightly.
Argument structure. This section has covered the semantic area usually referred to
as “argument structure”. I believe that this is a better account of it. The structure de-
scribed in argument structure theory is cognitive rather than linguistic, since the
number and nature of the “arguments” are held to be inherent in the “verb”; but, to
take an instance, the thing eaten is inherent in eating only in our understanding of the
world – linguistically, we can certainly have eating without a thing eaten, as in “He ate
ravenously”. Linguistically, we must say that speakers choose which real-world things
to lexicalise as Participants – under the constraints of what will make sense in their
hearers’ understanding, of course. It is not true to say that (in modern English) speak-
ers’ choice of which things to lexicalise is controlled by what is inherent in the “verb”,
although that may have been true in the past. (See also §5.4.5 and §7.4.5.)
Intervention by interpersonal functions. As speakers structure the groups, interperson-

al functions “intervene”, and sometimes overrule the ideational decisions. In our pas-
sage, the ideational message, “Very large dairy farmers kept their animals indoors …”
became “There were very large dairy farmers who kept their animals indoors …” for
the information-structure purpose of making the sentence introduce a new Topic.
‘Utah’ was realised as Theme, breaking its cognitive link to ‘living’.
Conclusion. Assignment of new semantic class and development of Process rela-

tions both entail the emergence of “new” meaning, as discussed in the introduction
(§12.4.1).
12.4.3 Ordering clause units
The Participants, Processes and Circumstances must be put in order. The position
of Participants and Processes is determined as part of the assigning of roles: a group
assigned the Actor role in a material Process is thereby made Subject, and must come
first in unmarked statements. What is more, the rules for order are not part of auton-
omous syntax: they are semantic in that the order has meaning, as a signal. In “I came
back to Steve”, the initial position of “I” signals its being Subject, along with the case
form, “I” not “me”; the order in “Did I come …” would signal questioning.
The positioning of Adjuncts realising Circumstances is quite different. As we have
seen, the syntactic relation of Adjuncts to the rest of the clause is not well-defined,
and the semantic relation is variable; commonly, those relations are not affected by
the position of the Adjunct; and accordingly the position of Adjuncts has motivations
other than that of the semantic class, information structure being the dominant one.
(See Chapter 7, §7.5, for discussion.) Our passage provides an instructive example. In
“[They] kept their animals indoors the whole time” ‘indoors’ is predicated of ‘keeping
their animals’; and ‘the whole time’ is predicated of ‘keeping the animals indoors’;
that is, the speaker is asserting that they kept the animals indoors the whole time,
not just that they kept the animals the whole time. The semantic structure is ‘[[[They
kept their animals] indoors] the whole time]’. (The structure of the syntax may be a
different matter.)
One other point may be noted here. In “who kept their animals …”, who lexical-
ises ‘farmers’ – because of the higher-level decision to make the clause dependent. It
is generally the information to be expressed which controls lexicalisation, but here, in

the replacement of farmers by who, morphosyntax controls it.
12.4.4 Signalling the structure
Introduction. This section deals with realising the grammatical meaning which guides
hearers to the relations among groups that constitute figures and clauses – the provi-
sion of signs which will signal their structure.
As explained in §12.4.2.5, on the links between groups, the structural relations
between Participants and Process are inherent – to be a Participant is to be involved in
a Process; and since Participants are realised as Subject or Complement, that syntactic
status signals the semantic structure. Consequently, little needs to be said here; the
section adds some extra points. (The signs of being a Subject or Complement are not
directly our concern, being part of syntax.)
Signalling Process type. In being a Predicator, an Event group signals that it is a Pro-
cess, but it does not thereby signal the Process type. As noted earlier (§7.2), Process
type is not fully grammaticised; the distinctions are not strict, morphosyntactically,
and the signs are indirect and not fully reliable. The principal signal is the conceptual
meaning of the head of the Predicator, as set out in §7.2 and summarised in §12.4.2.2
above. Interpretation of that signal often requires consideration of the Participants,
since Predicator words are often polysemous or vague, and since Events may be con-
strued as a Process type which is not their congruent realisation. In our passage, for
example, ‘live’, in “where we were living”, is a state not an Event (SOED’s <1>), which
is marked (not very reliably) by its having no Complement; “live on [food etc.]” would
signal a material Process (SOED’s <2>).
Signalling Participant type. Participant type is signalled partially by Process type, since
the Participants for a relational Process, for example, must be Carrier and Attribute, as
the two terms of the relation. The rest of the signal is the syntactic role: the Subject will
be the Carrier, and the Complement will be the Attribute, for example. The semantic
class of the Entity is an additional marker, since abstract Entities are usually terms of
a relation, not Actors or Undergoers, for instance. The sign here is thus a composite
one, composed of the Process type and the syntactic role; its meaning is “Relate this
Entity to the other items in the figure in the Participant role of Attribute.”
Signalling the function of the other groups. The function of a group as a Circumstance
is marked by its being syntactically an Adjunct. However, status as an Adjunct is not
wholly clear-cut in traditional thinking, as the debates on “objects”, “complements”
and “adjuncts” show. Part of the problem is that the syntactic form of the group is not
criterial, since prepositional phrases, Entity groups and Property groups can all be
either Complement or Adjunct; so that prepositions are not reliable indicators of Cir-
cumstance. Another part of the problem, I believe, is confusion about the relationship
between syntax and semantics. The matter is more straightforward if we take syntax
as realising semantics; here, that implies making the decision depend on the nature of
the Process. For example, material Processes are fully transitive and take Undergoer
Participants, realised as Complements; relational Processes are not transitive seman-
tically, the syntactic Complement being a term of the relation, not an Undergoer. Thus
in “[They] kept their animals indoors the whole time,” ‘indoors’ can be seen as a Com-
plement or an Adjunct in traditional thinking; however, if we accept that the transitive
form of the clause does not indicate semantic transitivity, but is a rather incongruent
realisation of a stative meaning, we can see that indoors realises a Circumstance.
12.4.5 Conclusion: Syntacticisation into clauses
As with groups, syntacticisation into clauses entails structuring, ordering, and sig-
nalling the structure. The structuring relation, however, is complementation, not de-
pendence. The signs which realise grammatical meanings are similar, but rely more
on syntactic status and less on order.
As in the first stage of syntacticisation, meaning has been added to the meaning
which was the input – meaning as links between concept nodes, and realised as gram-
matical meaning. Those conceptual links come from two stages back – the intended
information – not just one stage back, as in the previous level of syntacticisation.
As noted in the introduction to the chapter, the section has not covered realisa-
tion of the figures into nonfinite clauses, or into groups, and so on.
12.4.6 Grouping figures in a clause complex
As stated in an earlier section (Chapter 7, §7.7), discussion of clause complexes ex-
cludes both “nominal clauses” and defining “relative clauses”, since they are rankshift-
ed constituents of a clause, not elements of a clause complex.
12.4.6.2 Structuring a clause complex

Semantic co-ordination typically realises logical relations such as equivalence; sub-
ordination typically realises logical relations of dependence, such as conditionality or
causation. (I say “typically”, because, as we saw in the section on interpersonal real-
isation, clause co-ordination and subordination may be the speaker’s signal of equal
or inferior information importance, regardless of the narrative or logical relation.)
Clause complexes realise figure complexes directly, both as to structure, and as to the
relationships of co-ordination and subordination. Those relations will be familiar to
readers.
That will help the unempoyment rate

which is currently 9.6 %.
Diagram 2. Semantic overlap created by a relative pronoun (which)
12.4.6.3 Signalling clause-complex structure

As at the lower levels of morphosyntax, the semantic structure is signalled by the
syntactic structure. That in turn is signalled straightforwardly.
One main type of structural signal is the conjunction, operating as prepositions
do for groups, realising both a grammatical meaning (such as “Relate these two fig-
ures/clauses as …”), and a conceptual meaning which specifies the relation, such as
‘cause’ and ‘result’. The other type of sign is the relative pronoun. Semantically, a rela-
tive pronoun creates an overlap between the figures, since it denotes the same Entity
as the nominal it depends on. That can be illustrated in Diagram 2: rate and which
refer to the same thing.
12.4.6.4 Conclusion: Clause complexes

Clause complexes are markedly different in their realisation from words, groups and
clauses. The realisation is much simpler, without change in internal structure as Enti-
ties become Participants, for example. In being signalled by conjunctions and relative
pronouns, the structure is realised more obviously and in a more intuitive way: those
excellent devices realise simultaneously both the guiding instruction and the concep-
tual relation we need, and occur in just the right, logical places. That fits the nature of
their use: they are controlled relatively consciously, and realise relatively intellectual
elements of meaning.
12.5 Physical realisation
12.5.1 Introduction
This section deals with the last stage of realisation, as the mental forms of syntax and
semantics are given physical reality in speech or writing. It concerns the ideational
function, complementing parallel sections in Chapter 11 on the phonological and pho-
netic realisation of interpersonal functions. Separate subsections are devoted to reali-
sation into sound – first phonological, then the phonetic – and realisation in writing.
Lexicalisation was the conversion of the concept to be expressed into a radically
new form, that of linguistic units. Syntacticisation converted the words into another
radically new form, that of linguistic structure. In this stage, the units and their struc-
ture are converted into more new forms: into sound waves, by “phonologisation” and
“phoneticisation”, perhaps, or into ink marks on paper, and so on.
12.5.2 Realisation in sound
12.5.2.1 Phonological realisation

Introduction. In this section, we consider the conversion of the lexicogrammatically
realised meaning into patterns which will be realised as sound waves. It makes a single
stage to the extent that there is a single output, but to some extent there is a double
process, as follows. Words, formulated in a much earlier stage, are realised as pho-
nemes; the overall pattern (syntax, approximately) is realised in intonation.
There is little to be said here. The expression value of phonology is almost en-
tirely for the interpersonal functions. For the ideational function, phonology simply
embodies the meaning, without adding to it or realising elements that were missed at
earlier stages.
We begin with realisation into figures, going on to the figure complexes which
organise them into larger structures. (The difference between the two is not great
enough to justify the term “realisation”.)
Figures. In formal speech, figures are realised phonologically by intonation groups,

and signalled by the tonality contour, the single tonic stress, and the tone at the end;
the three phonological systems of tonality, tonicity and tone combine and reinforce
each other, as if to ensure that this crucial grammatical signal will not be missed.
(In informal speech, the intonation groups serve information structure, not idea-
tion.) The boundary between figures (again in formal speech) is frequently realised
by pause, in the rhythm system. Pause, however, is an unreliable marker, because it is
often omitted, sometimes has a different significance (as with dramatic pauses), and
sometimes has no significance at all.
All of the significance, however, is fairly well signalled by the syntax. The figure
structure is signified by the syntactic structure, and the boundaries are signalled by
the completeness of one structure and the beginning of another, before they are sig-
nalled (somewhat redundantly) by the phonology. However, the phonology allows the
ideational structure to be made clear when the syntax is defective or absent, as it often
is in informal utterances such as single-group or single-word answers to questions.
One specific ideational function served by the tone system is that of the list-
ing tone (Halliday and Greaves 2008), which is a slight rise, repeated at the end of
successive groups. It signifies that the groups form a list structure, not a structural
complementation.
Group structure has little realisation directly into phonology. It is limited to the
fact, noted in previous sections, that the cognitively most important word is marked
by the phrasal stress.
Figure complexes. Section 4.6.3 set out morphosyntactic realisations of figure-level

grammatical meanings – conjunctions and relative pronouns. They are aided by pho-
nological patterns. The grammatical meaning that structures a sequence of figures
as main and subordinate figures and clauses is realised by tone sequence, which is
the pattern of rising tones and falling tones (Halliday 2014, §7.6). The instruction to
interpret a figure as having the same status is realised by tone concord, in which an
intonation group has the same tone as the preceding one (Halliday 2014, §7.6). Those
signals may be reinforced by a change in rhythm, and occasionally by a pause. Those
points apply also to paratones (Chapter 11, §11.5.4.5): there is very little difference
between the realisation of figure complexes and that of paratones.
Those phonological features can serve interpersonal functions, as shown in
Chapter 11; but speech that has such complex structure is almost always formal and
ideational. The features then serve the ideational function of marking off conceptual
units such as propositions.
The boundary between figure complexes has no specific sign; there are various
markers, of which the clearest is the tone on the focal stress – certain tones signal
finality, thereby implying that what follows will be a new figure. Pauses do not mark
figure boundaries reliably, there commonly being no pause between figures. For all
these points, see Halliday (2014, §7) and Halliday and Greaves (2008); for pause, see
also Cruttenden (1997, p. 36).
Conclusion: phonological realisation of the ideational function. The morphosyntactic

form of messages is realised into a phonological structure. That stage of realisation is
very different from that of lexical and syntactic stages. At those lower levels, meaning
is changed: cognitive organisation is partially replaced by linguistic organisation at
the lexical level; definiteness and reference are added to the word senses at group level,
for example; Participant-Process meaning and transitivity are added at figure level.
Realisation into phonology, however, does not change anything essential. In formal
speech, it is simply a transmutation of form; in informal speech, it is very much more
important.
That is presumably a result of the way that language has evolved, which I take to
be indicated in language development in children. Before they learn to speak, chil-
dren learn a wide range of pitch contours, and learn some control of volume, speed
and so on. They also learn most of the phonetics of the language. Accordingly, they
use phonology for almost all linguistic functions while they are still learning the vo-
cabulary and grammar; it functions fairly independently, and lexicogrammar is at first
quite subordinate.
12.5.2.2 Phonetic realisation

This section deals with “realisation” in its narrowest sense, that of phonological struc-
tures being given physical form – “reality” – in sound waves; phonetics is regarded
here as distinct from phonology. As noted in Chapter 11, phonetics, like phonology, is
important for the interpersonal functions. For the ideational function, it has even less
importance than phonology does.
The only way I am aware of in which phonetic realisation can serve the ideational
function is through the minor phenomenon of onomatopoeia. When we say, “It fell
with a loud bang,” or “The cat meowed,” the phonetic features of the words are index-
ical signs of the sounds denoted. They are referential, which is an ideational function.
12.5.3 Realisation in writing
Introduction. Graphology, as the realisation of language in writing, is a specialised

topic, and outside the focus of this book, so it will be given little space here.
Spelling. Spelling, of course, replaces phonetics as the physical realisation of words.

(The commonplace that writing is secondary to speech is manifest here, since the
realisation of senses is through letters that basically symbolise phonemes; it could use
pictographs or ideographs which have no connection to sound.)
Punctuation. To some extent, punctuation likewise replaces phonological features, re-

alising some semantic structures, as well as the interpersonal functions discussed in
Chapter 11. The full stop, for example, typically signals, “You’ve reached the end of the
figure or figure complex”; so that when the semantics is being used to convey infor-
mation, the full stop is being used for the ideational function. The syntactic structure
does that also; the two work in parallel.
Layout. Paragraph layout has usually been thought to have no correlate in phonolo-
gy, but recent research has shown that to be not so. Intonation group complexes are
realised by paratones (see Chapter 11, §11.5.4.5), which are realised in speech by par-
agraphs. In informational language, they realise a section of the content.
Other elements of layout certainly do go beyond any phonological equivalents,
and correspondingly seem to be non-linguistic. But section and chapter breaks, head-
ings, and the type size used in the headings all help signify semantic relations being
expressed in the ideational function, as do the distinction between italic and roman
lettering. (Even a specific font can have significance: a romantic novel would not be
printed in square-set, no-nonsense Times New Roman; but that is an interpersonal
matter.) Thus, having become conventionalised signs for meaning relations, they can
validly be seen as linguistic realisations.
Conclusion. While graphology is largely a mechanical transmutation of information

already fully realised in lexicogrammatical form into the form of ink on a page, it is
also a medium of its own, with its own semiotics.
12.6 Discussion: Realisation of ideational function
12.6.1 How we conceptualise realisation
“Realisation”. The use of the word “realisation” has more important implications. Al-
though it is an apt word in standing for a complex set of interacting processes – and
a complex relationship between the meaning and form – it is in one way potentially
misleading. It can easily suggest that the meaning is real in the expression, and that
there is a one-to-one relationship between elements of meaning and elements of form.
This section has shown, however, that that is not so. Many of the signs – including
words, word order and so on – represent grammatical meaning not content; so they
do not realise the conceptual meaning to be conveyed, so much as realise instruc-
tions for assembling it. The intended meaning and the resultant hearer meaning are
both seamless networks; but the form that mediates between the two consists of small
spurts of ink that are too discrete to represent a network, or it consists of a continuum
of sound that has no discrete units that could represent either nodes or links in a net-
work. Since hearers must construct the meaning from signs of many types, operating
on different levels, a text is at best a guide to navigating through the hearers’ own
cognitive network; and a more appropriate comparison is perhaps that of dots to be
joined, to make a drawing.
“Underspecified” realisation. Meaning is often said to be “underspecified”, and I have

used the term myself. But it is tendentious if it implies that linguistic expression can
be fully specified, since the boundaries of conceptual senses used descriptively are
not so well defined that we could determine whether the sense has been fully speci-
fied, and the term does not apply to nondescriptive meanings such as emotion. It is
particularly inappropriate when its use implies that expression should be specific, in
principle.
12.6.2 Incongruent realisation: Grammatical metaphor
Incongruent realisation seems unsystematic; but a good deal of it can be understood

as systematic through the concept of grammatical metaphor. In grammatical met-
aphor (Chapter 6 §6.4.4), a grammatical form is represented – that is, realised – by
another similar to it. Chapter 6 examined modality; another instance which we have
met several times is nominalisation: a sublexical event sense such as ‘destroy’, which
would be realised congruently as destroy, is realised by grammatical metaphor as
the Entity word destruction; one semantic class is represented as another. Another
common instance is rankshift: a Subject, which is congruently realised as a group, is
realised as a clause; one syntactic structure is represented as another. A command,
such as ‘Get out!’ can be realised incongruently and metaphorically as a statement,
“You will leave now” (with appropriate intonation). Other instances include: linguis-
tic deixis for strict deixis; grammatical modality for cognitive modality (Halliday
2014, §10.4); stative Processes for Properties (above, Chapter 8); and “type coercion”
(Pustejovsky 1995).
The concept of grammatical metaphor, like that of marked usage in general, al-
lows us to generalise about many forms of realisation which otherwise seem diverse
to the point of being almost arbitrary. For detail, see Halliday (2014, Chapter 10) and
Simon-Vandenbergen (2003).
12.6.3 Construal
The concept of construal has been used periodically so far, especially in the early
sections of this chapter, but without a full explanation. This section describes it more
formally, and sets out the main varieties of construal that occur in the realisation of
English.
The varieties we have met in this chapter are the following: construal of a concept
as a sense, and reconstrual of such an entity sense as an Entity, then a Participant;
construal of closely related senses as integrated semantic groups, and of a string of
groups as figures.
Thus, as used here, to “construe” something is to shape it into a new form which
includes the original but transcends it, forming a gestalt by relating it to other things.
The other things to which it is related may include other things of the same kind, and
other dimensions in which it can operate. The term is used for building something
into a pattern or structure on a higher level, and, in the context of an existing con-
strual, for developing an alternative form.
Construal generally entails changing the category (including “class”) of the sense
or expression. As noted in §12.2.2.5, senses and expressions are re-categorised several
times during realisation.
The concept of construal is valuable in several ways. It explains the origin of many
linguistic phenomena, thereby illuminating their nature. It shows something of the
coherence and integrity of English, since it characterises many phenomena. It relates
the use of language to other mental activities, since thinking, imagining, and planning
also involve construal. Its value is limited, however, by the inconsistency of past uses
of the words, construe and construal. See, in particular, Langacker (1987a, 1987b),
Croft and Cruse (2004), and Cruse (2014).
12.6.4 Statives
The so-called “stative verbs” were discussed in §7.3.5 and §8.4.4.3, where it was noted
that “stative” has often been applied also to “adjectives”, particularly to “predicative
adjectives” such as asleep, aware, and awake. It was asserted in those sections that the
concept is unsatisfactory, since it does not characterise any coherent linguistic class.
There is no particular morphological or syntactic form or behaviour that expresses
stativity. Moreover, a better explanation of the phenomenon was given in Chapter 8,

§8.4.4.
The explanation of the confusion lies, I believe, in the fact that state is a cogni-
tive category, which is not realised in a corresponding linguistic category. Concepts
that are cognitively stative are construed in different ways in English. In general, such
concepts seem to be properties cognitively, such as being in a state of sleep. When
semanticised, that can be construed as the (stative) Property ‘asleep’ – which becomes
the “predicative adjective” asleep – or as the (stative) Event ‘to sleep’, which is rendered
as the (stative) Process slept, sleeping and so on. That Process can be reconstrued as a
(stative) Property, as in “a sleeping child”. Other cognitive stative Properties, such as
BE GREAT, are often reconstrued cognitively into a relation, as in GREATER THAN. That
can be semanticised as ‘greater than’, and lexicalised in “greater/bigger/more extensive
than” (often used with the stative relational Process, ‘to be’), or in the (stative) Process,
“exceeds”. Thus the linguistic distinctions are between Property, Entity and Event, and
between eventive and relational Processes; the cognitive distinction between state and
happening is not grammaticised.
That, I suggest, resolves most of the problem with stativity in English. (I acknowl-
edge that there are other issues, such as the historical relic of the prefix a-, derived
from the Old English preposition an (in aboard, awake, aware, awash, etc.); thus
aboard derives from “on board”, and retains both the syntactic behaviour and the sta-
tive cognitive meaning of the prepositional phrase; but those words and phrases are
idiomatic as relics, not systemic.
12.6.5 Domains offering alternative realisation
12.6.5.1 Irrealis
Chapter 11 §11.5.5 discussed irrealis (including modality) as warning the hearer that
the content being stated should not be relied on at face value, for one of several reasons;
such expressions guide the hearer towards taking a cautious attitude to the relevant
content. Chapter 11 also noted that some utterances in the same form are functionally
quite different, being not interpersonal warnings, but ideational statements of truth.
The example given there was, “If you throw two dice, you will probably not turn up
two sixes;” probably states statistical fact, so it gives information; the speaker is not
asserting something hesitantly. Such uses are sometimes denoted by “modality”, but
probability of that sort is a cognitive issue, and does not call for linguistic analysis.
Like that form, this ideational form of “modality” is realised in a wide variety of
lexicogrammatical ways: as subordinate clauses, as Adjunct groups, as premodifiers
and postmodifying groups in Event groups, and as premodifiers in Entity groups.
It is also realised by punctuation, as in “scare quotes”, and by intonation – typical-
ly by tones 1 and 4 in combination, or by tone 4 on its own (Halliday and Greaves
2008, §5.2.4).
12.6.5.2 Time
Klein (2004, pp. 40–41) lists six ways that are commonly used in language to “encode”
some aspect of time. They are as follows.
– Morphological realisation, in tense: was / is / will be.

– Morphosyntactic realisation, in aspect: “Eva was closing the door” / “Eva closed
the door.”
– Lexical realisation in “temporal adverbials”: now, soon, when, after.
– Lexical realisation in “temporal particles”: the Chinese particles guo, zhe, le.
– Realisation by “discourse principles”: the iconic principle by which the order of
statements indicates the order of events.
– Semanticisation, as “Aktionsart”: the concept CLIMB can be given different se-
mantics, even when lexicalised the same way, as climb: “They climbed Mount
Everest” has the concept specified as having an end point, but “They climbed for
three hours” does not.
The listing reinforces the main point being made in this chapter, that ideational mean-
ings are realised in many ways. However, it misses one form of “encoding”, namely
the use of premodifiers such as former, recent and ex (or prefix, ex-). I will restate
Klein’s analysis, according to the strata of language assumed here, and for English, as
follows. Meanings in the domain of time are sometimes realised lexically: overtly, in
“adverbials” like now and soon, “conjunctions” like when, and after, premodifiers like
former, and prefixes like ex-; and covertly, in Aktionsart. They are sometimes realised,
morphologically, in tense. They are sometimes realised by information structure, i.e.
by the iconicity of content order, as in “I came; I saw; I conquered.”
12.6.5.3 Figurative realisation

The discussion of figurative language in Chapter 11, §11.6, asserted that figurative
language such as metaphor does not have conveying information as a basic or normal
function. It is occasionally used that way successfully, however. Example (9) is from
a newspaper article discussing the American Republican Party, and the opinion of
moderates that the party was moving too far to the political right.
(9) “The search is on for a political star who can appeal to a broader audience and
keep the party from disappearing down its ideological rabbit holes.”
(New Zealand Herald, Nov 4th 2013)
A literal statement of the meaning would be something like, “going to such an extreme
in its ideology that people would regard it as petty and absurd, and stop taking it se-
riously.” The metaphor makes that meaning clearer, in an intuitive if not conceptual
way, as well as being admirably brief and expressing derision. Figurative language can
sometimes realise information effectively.
12.6.6 Comparison between morphosyntactic layers
In realisation of descriptive meaning, words are arranged in groups, and groups are
arranged in figures. Since both forms of expression are matters of syntax, one might
expect them to be much the same. In fact, however, they are remarkably different.
The relation between items in a group is that of dependence, which allows var-
iation in unmarked order; the relation between items in a realised figure is that of
complementation, which allows no variation in unmarked order, apart from that of
the Adjunct. In groups, a single item (the head) controls the position and order of all
the other items; in realised figures, there is no dominant item, the items being inter-
dependent for their order; the weakness of links between the elements in a group, and
the strength of the links between elements in a realised figure correlates with that.
Functionally, groups are strongly cognitive and ideational, identifying and describing
the things and happenings which the utterance refers to; realised figures are also idea-
tional in formulating a message, but can reasonably be described as dominated by the
interpersonal function of asserting or exclaiming and so on.
In comparison, there is very little difference between a figure and a figure com-
plex, which is simply a series of figures, but quite acceptable as such. In contrast, a
mere series of groups is unacceptable and almost meaningless; the Subject-Predicator-
Complement series of groups, however, constructs a gestalt-like structure which is
quite unlike a group.
12.7 Conclusion: Realisation of ideational meaning
Summary of this chapter. The first step given in this chapter, on the realisation of idea-
tional meaning, was that the knowledge to be expressed is conceptualised and formu-
lated as a message; that is a cognitive process. Next, its concepts are semanticised into
sense elements, which are conceptual, ranging between the relatively concrete and the
relatively abstract, with the characteristic linguistic dimensions; in brief, they are se-
manticised. Sense elements are then lexicalised; that is, they are combined in various
ways into senses, and associated with morphemes which symbolise them. Those mor-
phemes have very strong links to spoken and written physical forms; speakers and
writers normally have no choice of the physical form after choosing the word sense.
The morphemes have links, of varying strengths, to structural uses: dangerous has a
strong link to modifier use; but ‘sign’ has weak links to Entity modifier use (“a sign
board”), Entity head use (“a sign”), and Event head use (“ He signed his intentions”).
Those links are potential; the senses do not belong to semantic classes, at this stage.
They have weak links to linguistic structure, and strong links to cognitive structure.
The word senses are grouped, and structured semantically – primarily by subordi-
nation, in the head-modifier relationship. That actualises certain potential links; senses
thereby gain the status of semantic class membership – Entity, Event or Property. At
the same time, grammatical meanings are formulated and symbolised, syntactically
Table 1. Summary of the realisation of information

Process Items produced Name of process Layer/stratum
The state of affairs is formulated. Concepts Conceptualisation Cognition
Concepts are construed into Sense elements Semanticisation Semantics
linguistic elements.
Sense elements are grouped. Word senses Lexicalisation Lexis
Words are grouped, structured and Groups Syntacticisation Morphosyntax
ordered. (subordination)
Groups are structured and ordered. Figures Syntacticisation Morphosyntax
(complementation)
Figures are assigned phonological Intonation Phonologisation Phonology
form. groups
Phonological forms are spoken or Spoken or writ- Phoneticisation Phonetics
written. ten utterances
as word order for example, or morphologically as inflections, or lexically as free mor-

phemes such as determiners. The grammatical meanings are not easily thought of as
part of the content meaning which the speaker sets out to realise, and which is lexical-
ised in the first stage, but the content cannot be realised without them. All this seman-
tic structure is realised as the corresponding the syntactic structure.
In a second stage of syntacticisation, the groups are themselves grouped, struc-
tured (primarily by complementation) and ordered, in ways corresponding to those
in the first stage; the structures formed are figures (and then clauses). In default
realisation, Entities are developed into Participants, Events developed into Processes,
and Properties developed into Circumstances.
In a further stage, the lexicogrammar is linked to a phonological form, with its
four layers, tonality, tonicity, tone and rhythm. In the last stage, the whole structure
developed so far – words in syntactic order and the phonological patterns they are
linked to – are spoken in phonetic form, or written in graphological form.
That outline is summarised further in the Table 1.
Discussion. The account just given is an analytic linguistic one, not a psycholinguis-
tic one. In psychological operation, the processes often overlap and interact (Lamb
1999), and consequently may cycle until a satisfactory form is decided on.
The semantic and syntactic lines of development each have a continuity and in-
tegrity of their own, but also interact at each main stage. That is, senses are developed
into more complex semantic classes, and related by the addition of links, forming se-
mantic groups and then figures; those are strictly semantic structures, and follow se-
mantic “rules”. Morphemes and word forms are similarly related by links and formed
into groups and clauses, by syntactic rules and into syntactic structures. However, the
syntactic steps are always motivated by the semantics, and are controlled by it.
The links just referred to must be understood from the three semantic aspects:
those of the speaker, the hearer and the system (Chapter 4, §4.1). From the speaker’s
point of view, linking is indicating links such as those of modifier and head, and of
Participant and Process, by adding one level of grammatical meaning to the concep-
tual meaning which is the input to the first syntacticisation stage, and a further level
to the input to the second syntacticisation stage. From the hearer’s point of view, the
linkage consists of the linguistic and conceptual links between content items which
are to be constructed according to the procedures specified by grammatical meanings.
From the system point of view, the links are linguistic and cognitive semantic links be-
tween the nodes of meaning, commonly referred to by grammatical terms for category
or status, like “noun”, “Subject”, “finite” and “Goal”.
The semiotic relation of form to meaning varies with the different stages. As has
often been remarked, phonemes, as the most basic element of form, are not mean-
ingful individually; they are semantically neutral – that is, realisation of a sense into
phonemes does not add meaning. Word order need not be meaningful either; uttering
words in order is merely a practical necessity; it is like a low frequency radio wave
which is meaningless in itself, but carries a higher frequency wave within it (“syntax”),
which constitutes the meaningful signal, as with the order of Subject and Predicator.
Both the phonological and phonetic forms are also “carrier waves”, necessary for the
“transmission” of the word senses, and having meaning of their own only when some
aspect of the sound has been conventionalised into a sign.
I have been emphasising the changes that occur in the apparently simple process
of the realisation. I wish to emphasise one change particularly, because it has been
bypassed in most linguistics. Semiotically, the process is one of creation, rather than
change. The grammatical meanings are not part of the message; their nearest equiv-
alents there are logical relationships. Speakers create the grammatical meanings, and
their signs, adding them to the content. That is relatively clear in the insertion of de-
terminers and inflections, but it applies also in syntax: putting the Predicator before
the Subject creates the sign for the interrogative.
12.8 Conclusion: Both forms of meaning realisation in English
12.8.1 Relation between the two forms of realisation
Introduction. This section relates to realising both the interpersonal function in Chap-
ter 11, and the ideational function in Chapter 12.
The two chapters have followed the useful expository procedure of presenting
realisation as two contrasting processes: a process of differentiation for interperson-
al functions and a process of combining elements for the ideational function. This
section will qualify that generalisation, and present the processes as complementing
each other.
Independence. The two processes are only partly independent. They usually occur to-
gether, since when speaking interpersonally we usually express some content, even
if we do not intend hearers to remember it; and even when focused on passing on
knowledge, we usually have a personal motive. Indeed, each metafunction facilitates
the other. Moreover, differentiation is important in the development of ideation; in the
new-born’s perception, there is a gradual differentiation of things from happenings,
and of participants in a situation from one another; even in language, Participants are
distinct only as a matter of degree (Kemmer 1994).
For holophrases, and for very simple utterances like one-word replies and excla-
mations, realisation is so simple that it can hardly be characterised as being of one
type or the other; the intention (at the top) and the wording (at the bottom) can be
decided simultaneously. In other utterances, the two are complementary, since the
overall planning and the choice of words need each other, especially when informa-
tion structure is important to the speaker. Moreover, the same strata and the same
layers within them are used in each form of realisation.
On the other hand, the two can be distinguished even when they are integrated.
Utterances inevitably belong to some register, which alters the mode of realisation in
each metafunction. In formal registers, affective meanings are typically realised lexi-
cally, emphasis is achieved syntactically, and lexical and syntactic forms are preferred
for grammatical meaning, for example. In informal registers, both affective meanings
and emphasis are typically realised phonologically, and phonology is preferred for
grammatical meaning, when it is available.
Complementary relation. The general relationship is that the two forms of realisation
are complementary, not only in form, but in being used simultaneously, each com-
pleting what the other does only partially. Our communicative intentions usually
combine a personal motive and an intention to convey information. Indeed, it is ar-
guable that that is always so, since speakers and writers always have a personal reason
for delivering even the purest and driest information.
If we consider realisation as the relation of meaning to text, the lexical level serves
primarily for information, as descriptive meaning. The phonological level serves in-
terpersonal functions primarily, with intonation and stress, for example, conveying
feeling. Syntax, in the “middle”, serves both metafunctions. If we consider realisation
as a process, there is a similar meeting in the middle: as soon as we differentiate out
our primary and secondary intentions, we make the high-level decision as to whether
those intentions need statement or question and so on, and the low-level decision as
to what senses we will need to lexicalise; the statement or question will then need a
syntactic structure, which the word senses will also need. Fortescue (2009, p. 182)
supports that psycholinguistically: utterances are initiated by “a combination of high-
er order intentions and specific semantic elements.”
12.8.2 Sequence of steps
Introduction. Chapters 11 and 12 have also simplified the sequencing of steps. Steps
that have been presented as late in the process must often come before an “earlier”
one. For example, decisions to ask a question or make an exclamation routinely lead
to the choice of an interrogative or exclamatory syntactic structure. Those structures
often require a particular word such as what or which; the lexical step must follow the
syntactic one, not precede it, as in the canonical process set out above. Also, steps are
sometimes repeated. For example, the step of providing emphasis is often carried out
in the lexical stage, by the choice of very or extremely; but providing emphasis is then
repeated at the phonological stage by the selection of marked stress. Psycholinguis-
tically, the pathways of the brain have “bidirectional information flow” (Spivey et al.
2005, p. 249), allowing the sequence to be reversed, or to be repeated cyclically.
Ideational meanings. There is relatively little choice of “route” for informational mean-
ing – little variation from what is set out above. One variation is implicit in what was
set out in Chapter 8, §8.2, on Entity groups with complex premodification. There,
realisation of different types of conceptual meaning is done in steps, and the types
assigned to different syntactic positions – and grammatical meaning is fitted into the
same pattern. Grammatical meanings are realised in the determiner position, abstract
descriptive meanings in the Epithet position, concrete descriptive meanings in the
Descriptor position, and referential meaning in the Head. The “steps” are parallel,
rather than sequential.
Interpersonal meanings. Interpersonal meanings offer far more variation, contrasting

with the realisation of conceptual meanings (in ideation), which are realised in lexis,
which is realised in syntax, which is realised in phonology, which is realised in pho-
netics. However, some emotive meaning is realised in lexis, but is unaffected by the
syntactic and phonological steps, the lexical form being realised directly in phonet-
ics – its spoken form. On other occasions, emotive meaning “skips” the lexical step,
and is realised in syntax (by the exclamatory structure), then phonology and phonet-
ics. On yet other occasions, emotive meaning skips lexis and syntax, being a realised
in phonology (by marked intonation contour). There are similar variations for social
meanings and information structure.
Conclusion: sequence. Realisation could thus be described as a network of procedures,

rather than as single procedures with variations in the steps. In that description, we
would say that there are alternative routes available, with parallel processing and cy-
cles of processing being quite normal. (The depiction of realisation as routes through
a network will be expanded in Chapter 14, §14.3.)
12.8.3 Realisation strategies
There are several patterns of expression which occur repeatedly in different situa-
tions, which amount to strategies for the realisation of meaning. That is, they form
a cluster of lexical, syntactic and phonological forms of expression, serving a single
purpose each time it is used, but capable of serving different purposes. For example,
a descriptive strategy may be used for no further goal than being descriptive – vivid
or imaginative – but may also be used to convey information. The strategies are not
part of the grammar of English, but they are real as patterns, like registers and genres.
The descriptive/referential uses and the figurative/literal uses (Chapter 4) are parts
of these strategies. What follows is a description of some frequently used strategies,
arranged as complementary pairs.
There is an understating strategy, and an “expressive” one, where “expressive”
means being personal, forceful and vivid, and tending toward overstatement. The un-
derstating strategy typically uses descriptive meaning in preference to affective mean-
ing, referential use rather than descriptive use, and factual content; it may use irony
and litotes, but not usually other figures of speech; if it indicates emotion, it does so
through stating its causes or outcomes. The expressive strategy may use Expression in
the technical sense; it regularly uses emotive and attitudinal meaning, marked uses,
and figurative language; in phonology, it uses high pitch and stress levels, and marked
forms. Example (4) in Chapter 11 used it: “The whole farm was just a sea of ragwort!!
It was everywhere! You just couldn’t believe how much ragwort could grow in one
place!” – The excitement is perhaps overstated. Example (3) in this chapter used the
understating strategy in reporting excitement: “I came back all excited to Steve and
he didn’t quite know what to think” – the speaker was “excited”, she says, but she un-
derstated it.
The referential strategy employs the referential or defining use, which involves
names, acronyms, deixis, little modification, and restrictive syntactic structures; it en-
tails using words with no more than than their necessary meaning; example (1) illus-
trates an extreme form, with “29Aug 4of12”. The complementary descriptive strategy
invokes possible meanings plentifully, uses non-restrictive syntactic structures, and
much modification, as in romantic fiction and travel journalism.
In the final pair, assertion contrasts with suggestion in putting forward infor-
mation. The assertive strategy uses the imperative and exclamatory moods readily,
minimises modification and especially qualifying elements, and in phonology uses
low, even intonation and falling tones. The suggestive strategy avoids imperatives and
exclamations, but readily uses the interrogative to make statements; it uses qualifi-
cation heavily, through modifiers, modals and other irrealis such as the subjunctive
mood; in phonology, it uses high, light intonation, and prefers rising tones. “It is high
time you went,” is assertive; “Perhaps it’s about time to go?” is suggestive.
12.8.4 Adequacy of realisation as a semantic concept
Introduction: realisation and composition. According to Halliday and Greaves (2008,

p. 14), “the sound of language is a resource for the construction of meaning”; but
according to what has been said so far, that is back to front, because meaning is the
source of – or resource for – the construction of the sound of language. The appar-
ent contradiction comes from different uses of “meaning”, of course; Halliday and
Greaves are thinking of meaning as what the hearer takes from the sound. To that ex-
tent, they are correct: the meaning is constructed; it is compositional; pieces of mean-
ing, as in words, are put together according to recognized and definable procedures,
which are indicated in the utterance, particularly by the meanings. Hearer meaning is
not realisational, since its input is physically real but its product is mental; rather, as a
process, it is fundamentally compositional.
Speaker meaning, on the other hand, can hardly be said to be realisational either,
since it is potential, which the speaker may or may not realise. Nor can it be said to
be compositional, strictly, because it is a pattern of activation in the mental network,
having no definable parts that can be put regarded as put together, stage by stage.
“Under-realisation”. System meaning is different again, and the adequacy of explain-

ing it as realisation needs fuller discussion. In one respect, system meaning is not very
fully realisational, since there are many ways in which meaning is underspecified, or
“under-realised”. In the area where linguistic study borders on conversation analysis,
speakers give hearers cues for action, as for turn-taking; that is, speakers signal, but
do not specify, the action intended. Many “discourse markers” and marked uses of
phonology likewise signal an intention or desired response, without setting out the
specifics. Holophrases are obviously an under-specification of meaning. Pragmatic
meanings such as implicatures are left for the hearer to infer, so are not specified.
There are many small ambiguities and instances of vagueness which hearers must
resolve from the context, as with the relation of circumstances to the rest of the figure.
More fundamentally, the fact is that the text realises only the signs from which hearers
will construct the meaning; it does not realise the meaning, itself.
“Over-realisation”. On the other hand, some expressions “over-realise” meaning; that

is, they specify elements of meaning which are needless, and which the speakers does
not intend. Some uses of articles are meaningless, as with “Play those notes up the
octave,” and “He’s on the move again.” So are some uses of prepositions; we can say
“Give him the book”, but must insert a conceptually redundant preposition in “Give
the book to him.” Tense must specified in every clause, even if time is stated adverbial-
ly, or is irrelevant. Moreover, some pieces of over-realisation are confusing. Examples
include specifying the tense of eternal truths, as though they were true in only one
time frame. Expressions like “The committee are thinking of …” are morphologically
singular and plural simultaneously; so is “a pair of scissors”. In all these ways, the
grammar or idiom of English requires meanings to be realised which are not in the
speaker’s meaning.
Conclusion: realisation as a concept. In those instances of under- and over- realisation,

it is better to describe utterance meaning as being compositional. Arithmetic pro-
vides a useful analogy. Its “language” is compositional: the numerals have content; the
operators signal how the content is to be combined, without realising any content of
their own; the hearer, not the written or spoken expression, “realises” the meaning. In
some respects then, English is better explained as both compositional and realisation-
al – and I have shown earlier that it is far more compositional than has been shown
previously.
On the other hand, the whole of these two chapters has shown that the concept
of realisation is very useful in explaining the structure of semantics in English utter-
ances, provided that we think of realisation as giving form to the message, rather than
changing its substance. Part of that usefulness is that it forces us to go as far as we can
in finding meaning in the utterances, not implicated or suggested by it, and it forces
us to justify whatever exceptions we allege. Moreover, when we think of speech and
writing as physically real, we cannot avoid the concept of realisation.
Realisation, like composition, is a useful concept in the study in language, but not
fully satisfactory.
12.8.5 Constraints on realisation
It is not obvious that there are any constraints on the realisation of meaning in Eng-
lish, but reflection on the history of English suggests that there probably are. The
size and the structure of nominal groups have grown greatly since Middle English,
and the complexity and importance of information structure have grown similarly, as
noted several times previously. The language will presumably keep on changing. The
analysis in the previous chapters, especially Chapter 8, suggests two areas of syntactic
structure where scope for development is fairly clear.
First, Event groups are much simpler than Entity groups; they could develop more
scope for choice between premodification and postmodification, and could develop
zones of modification comparable to those in Entity groups. That would give English
greater variety and precision in the realisation of Events and of the Properties that
modify them (“adverbial” meanings). Second, perhaps the vagueness of the relation
between Circumstances and the rest of the figure is also a constraint, which could be
eased by greater precision, and a paradigm of options.
An instructive precedent exists for such a development. The Classifier construc-
tions, and the choice between constructional and constructionless Classifiers, have
developed since the 18th century (Feist 2012a, §8.5). That development has facilitated
expression greatly. It provides a choice between one of many precise meanings on the
one hand, and a wide open construction where you can put a nominal with any cogni-
tive-semantic relationship at all (Chapter 8, §8.2.2). The lack of that choice would not
have seemed a constraint to 18th- and even 19th-century speakers, but now its being
a constraint is clear; so the lack of other such choices in the semantic-syntactic system
of English is presumably a constraint.
12.8.6 Looking forward
We have now completed the exposition of semantic structure in English, having seen
its basis and elements in Chapters 3 and 4, the various main and secondary structures
in Chapters 5 to 10, and the structure of their realisation in Chapters 11 and 12. The
next two chapters conclude the book, giving discussion and conclusions in turn.
Chapter 13
Discussion
Introductory note
This chapter discusses issues which have arisen during the book, but which are not
central to its intention or the conclusions to be drawn in the next chapter. The issues
chosen are mostly controversial, and are those in which the book goes significantly
beyond what is presented elsewhere, or implies conclusions which are not accepted
elsewhere.
13.1 Compositionality
13.1.1 Introduction
According to Cruse (2011, §4.1), compositionality is the principle that “the meaning
of a grammatically complex form is a compositional function of the meanings of its
grammatical constituents”; that entails the claims that the meaning of the whole form
is determined by its constituents, that it is completely predictable by general rules
from the constituents, and that every constituent contributes to the whole. There are
many variations on that principle – generally weaker – but there is now consensus
that “While compositionality is clearly a fundamental component of cognitive activ-
ity, it is equally clear that compositional behaviour is neither perfect nor unlimited”
(Stewart and Eliasmith 2012, p. 598). That acceptance of compositionality is based on
such needs in linguistic theory as the following, given in Chapter 5, §5.2.5: to explain
errors in the choice among closely related words, to explain sense relations among
words, especially synonyms, and to explain lexical access. I accept the consensus.
Accordingly, this section will be limited to comments that arise from the preceding
chapters, adding to the consensus view without debating it. It is also limited to con-
sidering compositionality in figures and smaller structures, although some writers
consider that it exists in larger structures; see Janssen (2012, pp. 41–42).
A further reason for limiting the coverage here is the fact, established in Chap-
ter 2, §2.5.7.4, that compositionality of meaning is not important fundamentally, for
two reasons. First, since language is functional, the fundamental issue is whether the
meaning of an utterance is effective, not its completeness or any thing else structural.
Second, since the meaning of each word and utterance is part of a network, with links
to other areas of the network, compositionality is not a fully appropriate concept.
13.1.2 Full compositionality
I suggest that the book has provided an important contribution to the understand-
ing of compositionality, by specifying the nature of the “compositional function” and
“general rules” referred to above as constructing the meaning. They are given little
attention in the literature; it is widely assumed that they are like mathematical func-
tions, being adequately indicated by concepts such as addition, sets and their combi-
nations, and such notation as “x(y,z)” for combining something with two others as its
arguments. The book has argued implicitly that linguistic composition is specifically
linguistic, and has given the “compositional functions” or “mode of combination” ex-
plicitly as grammatical meanings.
Those meanings specify combining as (1) making units of equal status, either
(a) coordinate, with the same function, or (b) complementary, with different func-
tions; or as (2) making the units of unequal status, one dependent on another. The
modes of dependent combination, i.e. modification, were listed in Chapter 6, §6.2.3,
as adding, specifying, adjusting, determining and negating. The steps required in
those compositional processes are more complex in figurative and other marked uses
(Chapter 2, §2.4.2.3, Chapter 4, §4.7), which require extra steps, sometimes recur-
sively. In restrictive use, the hearer must, rather oddly, compose meaning elements by
adding them (to identify a referent), but must then remove them (Chapter 8, §8.2.9.5);
for example, in “Press the <Yellow> or <OK> button to enter the timer edit window,”
the underlined modifiers are added to their head, to identify the head’s reference
(which button, and which window), but then are discarded in favour of ‘button’ and
‘window’, because the intention is simply referential, not descriptive. We have seen
that it is not only words, figures and paragraphs that are composed, but also informa-
tion structures, in their own hierarchy (Chapter 9), parallel to the hierarchy built by
lexicogrammar.
Composing units into a structure requires a mechanism which bonds the units;
a steel construction, for example, must be either welded, bolted or riveted. We have
seen the nature of the semantic bonds in the previous chapters – the transitive rela-
tion which binds elements in the Participants and the Process, for example, and the
attribute-value relation binding modifiers and head – and the precise location of the
bonding elements, such as a dimension or a meaning type.
These procedures, specified and illustrated throughout the book, are clear and
specific, giving a much better understanding of how meaning is composed in English
than is given by the traditional accounts as set out in Cruse (2011, §4.2), for example.
Further, they reveal a degree of rigour and complexity in compositionality that has
not been shown in the literature, in Entity group structure in particular (Chapter 8,
§8.2), especially for Classifiers §8.2.2).
Further, the literature gives little indication of how compositionality relates to the
structures composed. This book has set out both the syntactic and the semantic struc-
tures which are built up, relating them explicitly to the grammatical meanings which
build them. It has also explained why these rather elaborate procedures are necessary,
Chapter 13. Discussion 387
in describing the constraints under which language operates (Chapter 3, §3.4); that is,
language is heard or read in temporal order, but hearers replace that linear structure
with hierarchic and network structures outside time, so fairly fully specified compo-
sitional procedures are essential.
13.1.3 Limited compositionality
Compositionality is important and problematic only if you assume that language is

compositional fundamentally, consisting simply of constituents; but that assumption
is made by those who restrict their view to language as a formalisable means of repre-
senting concepts. In the view taken here, in which stating concepts is only one form of
English, and language is fundamentally functional not representational, there is sim-
ply no need to see language as inherently compositional. “Individual meanings are not
basic and prior to complex meanings, as atoms are in relation to molecules”; they are
“functional components”, like the handle and blade of a knife (Harder 1996, p. 285).
Indeed, the book so far has shown, I believe, the importance of several facts about
meaning in English, not generally accepted, which limit compositionality sharply.
They are as follows.
– Meaning is usually contextual; the extensive underspecification of English is

made good from linguistic and cognitive context; frequently, there are insufficient
linguistic elements for compositionality to work on.
– Composition requires “blocks” or “particles” as constituents; but meaning is often
structured as waves or fields (Chapter 10).
– Meaning is sometimes holistic, either using units whose structure is indetermi-
nate, such as holophrases and idioms, or units that form gestalts – the nature of
figures and groups as units is not composed from the units, but emerges from
them.
– Meaning is often associational, as in constructionless Classifiers (Chapter 8,
§8.2.2), and in the link between descriptive and nondescriptive meanings within
a word sense.
– Descriptive meaning is relational, in the system meaning of utterances, in the pre-
linguistic meaning the author intends, and in the final meaning the hearer under-
stands; relations cannot be reduced to constituents, and are not compositional.
– Meaning is fundamentally the fulfilment of intention, rather than the statement
of content; as the traditional term “speech act” indicates, the crux of meaning
is action, from the speaker’s point of view, and effect or result from the hearer’s
point of view. The top-down understanding is essential, as well as compositional-
ity’s bottom-up view.
– Compositionality must be understood as not only a static relation of the parts in
a whole, but also as a dynamic one – the hearer composes the meaning.
Finally, we may note that compositionality is sometimes counter-productive, since it

requires all meaning to be stated explicitly. In an emergency, that would slow us down
crucially. In close personal relationships, compositionality prohibits the assumption
of shared experience and feeling that is essential to intimacy. In expressive language,
it dilutes forcefulness. In imaginative language, it prevents the essential creative spark
between images.
13.2 Word classes
13.2.1 Introduction
There is no consensus on how to define word classes or “parts of speech” (which I take
to be simply an older term) or “lexical categories”. Current accounts do not cohere
well internally or with other theory; they are based on the unquestioned assumption
that such classes do exist; and no linguistic work that I have consulted even tries to
state firmly the word class of a range of difficult instances. In the argument of Chap-
ter 8, §8.3.8.5, there is no such word class as verbs, in particular. This section argues
that the very concept of such classes is unsatisfactory – the concept, that is, of a clas-
sification of words (not of senses or uses or functions), words being at once semantic,
morphological and syntactic. The section also suggests a better understanding of the
phenomena concerned.
Historically speaking, the reason why the concept is unsatisfactory is very simple:
English has changed. The classes applied fairly well to Old English; but the radical
changes in syntax, morphology and semantics have changed the “parts of speech”
radically, as well. In particular, Old English nouns and pronouns, with their distinc-
tive pattern of inflection, regularly named people, places or things, and functioned as
Subject or Object. But their function has come to be dominated by information-struc-
ture function – hence the need for the term, “nominal”; we have dummy nominals
as in “there is…” and “it is raining”; nominals primarily realise large- or small-scale
topics in language. Similarly, verbs can now be Predicators without naming actions or
events, while retaining the old inflections. The links between meaning, morphology
and function have loosened.
13.2.2 Needlessness and unworkability
Synchronically speaking, it is significant, I suggest, that we have not needed the parts
of speech at all in this book, although we have seen a full explanation of the structure
of semantics and a partial explanation of the structure of syntax and morphology. (I
have used the terms “noun”, “adjective”, “verb”, “adverb” and so on, but only to keep
to familiar wording, where more exact terms are not important.) Indeed, the need for
the concept is obscure, and is very seldom stated. The only “need” I can identify is for
help in teaching; but there it is like the “rules” of spelling – a useful approximation
for beginners.
I suggest that we need linguistic concepts such as those of word or semantic class-
es only where they help explain how we produce and understand utterances in the
language. As Wierzbiecka (1990, p. 365) says in discussing prototypes, a linguistic no-
tion “has to prove its usefulness through semantic description, not through semantic
theorising.” We need to explain how speakers decide what form of a word to use, for
example, and how hearers decide which sense is the semantic head of a group. Now
every current “definition” of nouns, for instance, takes the form of a set of criteria such
that a word is a noun if it fits most or all of the criteria. Since many nouns do not fit
them all, their being a noun is a matter of probability; and they have some possibility
of being a verb or an adjective. When hearers encounter a word, then, they must cal-
culate probabilities of its word class in that use, which must be quite complex, since
there are from 3 to 5 criteria (the number varies with the authority); and, because
words are now used in various roles fairly freely, the calculation may well involve all
four major classes. But as we listen, we are processing three or four words a second,
and are dealing with the overall structure and discourse relevance of the past clause
or two while we process the current one; and we do that with great accuracy, as well
as great speed. It seems that real-life processing is impossible with such probabilities –
with such criteria – and that the word-class concept is unworkable.
Nor do we need to explain any systematic linkage of semantics, morphology and
syntax: there simply is none in modern English. Hearers and speakers can make de-
cisions about a word’s syntactic role, the interpretation of its morphology, and its se-
mantic significance, separately. (In the linguistic system, those three have separate
signs, but on occasion one may be used in understanding another, English being fre-
quently underspecified.) The word-class concept is also needless.
13.2.3 Lack of explanatory power
The word-class concept lacks explanatory power: it does not explain the pattern of
semantic, morphological and syntactic phenomena in English, because those phe-
nomena do not inhere in words: any classes to be identified are not word classes.
The arguments I have given so far perhaps build only a high probability that that
is so. Similarly, proof does not come from listing some of the myriad instances where
modern English uses Event words without tense (“a travelling salesman”), concrete
Entity words without number (“lawyer privileges”), Property words as heads (“the
Greens”), and other “confusing” expressions (e.g. Chapter 12, §12.3.5.3, §12.3.6) –
although the reader will, I presume, recognise that semantic, morphological and syn-
tactic properties are now much more loosely linked than they were in past centuries.
The point is established conclusively, I believe, by the following two arguments.
First, for “nouns” or “nominals”, the word on its own carries descriptive and nonde-
scriptive meaning; but it is a larger unit, the “nominal phrase”, which identifies the
person, place or thing – which “refers”; it is only the phrase as a unit in the clause
that can be the grammatical Subject and the Participant which relates to an action,
and only a unit in information structure can be the “logical subject” or Topic of dis-
course. Similarly, “verbs” carry conceptual and other meaning at the semantic level,
but only “verb phrases” at group level can identify actions, and only verb phrases
(or Predicators) at figure level that can be “the verb” in a sentence i.e. can predicate.
These points were established in different forms in the sections on semantic class,
particularly Chapter 8, §8.5.3, and on the indeterminacy sublexical senses, especial-
ly Chapter 5, §5.4. Those sections showed that morphosyntactic and even semantic
properties are gained in stages as meanings are realised at successive levels; they do
not all apply to words or at any other individual level. See Lamb (1999, pp. 236–238)
and Smith (2010) for supporting argument. Some of the “word class” properties, then,
are properties of phrase and clause constituents.
The second argument is that even at word level, the properties do not, in present
day English, inhere in words as such. That was shown in the analysis of premodifiers
in Chapter 8, §8.2. We saw there that the same sense can occur in different zones – and
even that speakers can (within limits) move a word between zones at will – and that
as a word changes zone, its morphological properties and its semantic nature change,
as well as its syntax: that is, it changes in all of the properties which constitute “word
class”, while remaining the same word. An example is the difference between the two
uses of slow when people are filmed “in slow slow motion” (COCA). In the first oc-
currence, the word can be graded, and can inflect, but in the second occurrence, it
cannot; in the first use, it has a property sense, but in in the second use it has an Entity
sense (a type of camera action). In “Oh no, don’t film it in ordinary motion, film it
in slow!”, it would have the same Entity sense as in the second use but be a head not
a modifier. The point is that the word-class criteria change as the word use changes;
they are controlled by the use – or one might say that they are dependent on the
construction.
The point is demonstrated in detail in the full set of possibilities for premodifiers
set out in Table 1, which is based on Chapter 8, §8.2 (especially §8.2.9), and which
illustrates pure used in all four pre-modifying zones, all uses having the fundamen-
tal meaning, ‘not mixed with anything else’. (The examples are from COCA; (d) is
from science fiction, with “electric” meaning a type of electric vehicle; words in square
brackets have been inserted to help make clear the zoning of the other words.)
Table 1. Pure in all four pre-modifying zones

Example Determ. Reinforcer Epithet Descriptor Classifier Head
(a) pure [obnoxious] local ethnic constituency politics
(b) a high[,] pure shining [vocal] sound
(c) uncommon pure elliptical galaxies
(d) a [very smart] three-year-old pure electric
It is only the Epithet use, in (b), that has a gradable sense and can be inflected
for comparison: we can have a purer sound, but not purer obnoxious local politics,
or purer elliptical galaxies, or purer electrics [i.e. electric vehicles]; the Epithet con-
struction specifies semantic and morphological gradability; the other pre-modifying
constructions disallow it. The other uses of pure also have their own semantic and
syntactic characteristics, as set out in Chapter 7, §7.2.
The semantic, morphological and syntactic phenomena which the concept of
word class used to describe with a worthwhile degree of validity do not now cohere at
all tightly with each other. The properties concerned are not inherent in words, how-
ever that be defined, or in senses, but are produced when senses are realised in use,
and apply at different linguistic levels.
It may be true that certain words are always used with the same morphological
and syntactic properties, as if they belonged to a class. However, the grammatical sys-
tem of English allows them to acquire new ones in new conventional uses, and allows
users to give them new ones in nonce uses, as Shakespeare did centuries ago, in “But
me no ‘but’s”, for example. We switch words’ properties much more freely now than
even Shakespeare could: Event words become Entity words (“a rebuild”, and a film is
“not a pretty watch”); Entity words become Event words (“he backpacked…”); modi-
fying words become heads (“Healthy reigns supreme”, “They switched to solar [ener-
gy]”, and “Have some brazils”); a whole Entity group becomes a whole Event group,
with the “adjective” as “adverb”, and the “noun” as “verb” (“Production has flat lined”;
“You can’t green screen an eel”; “You have been dangerous driving”). Such changes are
not novel, in fact: we have been creating intransitive uses for transitive verbs – and
vice versa – for centuries, and converting other parts of speech into new ones, with or
without affixes. Those were marked uses at first, being accepted into standard English
only gradually; now, the processes of conversion have become systematic and regular;
the uses are now not marked ones.
We conclude that, for English at least, linguistics should do without the concept
of word classes.
13.2.4 Support
There is now a good deal of support for this understanding in the typological liter-
ature, where words in the lexicon are said to be “pre-categorical” e.g. Bisang (2008)
and Donohue (2008); or classes are said to be “flexible” e.g. Hengeveld et al. (2004),
or there are said to be no “pre-established categories” e.g. Haspelmath (2007). Hopper
and Thompson (1984) long ago set out this view for English (and other languages),
in a somewhat different form; Wymer (2008) discusses flexible types (adverbs in par-
ticular) as having developed in English; de Smet and Heyvaert (2011) see English
participles as ambivalent. Croft (2005, p. 283) says that categories, which includes
word classes, are “defined by constructions, that is the elements that can fill the roles
[are] defined by the components of a construction…. Syntactic categories exist, but

only derivatively so.”
The psycholinguistic literature appears to support the reality of word classes,
since it reports that they are processed in different parts of the brain; see Pulvermüller
(1999, p. 297) and Citron (2012), for example. However, I know of no psycholinguistic
evidence that word classes exist as independent neurolinguistic entities; and the work
just cited can be interpreted naturally as applying to word uses, not word classes; the
researchers worked with the old understanding of “word classes”.
13.2.5 Discussion
The alternative to word-class theory that the book leads to is simple. In outline, it is
as follows. Syntax, morphology and semantics are now in English independent to a
degree that requires us to define their classes independently (if we want classes at all,
in our linguistic theory). In semantics, there are the basic classes of Properties, Events
and Entities (and corresponding classes at higher levels), as set out in earlier chapters.
In syntax, there are heads and modifiers (and larger units at higher levels). In mor-
phology, there are forms that inflect for tense and person, for number and possession,
and for degree of comparison. In principle, the features in each of those three strata
may combine in any way; in practice, the combination is constrained by our experi-
ence of the world, and by other linguistic issues such as information structure. An
analogy may be drawn with furniture. It may be constituted of wood, steel or plastic;
its function may be for sitting on or putting things on; structurally, it may have three
or four legs. We keep the constituency, functional and structural categories distinct;
we do not have a class for every combination.
The explanation has been in terms of classes, because linguistics has been so often
understood that way that I have often kept to familiar terms. The explanation itself
was awkward in implying that words belong to several classes simultaneously (syntac-
tic, morphological and semantic). More important, we understand the issues better if
we use concepts such as function, form, and use, in preference to class, as I trust the
previous chapters have shown – if we say, for example, that in “He boated down the
Rhine,” boated functions as a head (syntactically), has the -ed form (morphologically),
and is used with a dominant Event sense and a subordinate Entity sense (semantical-
ly). That is better than saying that it belongs to the head, -ed, and Entity classes all at
once; it is very much better than saying it is partly nominal and partly verbal, or that
it is a central verb and also a peripheral noun.
There are a couple of other elements of explanation that will make the argument
above clearer and more credible. First: we should not assume that the word deter-
mines the use (through belonging to a word class). Rather, our intention determines
structure and use, and we find a word that fits easily, or “bend” one until it does fit;
for example, we put what we want to talk about in Topic position, and select an Entity
word accordingly; if we want movement down the Rhine in Predicator position, not
focus, we say “He boated down the Rhine,” not “He went / travelled down the Rhine
by boat”. “Healthy reigns supreme” came from the need for a striking brief slogan, and
“They switched to solar” comes from the desire to avoid what is felt to be redundancy,
using neologistic but quite grammatical usage.
A second explanation lies in the twin concepts of reduction and individuation.
We have seen that all three of the basic semantic classes are individuated to varying
degrees when they are realised; that can be stated conversely by saying metaphorically
that relatively fully individuated senses are “reduced” semantically, morphological-
ly and syntactically in some contexts, such as when they are incorporated or used
as premodifiers. (See Chapter 8, §8.2.2 for Entities, §8.3.8 for Events, and §8.4.2 for
Properties, and §8.3.6 for incorporation.) The crucial point is that when “reduced”, i.e.
low in individuation, they have fewer of the putative word class characteristics. (Many
linguists would say that “nouns” are then “less nouny”, or that they are no longer “pro-
totypical nouns”.) That demonstrates again that they have those characteristics only
when realised, not as “words” in the “lexicon”.
13.2.6 Conclusion
Several things have contributed to the persistence of the word-class view of grammar,
in addition to the natural lag in recognising historical change. One is a powerful pre-
disposition to describe language phenomena as things, and to then classify them –
even when the resultant class obscures understanding, as when people grouped
shellfish, crayfish, sharks and whales together as all being “fish”. A second is the confu-
sion of linguistic with cognitive semantics, whereby ontology slides from metaphysics
into linguistics. A third is lack of awareness of grammatical meaning, especially in
syntax and phonology, resulting in the belief that structure must arise from content
words and their properties; for example, it is thought that the head-modifier structure
arises from words’ belonging to the noun and adjective classes, not from speakers’
intentions and the grammatical meaning which implements them.
13.3 Prototypes
13.3.1 Introduction
The goal of this section is to show that the common view of “prototype categories” as
real and important in English semantics – and by extension, in English grammar – is
wrong. What is meant by “prototypes” varies a little; here prototype categories will be
those in which some instances are better examples of the category than others, and in
which there are unclear or uncertain cases, and for which non-members vary in how
different they are from members (Barsalou 1983, p. 260).
Among exclusively semantic categories, polysemous words are seen as constitut-

ing prototypes by Fillmore and Atkins (2000, pp. 99–100). Categories with a semantic
element identified as prototypes include “parts of speech” (Jespersen (1924, Ch. 4),
“word classes” (Davie and Crystal 1967), Subject (Bates and MacWhinney 1982), co-
ordinated and subordinated items (Quirk et al. 1985, §13.6), “noun phrases” and “ad-
jective phrases” as Complement of a verb (Leech and Li 1995). Indeed, according to
Newmeyer (2000, p. 488), virtually all of “the constructs of grammar” have been seen
by someone as prototypes.
Grammar must recognise some situations in which members of a category do not
instantiate it exactly, and this book has identified a couple of them, such as Processes
in present-day English (Chapter 7, §7.2.1), and word classes (this Chapter, §13.2) in
Middle English and perhaps Early Modern English. Those are categories which are or
were in the process of historical change, with some instances following the old pat-
tern, some following the new one, and some that should probably be rated as obscure
or exceptional. However, those categories do not constitute prototypes as understood
usually, or as defined here; rather, there is an overlap between the use of an old cate-
gory and the use of a new one.
13.3.2 Needlessness of prototype theory in English grammar
Prototype theory is simply not needed in English grammar. The reason is that, in
most instances, when necessary distinctions are made, the classes set up are strict
categories. Section 2, on word classes, illustrated that: to understand lexical forms and
relations, and semantic structure, we must distinguish syntactic heads and depend-
ents, morphological paradigms, and certain semantic structures; and when that is
done the whole situation is accounted for, and nothing is left for a theory of prototype
word classes to explain. A second example is provided by the concept of subject. Some
decades ago, it was understood as a prototype, although that term may not have been
used; there is now agreement, I believe, that the situation is resolved when we distin-
guish such elements as morphological, syntactic, and logical subjects, and “discourse
subject” or Topic.
The instances where “fuzzy grammar” does not seem to resolve into strict catego-
ries are as follows. Polysemous senses cannot all be separated rigidly, but have over-
laps, and sometimes have indefinite boundaries: they are not all categorial (Chapter 5,
§5.6.3). Adjuncts seem to vary in how they are related semantically to the rest of the
clause, so that some seem more typical than others (Chapter 5, §5.5.1). Epithets seem
to vary in their semantic relation to other words, so that again some may seem more
typical than others (Chapter 5, §5.5.2). In all three instances, however, the semantic
structure is a network, not one of classes which could be prototypes, as the sections
cited show. Moreover, Adjuncts seem to vary semantically because we use the syntac-
tic term “Adjunct” (which is categorial) while we are thinking of (semantic) Circum-
stances, which are categorial in their (functional) definition, which excludes the issue
of semantic relations. Likewise with Epithets, the features that vary are not part of
the definition (which is by meaning type, not semantic relations), so that the class is
categorial. All falls into place when proper distinctions are made: the linguist’s slogan
should be, “To resolve a contradiction, make a distinction”.
13.3.3 Unworkability of prototypes in grammar
The structure of modification in English Entity groups is a crucial case for prototype
theory. It clearly involves syntactic structure, semantic relations and morphological
behaviour; so that prototypes, which can combine all three of those elements, should
be just right to explain it. However, no-one has been able to explain it that way, and it
is not hard to see why. So-called “adjectives” (defined as prototypes) occur in all four
positions, as we saw with pure (Table 1, in §13.2 above); “verbs” such as sleeping occur
in three positions; and so on. The nature of the structure, as given in Chapter 8, §8.2,
becomes clear only when we distinguish syntax, semantics and morphology from
each other, and when we make further distinctions within each of those fields (as
between the different types of meaning, and different syntactic relations), recognising
that the distinctions apply semi-independently. Prototypes fail in the very case where
they should be most effective: they are unworkable, in practice.
Prototypes are also unworkable in principle. As Fodor notes (as cited by
Armstrong et al. 1983), the overwhelming problem with prototype concepts is that
they cannot be systematic and productive in a theory. In language use, concepts must
be used to make binding decisions rapidly, but that is fairly clearly psychologically
impossible, as was argued in §13.2 on word classes. It also seems logically impossible
to make categorial linguistic choices on the basis of probabilities.
13.3.4 Confusions in acceptance of prototypes
Readers will perhaps find the case against prototypes in grammar more convincing if
they can see why so many scholars have accepted them. This section will outline the
reasons: natural confusions that have occurred, and traditional but false assumptions
that have been made.
Names versus categorial words. In a celebrated experiment, Labov (1973) had his sub-
jects identify various drawn cup-like objects; the resultant classes – “cups”, “mug”,
and “vases” – overlapped, and Labov concluded that they are prototype categories,
with prototype concepts for them. That conclusion is not a necessary one, however.
Since the objects were presented simply as images, with no context to imply function,
use, or other properties, we should conclude that the words which the experiment’s
subjects assigned to them were being used simply as names – words which do not
carry content but refer directly to an object or class of objects. Argumentation such as
Labov’s confuses different types of meaning (naming and describing) with each other,
and meaning with concepts. Names (proper nouns and words used referentially) refer
to things, happenings and qualities without carrying descriptive meaning; the things
referred to will vary, and naturally seem gradient; but gradience of the instances does
not constitute gradience of any concept or of the name’s meaning.
Argumentation such as Labov’s also makes no distinction between categories and
two other kinds of class, although there is a clear and vital difference. Some classes
can be defined abstractly and rigorously; there is no uncertainty about membership or
criterial features; they are categories. Classes such as school or university classes and
the class of Charles Dickens’s novels are strict, being definable rigorously by listing,
though not by concepts; also, they have a reasoned rationale; they are strict classes.
Loose classes are arbitrary in rationale, and may have no fixed membership. The class-
es identified by Labov’s subjects were loose ones, especially since many of them had
shapes the subjects cannot have seen previously.
Polysemous words. Armstrong, Gleitman and Gleitman (1983) investigated “proto-

type classes” also. Their subjects identified grandmothers as a class, and felt that some
instances were more typical or “central” than others, since some grandmothers are
more grandmotherly than others. Clearly, grandmotherly was taken as a word with a
descriptive meaning, whereas grandmothers was taken as a word whose meaning de-
noted a class; the instructions, standard in prototype research, did not prevent them
from confusing the meanings with the things meant. (The authors concluded (1983,
p. 341) that the research into prototypes has been unsound methodically.) That con-
fusion between linguistic meaning and cognitive concept has vitiated discussion of
prototypes. If we are effectively taking the meaning of the word as the members of the
class, then the number of prototypes will be vast.
For example, cliff means <1> “A steep rock face”; cliffs vary in how steep they are,
and often have some quantity of vegetation. In prototype thinking, there will accord-
ingly be a cliff prototype; but “steep rock face” is not gradient in steepness or being
vegetated, since the degree of those qualities is simply not specified. There exist a
somewhat “underspecified” meaning, and a moderately strict class, but no category;
the underspecification of many English word meanings does not create prototypes.
(Wierzbicka (1990, p. 365) makes that point, using “vague” for “underspecified”.)
That confusion, I suggest, lies behind the much discussed “bird prototype”. SOED
specifies that bird means <1> “Nestling (obsolete)”; <2a> “… animal of the class
Aves …”; <2b> “Sport: a gamebird; [<2c>] spec. a partridge.” The classes denoted by
those definitions are for <2a> and <2c>, well defined categories, and for <1> and
<2b>, fairly strict classes; there are no prototypes in the definitions. (The elements
included in the “prototype concept” of a bird are included as encyclopaedic details
in the full sense <2a>: “… characterised by … wings … oviparous reproduction …”).
Subjects in the bird-prototype experiments will have known the biological meaning
(in <2a>), as well as the encyclopaedic elements that appeared in their responses; they
had several “concepts” of ‘bird’: to regard one of them as the concept of bird, i.e. the
only concept, is fallacious.
Cognitive versus linguistic use. The work by Armstrong et al. just discussed has a fur-
ther important implication. Some of the conceptual elements which their subjects
were using are elements of linguistic meaning (as in sense <1>, and the ‘sport’ element
of <2b>); other elements are cognitive, that is, parts of cognitive meaning, and con-
cepts used in both everyday and biological cognition. Concepts, meanings, categories
and prototypes all exist for use, and we should consider what prototypes are used for.
(I accept that there are prototypes in human thinking, including a prototype for birds
along with a categorial concept of birds and several meanings for the word, bird.)
Armstrong et al. showed that people have a prototype concept of odd numbers
(along with the categorial concept, clearly); they evidently use it as a mnemonic short-
cut in handling a concept they find hard to define. Other uses for prototype concepts
appear to be: giving a quick explanation to other people, especially of concepts not
easy to define; quick and approximate recognition of objects; and making quick in-
ferences about things where rigour is not crucial. Also, people use the name of the
prototype, e.g. the word bird, in reference, where the identity of the referent is al-
ready known, and thing or thingummy would also serve. For rigorous thought, such as
drawing reliable influences, we use categorial concepts; for descriptive speech, we use
descriptive concepts with many minor elements that can be evoked. English provides
different senses of ‘bird’ and other “prototypes” for the different uses.
Gradience. It is generally assumed by those who accept that the use of prototypes is
widespread that gradience of features constitutes prototype structure. That is a false
assumption. The concept of a rare disease is gradient, because rarity is gradient, and
such a class could naturally be conceptualised as a prototype; but to the National Insti-
tutes of Health, it is a category, being a disease that affects fewer than 200,000 people
(The Economist, May 23rd 2015, p. 68). Gradient phenomena can be differentiated
into categories by imposing either motivated or arbitrary boundaries. The absolute
distinction between 0 and 1 in computing relies on imposing an arbitrary boundary
on the gradient of voltage which represents them; a current of .5 V or more is inter-
preted as 1, and a current of less than that is interpreted as 0. We have seen the same
thing in the distinction between Epithets and Descriptors (Chapter 8, §8.2): although
they denote a gradient of abstractness, they are distinct in English grammar according
to gradability – where the word may be used either way, speaker simply choose, at
their own discretion, whether to treat the quality as gradable.
That argument applies even to the account of gradience, and of prototypes gener-
ally, given by Aarts (2007), which is much better argued than most accounts. Further-
more, the fundamental flaw, for which he faults Jakobson, is true of him: “he fails to
… question the taxonomy of categories he started out with” (2007, pp. 201–202); he
never justifies the use of such gradient, prototype “categories” as noun and adjective.
Passages illustrating muddled thinking. The confusions just described, and one or two
others, can be illustrated from sample passages.
In the often cited passage on the “family resemblance” of games, Wittgenstein
(as cited by Hanks 2013, p. 325) denies that there is one thing in common among
all games, but only “a complicated network of similarities”; nevertheless, he affirms

that there is one thing in common, in asserting that games form “a family” – a single
family, which unifies them. The contradiction may appear to be a negligible quirk of
Wittgenstein’s choice of words; but the fundamental intention of the passage is to
show that readers’ impression of games’ diversity is rather superficial and wrong; he
relies on words (as in the metaphor “a family”) to present the unity which his argu-
ment is denying. Wittgenstein was confusing meanings and concepts.
Hanks’s exposition of Rosch’s concept of prototypes (Hanks 2013, pp. 340–341)
begins from “perceptually based categories” such as those of red and yellow (page 340,
last paragraph); but they are then (in the next sentence) referred to with the phrase,
“such concepts.” The reference back to “categories” suggests that concepts here simply
stands inaccurately for perceptual categories, not for abstract and general categories,
which concepts usually stands for. However, Hanks goes on (in the same sentence) to
say that such concepts have “central and typical ‘prototypical’ members”, illustrating
the concepts (in the next sentence) from people evaluating the truth of the statement,
“An X is a bird’,” where the immediate significance of bird is a meaning. Hanks plays
on the ambiguity of concept; he equate categories, percepts, concepts and meanings,
without saying so, and without justifying the equation; and we have seen in the previ-
ous paragraphs that those things must be distinguished.
De Deyne et al. (2014, p. 568) provide an example of the same sort of confusion
in explicitly semantic discussion. “Categories of objects … in the real world” (first
sentence of the article) become “words” (second sentence), and then “concepts” (third
sentence), one class of which “are adjectives” (fourth sentence). The article goes on to
demonstrate various things about category structure as graded, having begged the key
question of whether adjectives are or belong to a category, through the confusions in
its first paragraph.1
13.3.5 Conclusion: Prototypes
We conclude that prototypes do not occur in English semantics. The argument here
has concentrated on semantics, but I believe that it has applied widely enough for it
to be accepted for English grammar in general (i.e. the system of language, not the
study). Belief in the widespread occurrence of prototypes in grammar, and in rigorous
thinking of many sorts, has come from overextending their real use (which seems to
be in practical thinking), through a number of confusions (illustrated just above) and
invalid assumptions. I believe that Wierbicka (1990, p. 366) is right in identifying two
of those assumptions as the behaviourist one that “necessary and sufficient features
should correspond to measurable, objectively ascertainable aspects of external reality”,
1. It is striking that, as noted by Aarts (2007, p. 89 footnote), Taylor (2002) referred to catego-
ries such as “BIRD” as “semantic”, but in the second edition changed that to “cognitive” (2003,
p. 202), indicating his awareness of the confusion being described here.
and the more general assumption that we can theorise usefully about the concepts in
language on the basis of empirical work on them which is as yet relatively slight.
13.4 Lexicon
Introduction. It is generally held that there is a store of words in the mind, “the lex-
icon”. This section will bring forward conclusions from earlier parts of the book to
show that some conceptions of it are unsatisfactory – conceptions that are still com-
mon, although not held by all linguists. It will also suggest that the very concept of
“lexicon” as a mental entity is unsatisfactory.
Variety of word structure. In the traditional view, word meanings are treated as units,
with clear boundaries and a strong identity. But we have seen (Chapter 5, §5.2.1) that
that is not so; although word meanings have a cluster-like core that gives them a rough
identity adequate for linguists’ practical purposes, they extend outwards through the
semantic network, with no definite boundary. Second, we have seen in Chapter 12 that
while meaning elements are often lexicalised (assigned word forms which symbolise
it, as when as when PAST becomes the past), they are in a few instances grammat-
icised (realised in a grammatical form, as when as when PAST becomes the signif-
icance of the past tense form). Moreover, lexicalised content words are sometimes
grammaticalised (converted partially or wholly into grammatical items). Theories of
a unitary lexicon do not account for that variety of words, and it is difficult to see that
they could do so. The theories are written as if based on a simplified understanding,
in which using language consists of assembling concepts and assembling the corre-
sponding words.
Relation of semantics to cognition. A further problem with the assumption that word
meanings are units lies in the relation between semantics and cognition. Even in my
account, which distinguished between the cognitive and the linguistic as parts of
meaning, meanings are not unitary, since there are conceptual elements with strong
links to other elements of both semantics and cognition. For example, the opposition
between ‘hot’ and ‘cold’ is conceptual, lying in the concepts HOT and COLD (Murphy
2003, §7.1), which constitute its descriptive meaning; so the antonymy is cognitive,
rather than linguistic. The problem is much greater for theories which equate mean-
ing with conceptual structure (e.g. Jackendoff 2011).
New and re-structured words. The concept of a lexicon relies on the assumption that
there is a fixed list of lexical items or “lexical entries”; see Bierwisch and Schreuder
(1992, §2.1), for example. But as Levin notes (1993, pp. 2–3), English has, as part of its
grammar, processes for producing new words and meanings impromptu, on particu-
lar occasions; she gives the example, “I’ll modem him tomorrow”; conversely, hearers
understand words they have never met before. That fits the repeated observation in
this book that there is an indefinitely large network of potential meaning, some of
which is used regularly and is entrenched as established meanings, but all of which is
available if the speaker chooses to call on it. Indeed, new uses not only fit the network
structure, but require it: they could not occur otherwise.
Storage. The words and meanings are treated as entities that are “stored”; they are “in”
the lexicon, which is a module distinct from the grammar. The nature of language
structure as network makes those metaphors quite misleading, since there is no dis-
tinct part of it that can be marked off as “the lexicon”. Neurolinguists such as Fortescue
(2009; e.g. 175) report that the processing of word-meaning types, and even of differ-
ent nouns and verbs, is widely distributed through the brain; even the sense (and the
syntax, and the sound) of a single word are processed in different parts of the brain,
for production and comprehension. Moreover, the reality of meaning is momentary
activation of neurons in the brain; there is nothing permanent that could be “stored”.
(The analogy with computing is misleading, since computers do have memory loca-
tions.) Finally: since activations of a word sense on different occasions regularly vary
slightly in which elements are activated, meanings hardly have enough identity to be
treated as entities.
Structure. The lexicon is generally thought to have no structure. To Bloom (1933,

p. 274), for example, “The lexicon is really an appendix of the grammar, a list of basic
irregularities.” According to Miller (2014, §1.10), recent authors have believed it to
be “lawless”, or to contain items that are “arbitrary”, or “idiosyncratic/unpredictable”
(2014, p. 23). Those views are associated primarily with formal linguistics, but Func-
tional Discourse Grammar, as in Hengeveld and Mackenzie (2008), holds a similar
view. But we have seen complex structures of lexical relations between word sens-
es, and of structures within each sense – that of meaning types and dimensions. We
have also seen that “features” are not merely listed in a sense, and attached by asso-
ciation; each feature is related as a value of an attribute; for example, ‘female’ is not
merely “tacked” on to ‘ewe’, but is a value of the sense’s attribute, sex. A great deal of
the lexicon shares this structure with cognitive concepts; see, for example, Barsalou
(1992a, p. 251).
Word use as behaviour. When we consider language as it is used, we see it as behav-

iour, occurring in time. Psycholinguistics shows that the processing of language does
not follow a single invariant sequence, but varies the order of steps, and sometimes re-
peats steps in a cycle. Pickering and Branigan (1998, p. 633) reported strong evidence
from many types of study that semantic and syntactic properties of words are accessed
separately from their phonological and morphological properties; but if words con-
sisted of entities in a lexicon, accessing a word would access all its properties. See also
Lamb (1999), for example. The lexicon, however, is traditionally conceptualised as a
module, with a set place in processing; so it misrepresents the reality. Psycholinguis-
tics, also, shows that there is no unity of linguistic function, in the brain. Different
biological and social functions are processed in different lobes; for example, activity
related to personality and social behaviour is processed in the frontal lobe, whether
or not it is linguistic (Pennebaker 2011, Ch. 2). That is the psycholinguistic substance
behind the differentiation of the “lexicon” into five “lexicons”, given by Wray (2002).
That in turn fits very closely with the variety and the independence of the functions
we have seen in language: her lexicon I fits my grammatical system (Chapter 6); her
lexicon II fits descriptive meaning and the referential function; lexicon III fits the
interpersonal function; and lexicon V fits the Expressive function. (Lexicon IV has
memorised formulae, which have not been discussed in this book.)
Conclusion. The grammatical term “lexicon” is derived metaphorically from “lexicon”

as ‘dictionary’, and the concept has not escaped the constraints of the metaphor. We
have seen repeatedly that semantics, along with the other areas of language, is more
accurately represented as a network than as a container, even such a “container” as a
book; see Roelofs (1993) on the lexicon as network, characterised by spreading ac-
tivation. Elman (2009) also presents a network, concluding that there is no lexicon.
We have better ways of explaining remembered meaning. We should do without the
concept of a lexicon.
13.5 Systematisation
This section takes an abstract view of semantic structure, considering the extent to
which English semantic structure is systematic. A system, for this purpose, is a set of
interdependent parts serving a common purpose. To be more systematic, in language,
is to function on more levels, with more choices, or with the choices logically related,
forming paradigms. English semantics has become fairly fully systematised, but cer-
tainly not fully so.
English semantics is not fully systematic in allowing meaning for holophrases
and idioms, which have semantic elements with no organised or operating structure.
Nor is its semiotics systematic. For example, the elements of sound, as perceived by
our hearing, are pitch, loudness, timbre and duration. Music uses all of them as fully
significant. However, linguistic speech uses pitch fairly fully and loudness to some
extent (in stress and intonation contour); but it does not use timbre, and uses duration
only slightly (as an element of stress) – although paralinguistic effects do. The use of
other types of sign is similarly variable.
In some respects, English semantics is quite fully systematic. In the basic struc-
ture of ideational semantics, for example, each stratum of expression simply realises
the one below – lexis realises the conceptual meaning, syntax realises the lexis, and so
on. But in the details of its structure, English semantics uses each stratum twice over,
as we saw in Chapters 11 and 12. For example, syntax carries meaning of its own, as
well as realising the lexically carried meaning.
Some of the strata are subdivided into layers. Lexis has layers of descriptive,
emotive and other meaning types. Syntax has layers for morphemes, words, groups
and figures; pidgin languages, by contrast, usually have only two layers, those of the
word and the figure. Those layers – and the strata – sometimes work in parallel, re-
inforcing or complementing each other; in one respect, however, they are restruc-
tured, the layers of morphosyntax becoming hierarchic, which is crucial to semantic
expressiveness.
English semantics, then, is fairly fully systematised, and in some ways remarkably
so, using devices that do not appear to be used in any other natural system, such as
recursiveness and doubling a form’s functionality through marked use. The benefits
are greater precision, clarity, economy, learnability and expressiveness.
13.6 Minor topics for discussion
13.6.1 Semiotics
The book has shown, I suggest, that the importance of semiotics has been seriously
underrated in the study of semantics. Lack of awareness of semiotics has brought with
it identification of meaning with descriptive meaning (i.e. concepts, what can be de-
fined in a dictionary entry), which is carried by symbols; but we have seen repeatedly
that indexical and iconic signs are important in English semantics. Moreover, many
symbols carry non-descriptive meaning, such as emotive and social meaning, and
other symbols do not so much carry “meaning” as have a function: description is only
one of the three semiotic processes (§4.3.2).
Of the other two functions, pointing or “deixis” has been studied a good deal,
though often in isolation from the study of meaning. But we have seen that nam-
ing, by contrast, has been misunderstood by confusion with the use of proper nouns.
Similarly its importance in the following topics has been missed: the use of acronyms
(§3.4.3.3); the development of meaning, both historically and in children (§4.2.1); the
use of common nouns (§4.4.1); the structure of senses (§5.2.3.2); and the nature of
realisation (§12.2.2.6). Finally, unawareness of the role of naming has contributed to
the confusion about prototypes (§13.3).
13.6.2 Distinction between descriptive and referential use
As with the distinction between semiotic processes, the distinction between descrip-
tive and referential use of meanings has been underrated. We have seen it to be rele-
vant to the following: the structure of individual senses (§5.2.3.1); dictionary entries
(§5.2.3.2); the development of meaning, historically and in children (§5.2.5); meta-
phor (§5.3.4.4); the distinction between nouns and pronouns (§8.2.9.2); and modals
like fake (§8.2.10.4). It is necessary to understanding the anomaly that proper nouns
can be modified, resolving the controversy concerning holistic and compositional
meaning, and reconciling polysemy and monosemy (all in §4.7.3). Finally, it is vital
to understanding the fundamental issues of how language is grounded in experience

(§9.3 .3 .2), and the nature of “speech acts” and “sub-acts” (§11.5.6.2).
13.6.3 Distinction between grammatical and content meaning
The distinction between “grammatical items” and “content items” is always recog-
nised, but the distinction has been poorly understood. The book has shown that the
distinction is not between classes of items, but between types of meaning which com-
monly occur together in a single item. Content meaning has been shown to be wider
than conceptual meaning, with which it is usually identified, and to be functional, in
that it is often a means to the end of identifying a referent, without being of value in it-
self. Grammatical meaning has been shown to be wider than that of “operators”; while
it can usefully be represented as grammatical status such as headship or dependency,
it can also be usefully described as procedural – it may even amount to a request such
as “Please wait, since I haven’t finished yet” (§2.5.3). It has its own types (§4.5.2) and
dimensions (§4.5.3), and it creates information structure, as well as the semantic and
syntactic structure of groups and clauses.
Further, the importance of the distinction has been underestimated, as with the
distinctions discussed above. It is ubiquitous: every word in an utterance either car-
ries grammatical meaning or is subject to it (because of its position, if not because
of the word itself); and it occurs at all levels, not just at the lexical level as the term
“grammatical item” implies. Grammatical meaning controls not only the relation of
concepts, but all grammatical status, and all grammatical relations; without it, even
full explicit statement of content is meaningless (see §8.7, for example). Hearers use
it to build semantic and syntactic structure; those structures are not inherent in the
intended meaning.
13.6.4 Distinction between cognitive and linguistic areas of meaning
The difference between cognitive and linguistic areas of meaning is another dis-
tinction whose importance the book has emphasised. It helps explain the stages of
comprehension (§4.5.5.1). It has been needed to resolve problems concerning sense
relations such as antonymy (§5.3.2), proper nouns (§4.2.1), metaphor, uncertainty of
interpretation (e.g. with adjuncts §5.5.1.1), constructional meaning (§7.8.2, §8.2.2.4),
and the distinction between Descriptors and Epithets (§8.2.4.4). It is crucial to under-
standing the problems with prototypes (§13.3). The distinction is fundamental: it un-
derlies the distinction between content and grammatical meaning; and the cognitive
area correlates with the ideational function and syntactic structure, while the linguis-
tic area correlates with the interpersonal function, and with information structure.
13.6.5 Formalisation of linguistic description
Most semantic work has been oriented either to formalisation – following a logical
or mathematical tradition, or aiming at computerisation of language – or to descrip-
tion – including the cognitive and systemic traditions. This book has followed the
descriptive tradition, but its analysis seems to provide for formalisation of mean-
ing to a considerable extent. That was suggested explicitly for the types and dimen-
sions of meaning in §5.2.3.4, but it applies more widely, because of the nature of the
analysis given.
Meaning has been analysed into units which are defined by their type (§4.4 and
§4.5), dimensions (§4.3), and semantic class (§4.6). The units, in utterances, draw on
a particular area of meaning (§4.2), are employed in particular uses ((§4.7), and have
particular functions (§3.3.1); they build specific structures (Chapters 5–10). Crucial-
ly, those descriptive categories are categorial, forming paradigmatic alternatives; it has
been shown that even gradients may be divided categorially, as with gradability in De-
scriptors and Epithets (§8.2), just as electronic systems create the distinction between
0 and 1 out of the gradient between full voltage and no voltage. Being categorial and
paradigmatic lets them be formalised.
13.6.6 Concept of classes
Linguists’ faulty use of classes for analysis has been a recurring theme of the book. The
“classes” of grammatical and content “items” has confused how meaning functions.
The class of “nominalisations” has confused the origin of those senses with their actu-
al meaning and function. The “parts of speech” have classified words or meanings by
three or more criteria at once, like classifying furniture simultaneously by their ma-
terial, their function and their origin. “Prototype” categories have confused referents,
concepts and meanings.
Using classes such as nominalisations has a less obvious disadvantage. It com-
monly obscures how change occurs, since it is hard to see how something can switch
from belonging in one category to belonging in another; if we speak of words or sen-
tences with certain properties and uses, then we can describe change easily as a grad-
ual addition or loss of the various properties and uses. Similarly, using classes such as
“adjective” or “describing word” obscures the wide range of functions and properties
that we saw in the description of premodifiers in Entity groups (§8.2).
We need to use one criterion at a time in classification, keeping the class quite
simple, and to distinguish the things we are classifying from their functions and uses.
13.6.7 Philosophical tradition in linguistics
The abuse of classes just discussed is one instance of what this book has represented
as another recurring problem in linguistics, the inappropriate use of philosophical
traditions. It has brought too close a connection between meaning and knowledge
(and sometimes even identification of the two), with linguistic differences identified
with real-world differences, and a serious under-evaluation of the role of construal
(§12.6.3); utterances have been taken to be essentially propositions. The problem is
not only that conceptual meaning, and the relations among language, cognition and
the external world have been misunderstood, but that other important things in lan-
guage have been ignored. They include non-conceptual meaning, distinctions in the
use of language (such as descriptive/referential and literal/figurative), and linguistic
function, particularly interpersonal and social function. That refusal to consider lan-
guage as functional has, according to the argument of this book, caused fundamental
distortion of semantics.
Chapter 14
Conclusion
14.1 Introduction
Primarily, this chapter sums up what the book has to say about the semantic structure
of English, and draws some conclusions on the related matters of the nature of mean-
ing and the stratification of meaning. Secondarily, the chapter suggests what has been
added to our understanding of semantics, and suggests areas for further research.
There is no chapter-by-chapter summary, since the substance of the book lies in the
detailed analysis, not in any overall argument.
14.2 Nature of meaning
14.2.1 Introduction
This section sets out to draw conclusions about the nature of meaning, from the body
of the book, not to provide a general theory of meaning.
A murmured “Mmmm” or “ah” is meaningful, but neither the form nor the mean-
ing is well-defined, so there is no distinct linguistic sign. Rather, each such utterance
is a symptom of something in the speaker’s mind. Symptoms work in patterns, and
are very dependent on situational context; the “ah” is to be interpreted along with
facial expression and sometimes with gesture. Meaning of this kind may be called
“symptomatic meaning”. Other meanings are indeed carried by signs; the interpreta-
tion may be affected by context, but the signs have core meaning independent of any
context. They are highly conventionalised and are generally arbitrary in synchronic
use. Since they are produced by sign systems, they will be called “semiotic meaning”.
The relation between semiotic and symptomatic meanings is that the latter are car-
ried by some feature of how people use the sign system, not by the sign system itself;
for example, “mmmm” uses uses the [m] sound independently instead of as part of
the phoneme system. If a restricted use of the words “semantic” and “semantics” is
wanted, both words could be restricted to semiotic meaning; the word “meaning”
could be restricted by equating it with “semantics”, and using “significance” as a more
general term.
14.2.2 Symptomatic meaning
Language as Expression. With “mmmm” and “ah”, it is the fact that we say something
that counts, rather than the exact nature of what we say. Even “hi” and “ta-ta”, which
are established words, are not symbolic signs. In both cases, the meaning is the inten-
tion, such as being attentive, being sociable, and giving a farewell; they are instances
of language as Expression, not of language as communication.
Another form of language as Expression is illustrated in example (1). It is from
a man’s narrative of spending two winter days unable to move from a ledge on a
sheer cliff.
(1) “Everything was blurring… Sleep was out of the question… and food wasn’t
an issue because there wasn’t any. It was about trying to stay warm.”
(New Zealand Herald, June 30th 2014, A3)
The intended meaning lies below the meaning of the words; it is the experience which
prompted them, and which we should recreate mentally, and share vicariously. In all
these instances, the meaning is the intention, or the experience; it is highly dependent
on the circumstances; it is centred on the speaker.
Language as action. With utterances such as Lincoln’s Gettysburg speech cited earlier
(§10.5.3), and with the short passage from Shakespeare’s Antony and Cleopatra, also
cited earlier (“Age cannot wither her…”; §5.6.2), even the meaning is a means to an
end, which is the hearer’s mental response. The same is true in a number of other situ-
ations, such as that of humour, and of nursery rhymes. The significance (the response)
is evoked, rather than carried, by the style of the language – that is, by characteristics
of the way the linguistic signs are used, particularly their imaginative qualities. Here,
language is stimulus – and a study of it easily shifts from linguistics to literary studies.
Speakers often signal that they want the hearer to respond, by ending a clause
with the rising intonation and slowing down which we call “trailing off ”, then pausing,
and looking at the hearer in a certain way. The meaning maybe seen as the speaker’s
intention, but may also be seen as the hearer’s response. Here, language is interaction.
Study of it becomes conversation analysis or discourse analysis when the signals are
relatively conventionalised, and are as much social as linguistic.
With the “speech acts” of promises, the naming of a ship, declaring two people
to be married, and so on, other people can act in reliance on the “act”; its meaning
(though not its linguistic meaning) lies in the changed state of the world. Language
here is social action. Study of it is often outside linguistics proper, lying in pragmatics
(according to how that term is understood); it becomes linguistic when the speech
acts are fully conventionalised and are represented by the purely linguistic signs
which constitute such constructions as the imperative and indicative.
This form of meaning – that of language as action – is dependent on social con-
text. It is centred on the hearer. As with the speaker-centred meaning discussed in the
Chapter 14. Conclusion 409
previous section, there is no significant system meaning, simply because these forms
of meaning are not systematic.
14.2.3 Semiotic meaning
We have defined semiotic meaning as meaning carried by intentional signs, not in-
dexes or signals. It is conventionalised, and usually arbitrary in synchronic use. (Ide-
ophones and other sound symbolism are examples of semiotic meaning which is not
arbitrary.) It is functional; that is, it has a result, and it operates in a system. Being func-
tional, it is real socially, not theoretical and existing in a static world of its own. Nor
is it only conceptual, but includes emotion and guidance to the hearer. This section
outlines its nature – but not its structure, which is dealt with in the following section.
Semiotic meaning is abstract in varying degrees. We have seen that in the origin
of descriptive meaning in perception, and in the degrees of abstraction that distin-
guish the semantic classes, and distinguish grammatical from content meaning.
Semiotic meaning is very varied. We have seen a fundamental distinction be-
tween the sort of meaning that makes up what is conveyed (“content meaning”), and
the meaning which sets out the relations and use of the items conveyed (“grammatical
meaning”). The content meaning varies in type, and in the dimensions on which it
operates and which define its senses. Grammatical meaning varies in the level on
which it operates, and in its mode of operation. All of that was set out systematically
in Chapters 4 and 6, and has been illustrated in subsequent chapters.
Semiotic meaning is unconstrained, although its expression is subject to con-
straints (see Chapter 3, §3.4). It can be derived, it seems, from any mental faculty, such
as those of cognition, affect, and conation. There is no apparent limit to the types of
relation that can hold within it, or to its internal structure. It is not fixed in the way it
is used, but varies in the descriptive/restrictive alternation, the figurative/literal alter-
nation, and the marked/unmarked alternation (Chapter 4, §4.7). The marked use is
particularly remarkable, I suggest, because in it language overturns its own principles,
providing the rule that you may break any rule.
Meaning is distinct from other products of mind to varying degrees. We have
seen repeatedly that it is distinct from knowledge, although it overlaps it (see Chap-
ter 4, §4.2 for direct exposition); the conceptual elements that make up descriptive
meaning are not distinct from those of knowledge, but become linguistic by being
incorporated into linguistic structures. At the other end of the scale of cognitive de-
pendence (Gentner and Boroditsky 2001; Chapter 4, §4.2.1, above) are grammatical
items such as the copula is and wholly grammatical meanings such as definiteness
of reference. There are meanings at very many points along that scale. In being thus
distinct from other mental products, meaning is autonomous, to an important degree.
There are other things that could be said about the nature of semiotic meaning,
such as the “design features” of “displacement” of meaning; but, as noted above, this
section deals only with features that have been important in the preceding chapters.
14.3 Nature of semantic structure
14.3.1 Introduction
In the book so far, the discussion of semantic structure has been arranged to suit read-
ers’ familiarity with ideas, and in accordance with the various structures’ importance
in use. Here, it will be summarised systemically, according to structural type.
The main distinctions being made here are not used in traditional semantics, but
have been developed through the book. At the top level, the distinction is in the view
taken: first, the structure will be outlined analytically, as parts and wholes, in static,
non-functional relationships that suit diagrammatic presentation – as trees, for ex-
ample (as in Chapters 5 and 7–10); then the structure will be outlined functionally,
as a system with parts that work together dynamically and in sequence, according
to how people use them (as in Chapter 6). The analytic view is presented in §14.2.2
and §14.2.3 below; the functional view is presented in §14.2.4. At the second level, a
distinction is made between structures that consist of units such as words and groups
(§14.2.2), and and those that subsist in a medium, and consist of variation within
the medium, i.e. waves and fields (§14.2.3). The two views are complementary, each
having something of value; the analytic view fits traditional linguistics better, and the
functional view fits the exposition in this book better.
14.3.2 Analytical view of semantic structure (1): Structures of units
Structures with one or two units. The simplest semantic structures are those with mul-
tiple functions but no definite internal structure at all – holophrases and ideophones
(Chapter 10, §10.3.2, §10.3.3). They evoke varying and sometimes numerous mean-
ings, which the speaker has not structured; hearers may or may not impose a structure
of their own.
Structures with two units are relations, consisting of two unitary nodes and the
link between them. Clearly, they are the constituents of hierarchies and networks –
most of the structures to be dealt with in this section. Relations of a more abstract
kind can be seen in waves (relation between peak and trough) and in fields (rela-
tion between the source of energy, e.g. a magnet or a social meaning, and the thing
influenced).
Structures with three or more units: (1) hierarchies. Hierarchies, such as taxonomies
and standard organisational structures, have an inherent internal structure, with a
top and a bottom, which constitute entry points for understanding the pattern. That
contrasts with networks, which have no set entry points, and no top or bottom. The
most prominent hierarchy in semantics is that of the figure, group and sense (Chap-
ters 7 and 8); second comes information structure (Chapter 9). Linguistic hierarchies
consist of classes (if we consider structure in the abstract) or of members of particular
classes (in particular utterances), with a status according to their level.
Semantic hierarchies have four organising principles. First and dominantly, they
are organised with heads and dependents, creating the layers. Second, the units on
each layer are not simply equally dependent on a unit in the layer above, but togeth-
er make a syntagm, such as a figure, a group, and a word (when the morphology is
transparent, as with millimetre, centimetre and kilometre). In the unmarked structure,
the units within a syntagm have complementary functions – those of Participant and
Process in the figures, for example, and of Epithet, Descriptor and Classifier in Entity
groups. In a marked structure, the units have the same function, so are coordinate – as
with the figures in two main clauses, and two heads in a Participant e.g. “Jack and Jill”.
(That distinction between coordinate and complementary is distinctive of linguistic
hierarchies.) Third, each unit in the syntagm is an item in a paradigm – either of class-
es, such as quantifiable, countable and abstract Entities, or of instances, such as ‘book’,
‘magazine’ and ‘pamphlet’ within the class of countable Entities. Finally, the whole of
the syntagm constituting a layer or “rank” may in marked use be used on a lower rank,
as when a rankshifted figure is used as a Participant (Subject or Object) in another
figure – a “noun clause”. (That is another distinctive feature of linguistic hierarchies.)
The structure and its organising principles are the same as those usually taken
as building the syntactic hierarchy. The relation between the syntactic and semantic
hierarchies will be considered further in §14.3.5 below.
The layers in linguistic hierarchies are different enough to be marked with the dis-
tinct term, “strata”; they differ in constituents and internal relationships, and are levels
of realisation (see Chapters 11 and 12). The dominant characteristics of the linguistic
strata are the correlation between their units, and the contrast in their substance. Sense
units in the semantic stratum correlate with word forms in the morphosyntactic stra-
tum, and with morphemes (as structures of phonemes) in the phonological stratum;
their nature is contrasting, ranging from an unordered conceptual nature, through
conceptual order, to the physical nature of audible speech. (Further discussion is de-
ferred to §14.4 below, since there is considerable detail, and it is controversial.)
Structures with three or more units: (2) networks. As already noted, linguistic networks
have no fixed entry points, and are not only interconnected but inter-communicating.
They are unlike hierarchies in that respect, but like them in consisting of relations,
fundamentally.
The network nature of English semantics is crucial to it. Since semantics is rela-
tional, there are no tightly defined units whose combination will define the identity of
a larger unit; the larger unit is defined by its dimensional structure, just as a physical
shape is defined by its dimensions of length, breadth and height. (Consequently, using
the term “class” can easily be misleading.)
A second outstanding characteristic is the number and complexity of the dimen-
sions. Let us note some of the instances that have been presented in earlier chapters.
First, there are the dimensions of meaning (Chapter 4, §4.3), such as specificity and
vagueness – to which the term “dimension” has been largely restricted. They have
been seen most often in the descriptive element of word senses, but they occur also
grammatical meaning, and in the meaning of phonological structures; and attitudinal

and social meaning have their own sets of dimensions. Some dimensions are very
general, as with intentions, where the secondary intentions form a separate dimension
which modifies the primary dimension (as with the intentions of being precise and
unambiguous modifying the primary intention of conveying information). Others are
more specific, as in information structure (relevance, orientation and salience), and
in the uses of semantic forms (descriptive/restrictive, literal/figurative, and marked/
unmarked). In being dimensional, semantics is thus like other areas of language: di-
mensions define consonant sounds (place of articulation and so on) in phonetics, and
the dimensions in phonology (pitch, volume and duration) define tonicity and so on.
Most networks exist in one dimension, including literal networks (i.e. nets) and
those of roads and friendships; the semantic network is very much more complex.
For descriptive meaning in particular, nine dimensions were listed (Chapter 4, §4.3),
some with several values and some scalar; that set of dimensions has some corre-
lations with the dimensions in other structures, such as those of intention and of
information structure.
14.3.3 Analytical view of semantic structure (2): Structures in a medium
Wave structure. When we analyse structure, we expect to find parts and wholes. But,
as we have seen, phonological patterns are structured as rise and fall in pitch and
stress, and its meanings rise and fall also, in degree of feeling or importance, in a wave
structure (Chapter 10, §10.5).
We have seen that the rhythm system varies in degree of regularity and speed,
embodying varying degrees of emotion and of imaginative excitement. The system of
tonality varies in degree of pitch over the whole intonation unit, embodying degrees
of salience and information structure, chiefly. Tone varies in degree of pitch over the
tonic syllable, embodying chiefly guidance of the hearer’s response – to statement,
question and so on. The tonicity system varies the degree of stress, signalling degree
of salience chiefly.
Field structure. Semantic field structures consist of an area of meaning in which one
unit gives its character to all others, just as gravitational fields influence everything
within them. The two forms we have noted (Chapter 10, §10.4) are those of emotive
meaning and social meaning. For example, colloquial syntax such as ellipsis, and col-
loquial lexis such as gidday, make the whole utterance colloquial.
14.3.4 Functional view of semantic structure
The preceding sections viewed semantic structure analytically, as static structures

which may aptly be represented in two-dimensional spatial diagrams; but language
exists in time not space, as human actions, not objects, and with motivating energy
and function. Seen functionally, it is a system. That is, it is like a computer system or
the human nutrition system, with input, processing and output. In outline, intentions
and content meanings are the input to the semantic system, the building of struc-
tures in accordance with grammatical meanings is the processing, and the meaning
of utterances and texts is the output. (There is a corresponding system for the hearer’s
understanding, of course.)
The system has been presented conceptually (rather than psycholinguistically
or as “processing”) in Chapter 6. It is summarised here, from the speaker point of
view. There are several stages, each using input from a previous one. In the first main
stage, content word senses are produced, with input from such primary intentions as
giving information (descriptive meaning) and conveying feeling (emotive meaning),
and grammatical word senses are produced from the secondary intention of guiding
the composition of content; simultaneously, non-descriptive senses are related to the
descriptive ones, and the whole word senses are linked to word forms. In the second
stage, words are processed into groups with more input from intentions, and with the
inclusion of grammatical signs other than grammatical words, such as word order.
Then groups are processed into figures, as structures of morphosyntactically defined
units. In a parallel operation, the intention to help hearers structure the content and
relate it to existing knowledge motivate the building of information structure, as a
structure of content units. Sometimes, later processes build larger structures – those
of the figure complex, and the paratone in speech and the paragraph in writing.
14.3.5 Conclusion: Nature of semantic structure
Semantic structure in English is very diverse; structures overlap, and seem inconsist-
ent (as with hierarchy and network). That has resulted, I believe, from its evolutionary
history, where – as with the evolution of other human systems – structures have been
turned to new functions, and new needs have produced new and different structures.
But, although it has no overall character or unity of structure, its variety and flexibility
seem to make it very well adapted to human needs.
14.4 Stratification of the semantic structure
14.4.1 Introduction
The semantic structure outlined in §14.2 is realised in the strata of English, so the
relationship between semantic structure and the strata is a natural topic for discus-
sion. There are further reasons for discussing it, however. The strata all have their own
structure, which is necessarily semantic as well as morphological, syntactic or phono-
logical, and there are complex (semantic) relations among the strata. Those structures
and relations form the subject of this section.
The concept of stratum has been taken for granted so far, with illustrative detail
(especially in Chapters 11 and 12) replacing definition. It is time that it was given at-
tention. As a specifically linguistic concept, it was developed in the 1960’s by Hockett,
Gleason and Lamb independently (Lamb 1966), and developed in somewhat various
ways. I take a stratum to be one of the layers in which language is realised from in-
tended meaning to physical utterance, having a kind of unit and a kind of structure
of its own. In a simple instance, it is fully dependent on the stratum below, being its
conversion into a different medium or form.
There are many things in English which may reasonably be called layers or levels,
and it is not obvious which of them should be considered strata; moreover, there is no
consensus in the literature on what are strata. Accordingly, the rest of this section will
consider likely candidates in turn. As we will see, there are no categorial distinctions
to be made, but the discussion should be useful as a summary of parts of the book,
and for what it shows about the semantic structure of English.
14.4.2 Semantics
Chapter 4 introduced the distinction between the cognitive and the linguistic in se-
mantics, portraying them as two areas, with an overlapping area consisting of con-
cepts which are used by both cognition and language. That understanding has been
developed in subsequent chapters, to include the following. Cognition is much wider
than concepts, including images and perceptions as well; it is a processing faculty
independent of language. The concepts in the overlap area mediate between language
and cognition: they are related to the rest of the semantic stratum by linguistic links,
but related to the rest of cognition by links of knowledge and logic; and they are used
in both linguistic and cognitive processes, which are distinct.
The distinction between the cognitive and the linguistic on the semantic level is
reinforced by the facts that semantics draws on speaker intention, as well as on cogni-
tion, and that cognitive concepts do not correlate one-to-one with semantic concepts,
as when there are several synonyms for one cognitive concept.
We conclude that the cognitive semantics and the linguistic semantics referred to
throughout the book are distinct in principle, while closely related. Linguistic seman-
tics is therefore the fundamental stratum of English. The following sections will show
that it differs from the levels above, and confirm that it is in fact a stratum distinct
from them. It is sublexical, consisting of indeterminate sense elements; its nature has
been set out above, particularly in Chapter 5.
14.4.3 Lexical items
Words as used in utterances have been presented throughout the book as symbolising
word senses, which are built from sublexical sense elements, partly by making them
determinate and partly by combining them. Words, as forms which can be realised
as spoken sounds or as written letters, exist in a different form from purely abstract
semantic elements, and have units of a different nature. Semantically, they are struc-
tured differently as well, since their relationships are the tightly structured, fairly well
defined relationships synonymy, antonymy, and so on, which differ from the relation-
ships among sublexical sense elements, some of which are loose and associative, while
others are tight and logical.
Making the senses determinate changes them appreciably (see Chapter 8, §8.5,
and Chapter 12, §12.2.2). They become complex in structure, and sense elements be-
longing to the cognitive classes of object, quality and happening become the senses
belonging to the semantic classes of Entity, Property and Event. Non-descriptive ele-
ments are now included in them.
In all those ways, the senses of lexical items are elements of a different stratum.
I will refer to it as the “lexical stratum”. (It is roughly equivalent to the “lexicon”, dis-
cussed in Chapter 13, §13.3.)
14.4.4 Morphology
Morphology does not fit straightforwardly into the pattern in which one stratum is
realised simply and directly into the next stratum, because of the familiar fact that
English provides, for some meanings, a choice between synthetic expression through
morphological inflection, and analytic expression through the lexicon.
Realisationally, morphology has some characteristics of a stratum. It realises se-
mantics; it feeds into the stratum above, i.e. syntax; and it has its own structure, that
of root and affix. On the other hand, it does not form a distinct kind of semantic
structure, since affixes may fairly be said to modify the root as head. Morphology is
thus more accurately described as a “complexity level”; to an important extent, that is
a better term for it than “stratum”.
We conclude that morphology has some characteristics of a stratum, but that,
when related to syntax and the lexicon, it does not qualify as a stratum fully – it is a
mezzanine floor, as it were. It will be combined hereafter with syntax, as morphosyntax.
14.4.5 Morphosyntax
Introductory note. Here, as in previous parts of the book, comments on morphosyntax

apply to the syntax of formal English, and often do not apply to informal English. The
level of sentence and figure complex are passed over, as not involving other issues
important to the present purpose.
Meaning in the morphosyntactic stratum. In part, morphosyntax simply realises what

the speaker has put in the lexical stratum, without changing the meaning. That con-
sists, at group level, of putting the lexical items in order, and giving them absolute
position such as first or last, or relative position such as before or after the head. At the
figure level, it consists of ordering the groups.
In part, however, syntax changes the meaning of what it puts in order. Assigning
a word or group to a position gives it grammatical meaning, generally speaking. In
system terms, that is setting status, such as being head (semantically or syntactically),
or being a Participant semantically, or a Subject syntactically. In speaker terms, it is
inserting an instruction to the hearer on how to relate the items in composing the
full meaning. Thus, content meaning, realised from semantics through the lexicon,
and grammatical meaning, realised from the speaker’s intentions, are now combined.
Finally, any prepositions inserted add cognitive relations such as those of place and
time, as well as grammatical meanings.
The new units formed – words and groups as structural elements – belong to
new semantic classes: Entities, Properties and Events at the group level, and Partic-
ipants, Circumstances and Processes at the figure level, and the corresponding syn-
tactic classes. Those semantic classes do not account neatly for prepositions, which
therefore seem anomalous, just as prepositional phrases are anomalous in syntax, in
being Entity groups substantially but not wholly. The apparent anomaly results from
equating the epistemological view of knowledge as consisting of things, happenings
and qualities with the semantic classes of Entities, Event and Properties. We must see
the semantic level as sublexical, with sense elements which may be realised various-
ly; for instance, the element that can be indicated as “CONTAIN” does not correlate
one-to-one with any word, and may be realised in the Event ‘to contain’, the Entities
‘contents’ and ‘absorb’, and the prepositional meaning, ‘in’.
All of those changes carry meaning; the meaning of the input stratum is trans-
formed; what was, at that lexical level, a mere bunch of unordered words is now effec-
tively an utterance. Indeed, one might say that real meaning has appeared, where only
the potential for it existed in the lexical input.
Semantic structure in the morphosyntactic stratum. As well as having extra meaning,

the syntactic form has different semantic structure. As explained earlier (see Chap-
ters 7 and 8), syntax is structured as a hierarchy, not as a loose network like the lexical
stratum. However, the differences go further.
The structure is constrained by the linear nature of language (see Chapter 3, §3.4),
where semantics and the lexicon are not constrained in that way. Further, there is
duality of structure, in two respects. First, the structure set out so far is the unmarked
structure; most elements of it can be inverted in some way, in one or more marked
structures with a quite different and generally incompatible meaning. Second, content
and grammatical meanings alternate, as illustrated in example (2). It uses boldface for
words with content meaning and italics for words with grammatical meaning; “new
figure” represents the grammatical meaning of the comma.
(2) “They arrived at dawn, [new figure] strapped with explosives and armed with
AK-47s.” (New Zealand Herald, April 4th 2015, B1)
Third, there is an intangible structure of alternating overt and covert signs – of

visible words as signs for content meaning, and invisible (usually syntactic) signs for
grammatical meaning. That alternation is shown in example (3), which sets out the
Entity group, “a quite different and generally incompatible meaning”. In that group,
the position of different is a covert sign that the word modifies meaning, and the po-
sition of quite is a sign that it intensifies different. Those covert grammatical signs are
represented in the example by italics in square brackets, following the overt signs to
which they apply. (The sign itself is not spelled out as “position of the word”, etc.) The
example is hard to make sense of; but I intend that difficulty to make a point, stated
below the example. The first part of the example is to be read as: “In reading ‘a quite
different…’, you must use a to determine the headword i.e. meaning; you must use
quite to intensify the following word i.e. different.”
(3) “A [determine “meaning”] quite [intensify “different”] different [modify

“meaning”] and [link “different” and “incompatible”] generally [qualify
“incompatible”] incompatible [modify “meaning”] meaning [make that the
head]”. (Invented)
The point being made is that the morphosyntactic stratum of an English phrase does
not simply realise its content meaning – and that even the conception of it as “compo-
sitional”, according to general rules, is seriously misleading.
Discussion of the morphosyntactic stratum. There has been a great deal of discussion
of the “autonomy” of syntax; but it is not an issue here, because it arose from assump-
tions made in generative grammar which are not made here; besides, the concept
is confused, I believe. However, it is worth making a couple of comments, because
some of the issues are relevant, and because readers may want to know what view is
taken here. Syntax as described here is autonomous to the extent that it has structure
or “rules” that are distinct from those of all other strata. On the other hand, it is not
autonomous, to the extent that it is dependent – dependent on semantics and speaker
intention – in the sense that it exists to realise them. That view is common in func-
tional linguistics e.g. Boye and Harder (2009, p. 39) and Allan (1986, p. 391); as far
back as 1966, Weinreich – from a rather different school –saw the relation of seman-
tics and syntax as “deep interpenetration” (1966, §4.1).
Some schools of linguistics put the lexicon and syntax together as a single stra-
tum, “lexicogrammar”, which is a continuum having the lexicon as its most “deli-
cate” level of analysis. Examples include Systemic Functional Grammar, e.g. Halliday
(2014), and some construction grammar e.g. Croft (2001, p. 75); Langacker (1987a,
p. 3) and Murphy (2003, p. 240) have the same view. That concept of is suited to Sys-
temic Functional Grammar’s top-down analysis of broad systems such as modality
and transitivity, but is not applicable in this analysis of strata, which is specifically
semantic, rather than generally grammatical.
The relationship between syntax and semantics has been given very extensive
discussion in the literature without consensus as to what it is, as in the series, “Syntax
and Semantics” (Academic Press), with over 30 volumes. The problem, I believe, has
been that writers have not had, and used, a satisfactory understanding of the strata of
English. In particular, in most work in the literature, semantics is not distinct from
knowledge; it is conceptual, with neither nondescriptive meaning nor grammatical
meaning included as semantic realities. Consequently, key general concepts are treat-
ed as simply syntactic, although they include elements of meaning; examples are the
concepts of grammatical relations such as transitivity, and of semantic roles (or “the-
ta roles” and so on), and of direct/indirect objects. Those concepts entail meaning,
because they entail conceptual relations that underlie the relations of the words as
word forms.
To outline a resolution of the confusion, I will begin with the apparent duality of
structure referred to above, in §14.2.2, where it was noted that there are apparently
parallel hierarchic structures, namely those of syntax and semantics. We may under-
stand that as being because the syntactic structure realises the semantic structure; in
that view, the syntactic structure is a distinct stratum, “above” semantic structure. On
the other hand, we may see the apparent duality as two views of a single thing, whose
substance is content – semantic – and whose organisation or structure is syntagmatic
(“syntactic”).
This view narrows down the usual meaning of “syntactic”. It is now generally
agreed, I believe, even among linguists in the generative tradition, that the original
Chomskyan analysis over-extended “syntax”. There, syntax was almost equivalent to
structure – all linguistic structure outside phonology – including (especially in “deep
structure”) elements that are better seen as semantic, since they consist of conceptual
relations. I suggest that we should go further in reducing syntax than the consensus
view does, seeing syntax as dealing with only the arrangement of word forms, rather
than with word meaning. The arrangement of word forms consists of their presence
or absence, their grouping (as evidenced by being omitted together and moved to-
gether as units), and their order. The relations among word meanings are those of
dependence, coordination and complementation; those relations therefore belong
in the semantic domain. Syntax concerns the syntagmatic relations of word forms;
the syntagmatic relations of meanings is part of semantics. As Matthews says (1981,
p. 124), “syntax has its basis in codification of semantic relationships”; Givón (1988,
p. 278) and Harder (1996, p. 95) support that.
Conclusion: morphosyntactic stratum. Morphosyntax is a stratum in every respect. It

realises word senses, in giving them form, directly into syntax, or indirectly through
morphology. It realises some intentions directly, in the form of grammatical meaning.
Syntax has two layers – those of the group and of the figure/clause – which are like
strata in that one realises the other, but unlike strata in that they have the same kind of
constituent and of essential structure; they accordingly should not be rated as strata.
14.4.6 Information structure
Information has been represented in this book (see Chapter 10 in particular) as a

structure of content with chunks of content per se as the units; in that respect, it par-
allels syntax, which is a structure of content as syntactic units. It has been analysed as
having three dimensions, those of relevance (with Topic and Comment), of orienta-
tion (with Theme), and of salience (with Rheme).
Information structure has a number of the characteristics of strata. It realises the
interpersonal function of guiding hearers in relating the content to the rest of the pas-
sage, and to what they know; in that way, it is like syntax, which realises the ideational
function of conveying content. It has its own kind of structure, such as that of Topic
and Comment, and thus its own semantic relations. Those are built with a set of gram-
matical meanings distinct from those used in syntax, although similar. It is realised
with its own phonological forms, rather than those of syntax.
However, is unlike a stratum in some respects. First, it is dependent on syntax for
part of its structure: it is “superimposed” on syntax, sometimes depending on the syn-
tactic order of senses for its significance, as with Rheme being what it is syntactically
first. To some extent, then, it is a function of syntax. More significantly, information
structure does not fit into the chain of realisation, which runs from speaker inten-
tion through the semantic, lexical and syntactic strata to its completion in phonology.
Information structure realises its speaker intention directly, not through any lower
strata. Those criteria of realisational relationships are crucial to the nature of strata.
We conclude that information structure should not be rated as a stratum, since it
lacks the crucial characteristics. However, I draw attention to this account of its rela-
tion to the strata, since I believe that it makes clear the place of information structure
in the overall structure of English, which is not made clear in the literature, as far as
I am aware.
14.4.7 Phonology and graphology
This section takes phonology and graphology together, since they are closely related.
It will deal with phonology chiefly, since it is primary to writing; it will include de-
scription of graphology where appropriate.
Form. Phonology gives a new kind of form to its input of meaning in morphosyntactic
structure – the incipient utterance. The form given is more concrete and is fully linear,
as with the sequence of phonemes and of weak and strong stress in rhythm; syntax
and information structure, which it realises, are basically hierarchic, with only the po-
tential for sequence. Phonological forms are those of a distinct medium, sound. They
use the physical properties of pitch, volume and duration to structure the systems of
tonality, tone and so on (see Chapter 2, §2.5).
Meaning. Giving the utterance that new form is its prime function, which does not
entail any change in the utterance’s meaning. In a sense, then, it is meaningless: the
individual phonemes /p/, /u/ and /t/ have no meaning, although the morpheme they
make up is meaningful. However, phonology has other functions in rendering mean-
ing, as syntax does.
One function is to give a second realisation of meanings already realised in oth-
er strata. For example, the expressive function of interjections such as “Christ!” and
“Shit!” is already realised in the blasphemy and indecency of the words; phonology
adds exclamatory intonation. Guidance of the hearer’s response (commands, and so
on) is usually already realised in syntax (as declarative/interrogative mood etc.), but
it is realised in phonology also, by tone. Where the syntactic and phonological mean-
ings concur, the duplication may serve as a backup to ensure that the significance is
picked up. Where the two meanings differ, phonology overrides syntax. For example,
if the utterance is syntactically a statement but phonologically a question, as in “You’re
ready?”, hearers take it as a question; but the statement syntax is taken to carry the
secondary meaning that the speaker accepts or expects the statement to be true – that
the hearer is ready, in the example. Various elements of meaning are thus realised in
both syntax and phonology.
Some elements of meaning are realised in the phonology as well as both syntax
and lexicon, providing three alternatives, not just two. For example, emotion is real-
ised in emotive lexical meaning, and exclamatory syntactic structure, and in exclam-
atory tonality; social meaning is realised in lexical meaning, syntactic forms such as
elision, and intonation patterns such as those in the Welsh and Indian dialects of Eng-
lish. Phonology, then, has repetitive realisations, which may be apparently redundant,
may reinforce another realisation, and may overrule it.
Another function of phonology is to realise meanings which have no realisation
in the lower strata, realising them direct from the speaker’s intention. It gives mean-
ing to meaningless “words”; that is, certain sound patterns become words by being
given appropriate intonation. For example, “heck” and “shucks” become interjections;
forms like “aw” become discourse particles, forms like “oh” become whole exclama-
tory utterances. Phrases can also become utterances, when assigned the intonation
of statements, questions and so on – a significance otherwise supplied by syntactic
structure. Even some elements of a phoneme can gain a significance: the allophones of
/r/, for instance, often signify the speaker’s regional dialect. Finally, meaning is added
recursively – recursivity is not unique to syntax. For example, secondary tones modify
the primary tones, realising finer shades of emotive, attitudinal and modal meanings
than are realised in lexis (Halliday and Greaves 2008); the modification of a phono-
logical question by a syntactic statement (just above) is another example.
Thus, in the meaning which it carries directly – distinct from the meaning from
the lower stratum which it simply gives a form to – phonology is like syntax in car-
rying a great deal of meaning; but it is very different from syntax in the nature of that
meaning. Syntactic meaning is grammatical, and accordingly is systematic and para-
digmatic; it is mostly devoted to organising content, and thus to serving the ideational
function. Phonological meaning is seldom grammatical; it is not systematic, and even

allows idiosyncrasy; it is wholly interpersonal.
Semantic structure. The structures of the semantic stratum are networks, and the se-
mantic structure created by syntax is hierarchic, both having constituents and their
relations. Phonology, on the other hand, has wave structures itself, and creates se-
mantic wave structures. Tone, tonicity, tonality and rhythm all consist of patterns of
rise and fall. The patterns are all suprasegmental, running across and “ignoring” the
segments that constitute lexis and syntax. The semantic structures are of rising and
falling expressed feeling, informational importance, and so on.
Graphology. Graphology resembles a separate stratum in some ways: spelling primar-

ily realises phonemes, and punctuation largely realises tone, as if graphology realised
phonology; but word divisions, for example, do not. Punctuation also seems to realise
syntax directly, and word divisions certainly realise morphosyntax. Further, it is in a
separate medium, distinct from those of both syntax and phonology. On the other
hand, it is an alternative to phonology, so it cannot constitute a stratum as a distinct
realisational level.
Conclusion: phonology and graphology. We conclude that there is a distinct stratum

which includes phonology and graphology, and that graphology on its own is not a
separate stratum.
14.4.8 Phonetics
Phonetics here is the realisation into sound waves of all elements of phonology, not
only the realisation of phonemes, as “phonetics” is sometimes taken to be. That con-
version of phonology as mental form into physical form is obviously characteristic
of an independent stratum. So is its difference in semantic structure: it carries very
little semantic significance. In general, the phonetic patterns of sequences of pitch
and stress levels are simply the physical realisation of phonological patterns. They
realise the patterns of tonality, tonicity and so on that were assigned to figures and
clauses; but they also realise the patterns of phonemes that were assigned to senses at
the lexical level.
There is one respect in which phonetics does carry a meaning of its own, which
occurs with the phonic use of sound – contrasted with the phonemic use (Chapter 10,
§10.2.4, and Chapter 11, §11.2.5) – commonly referred to as “sound symbolism”. As
noted in the sections referred to, it occasionally has descriptive meaning, in onomat-
opoeia, but it normally serves an interpersonal function not the ideational one, being
playful or imaginative, for example. It thus usually realises meaning directly from the
speaker’s intention, not from a lower stratum, as when aspiration of a stop consonant
realises the intention of emphasising a meaning.
Phonetics therefore constitutes an independent stratum, although that has gen-

erally not been accepted.
14.4.9 Discussion: Stratification
Other levels. Cohesion, as a “level” of language, could be considered as a stratum, since

it has a distinctive structure of significance, and cannot be placed within any one oth-
er stratum. However, it is not in any way a realisation of another stratum; rather, it is a
pattern spreading across strata, and seems to be stylistic rather than strictly linguistic.
Paratones (Chapter 11, §11.5.4.5) have a better claim to forming another stratum,
since they occur on a level of their own; but their structures are those of syntax and
phonology.
Other views. There are several other accounts of the strata. Lamb, in the 1960’s, posit-
ed three strata – concepts and deep grammar, lexicogrammar, and phonology; but in
the 2000’s posited four strata – the conceptual system, lexicogrammar, phonology and
phonetics. To Lambrecht (1994, p. 25), the “components” of English are semantics,
morphosyntax, information structure, and “prosody” (i.e. phonology). In Functional
Discourse Grammar (Hengeveld and Mackenzie 2008), the “contextual component”
has an “interpersonal” (or pragmatic) “level” (or stratum), and a “representational”
(or semantic) level, which formulate knowledge and intentions from the “conceptual
component”; the contextual component also has morphosyntactic and phonological
levels, which encode what the lower levels have formulated; the “output component”
produces sounds or writing. Mel’cuk (2012, pp. 100–101) lists semantic, syntactic,
morphological and phonological representational levels, with semantics, syntax, mor-
phology and phonology as interfaces between them. In Systemic Functional Gram-
mar (Halliday 2014), there are effectively four strata: graphology and phonology, as
modes of expression; and lexicogrammar and semantics, as levels of “content”. I be-
lieve that, although they differ in detail, those accounts give strong general support to
what has been said here, but need a better basis in semantics.
14.4.10 Conclusion: Stratification
This section has rated semantics, the lexicon (lexical items), syntax, phonology and
phonetics as strata of English. Morphology, information structure and graphology had
some of the characteristics, but lacked crucial ones. The section also showed that there
was no consistent pattern in which characteristics were relevant to a strict definition;
nor was there a clear distinction between a layer and a stratum. The characteristics of
a “stratum”, then, do not cohere to form a category, but only a loose class; the concept
will not classify phenomena strictly, and does not have great explanatory power.
The concept is useful, however. It has helped illuminate several phenomena in
the semantic structure of English. First, it has highlighted the variation in the role of
the different systems (the “strata”) in the realisation of speaker intentions. Different
elements of meaning have different “routes” through the strata as they are realised.
That is shown diagrammatically in Diagram 1, which represents the utterance “What
nice new knickknacks!” – invented to give maximum routes in a minimal utterance.
The circles at the bottom represent intentions; the boxes represent the strata; the ar-
rows represent the path of an element of meaning as it is realised, from the bottom
[What nice new knick-knacks!] as audible speech
PHONETICS
Sharp [k,n]
[What nice new knick-knacks!] as phones
Short [i,a]
Exclamatory . PHONOLOGY
tone. /What nice new knick-knacks!/ as phonemes
MORPHOSYNTAX
“What nice new knick-knacks!” as clause
“What...!” construction. Word order. Plural inflection
LEXICON
knick-knack new nice
SEMANTICS
OBJECTS NEW
Guide
Be Give composition Secondary
playful information of intentions:
meaning
Expressive Emotive Primary

function function intentions:
Diagram 1. Realisational routes through the strata, for “What nice new knickknacks!”
upwards. For example: at the bottom of the diagram, the speaker’s primary intention
of conveying feeling – the emotive function – activates the secondary intention of
giving information about the “knick-knack” objects to be admired; that activates the
concepts of OBJECT and NEW in the semantic stratum; under the influence of the sec-
ondary intention of being playful, the concept OBJECT is realised in the lexical stratum
by the word, knickknacks. Again, the intention of guiding the hearer in composing the
meaning of the words into a whole utterance leads to setting the word order in the
morphosyntactic stratum, and inflecting knick-knack into the plural – as two gram-
matical meanings.
The diagram is very complex. That makes the main point: in English, meanings
are realised in many strata, not only in the lexicon; realisation goes through many
stages, the number varying according to the nature of the meaning and the speaker’s
choice; and there are many variations in the process.
A second phenomenon illuminated by the concept of strata is the systematic na-
ture of English semantics. Like syntax, semantics has layers with different units and
different structural relations. Each builds on those below; but the relations between
them are not so simple: their complementarity is not only that each can supply some-
thing that the others lack, but, in a variant of the marked/unmarked alternation, each
can be used in an unusual way to supply what is normally supplied by another, as
when syntax is used to embody emotion, although its domain is generally descriptive
content, and although emotion is a domain shared by lexis and phonology.
That entails a third issue, that of “coding”, as in Morse coding of the letter S as
three long sounds. Meaning is converted into a new form, just as the word, “meaning”,
takes new typeface forms in, “meaning → meaning → MEANING”. But as we have seen
repeatedly, meaning is not simply converted mechanically and one-to-one into a new
form (lexicon, syntax etc.), as in that analogy. To use a different analogy, it is rather
as if (thinking of the meaning in the bottom stratum as “meaning A”): meaning A →
meanings A + B → meanings A + B + C. Realisation of meaning does not resemble
coding in any way, although it is often said to be coding. See Kravchenko (2007) for
further discussion.
The discussion of strata also brought out specialisation of semantic function. Ac-
cording to DuBois (2003, p. 49), “Grammars code best what speakers do most”. Pho-
nology “codes” or “signals” the interpersonal functions best; lexis signals individual
concepts best; and syntax signals the articulation of concepts into complex structures
best. In DuBois’s ‘coding best’, there is specialisation of labour.
Finally, the analysis has brought out significant incongruity in the form and func-
tion of the systems that realise meaning. Phonology works in quite a different way
from morphosyntax and lexis, being without the constituent units and well-defined
relations and structure that characterise the other systems. I rather speculatively sug-
gest, as I have earlier in the book, that the explanation lies in the evolution of lan-
guage: descent from primate calls has developed phonology; descent from conceptual
thought has developed lexis and syntax, converging with phonology. There is substan-
tial argument favouring that hypothesis in the literature of linguistic evolution.
14.5 Contribution to semantics
The book has been intended primarily to be a contribution to our understanding of

semantic structure in English. I regard it as giving an outline of a field which has been
given very little attention, and as providing useful detail within the field.
Further, I suggest that it provides a basis for coordinating the schools of seman-
tics. First, work in the generative grammar tradition, the generative lexicon tradi-
tion, and computational semantics may be characterised as oriented to formalisation,
whereas cognitive semantics is not; but all are concerned with semantics as a study
of conceptual meaning. My account shows, I believe, the possibility of rendering the
substance of cognitive semantics in formalisations, as well as the substance of formal
semantics, as argued in Chapter 13, §13.6.5, so that formal and descriptive presenta-
tions are complementary. Second, this account has also been built on the pairing of
form (as structure) and function; that pairing can reconcile formal and functional
linguistics, again as complementary. I include the formally oriented groups listed
above in the formal camp; and Functional Discourse Grammar, Systemic Functional
Grammar, and Meaning-Text Theory e.g. Mel’cuk (2012) are included in the function-
al camp. The divide between them has, after all, been one of the linguists’ interest and
focus, rather than the belief that the other is in error.
I draw attention to the discussion in Chapter 13 of wider issues arising from the
explanatory power of concepts developed earlier in the book. I believe that several
long-standing problems have been illumined (I would like to say “resolved”). They
include the following: word classes (§13.2); prototypes §13.3; the place of information
structure in the grammar (§14.3.6); the relationship between syntax and semantics
(many sections, but §14.3.5 above in particular); the nature and extent of composi-
tionality (sections in various chapters: §7.8.3, §8.6.2, §9.5.2, and §13.1).
I trust that the book has shown the value, for a complete understanding of mean-
ing in English, of approaching the study not only from a strictly linguistic approach,
but from semiotic, psycholinguistic and developmental approaches – although I have
given very little from that last-mentioned approach.
I hope to have contributed to filling some of the “serious black holes of ignorance”
which Cruse (2011, p. 449) said were abundant in all domains of meaning phenom-
ena. Lyons (1977, p. 138) said that we need a comprehensive and unified theory of
meaning, but was unsure even whether it is possible; I believe that the book shows
such a theory to indeed be possible, and that it makes a contribution to the theory, by
building on the work of Halliday partly, but chiefly that of Cruse.
14.6 Further research
Cruse (2011, p. 449) concludes with the assertion that, in all domains of meaning
phenomena, “serious black holes of ignorance abound”. I hope to have contributed
to filling some of them, as sketched in §14.5 just above, but acknowledge that many
remain. In this book, I have said almost nothing about the semantics of graphology,
and very little about the relationship between phonological meaning and graphologi-
cal meaning. I am not confident of the account I have given of sublexical meaning – of
its nature, the relations among its elements, and of their relation to cognition – nor
of semiotics. They all need further research, and I expect that a number of other top-
ics do too.
References
Aarts, Bas. (2007). Syntactic gradience: The nature of grammatical indeterminacy. Oxford:
Oxford University Press.
Aarts, Bas, & Meyer, Charles F. (Eds). (1995). The verb in contemporary English: Theory and
description. Cambridge / New York: Cambridge University Press.
Adamson, Sylvia. (1999). Literary language. In Roger Lass (Ed.), Cambridge history of the Eng-
lish language, Volume III, 1476–1776 (pp. 187–331). Cambridge: Cambridge University
Press.
Aitchison, Jean. (2012). Words in the mind: An introduction to the mental lexicon. 4th edition.
Chichester, UK / Malden, Ma.: Wiley-Blackwell.
Allan, Keith. (1986). Linguistic meaning. Volume I. London / New York: Routledge & Kegan
Paul.
Allan, Keith, & Jaszczolt, Kasia M. (Eds). (2012). The Cambridge handbook of pragmatics.
Cambridge: Cambridge University Press. doi: 10.1017/CBO9781139022453
Allen, Cynthia L. (1995). Case marking and reanalysis: Grammatical relations from Old to Early
Modern English. Oxford: Oxford University Press.
Ameka, Felix. (1992). Interjections: The universal yet neglected part of speech. Journal of Prag-
matics, 18(2–3), 101–118. doi: 10.1016/0378-2166(92)90048-G
Andreewsky, Evelyne; Rosenthal, Rosen; & Bourcier, Daniele. (1985). Meaning without lexi-
con: A computational model for the resolution of lexical ambiguities. Psychological rele-
vance. In Hoppenbrouwers et al. (Eds.), Meaning and the lexicon (pp. 380–383).
Armstrong, Sharon Lee; Gleitman, Lila R.; & Gleitman, Henry. (1983). What some concepts
might not be. Cognition, 13, 263–308. doi: 10.1016/0010-0277(83)90012-4
Asp, Melissa. (2013). The twin paradoxes of unconscious choice and unintentional agents.
In Lise Fontaine (Ed.), Systemic functional linguistics: exploring choice (pp. 161–178).
Cambridge: Cambridge University Press. doi: 10.1017/CBO9781139583077.011
Barsalou, Lawrence W. (1983). Ad hoc categories. Memory and Cognition, 11(3), 211–227.
doi:
10.3758/BF03196968
Barsalou, Lawrence W. (1992a). Cognitive psychology: An overview for cognitive scientists.
Hillsdale, NJ: Lawrence Erlbaum and Associates.
Barsalou, Lawrence W. (1992b). Frames, concepts, and conceptual fields. In Adrienne Lehrer, &
Eva Feder Kittay (Eds.), Frames, fields, and contrasts (pp. 21–74). Hillsdale, NJ: Lawrence
Erlbaum and Associates.
Barsalou, Lawrence W. (1999). Perceptual symbol systems. Behavioural and Brain Sciences, 22,
577–660.
Barsalou, Lawrence W. (2005). Situated conceptualisations. In Cohen & Lefebvre (Eds.),
(pp. 620–636).
Barsalou, Lawrence W. (2012). The human conceptual system. In Spivey et al (Eds.), (pp. 239–
258). doi: 10.1017/CBO9781139029377.017
Barsalou, Lawrence, & Wiemer-Hastings, Katja. (2005). Situating abstract concepts. In Pecher
& Zwaan (Eds.), (pp. 129–163).
Bates, E., & MacWhinney, Bryan. (1982). Functionalist approaches to grammar. In E. Wanner
& L. Gleitman (Eds.), Language acquisition: The state of the art (pp. 173–218). Cambridge:
Cambridge University Press.
Bauer, Laurie. (2004). Adjectives, compounds and words. Nordic Journal of English Studies, 3(1)
(special issue, World of Words: a Tribute to Arne Zettersten), 7–22.
Bencini, Giulia M. L. (2013). Psycholinguistics. In Thomas Hoffmann, & Graeme Trousdale
(Eds.), Oxford handbook of construction grammar (pp. 379–396). Oxford: Oxford Univer-
sity Press.
Biber, Douglas; Johansson, Stig; Leech, Geoffrey; Conrad, Susan; & Finegan, Edward. (1999).
Longman grammar of spoken and written English. Harlow: Longman.
Bierwisch, Manfred, & Schreuder, Robert. (1992). From concepts to lexical items. Cognition,
42, 23–60. doi: 10.1016/0010-0277(92)90039-K
Birner, Betty J., &, Gregory. (1998). Information status and noncanonical word order in English.
Amsterdam/Philadelphia: Benjamins. doi: 10.1075/slcs.40
Bisang, Walter. (2008). Precategoriality and syntax-based parts of speech: The case of late archa-
ic Chinese. Studies in Language, 32(3), 568–589. doi: 10.1075/sl.32.3.05bis
Blakemore, Diane. (2002). Relevance and linguistic meaning: The semantics and pragmatics of dis-
course markers. Cambridge: Cambridge University Press. doi: 10.1017/CBO9780511486456
Bloom, L. (1933). Language. New York: Holt.
Bohm, Isabel C.; Altmann, Ulrike; Lubrich, Oliver; Menninghaus, Winifried; & Jacobs, Arthur
M. (2013). When we like what we know – A parametric fMRI analysis of beauty and famil-
iarity. Brain and Language, 124(1), 1–8. doi: 10.1016/j.bandl.2012.10.003
Bolinger, Dwight. (1967). Adjectives in English: Attribution and predication. Lingua, 18, 1–34.
doi:
10.1016/0024-3841(67)90018-6
Bolinger, Dwight. (1986). Intonation and its parts: Melody. Stanford: Stanford University Press.
Borer, Hagit. (1998). Morphology and syntax. In Spencer & Zwicky (Eds.), (pp. 151–190).
Online: Chapter 8. doi: 10.1111/b.9780631226949.2001.00011.x
Boye, Kasper, & Harder, Peter. (2009). Evidentiality: Linguistic categories and grammaticaliza-
tion. Functions of Language, 16(1), 9–43. doi: 10.1075/fol.16.1.03boy
Braisby, N. R. (1990). Situating word meaning. In Robin Cooper, Kuniaki Mukai, & John Perry
(Eds.), Situation theory and its applications (pp. 315–342). Stanford: CSLI.
Brownell, Hiram H. (1988). Appreciation of metaphoric and connotative word meaning by brain-
damaged patients. In Christine Chiarello (Ed.), Right hemisphere contributions to lexical
semantics (pp. 19–31). Heidelberg, NY: Springer-Verlag. doi: 10.1007/978-3-642-73674-2_2
Bundegaard, Peer F.; Ostergaard, Svend; & Stjernfelt, Frederik. (2008). Waterproof fire stations?
Conceptual schemata and cognitive operations in compound constructions. Semiotica,
161(1), 363–393.
Büring, Daniel. (2007). Semantics, intonation and information structure. In Ramchand and
Reiss (Eds.), (pp. 435–473).
Burnley, David. (1992). Lexis and semantics. In Norman Blake (Ed.), Cambridge History of the
English Language, Volume II, 1066–1476 (pp. 409–499). Cambridge: Cambridge University
Press.
Butt, Miriam. (2006). Theories of case. Cambridge: Cambridge University Press.
doi:
10.1017/CBO9781139164696
429
References
Bybee, Joan. (2002). Sequentiality as the basis of constituent structure. In Talmy Givón, &
Bertram F. Malle (Eds.), Evolution of Language out of Pre-language (pp. 109–134).
Amsterdam / Philadelphia: Benjamins. doi: 10.1075/tsl.53.07byb
Chafe, Wallace L. (2001). The analysis of discourse flow. In Deborah Schiffrin, Deborah
Tannen, & Heidi Ehernberger Hamilton, (Eds.), The Handbook of Discourse Analysis
(pp. 673–684). Oxford / Malden, Ma: Blackwell.
Chen, Melinda. (2004). Affect in language interpretation. In Michael Achard & Suzanne
Kemmer (Eds.), Language, culture and mind (pp. 37ff.). Stanford Ca: CSLI.
Chiarcos, Christian; Berry, Claus; & Grabski, Michael. (2011). Salience in linguistics and be-
yond. In Christian Chiarcos, Claus Berry, & Michael Grabski (Eds.), Salience: Multidisci-
plinary perspectives on its function in discourse (pp. 1–28). Berlin / New York: De Gruyter
Mouton. doi: 10.1515/9783110241020.1
Citron, Francesca M. M. (2012). Neural correlates of written emotion word processing: A re-
view of recent electrophysiological and haemodynamic neuroimaging studies. Brain and
Language, 122, 211–226). doi: 10.1016/j.bandl.2011.12.007
Citron, Francesca M. M.; Weekes, Brendan S. & Ferstl, Evelyn C. (2014). Arousal and emotion-
al valence interact in written word recognition. Language, Cognition and Neuroscience,
29(10), 1257–1267. doi: 10.1080/23273798.2014.897734
Clark, Herbert H. (1996). Using language. Cambridge: Cambridge University Press.
doi:
10.1017/CBO9780511620539
Coates, Richard. (2000). Singular definite expressions with a unique denotatum and the lim-
its of properhood. Linguistics: An interdisciplinary journal of the language sciences, 38,
1161–1171.
COBUILD. (2001). Collins COBUILD English dictionary for advanced learners. 3rd ed. Glasgow:
Harper Collins.
Cohen, Henri, & Lefebvre, Claire. (2005). Handbook of categorization in cognitive science.
Amsterdam: Elsevier.
Collins, Peter. (2009). Information-packaging constructions. In Pam Peters, Peter Collins, &
Adam Smith (Eds.), Comparative studies in Australian and New Zealand English: Grammar
and beyond (pp. 73–88). Amsterdam/Philadelphia: Benjamins. doi: 10.1075/veaw.g39.05col
Couper-Kuhlen, Elizabeth. (1986). An introduction to English intonation. London: Arnold.
Couper-Kuhlen, Elizabeth. (1993). English speech rhythm: Form and function in everyday verbal
interaction. Amsterdam/Philadelphia: Benjamins. doi: 10.1075/pbns.25
Croft, William. (2001). Radical construction grammar: Syntactic theory in typological perspec-
tive. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780198299554.001.0001
Croft, William. (2003). Typology and universals. 2nd ed.; 1st ed. 1990. Cambridge University
Press.
Croft, William. (2005). Logical and typological arguments for Radical Construction Grammar.
In Östman & Fried (Eds.), (pp. 273–309).
Croft, William. (2007). Beyond Aristotle and gradience: A reply to Aarts. Studies in Language,
31(2), 409–430. doi: 10.1075/sl.31.2.05cro
Croft, William. (2012). Verbs: Aspect and causal structure. Oxford: Oxford University Press.
doi:
10.1093/acprof:oso/9780199248582.001.0001
Croft, William, & Cruse, D. Alan. (2004). Cognitive linguistics. Cambridge: Cambridge Univer-
sity Press. doi: 10.1017/CBO9780511803864
Cruse, Alan. (2000). Aspects of the microstructure of word meanings. In Ravin & Leacock
(Eds.), (pp. 30–31).
Cruse, Alan. (2011). Meaning in language: An introduction to semantics and pragmatics. 3rd ed.;
1st ed. 2004. Oxford: Oxford University Press.
Cruttenden, Alan. (1997). Intonation. 2nd edition; 1st edition 1986. Cambridge: Cambridge
University Press. doi: 10.1017/CBO9781139166973
Crystal, David. (1967). English. Lingua, 17, 24–56. doi: 10.1016/0024-3841(66)90003-9
Crystal, David. (1988). Language play. Chicago: University of Chicago Press.
Culicover, Peter W., & McNally, Louise. (Eds.). (1988). Syntax and semantics volume 29: The
limits of syntax. San Diego / London: Academic Press.
Dahl, Östen. (1976). Grammar and meaning: Papers on syntactic and semantic topics. Lingua,
39(1), 158–164. doi: 10.1016/0024-3841(76)90063-2
Davis, Barbara L., & MacNeilage, Peter F. (2002). The internal structure of the syllable. In Talmy
Givón, & Bertram F. Malle (Eds.), The evolution of language out of pre-language (pp. 135–
154). Amsterdam / Philadelphia: John Benjamins. doi: 10.1075/tsl.53.08dav
De Deyne, Simon; Voorspoels, Wouter; Verheyen, Steven; Navarro, Daniel J.; & Storms, Gert.
(2014). Accounting for graded structure in adjective categories with valence-based oppo-
sition relationships. Language, Cognition and Neuroscience, 29( 5), 568–583.
doi:
10.1080/01690965.2013.794294
de Smet, Hendrik, & Heyvaert, Liesbet. (2011). The meaning of the English present participle.
English Language and Linguistics, 15(3), 473–498. doi: 10.1017/S136067431100013X
DeLancey, Scott. (1981). An interpretation of ergativity and related patterns. Language, 37(3),
626ff. doi: 10.2307/414343
DeLancey, Scott. (1987). Transitivity in grammar and cognition. In Russell S. Tomlin (Ed.),
Coherence and grounding in discourse (pp. 57–68). Amsterdam / Philadelphia: Benjamins.
doi:
10.1075/tsl.11.04del
Diessel, Holger. (2006). Demonstratives, joint attention, and the emergence of grammar. Cog-
nitive Linguistics, 17 (4), 463–489. doi: 10.1515/COG.2006.015
Dik, Simon C., & Hengeveld, Kees. (1997). Theory of functional grammar, Parts 1 & 2. 2nd ed.
Berlin / New York: Mouton de Gruyter.
Dimroth, Christine; Gretsch, Petra; Jordens, Peter; Perdue, Clive; & Starren, Marianne. (2003).
A stage model for first and second language development. In Christine Dimroth, &
Marianne Starren (Eds.), Information structure and the dynamics of language acquisition
(pp. 65–94). Amsterdam / Philadelphia: John Benjamins. doi: 10.1075/sibil.26.05dim
Dingemanse, Mark. (2011). The use and meaning of ideophones in Siwu. PhD dissertation.
Nijmegen: Radboud University.
Dixon, Robert M. W. (1994). Ergativity. Cambridge: Cambridge University Press.
doi:
10.1017/CBO9780511611896
Dixon, Robert M. W. (2005). A semantic approach to English grammar. 2nd edition; 1st edition
1981. Oxford: Oxford University Press.
Donohue, Mark. (2008). Covert word classes: Making your own syntax in Tukang Besi. Studies
in Language, 32, (3), 590–609. doi: 10.1075/sl.32.2.03don
Downing, Pamela. (1977). On the creation and use of English compound nouns. Language, 53,
810–842. doi: 10.2307/412913
Dowty, David. (1991). Thematic proto-roles and argument selection. Language, 67(3), 547–619.
doi:
10.1353/lan.1991.0021
Du Bois, John W. (1987). The discourse basis of ergativity. Language, 63(4), 805–855.
doi:
10.2307/415719
Du Bois, John W. (2003). Discourse and grammar. In Michael Tomasello (Ed.), (pp. 17–88).
431
References
Elman, Jeffrey L. (2009). On the meaning of words and dinosaur bones: Lexical knowledge
without a lexicon. Cognitive Science, 33, 547–582. doi: 10.1111/j.1551-6709.2009.01023.x
Engelberg, Stefan. (2011). Lexical decomposition: Foundational issues. In Maienborn et al.
(Eds.), (pp. 124–144).
Ernst, Thomas. (2012). The syntax of adjuncts. Cambridge: Cambridge University Press.
Escandell-Vidal, Victoria; Jungl, Manuel Leonetti; & Ahern, Aoife (Eds.). (2011). Procedural
meaning: problems and perspectives. Bingley: Emerald. doi: 10.1163/9780857240941
Evans, Vyvyan. (2009). How words mean: Lexical concepts, cognitive models, and meaning con-
struction. Oxford: Oxford University Press.
doi:
10.1093/acprof:oso/9780199234660.001.0001
Fawcett, Robin P. (2000). The theory of syntax for Systemic Functional Linguistics. Amsterdam /
Philadelphia: Benjamins. doi: 10.1075/cilt.206
Feist, Jim. (2011). Compound constructions: Waterproof three-storey brick and tile fire sta-
tions? Semiotica, 187(1), 337–367.
Feist, Jim. (2012a). Premodifiers in English: Their structure and significance. Cambridge:
Cambridge University Press.
Feist, Jim. (2012b). What controls the ‘genitive variation’ in Present-Day English? Studies in
Language, 36(2), 261–299. doi: 10.1075/sl.36.2.03fei
Feist, Jim. (2013). Noun incorporation in English. Functions of Language, 20(2), 159–184.
doi:
10.1075/fol.20.2.02fei
Ferreira, Victor S., & Slevc, L. Robert. (2007). Grammatical encoding. In Gaskell (Ed.),
(pp. 453–170).
Ferris, Connor. (1983). The meaning of syntax: A study in the adjectives of English. London /
New York: Longman.
Fetzer, Anita, & Speyer, Augustin. (2012). Discourse relations in English and German dis-
course: Local and not so local constraints. Intercultural Pragmatics, 9(4), 413ff.
doi:
10.1515/ip-2012-0025
Fillenbaum, Samuel, & Rapoport, Amnon. (1971). Structures in the subjective lexicon. New
York: Academic Press.
Fillmore, Charles J., & Atkins, B. T. S. (2000). Describing polysemy: The case of ‘crawl’. In Ravin
& Leacock (Eds.), (pp. 91–110).
Firth, J. R. (1957). Modes of meaning. In Papers in linguistics 1934–1951 (pp. 190–215). London:
Fisher, Olga, & Nänny, Max. (2001). The motivated sign. Amsterdam / Philadelphia: Benjamins.
doi:
10.1075/ill.2
Flores d’Arcais, G. B., & Schreuder, R. (1987). Psychological Research, 49, 153–159.
Fodor, Jerry, & Lepore, Ernest. (2002). The compositionality papers. Oxford: Clarendon.
Fortescue, Michael. (2007). The non-linearity of speech production. In Hannay and Steen
(Eds.), (pp. 337–351). doi: 10.1075/slcs.83.18for
Fortescue, Michael. (2009). A neural network model of lexical organisation. London: Continuum.
Fortescue, Michael. (2010). A neural network approach to compositionality and co-composi-
tionality. The Mental Lexicon, 5(2), 180–204. doi: 10.1075/ml.5.2.03for
Fretheim, Thorstein. (2011). Description as indication: The use of conceptual meaning for a
procedural purpose. In Escandell-Vidal et al (Eds.), (pp. 131ff.).
Fries, Peter H. (1983). On the status of theme in English: Arguments from discourse. In János S.
Petöfi, & Emel Sözer (Eds.), Micro and Macro Connexity of Texts (pp. 116–146). Hamburg:
Helmet Buske.
Gagné, Christina L.; Spalding, Thomas L.; & Gorrie, Melissa C. (2005). Sentential context and
the interpretation of familiar open-compounds and novel modifier-noun phrases. Lan-
guage and Speech, 48(2), 203–221). doi: 10.1177/00238309050480020401
Garrod, Simon & Pickering, Martin J. (2007). Alignment in dialogue. In Gaskell (Ed.),
(pp. 453–452).
Gaskell, M. Gareth (Ed.). (2007). Oxford handbook of psycholinguistics. Oxford: Oxford Univer-
sity Press. doi: 10.1093/oxfordhb/9780198568971.001.0001
Geeraerts, Dirk. (2010). Theories of lexical semantics. Oxford: Oxford University Press.
Gentner, Dedre, & Boroditsky, Lera. (2001). Individuation, relativity, and early word learning.
In Melissa Bowerman & Stephen C. Levinson (Eds.), Language acquisition and conceptual
development (pp. 215–250). Cambridge: Cambridge University Press.
doi:
10.1017/CBO9780511620669.010
Gerdts, Donna B. (1998). Incorporation. In Spencer & Zwicky (Eds.), (pp. 84–100). Online:
Chapter 8. doi: 10.1111/b.9780631226949.2001.00007.x
Gibbs, Raymond W. (2005). Embodiment in metaphorical imagination. In Pecher & Zwaan
(Eds.), (pp. 65–92). doi: 10.1017/CBO9780511499968.004
Gibbs, Raymond W. Jnr; Wilson, Nicole L.; & Bryant, Gregory A. (2012). Figurative language:
Normal adult cognitive research. In Spivey & others (Eds.), (pp. 455–484).
Giegerich, Heinz J. (1999). Lexical strata in English: Morphological causes, phonological effects.
Cambridge: Cambridge University Press. doi: 10.1017/CBO9780511486470
Gil, David. (2005). Isolating-monocategorial associational language. In Cohen and Lefebvre
(Eds.), (pp. 348–380).
Gillon, Brendan S. (2005). Semantic categorisation. In Cohen and Lefebvre (Eds.), (pp. 167–
187). doi: 10.1016/B978-008044612-7/50062-7
Givón, Talmy. (1970). Notes on the semantic structure of English adjectives. Language, 46(4),
816–837. doi: 10.2307/412258
Givón, Talmy. (1979). On understanding grammar. New York / San Francisco / London: Aca-
demic Press.
Givón, Talmy. (1988). Pragmatics of word order: Predictability, importance and attention. In
Michael Hammond, Edith Moravcsik, & Jessica Wirth (Eds.), Studies in syntactic typology
(pp. 243–284). Amsterdam / Philadelphia: Benjamins. doi: 10.1075/tsl.17.18giv
Givón, Talmy. (1993). English grammar: A function-based introduction. Amsterdam / Philadel-
phia: Benjamins. doi: 10.1075/z.engram1
Givón, Talmy. (2009). The genesis of syntactic complexity: Diachrony, ontogeny, neuro-cognition,
evolution. Amsterdam / Philadelphia: Benjamins. doi: 10.1075/z.146
Goffman, Erving. (1981). Forms of talk. Philadelphia: University of Pennsylvania Press.
Goldberg, Adele E. (1999). The emergence of the semantics of argument constructions. In
Bryan MacWhinney (Ed.), The emergence of language (pp. 197–212). Mahwah, NJ /
London: Lawrence Erlbaum.
Gundel, Jeannette K. (2012). Pragmatics and information structure. In Allan & Jaszczolt (Eds.),
(pp. 585–598). doi: 10.1017/CBO9781139022453.032
Gussenhoven, Carlos. (2004). The phonology of tone and intonation. Cambridge: Cambridge
Haiman, John. (1983). Iconic and economic motivation. Language, 59(4), 781–819.
doi:
10.2307/413373
433
References
Halliday, M. A. K. (1977). Text as semantic choice in social contexts. In Teun A. van Dijk &
Janos S. Petöfi (Eds.), Grammars and descriptions (studies in text theory and text analysis)
(pp. 176–225). Berlin / New York: Walter de Gruyter.
Halliday, M. A. K. (1989). Grammar and daily life: Concurrence and complementarity. In
Jonathon Webster (Ed). (2012). On grammar (volume 1 of Collected works of M. A. K.
Halliday) (pp. 369–383).
Halliday, M. A. K. (2014). An introduction to functional grammar. 3rd ed.; 1st ed. 1985. London:
Hodder Arnold.
Halliday, M. A. K. (2000). Grammar and daily life. In Lockwood et al. (Eds.), (pp. 221–238).
doi:
10.1075/cilt.163.21hal
Halliday, M. A. K., & William S. Greaves. (2008). Intonation in the grammar of English.
London / Oakville, Ct: Equinox.
Halliday, M. A. K., & Hasan, Ruqaiya. (1976). Cohesion in English. London: Longman.
Halliday, M. A. K., & Matthiessen, M. I. M. (1999). Construing experience through meaning: A
language-based approach to cognition. London / New York: Continuum.
Hanks, Patrick. (2013). Lexical analysis: Norms and exploitations. Cambridge, Ma. /London:
MIT Press. doi: 10.7551/mitpress/9780262018579.001.0001
Hannay, Mike, & Steen, Gerard J. (2007). Structural-functional studies in English grammar: In
honour of Lachlan Mackenzie. Amsterdam / Philadelphia: Benjamins. doi: 10.1075/slcs.83
Hansen, Maj-Britt Mosegaard. (2012). A pragmatic approach to historical semantics with spe-
cific reference to markers of clausal negation in mediaeval French. In Kathryn Allan, &
Justyna A. Robinson (Eds.), Current methods in historical semantics (pp. 233–257). Berlin:
De Gruyter Mouton.
Hanson, Catherine, & Hanson, Stephen José. (2005). Categorization in neuroscience: Brain
response to objects and events. In Cohen and Lefebvre (Eds.), (pp. 119–140).
doi:
10.1016/B978-008044612-7/50060-3
Harder, Peter. (1996). Functional semantics: A theory of meaning, structure and tense in English.
Berlin / New York: Mouton de Gruyter.
Harder, Peter. (2007). Grammar flow and procedural knowledge: Structure and function at the
interface between grammar and discourse. In Hannay and Steen (Eds.), (pp. 309–335).
doi:
10.1075/slcs.83.17har
Hasan, Ruqaiya, & Fries, Peter H. (1995). Reflections on subject and theme: An introduction.
In Ruqaiya Hasan, & Peter H. Fries (Eds.), On subject and theme: A discourse functional
perspective (pp. xiii–xlv). Amsterdam / Philadelphia: Benjamins. doi: 10.1075/cilt.118.01has
Haspelmath, Martin. (2003). The geometry of grammatical meaning: Semantic maps and cross
linguistic comparison. In Michael Tomasello (Ed.), (pp. 211–242).
Haspelmath, Martin. (2007). Pre-established categories don’t exist. Linguistic Typology, 11,
119–132. doi: 10.1515/LINGTY.2007.011
Haspelmath, Martin. (2009). Terminology of case. In A. L Mal’chukov, & Andrew Spencer
(Eds.), The Oxford handbook of case (pp. 505–517). Oxford: Oxford University Press.
Hawkins, John A. (1994). A performance theory of order and constituency. Cambridge / New
York: Cambridge University Press.
Hawkins, John A. (2004). Efficiency and complexity in grammars. Oxford: Oxford University
Press. doi: 10.1093/acprof:oso/9780199252695.001.0001
Hengeveld, Kees; Rijkhoff, Jan; & Siewierska, Anna. (2004). Part-of-speech systems and word
order. Journal of Linguistics, 40(3), 527–570. doi: 10.1017/S0022226704002762
Hengeveld, Kees, & Mackenzie, J. Lachlan. (2008). Functional Discourse Grammar: A typologi-
cally-based theory of language structure. Oxford: Oxford University Press.
doi:
10.1093/acprof:oso/9780199278107.001.0001
Herbst, Thomas, & Götz-Votteler, Katrin. (2007). Valency: Theoretical, descriptive and cognitive
issues. Berlin / New York: Mouton de Gruyter. doi: 10.1515/9783110198775
Hinterhölzl, Roland, & van Kemenade, Ans. (2012). The interaction between syntax, informa-
tion structure, and prosody in word order change. In Terttu Nevalainen, & Elizabeth Closs
Traugott (Eds.), The Oxford handbook of the history of English (pp. 803–821). New York /
Oxford: Oxford University Press.
Hockett, Charles F. (1960). The origin of speech. Scientific American, 205, 88–96.
doi:
10.1038/scientificamerican0960-88
Hoppenbrouwers, G. A. J.; Seuren, P. A. M.; & Weijters, A. J. M. M. (Eds.). (1985). Meaning and
the lexicon. Dordrecht: Foris.
Hopper, Paul J., & Thompson, Sandra A. (1984). The discourse basis for lexical categories in
universal grammar. Language, 60(4), 703–752. doi: 10.1353/lan.1984.0020
Horgan, Kerry. (2012). Connectionism, dynamical cognition, and nonclassical compositional
representation. In Werning et al. (Eds.), (pp. 557ff.).
Hsu, Chun-Ting; Jacobs, Arthur M.; Citron, Francesca M. M.; & Conrad, Markus. (2015). The
emotion potential of words and passages in reading Harry Potter – an fMRI study. Brain
and Language, 142, 96–114. doi: 10.1016/j.bandl.2015.01.011
Huddleston, Rodney, & Pullum, Geoffrey K. (2002). The Cambridge grammar of the English
language. Cambridge: Cambridge University Press.
Hudson, Richard A. (1984). Word grammar. Oxford / New York: Blackwell.
Hudson, Richard, and Holmes, Jasper. (2000). Re-cycling in the encyclopedia. In Peeters (Ed.),
(pp. 259–290).
Inchaurralde, Carlos. (2000). Lexicopedia. In Peeters (Ed.), (pp. 97–114).
Jackendoff, Ray. (2002). Foundations of language: Brain, meaning, grammar, evolution. Oxford:
Oxford University Press. doi: 10.1093/acprof:oso/9780198270126.001.0001
Jackendoff, Ray. (1972). Semantic interpretation in generative grammar. Cambridge, Ma.: MIT
Press.
Jackendoff, Ray. (2011). Conceptual semantics. In Maienborn et al. (Eds.), (pp. 688–709).
Jacobson, Roman. (1936/1990). Contribution to the general theory of case. (Combining the
1936 paper with parts of the 1958 revised paper; translated from German). In Linda
R. Waugh, & Monique Monville-Burston (Eds.). (1990). On Language (pp. 332–385).
Cambridge, Ma.: Harvard University Press.
Jakobson, Roman. (1960). Closing statement: Linguistics and poetics. In Thomas A. Sebeok
(Ed.), Style in language (pp. 350–377). Cambridge, Ma.: MIT Press.
Janssen, Theo A. J. M. (2007). A speaker/hearer-based grammar: The case of possessives and
compounds. In Mike Hannay & Gerard J. Steen (Eds.), Structural-functional studies in Eng-
lish grammar: In honour of Lachlan Mackenzie (pp. 353–388). Amsterdam / Philadelphia:
Benjamins. doi: 10.1075/slcs.83.19jan
Janssen, Theo M. V. (2012). Compositionality: Its historical context. In Werning et al.,
(pp. 19–46).
Jespersen, Otto. (1924). Philosophy of grammar. London: Allen and Unwin.
Jung, Kiju; Shavitt, Sharon; Viswanathan, Madhu; & Hilbe, Joseph M. (2014). Female hurri-
canes are deadlier than male hurricanes. Proceedings of the National Academy of Sciences,
111(24), 8782–8787. doi: 10.1073/pnas.1402786111
435
References
Kamp, Hans, & Partee, Barbara. (1995). Prototype theory and compositionality. Cognition, 57,
129–191. doi: 10.1016/0010-0277(94)00659-9
Kamp, Hans, & Reyle, U. (2011). Discourse Representation Theory. In Maienborn et al. (Eds.),
(pp. 872–922).
Karmiloff-Smith, Annette. (1979). Micro- and macro- developmental changes in language ac-
quisition and other representational systems. Cognitive Science, 3, 91–118.
doi:
10.1207/s15516709cog0302_1
Karmiloff, Kyra, & Karmiloff-Smith, Annette. (2001). Pathways to language: From fetus to ado-
lescent. Cambridge, Ma. / London: Harvard University Press.
Katz, Graham. (2008). Modification of state verbs. In Louise McNally & Christopher Kennedy
(Eds.), Adjectives and adverbs: Syntax, semantics, and discourse (pp. 220–248). Oxford /
New York: Oxford University Press.
Kemmer, Suzanne. (1993). The middle voice. Amsterdam / Philadelphia: Benjamins.
doi:
10.1075/tsl.23
Kemmer, Suzanne. (1994). Middle voice: Transitivity and the elaboration of events. In Barbara
A. Fox & Paul J. Hopper (Eds.), Voice: Form and function (pp. 179–230). Amsterdam /
Philadelphia: Benjamins. doi: 10.1075/tsl.27.09kem
Kiparsky, Paul. (1998). Partitive case and aspect. In Miriam Butt & Wilhelm Geuder (Eds.), The
projection of arguments: Lexical and compositional factors (pp. 265–307). Stanford: CSLI.
Kirsner, Robert S. (2011). Instructional meanings, iconicity and l’arbitraire du signe in the anal-
ysis of the Afrikaans demonstratives. In Bob de Jonge & Yishai Tobin (Eds.), Linguistic
theory and empirical evidence (pp. 97–138). Amsterdam / Philadelphia: Benjamins
doi:
10.1075/sfsl.64.05kir
Klein, Wolfgang. (2004). How time is encoded. In Wolfgang Klein & Ping Li (Eds.), The expres-
sion of time (pp. 39–82). Berlin / New York: Mouton De Gruyter.
Koptjevskaya-Tamm, Maria. (2012). New directions in lexical typology. Linguistics, 50(3),
373–394.
Kravchenko, Alexander V. (2007). Essential properties of language, or, why language is not a
code. Language Sciences, 29, 650–671. doi: 10.1016/j.langsci.2007.01.004
Kuczaj, Stan A., & Hendry, Jennifer L. (2003). Does language help animals think? In Dedre
Gentner & Susan Goldin-Meadow (Eds.), Language in mind: Advances in the study of lan-
guage and thought (pp. 237–276). Cambridge, Ma.: MIT Press.
Labov, W. (1973). Boundaries of words and their meanings. In C. J. Bailey & R. Shuy (Eds.),
New ways of analysing variation in English (pp. 340–373). Washington, DC: Georgetown
University Press.
Lakoff, George, & Johnson, Mark. (2003). Metaphors we live by. 2nd ed.; 1st ed. 1980. Chicago /
London: University of Chicago Press doi: 10.7208/chicago/9780226470993.001.0001
Lamb, Sydney M. (1966). Outline of Stratificational Grammar. Washington, DC: Georgetown
University Press.
Lamb, Sydney M. (1999). Pathways of the brain: The neurocognitive basis of the brain. Amster-
dam / Philadelphia: Benjamins. doi: 10.1075/cilt.170
Lamb, Sydney M. (2004). Language and reality. London / New York: Continuum.
Lambrecht, Knud. (1994). Information structure and sentence form. Cambridge: Cambridge
Langacker, Ronald W. (1987a). Foundations of cognitive grammar, volume 1: Theoretical prereq-
uisites. Stanford, Ca.: Stanford University Press.
Langacker, Ronald W. (1987b). Nouns and verbs. Language, 63, 33–94. doi: 10.2307/415384
Langacker, Ronald W. (1999). Grammar and conceptualization. Berlin / New York: Mouton de
Gruyter. doi: 10.1515/9783110800524
Langacker, Ronald W. (2005). Dynamicity, fictivity, and scanning: The imaginative basis of logic
and linguistic meaning. In Pecher & Zwaan (Eds.), (pp. 164–197).
doi:
10.1017/CBO9780511499968.008
Langacker, Ronald W. (2008). Cognitive grammar: A basic introduction. Oxford: Oxford Univer-
sity Press. doi: 10.1093/acprof:oso/9780195331967.001.0001
Lascaratou, Chryssoula. (2007). The Language of pain: Expression or description? Amsterdam /
Philadelphia: Benjamins. doi: 10.1075/celcr.9
Leech, Geoffrey. (1974). Semantics. London: Pelican.
Leech, Geoffrey N., & Li, Lu. (1995). Indeterminacy between noun phrases and adjective phras-
es as complements of the English verb. In Aarts & Meyer (Eds.), (pp. 183–202).
Leffel, Timothy; Lauter, Miriam; Westerlund, Masha; & Pylkkänen, Liina. (2014). Restrictive vs.
non-restrictive composition: A magnetoencephalography study. Language, Cognition and
Neuroscience, 29(10), 1191–1204. doi: 10.1080/23273798.2014.956765
Lehmann, Christian. (2008). Roots, stems and word classes. Studies in Language, 32(3),
546–567. doi: 10.1075/sl.32.3.04leh
Lemmens, Maarten. (1998). Lexical perspectives on transitivity and ergativity: Causative con-
structions in English. Amsterdam / Philadelphia: Benjamins. doi: 10.1075/cilt.166
Levelt, Willem J. M. (1989). Speaking: From intention to articulation. Cambridge, Ma.: MIT
Press.
Levelt, Willem J. M. (1993). Accessing words in speech and production: Stages, Processes and
representations. In Levelt (Ed.), (pp. 1–22).
Levelt, Willem J. M. (Ed.). (1993). Lexical access in speech production. 2nd ed.; 1st ed. 1991.
Cambridge Ma. / London: Blackwell. doi: 10.1007/978-94-011-1840-8_11
Levin, Beth. (1993). English verb classes and alternations. Chicago / London: University of Chi-
cago Press.
Lockwood, David G.; Fries, Peter H.; & Copeland, James E. (Eds.). (2000). Functional approach-
es to language, culture and cognition: Papers in honour of Sydney M. Lamb. Amsterdam /
Philadelphia: Benjamins. doi: 10.1075/cilt.163
Lohndahl, Terje. (2012). Toward the end of argument structure. In Maria Cristina Cuervo &
Yves Roberge (Eds.), Syntax and Semantics 38: The end of argument structure? (pp. 155–
184). Bingley: Emerald.
Lüdtke, Jana; Meyer-Sickendieck, Burkhard; & Jacobs, Arthur M. (2014). Immersing in the
stillness of an early morning: Testing the mood empathy hypothesis of poetry reception.
Psychology of Aesthetics, Creativity, and the Arts, 8, 363–377. doi: 10.1037/a0036826
Lyons, John. (1966). Towards a ‘notional’ theory of the ‘parts of speech’. Journal of Linguistics, 2,
209–236. doi: 10.1017/S0022226700001511
Lyons, John. (1968). Introduction to theoretical linguistics. Cambridge: Cambridge University
Press. doi: 10.1017/CBO9781139165570
Lyons, John. (1977). Semantics. Cambridge: Cambridge University Press.
McClelland, James L.; Rumelhart, David E.; & Hinton, G. E. (1986). The appeal of parallel dis-
tributed processing. In David E. Rumelhart & James L. McClelland (Eds.), Parallel dis-
tributed processing: Explorations in the microstructure of cognition, volume 1: Foundations
(pp. 3–44). Cambridge, Ma.: MIT Press.
McGregor, William. (1997). Semiotic Grammar. Oxford: Clarendon Press / New York: Oxford
University Press.
437
References
McNally, Louise. (1988). On the linguistic encoding of information packaging instructions. In

Culicover & McNally (Eds.), (pp. 161–183).
Mcnamara, Timothy P. (2003). Semantic priming: The state of the art. New York: Psychology
Press.
MacWhinney, Bryan. (1999). The emergence of language from embodiment. In Bryan
MacWhinney (Ed.), The emergence of language (pp. 213–256). Mahwah, NJ / London:
Erlbaum.
MacWhinney, Bryan. (2005). The emergence of grammar from perspective. In Pecher & Zwaan
(Eds.), (pp. 198–223). doi: 10.1017/CBO9780511499968.009
Maienborn, Claudia; von Heusinger, Klaus; & Portner, Paul. (Eds.). (2011). Semantics: An inter-
national handbook of natural language meaning. Berlin / New York: De Gruyter Mouton.
Malinowski, Bronislaw. (1929). The sexual life of savages in north-western Melanesia: An eth-
nographic account of courtship, marriage and family life among the natives of Trobriand
Islands, British New Guinea. London: Routledge & Paul (1948).
Malinowski, Bronislaw. (1930). The problem of meaning in primitive languages. In C. K. Ogden
& I. A. Richards, The meaning of meaning: A study of the influence of language upon thought
and the science of symbolism (pp. 296–366). 2nd ed.; 1st ed. 1923. New York: Harcourt
Brace / London: Routledge & Paul.
Marantz, Alec. (2013). No escape from morphemes in morphological processing. Language
and Cognitive Processes, 28(7), 905–916. doi: 10.1080/01690965.2013.779385
Marchand, Hans. (1969). The categories and types of present-day English word formation: A
synchronic-diachronic approach. 2nd ed.; 1st ed. 1960. Munich: Beck.
Matthews, Gerald; Deary, Ian J.; & Whiteman, Martha C. (2003). Personality traits. 2nd ed.
Cambridge / New York: Cambridge University Press. doi: 10.1017/CBO9780511812736
Matthews, P. H. (1981). Syntax. Cambridge: Cambridge University Press.
Mel’cuk, Igor A. (2001). Communicative organisation in natural language: The semantic commu-
nicative structure of sentences. Amsterdam / Philadelphia: Benjamins. doi: 10.1075/slcs.57
Mel’cuk, Igor A. (2012). Semantics: From meaning to text. Amsterdam/Philadelphia: Benjamins.
doi:
10.1075/slcs.129
Meyer, Charles F. (1995). Grammatical relations in English. In Aarts & Meyer (Eds.), (pp. 27–39).
Miller, D. Gary. (2014). English lexicogenesis. Oxford: Oxford University Press.
doi:
10.1093/acprof:oso/9780199689880.001.0001
Miller, George A., & Fellbaum, Christiane. (1991). The semantic networks of English. Cogni-
tion, 41(1–3), 197–229. doi: 10.1016/0010-0277(91)90036-4
Mitchell, Bruce. (1985). Old English syntax. Oxford: Oxford University Press.
doi:
10.1093/acprof:oso/9780198119357.001.0001
Mithun, Marianne. (1984). The evolution of noun incorporation. Language, 60, 847–892.
doi:
10.1353/lan.1984.0038
Mithun, Marianne. (2012). Tags: Cross-linguistic diversity and commonality. Journal of Prag-
matics, 44(15), 2165–2182. doi: 10.1016/j.pragma.2012.09.010
Mosel, Ulrike, & Hovdhaugen, Even. (1992). Samoan reference grammar. Oslo: Scandinavian
University Press / Oxford: Oxford University Press.
Murphy, Gregory L. (1988). Comprehending complex concepts. Cognitive Science, 12, 529–562.
doi:
10.1207/s15516709cog1204_2
Murphy, M. Lynne. (2003). Semantic relations and the lexicon: Antonymy, synonymy, and other
paradigms. Cambridge: Cambridge University Press. doi: 10.1017/CBO9780511486494
Muysken, Pieter. (2008). A modular approach to grammatical categories: Evidence from lan-
guage diversity and contact. In Cohen & Lefebvre (Eds.), (pp. 46–69).
Newmeyer, Frederick J. (2000). The discrete nature of syntactic categories: Against a proto-
type-based account. In Robert D. Borsley (Ed.), Syntax and semantics, volume 32: The na-
ture and function of syntactic categories (pp. 221–250). New York: Academic Press.
doi:
10.1016/S0092-4563(00)80024-0
Nuckolls, Janis B. (1999). The case for sound symbolism. Annual Review of Anthropology, 28,
225–252. doi: 10.1146/annurev.anthro.28.1.225
Ochs, E. (1992). Indexing gender. In A. Duranti & C. Goodwin (Eds.), Rethinking context
(pp. 335–358). Cambridge: Cambridge University Press.
Őstman, Jan-Ola, & Fried, Mirjam. (Eds.). (2005). Cognitive grammars: Cognitive grounding
and theoretical extensions. Amsterdam / Philadelphia: Benjamins. doi: 10.1075/cal.3
Partee, (2010). Privative adjectives: Subsective plus coercion. In Rainer Bäuerle, U. Reyle &
Thomas Ede Zimmermann (Eds.), Presuppositions and discourse: Essays offered to Hans
Kamp (pp. 273–285). Bingley: Emerald. doi: 10.1163/9789004253162_011
Pecher, Diane, & Zwaan, Rolf A. (Eds.). (2005). Grounding cognition: The role of perception and
action in memory, language, and thinking. Cambridge: Cambridge University Press.
doi:
10.1017/CBO9780511499968
Peeters, Bert (Ed.). (2000). The lexicon encyclopedia interface. Oxford: Elsevier.
Pennebaker, James. (2011). The secret life of pronouns: What our words say about us. New York:
Bloomsbury Press.
Peirce, Charles Sanders. (1931–1958). Collected papers of Charles Sanders Peirce. Volume 2: El-
ements of logic. Cambridge, Ma.: Harvard University Press.
Petrova, Svetlana. (2012). The impact of focusing and defocusing on word order: Changes at
the periphery in Old English and Old High German. In Terttu Nevalainen & Elizabeth
Closs Traugott (Eds.), The Oxford handbook of the history of English (pp. 846–858). Oxford:
Pickering, Martin J., & Branigan, Holly P. (1998). The representation of verbs: Evidence from
syntactic priming in language production. Journal of Memory and Language, 39, 633–651.
doi:
10.1006/jmla.1998.2592
Pietroski, Paul M. (2012). Semantic monadicity with conceptual polyadicity. In Werning et al.
(Eds.), (pp. 129–148).
Prince, Ellen. (1981). Towards a taxonomy of given-new information. In Peter Cole (Ed.), Rad-
ical pragmatics (pp. 223–255). New York: Academic Press.
Prinz, Jesse. (2002). Furnishing the mind: Concepts and their perceptual basis. Cambridge, Ma. /
London: MIT Press.
Pulvermüller, Friedemann. (1999). Words in the brain’s language. Behavioural and Brain Scienc-
es, 22(2), 253–279. doi: 10.1017/S0140525X9900182X
Pulvermüller, Friedemann. (2002). The neuroscience of language: On brain circuits of words and
serial order. Cambridge: Cambridge University Press.
Pulvermüller, F., & Knoblauch, A. (2009). Discrete combinatorial circuits emerging in neural
networks: A mechanism for rules of grammar in the human brain? Neural Networks, 22(2),
161–172. doi: 10.1016/j.neunet.2009.01.009
Pustejovsky, James. (1995). The generative lexicon. Cambridge, Ma.: MIT Press.
Pustejovsky, James. (2012). Co-compositionality in grammar. In Werning et al. (Eds.),
(pp. 371–384).
439
References
Quirk, Randolph; Greenbaum, Sidney; Leech, Geoffrey; & Svartvik, Jan. (1972). A grammar of
contemporary English. London: Longman.
Quirk, Randolph; Greenbaum, Sidney; Leech, Geoffrey; & Svartvik, Jan. (1985). A comprehen-
sive grammar of contemporary English. London: Longman.
Radford, Andrew R. (1988). Transformational grammar: A first course. Cambridge: Cambridge
Ramchand, Gillian. (1997). Aspect and predication. Oxford: Oxford University Press.
Ramchand, Gillian, & Reiss, Charles. (Eds.). (2007). Oxford handbook of linguistic interfaces.
Oxford / New York: Oxford University Press. doi: 10.1093/oxfordhb/9780199247455.001.0001
Ravin, Yael, & Leacock, Claudia. (Eds.). (2002). Polysemy: Theoretical and computational ap-
proaches. Oxford: Oxford University Press.
Redeker, Gisela. (2006). Discourse markers as attentional cues at discourse transitions. In
Kerstin Fischer (Ed.), Approaches to discourse particles (pp. 339–358). Amsterdam: Elsevier.
Reid, Wallis Hoch. (1991). Verb and noun number in English: A functional explanation.
London / New York: Longman.
Reid, Wallis; Otheguy, Ricardo; & Stern, Nancy. (Eds.). (2002). Signal, meaning & message: Per-
spectives on sign-based linguistics. Amsterdam / Philadelphia: Benjamins.
doi:
10.1075/sfsl.48
Reverberi, Carlo; Kuhlen, Anna; Abutalebi, Jubin; Greulich, R. Stefan; Costa, Albert; Seyed-
Allaei, Shimas; & Haynes, John-Dylan. (2015). Language control in bilinguals: Intention
to speak vs. execution of speech. Brain and Language, 144, 1–9.
doi:
10.1016/j.bandl.2015.03.004
Rickheit, Gert, & Sichelschmidt, Lorenz. (2007). Valency and cognition – A notion in transi-
tion. In Herbst & Götz-Votteler (Eds.), (pp. 163–182).
Riehl, Claudia Maria, & Kilian-Hatz, Christa. (2005). Structure and function of incorporation
processes in compounding. In F. K. Erhard Voeltz (Ed.), Studies in African linguistic typol-
ogy (pp. 36ff.). Amsterdam / Philadelphia: Benjamins. doi: 10.1075/tsl.64.20rie
Rijkhoff, Jan. (2002). The noun phrase. Oxford: Oxford University Press.
doi:
10.1093/acprof:oso/9780198237822.001.0001
Rijkhoff, Jan. (2008). Descriptive and discourse referential modifiers in a layered model of the
noun phrase. Linguistics, 46(4), 789–829. doi: 10.1515/LING.2008.026
Roelofs, Ardi. (1993). A spreading activation theory of lemma retrieval in speaking. In Levelt
(Ed.), (pp. 107–142).
Ruhl, Charles. (2002). Data, comprehensiveness, monosemy. In Reid et al. (Eds.), (pp. 171–
190). doi: 10.1075/sfsl.48.11ruh
Sandra, Dominiek, & Sally Rice. (1995). Network analyses of prepositional meaning: Mirroring
whose mind – the linguist’s or the language user’s? Cognitive Linguistics, 6, 89–100.
doi:
10.1515/cogl.1995.6.1.89
Sanford, Anthony J. (2008). Context, attention and depth of processing during interpretation.
Mind and Language, 17(1–2), 188–206. doi: 10.1111/1468-0017.00195
Saussure, Ferdinand de. (1915). Charles Bally and Albert Sechehaye (Eds.), Course in general
linguistics. Trans. Ray Harris 1983. London: Duckworth.
Schreuder, Robert, & Flores d’Arcais, Giovanni B. (1989). Psycholinguistic issues in the lexical
representation of meaning. In William Marslen-Wilson (Ed.), Lexical representation and
process (pp. 409–436). Cambridge, Ma.: MIT Press.
Schreuder, Robert; Flores d’Arcais, Giovanni B.; & Glazenborg, Ge. (1985). Semantic decompo-
sition and word recognition. In Hoppenbrouwers et al. (Eds.), (pp. 108–114).
Schlesinger, I. M. (1995). On the semantics of the object. In Aarts & Meyer (Eds.), (pp. 54–74).
Schwanenflugel, Paula J. (1991). Why are abstract concepts hard to understand? In Paula J.
Schwanenflugel (Ed.), The psychology of word meanings (pp. 223–250). Hillsdale, NJ:
Lawrence Erlbaum.
Shorter Oxford English Dictionary. (2002). Shorter Oxford English dictionary on historical prin-
ciples. 5th edition; 1st edition 1933. Oxford: Oxford University Press.
Sinclair, John. (1998). The lexical item. In Edda Weigand (Ed.), Contrastive lexical semantics
(pp. 1–25). Amsterdam / Philadelphia: Benjamins. doi: 10.1075/cilt.171.02sin
Simon-Vandenbergen, Anne-Marie, & Taverniers, Miriam. (Eds.). (2003). Grammatical met-
aphor: Views from systemic functional linguistics. Amsterdam / Philadelphia: Benjamins.
doi:
10.1075/cilt.236
Slobin, D. (1970). Universals of grammatical development in children. In G. B. Flores d’Ar-
cais, & W. J. M. Levelt (Eds.), Advances in psycholinguistics (pp. 174–184). Amsterdam:
North-Holland.
Smith, Mark. (2010). Pragmatic functions and lexical categories. Linguistics, 48(3), 717–777.
doi:
10.1515/ling.2010.022
Spencer, Andrew, & Zwicky, Arnold M. (Eds.). (1998). The handbook of morphology. Oxford /
Malden, Ma.: Blackwell.
Spivey, Michael J.; Richardson, Daniel C.; & Gonzalez-Marquez, Monica. (2005). Perceptual-
motor and image schematic infrastructure. In Pecher & Zwaan (Eds.), (pp. 246–281).
doi:
10.1017/CBO9780511499968.011
Spivey, Michael J.; McRae, Ken; & Joanisse, Marc F. (Eds.). (2012). The Cambridge handbook of
psycholinguistics. Cambridge: Cambridge University Press. doi: 10.1017/CBO9781139029377
Štalnaker, R. C. (1978). Assertion In P. Cole (Ed.), Syntax and semantics: Volume 9: Pragmatics
(pp. 315–332). New York: Academic Press.
Stefanowitsch, Anatol. (2003). Constructional semantics as a limit to grammatical alternation:
The two genitives of English. In Gunter Rohdenburg & Britta Mondorf (Eds.), Determi-
nants of grammatical variation in English (pp. 413–444). Berlin / New York: Mouton de
Gruyter.
Štekauer, Pavol; Valera, Salvador; & Körtvélyessy, Lívia. (2012). Word formation in the world’s
languages: A typological survey. Cambridge: Cambridge University Press.
doi:
10.1017/CBO9780511895005
Stewart, Terrence, & Eliasmith, Chris. (2012). Compositionality and biologically plausible
models. In Werning & others (Eds.), (pp. 596–615).
Swee, Genevieve, & Shirmer, Annett. (2015). On the importance of being vocal: Saying ‘ow’
improves pain tolerance. Journal of Pain. Online: doi: 10.1016/j.jpain.2015.01.002.
Taylor, Ann, & Pintzuk, Susan. (2012). The effect of information structure on object position in
Old English: A pilot study. In Anneli Meurman-Solin, María José López-Couso & Bettelou
Los (Eds.), Information structure and syntactic change in the history of English (pp. 47–65).
Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199860210.003.0003
Taylor, John R. (2002). Cognitive grammar. Oxford / New York: Oxford University Press.
Taylor, John R. (2003). Cognitive grammar (2nd edition). Oxford / New York: Oxford University
Press.
Tench, Paul. (1996). The intonation systems of English. London: Cassell.
Thompson, Sandra. (1978). Modern English from a typological point of view: Some implica-
tions of the function of word order. Linguistische Berichte, 54, 19–35.
441
References
Thompson, Sandra A., & Hopper, Paul J. (2001). Transitivity, clause structure, and argu-
ment structure: Evidence from conversation. In Joan Bybee & Paul Hopper (Eds.), Fre-
quency and the emergence of linguistic structure (pp. 27–69). Amsterdam / Philadelphia:
Benjamins. doi: 10.1075/tsl.45.03tho
Timberlake, Alan. (1975). Hierarchies in the genitive of negation. Slavic and East European
Journal, 19(2), 123–138. doi: 10.2307/306765
Tobin, Yishai. (1990). Semiotics and linguistics. London / New York: Longman.
Toman, Jindrich. (2001). Word syntax. In Spencer & Zwicky (Eds.), (pp. 306–321). Online:
Chapter 15. doi: 10.1111/b.9780631226949.2001.00018.x
Tomasello, Michael. (Ed.) (2003). The new psychology of language: Cognitive and functional ap-
proaches to language structure. Volume 2. Mahwah, NJ / London: Lawrence Erlbaum.
Traugott, Elizabeth Closs. (2008). Syntax. In Richard M. Hogg (Ed.), The Cambridge history of
the English language: Vol. 1 The Beginnings to 1066 (pp. 168–289). Cambridge: Cambridge
University Press.
Trousdale, Nikolas. (2010). The event structure of perception verbs. Oxford: Oxford University
Press.
Tucker, Don M. (2002). Embodied meaning: An evolutionary developmental analysis of adap-
tive semantics. In Givón & Malle (Eds.), (pp. 51–82).
Ullman, Stephen. (1959). Principles of semantics. 2nd edition; 1st edition 1951. Glasgow:
Jackson. / Oxford: Blackwell.
Vallduví, Enric, & Vikunen, Maria. (1998). On rheme and Kontrast. In Culicover & McNally
(Eds.), (pp. 79–108).
Van de Velde, Freek. (2007). Interpersonal modification in the English noun phrase. Functions
of Language, 14(2), 203–230. doi: 10.1075/fol.14.2.05van
Van Linden, An. (2012). Modal adjectives: English deontic and evaluative constructions in syn-
chrony and diachrony. Berlin / Boston: De Gruyter Mouton. doi: 10.1515/9783110252941
Van Valin, Robert D. (2013). Lexical representation, co-composition, and linking syntax and
semantics. In James Pustejovsky (Ed.), Advances in generative lexicon theory (pp. 67 ff.).
Dordrecht / New York: Springer Science + Business Media. doi: 10.1007/978-94-007-5189-7_4
Vandelanotte, Lieven. (2002). Prenominal adjectives in English: Structures and ordering. Folia
Linguistica, 36, 219–259. doi: 10.1515/flin.2002.36.3-4.219
Vendler, Zeno. (1968). Adjectives and nominalizations. The Hague / Paris: Mouton.
Ward, Gregory, & Birner, Betty J. (2011). Discourse effects of word order variation. In
Maienborn et al. (Eds.), Ch. 73.
Warren, Beatrice. (1978). Semantic patterns of noun-noun compounds. Göteborg: Acta Univer-
sitatis Gothoburgensis.
Warren, Beatrice. (1984). Classifying adjectives. Göteborg: Acta Universitatis Gothoburgensis.
Weinreich, Uriel. (1966). Explorations in semantic theory. Reprinted in William Labov &
Beatrice Weinreich (Eds.). (1980). On semantics (pp. 99–207). Philadephia: University of
Pennsylvania Press.
Wennerstrom, Ann. (2001). The music of English speech: Prosody as discourse analysis. Oxford:
Werning, Markus; Hinzen, Wolfram; & Machery, Edouard. (Eds.). (2012). Oxford handbook of
compositionality. Oxford / New York: Oxford University Press.
doi:
10.1093/oxfordhb/9780199541072.001.0001
Wharton, Tim. (2012). Pragmatics and prosody. In Allan & Jaszczolt (Eds.), (pp. 567–584).
doi:
10.1017/CBO9781139022453.031
Wichmann, Anne. (2000). Intonation in text and discourse: Beginnings, middles, and ends.
Harlow / New York: Longman.
Wierzbicka, Anna. (1981). Case marking and human nature. Australian Journal of Linguistics,
1(1), 43–80. doi: 10.1080/07268608108599266
Wierzbicka, Anna. (1990). ‘Prototypes save’: On the uses and abuses of the notion of ‘prototype’
in linguistics and related fields. In S. L. Tsohadzidis (Ed.), Meanings and prototypes: Studies
in linguistic categorisation (pp. 347–367). London / New York: Routledge.
Wierzbicka, Anna. (1992). The semantics of interjection. Journal of Pragmatics, 18, 159–192.
doi:
10.1016/0378-2166(92)90050-L
Wierzbicka, Anna. (1996). Semantics: Primes and universals. Oxford / New York: Oxford Uni-
versity Press.
Wray, Alison. (2002). Formulaic language and the lexicon. Cambridge University Press.
doi:
10.1017/CBO9780511519772
Wunderlich, Dieter. (2012). Lexical decomposition in grammar. In Werning et al. (Eds.),
(pp. 307–327).
Wymer, Adam Zachary. (2008). Towards flexible types with constraints for manner and factive
adverbs. In Louise McNally & Christopher Kennedy (Eds.), Adjectives and adverbs: Syntax,
semantics, and discourse (pp. 249–273). Oxford: Oxford University Press.
Zaidel, Eran; White, Hedy; Sakurai, Eriko; & Banks, William. (1988). Representation, control
and interaction: Separate roles for the right and left hemispheres? In Christine Chiarello
(Ed.), Right hemisphere contributions to lexical semantics (pp. 71–88). Berlin: Springer-
Verlag. doi: 10.1007/978-3-642-73674-2_6
Zwaan, Rolf A., & Madden, Carol J. (2005). Embodied sentence comprehension. In Pecher &
Zwaan (Eds.), (pp. 224–245). doi: 10.1017/CBO9780511499968.010
Index
A association operator 65, 187

abstract noun 209 attitudinal meaning 92
accomplishments 233 realisation 318
achievements 233 Attribute 152, 153
acronym 36, 74, 328 attribute-value 386
active zones 210 auxiliary 325
activities 233 as Theme 271
adjectival clause 73 syntagmatic structure 224
adjectival phrase 178
adjective B
as prototype 395 basicness
use of term 388 introduced 20
vs. adverb 70 bond
Adjunct 10 defined 69
as realisation of Circumstance 148 in compositionality 386
adverb see also word class in semantic classes 77
use of term 388 of Entity 211
vs. adjective 70 of Events 70, 151
adverbial phrase 178 of modifiers 198
aesthetic use 31, 334 of Predicators 234
realisation 334 of Properties 70
affect 93, 121, 268, 294, 314 boundedness 71, 209, 233
defined 28 C
affective meaning 92 Carrier 152
realisation 316 categoriality
affordances 92 of concepts 397
Agent of linguistic choices 395
vs. Actor 172 of senses 394
Aktionsarten 233 of zones 192
alliteration 334 category
antonymy 7, 98 cognitive vs. linguistic 374
as cognitive vs. linguistic 399 covert 323
arbitrariness 21, 33, 36 prototypical 393
areas of meaning 200 vs. class 69
argument structure 177 central node 209, 213
article change dimension 91
as Theme 271 Circumstance
articulation 18, 317 introduced 11
aspectual features 233 realisation 140
aspectual structure 233 class
assertion 15, 136, 223 vs. category 68
defined 136 Classifier
vs. suggestion 381 described 180
Classifier construction 180 of Events 237

clause of semantic class 217
semantics of 148 varieties of 373
semantic structure in 10 construction 37
vs. figure 11 defined 180
closed class 67, 143 discussed 171
coding 309, 424 other views 144
cognitive dominance 95 Construction Grammar 144, 171
command 148 contextual variation 99, 103
incongruent realisation 372 conversation management 293
information structure in 259 co-ordination
realisation 267 as process 133
Comment as structure 216
discussed 265 copula 149
in figures 262 as relation 11
in paragraphs 263 count/mass distinction 71
common noun 34, 74, 209 covert sign 37
communication 31, 148
Complement 10 D
as realisation of Participant 148 declarative
semantic constituency 69 realisation 322
complementation 376 defining use 74, 142, 381
defined 134 defined 73
realisation 324 definiteness 141, 370
complexity deixis 66
as constraint on realisation 383 as semiotic process 34
in compositionality 386 dependency 10, 254
of dimensions 199 defined 11
of Event groups 236 dependency grammar 144
of information structure 383 dependent
of signs 141 defined 11
compositional functions 386 describing (semiotic process) 34
compositionality 425 descriptive meaning
general discussion 385 loss of 293
in field structures 303 realisation 380
introduced 13 subclasses 68
of Entity groups 216 descriptive use 73, 93, 142, 213
of figures 174 defined 73
of senses 93 Descriptor
compositional property of language 382 described 189
compound 134, 135, 188 internal structure 189
compound tones 15 determinateness
computational semantics 425 defined 236
conceptual metaphor theory 337 determiner 203
constraints 333 as function, not class 179
linguistic 32 internal structure 204
on realisation 383 semantics of 203
semiotic 33 syntagmatic relations 203
construal zone 203
defined 373 differentiation 26, 147, 149, 378
described 373
445
Index
dimensions Event
in grammatical meaning 65 described 232, 235
in synonyms 97 in figure 235
directive function 293 subclasses 235
discourse grammatical meaning 289 Event group 222
discourse marker 268, 293, 297, 382 as unit 236
discourse Participant head of 232
vs. situation participant 274 structure 224
discourse particle 9 exclamation 148
Discourse Representation Theory 1 information structure in 259
double articulation 37 realisation 267, 322
doubly bounded Events 233 exclusions from the book 1
dummy Subjects 206, 210 expectedness dimension
duration 233 in descriptive meaning 91
duration dimension 91 in synonymy 97
dynamic semantics 1 in use 100
Expressive function 32
E defined 28
effort code 314, 323 realisation 294, 312
emergent meaning 387 Expressive senses 93
entertainment 335
entity F
defined 69 field structure
described 69 defined 20
introduced 8 discussed 301
sublexical 71 in social meanings 321
Entity within hierarchic structure 308
defined 209 figurative use
described 208 as reconstrual 141
in figure 235 forms 334
subclasses 209 introduced 76
Entity group of grammatical meaning 142
as unit 211 figure
defined 179 defined 10
discussed 215 structure 10, 149
head of 206 final bound 233
structure of 211, 218 final-bounded Events 233
vs. noun/nominal phrase 178 Finite
zones 178 defined 223
Entity-head construction 180 internal structure 223
Epithet syntagmatic structure 223
described 193 finiteness 223
internal structure 193 semantics of 323
semantic class 193 focus
ergativity 172 defined 272
evaluative grammatical meaning 288 introduced 16
event realisation 326
defined 70 semantics of 16
described 69 frame 97
introduced 8 free order 220
sublexical 71 frequency code 314
functional H
versus representational 387 happening 69
Functional Discourse Grammar 143, 312, 333, head
400, 425 internal structure 207, 232
functional meaning 143 of Entity group 206
functional node 209 of Event group 232
hearer meaning 382
G hierarchic structure
generative grammar 425 construction of 139
generative lexicon 181, 425 figures 147
genitive groups 177
as determiner 203, 204 information structure 255
as premodifier 189 holophrase 308, 317, 336
gradience 327, 397 adults’ 296
grammatical auxiliary 224 children’s 295
syntagmatic structure 224 defined 9
grammaticalisation 142, 209 longer expressions 298
grammatical meaning realisation 379
as backgrounded 142 humour
auxiliaries 224 realisation 335
dimensions 65 hyponymy 99
in different uses 141
in Entity groups 215 I
in field structures 304 icon (type of sign) 335
in information structure 258 discussed 35
in morphology 10 in realisation 317, 375
introduced 8 ideational function 292
levels 321 defined 31
modal 219 realisation 372
of content words 218 identifying use 73, 74
other views 66, 143 ideophone 308, 335, 336
realisation 321, 323, 324 described 299
scope 218 imaginative meaning
signs for 64, 141 realisation 333
types 64, 288 imaginative use 31, 121, 299
vs. content meaning 8 realisation 333
grammatical metaphor imperative
as incongruent realisation 372 realisation 16, 37
defined 142 incongruent realisation 372
introduced 143 incorporation
grammaticisation 25, 154, 266, 399 for information structure 327
defined 256 of adjective into verb 230
graphology of adverbial 230
for information structure 263 of adverb into verb 230
in realisation 371 of noun 230
representing phonology 322 indefiniteness 203, 264
group index
vs. phrase 178 defined 33
group order discussed 34
semantics of 12 for Expression 312
for writer’s intention 261
introduced 21
447
Index
indirect speech acts 27 as mechanism of control 308

individuation as representation of concepts 387
and salience 274 as response 308
degree of 209 as situation expressed 26
in Event heads 234 as social communion 308
reduction of 230 layout 371
inference 1 lexical field 99
information lexical relations 95
defined 256 paradigmatic 7
information structure syntagmatic 8
defined 255 lexicon 391
other views 255 discussed 399
realisation 324 lexis
reality of 290 semantic structure in 7
status 257 linearity
types 255, 256 as constraint 33
information unit 257 exploitation of 332
marking 15 in realisation 311
initial bound 233 vs. co-ordination 216
intensity dimension linguistic dominance 95
in Epithets 194 link
in grammatical meaning 65 as bond 69
intention at group level 179
as basic to language 25 at sublexical level 69
formulated as functions 28 for entity 69
interjection 19 for Entity 179, 209
as Expression 29 for event 70
as indexical 35 for property 70
as item without internal structure 291 multiple, in network 140
internal structure literal use 76
explained 149 of grammatical meaning 142
interpersonal action 30 loose structures 261
interpersonal function
defined 30 M
introduced 29 mapping 309, 310
realisation 309 markedness
interpersonal Theme 267 discussed 339
interpretive grammatical meaning 288 marked order 220
interrogative marked pitch 16
realisation 35, 311 marked use 73
intersective modification 133 marker
irrealis defined 34
defined 330 discussed 36
realisation 330, 374 vs. symbol 141
mass 209
L mass noun 209
language Meaning-Text Theory 425
as action 12, 22, 293 meaning types 400
as communication 293 in grammatical meaning 64
as dots to be joined 372 medio-passive 172
as form for its own sake 334 meronymy 99
metafunctions node
as formulation of intention 28 semantic classes as 71
interaction 379 nominal group
listed 28 vs. Entity group 178
metaphor nominalisation 180, 208, 235
as information 375 non-specified Events 233
as reconstrual 141 noun
discussed 337 use of term 388
middle (voice) 172 noun incorporation see incorporation
misrelation 263, 270 noun phrase
modal 219 vs. Entity group 178
deontic 332 nuclear tones 15
epistemic 332
modal Adjunct 267, 303 O
modal auxiliary onomasiology 2
as representation of figure 142 onomatopoeia 295, 299, 370
for subordinate intention 27 open class 67, 143
vs. grammatical auxiliary 224 operator 68, 143, 224
modality order
as semantic category 331, 332 as sign 37
defined 142 iconic 35, 375
other views 331 marked 220
realisation 330, 374 of groups 12
modal modifier 141, 219, 333 of postmodifiers 206
mode of combination 386 of premodifiers 179
modification of words in a zone 220
introduced 10 unmarked 220, 376
(non)restrictive 73 orientation structure 266
structure of 13 other views 270
morphology realisation 325
semantic structure in 9 outline 209
overt sign 37
N
names P
as common nouns 74 paradigms
as proper nouns 73, 75 in grammatical meaning 262
in referential strategy 381 in phonological systems 21
naming 137 in social meaning 196
as semiotic process 34 of lexical items 7, 95
negation of signs 36
as field 303 of speech acts 136
functions of 338 paragraph
scope of 304 as realisation 371
negational descriptors 219 Topic structure in 263
negative polarity 303 paralinguistic features 1
negative polarity items 339 paratone
network as wave 305
among word senses 95 information structure 260
defined 8 introduced 15
introduced 5 realisation 329, 371
Topic structure in 263
449
Index
Participant introduced 10
introduced 11 non-eventive 11, 388
realisation 370 premodification
relations 135 vs. postmodification 218, 290
Participant-head construction 188 zones of 178
part of speech premodifier
discussed 141, 388 in Entity group 180
percept 69 subclass 192
perceptual node 209 vs. determiner 178
personal relationships (function) 30 preposition
phatic interjections 293 as Theme 271
Phenomenon 151 function of 325
phonaesthesia 19 semantics of 65
phonemic use prepositional phrase 11, 325
defined 295 primary tense 224
phonic use 299, 314, 315, 335 privative adjectives 219
defined 295 procedural meaning 66
phonological paragraphs 15 other views 143
phonology Process
semantic structure in 14 as category 153
phrase defined 148
vs. group 178 described 149
pitch 14 introduced 11
in phonic use 314 material type 151
playfulness 31, 336 mental type 151
pointing (semiotic process) 34 realisation 150
polarity 338 reconstrual 153
polysemy relational type 152
discussed 121 types 150
in discourse markers 300 vs. verb 149
in ideophones 299 Process-head construction 187
vs. monosemy 76 production phase code 314
postmodification proper names 75
vs. premodification 218 proper noun 34, 73, 209, 213
postmodifier of Entity property
internal structure 205 defined 70
syntagmatic structure 205 described 69
postmodifier of Event introduced 8
internal structure 231 sublexical 71
syntagmatic structure 231 prototypes
postposed particles 231 discussed 393
pragmatics 28, 298, 382 punctuation
defined 1 as sign 37, 263
predication 137, 149, 152, 236 discussed 371
modifiers as 289
nonfinite 331 Q
predicative adjectives 373 qualia
Predicator explained 181
and Event group 222 other views 181
as realisation of Process 150 qualia construction 180
determinateness 236 quality 69
question 148 rhythm
information structure in 259 as wave structure 19
realisation 268 basis of 17
defined 14
R function of 18
realisation semantics of 17, 305
as meaning-form relation 310 rhythmisation 17
as process 310
constraints on 383 S
defined 5 salience
figurative 375 defined 272
incongruent 372 described 272
into sound 369 in information item 273
into writing 371 in sense variation 103
of emotion and attitude 316 salience structure 272
of Expressive function 312 defined 257
of grammatical meaning 321 realisation 326
of imaginative meaning 333 scenario 97
of interpersonal functions 309 secondary tense 236
of personal relations 319 Seinsarten 209
of social relations 319 semantic
realisational property of language 382 term defined 1
realisation strategies 381 term discussed 1
reduction semantic class
of form 230 as role 209
of meaning 230 basic classes 69
of syntactic status 230 basis 69
reference (act of) 135 construal 217
referential use 73, 96, 211, 381 defined 72
Reinforcer discussed 71
described 200 historically 141
relations incongruent realisation 372
vs. semantic classes 71 introduced 8
relevance structure linguistic status 71
defined 256 outlined 68
discussed 258 psychological status 71
in figures 262 vs. cognitive class 8
in paragraphs/paratones 263 vs. word class 68, 388
realisation 324 semantic field 3, 99
reporting structure 260 semantic head 206, 209
representation semanticisation 375
language as 137 semantic relations
restrictive use 73, 142, 217, 386 as result of grammatical meanings 145
Rhematic structure vs. syntactic relations 211
introduced 16 semantic role 11
realisation 289 semantics
semantics of 16 approach to 2
Rheme other views 2
defined 272 semantic structure
discussed 272 in lexis 7
introduced 12 in morphology 9
451
Index
in phonology 14 in realisation 338

in syntax 11 in relation to signs used 37
introduced 1 layers in 401
other views 3 sub-act
outlined 4 in FDG 333
semantics vs. syntax sub-dimensions 209, 233
discussed 417 Subject 10
introduced 4 as realisation of Participant 148
semasiology 2 in relation to focus 261
semiosis 21 in relation to Theme 267
in information structure 332 in relation to Topic 258
semiotic constraints 33 semantic constituency 69
semiotic levels 34 subordination
semiotic processes 36 realisation 324
discussed 34 subsective modification 133
semiotic strategies 311 subzones
Senser 151 of Classifiers 182
serial tense 224 superordination
sign realisation 324
covert 37 symbol (type of sign)
for grammatical meaning 141 defined 33
forms in language 37 discussed 35
overt 37 for grammatical meaning 139, 141
types 33 for information structure 263, 271
types realised 311 introduced 21
signal 35 synonymy 7
situation participant defined 96
vs. discourse Participant 274 discussed 96
social meaning 92 syntactic head 206, 214
social relations 30 syntactic status 33
sociolinguistics 1 syntagmatic structure
sound symbolism 294 explained 149
speaker meaning 382 syntax
specificity dimension semantic structure in 10
in synonymy 97 syntax vs. semantics
speech act 12, 30, 322 discussed 417
defined 288 introduced 4
term discussed 36, 137 systematicity
speech participants 270 in information structure 289
spelling 371 in morpheme paradigms 141
statement 148 in semantics 424
information structure in 259 systematisation 401
subclasses 136 Systemic Functional Grammar 2, 28, 207, 266,
stative verbs 373 290, 310, 425
stativity system meaning 382
discussed 374 system operations 138
strata
as levels of realisation 309 T
emotive meaning in 316 tense
grammatical meaning in 323 in determinateness 236
identified 7 in grammatical meaning 65
realisation 224 transitivity
vs. time 66 as bond in figure 71, 151
Textual function 28, 267 realisation 370
textual Theme 267, 269 reconstrual of 151
Theme vs. ergativity 172
above the figure 269
defined 266 U
in groups 270 unbounded Events 233
interpersonal 267 Undergoer 151
introduced 12 underspecification 134, 307, 372, 387
other views 270 uniqueness 209
textual 267 unmarked order
topical 268 variation in 376
types 267 unmarked use 73
thing 69 uses of meaning 72
realisation 10
V
time
vagueness
realisation 375
as constraint on realisation 383
vs. tense 66
specification of 382
tonality
verb
defined 14
use of term 388
semantics of 14
viewpoint dimension
tone
in deixis 66
defined 14
semanics of 15
W
tone group 14
wave structure 304
tones
defined 19
basic 15
word class
compound 15, 16
discussed 388
significance of 15
other views 391
tonicity
word order
defined 14
semantics of free order 220
semantics of 16
semantics of marked order 220
Topic
absence from figure 259 Z
defined 258 zones 178
discussed 258 as construction 218
in figures 262 categoriality 192
marking 262 exceptions 219
realisation 325 free order within 220
scope 260 relationships 222
Topical Theme 268 subzones 182
Topic-Comment structure 262

Semantic Structure in English

Uploaded by

Copyright:

Available Formats

Semantic Structure in English

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Semantic Structure in English

Uploaded by

Copyright:

Available Formats

What are the main topics covered in the book?

What are the main topics covered in the book?

What are some of the linguistic levels discussed in relation to semantic structure?

What are some of the linguistic levels discussed in relation to semantic structure?

studies in functional and structural linguistics 73

John Benjamins Publishing Company

Founding Editor Honorary Editors

John Benjamins Publishing Company

theâ•¯American National Standard for Information Sciences – Permanence

© 2016 – John Benjamins B.V.

4.5 Grammatical meaning 61

5.5 Networks in other strata 116

7.2 Figure structures 149

8.2 Entity groups 178

10.3 Semantic units with multifunctional structure 295

11.5.5 Guiding the hearer’s attitude to content: Irrealis 330

12.6.3 Construal 373

1.1 Goals of the book

language itself is seen as functional in the evolutionary sense of being adapted to a

1.3 Argument of the book

1.4 Plan of the book

Semantic structures in the strata of English

2.2 Semantic structure in lexis

2.2.1 Paradigmatic lexical relations

We use whimper and blubber to convey disapproval: attitudinal “meaning” is another

2.2.2 Syntagmatic lexical relations

2.2.2.1 Grammatical and content items

2.2.2.2 Semantic classes

2.2.2.3 Semantic classes and cognitive classes

2.2.3 Words without paradigmatic or syntagmatic relations

2.3 Semantic structure in morphology

2.3.1 Semantic classes

2.3.2 Grammatical meaning

2.3.3 Dependency and modification

2.4 Semantic structure in syntax

2.4.1 Semantic structure of clauses: The figure

2.4.1.1 Constituents of figures

Adjuncts realise meanings which will be called “Circumstances” (“adverbials”, rough-

2.4.1.2 Relationships among figure constituents

2.4.1.3 Significance of group order

Conclusion. Order of words and of content generally is a necessity in speaking and

2.4.2 Semantic structure in groups

2.4.2.1 Groups as part of the semantic hierarchy

2.4.2.2 Modification structure

2.5 Semantic structure in phonology

All four are here included in “intonation”.

2.5.2 Semantics of tonality

2.5.3 Semantics of tone

Introduction. Phonetically, the system of tone is made up of seven tones (Halliday

2.5.4 Semantics of tonicity

2.5.5 Semantics of rhythm

North Manchester I knew nothing at all about Withington” . . .

Diagram 1. Structure of rhythm (Couper-Kuhlen 1993, pp. 95–96)

2.5.6 Semantic structure in phonetics

Meaningful phonetics in English. Differences in regional and social dialects consist to

2.5.7 Discussion: Semantic structure in phonology

2.5.7.1 Wave-like structures

2.5.7.2 Field-like structures

2.5.7.4 Compositionality in phonological semantics

2.5.8 Conclusion: Semantics in the phonological stratum