J. Blake, Barry. Case

This page intentionally left blank
CASE
Second Edition
Case is an accessible introduction for students of linguistics to the ways rela-
tions between words in sentences are marked in languages. Case is fundamen-
tal to the whole system of language. One of its most interesting features is the
recurrence of apparently idiosyncratic patterns and devices in otherwise un-
related languages. This book picks out these recurring strategies and explores
their signicance. It provides the background against which the case marking
of particular languages can be best understood.
In this revised edition, Blake renes and expands on his discussions of the
most important concepts in the study of case, taking into account recent devel-
opments inthe eld. It incorporates signicant additions tothe data andincludes
a thoroughly revised section on abstract case in the Chomskian paradigm.
CAMBRI DGE TEXTBOOKS I N LI NGUI S TI CS
General Editors: S . A N D E R S O N , J . B R E S N A N , B . C O M R I E ,
W. D R E S S L E R , C . J . E W E N , R . H U D D L E S T O N , R . L A S S ,
D . L I G H T F O O T , P . H . M A T T H E W S , S . R O M A I N E , N . V . S M I T H ,
N . V I N C E N T
CASE
Second Edition
In this series
P . H . M A T T H E W S Morphology Second edition
B . C O M R I E Aspect
R . M. K E M P S O N Semantic Theory
T . B Y N O N Historical Linguistics
J . A L L W O O D , L . - G . A N D E R S O N and O . D A H L Logic in Linguistics
D . B . F R Y The Physics of Speech
R . A . H U D S O N Sociolinguistics
J . K . C H A M B E R S and P . T R U D G I L L Dialectology Second edition
A . J . E L L I O T Child Language
P . H . M A T T H E W S Syntax
A . R A D F O R D Transformational Syntax
L . B A U E R English Word-formation
S . C . L E V I N S O N Pragmatics
G . B R O W N and G . Y U L E Discourse Analysis
R . H U D D L E S T O N Introduction to the Grammar of English
R . L A S S Phonology
B . C O M R I E Tense
W. K L E I N Second Language Acquisition
A . C R U T T E N D E N Intonation
A . J . W O O D S , P . F L E T C H E R and A . H U G H E S Statistics inLanguage Studies
D . A . C R U S E Lexical Semantics
F . R . P A L M E R Mood and Modality
A . R A D F O R D Transformational Grammar
M. G A R M A N Psycholinguistics
W. C R O F T Typology and Universals
G . C O R B E T T Gender
H . J . G I E G E R I C H English Phonology
R . C A N N Formal Semantics
P . J . H O P P E R and E . C . T R A U G O T T Grammaticalization
J . L A V E R Principles of Phonetics
F . R . P A L M E R Grammatical Roles and Relations
B . J . B L A K E Case Second edition
CASE
Second Edition
BARRY J . BL AKE
P R O F E S S O R O F L I N G U I S T I C S
L A T R O B E U N I V E R S I T Y
iuniisuio n\ rui iiiss s\xoicari oi rui uxiviisir\ oi caxniioci
The Pitt Building, Trumpington Street, Cambridge, United Kingdom
caxniioci uxiviisir\ iiiss
The Edinburgh Building, Cambridge CB2 2RU, UK
40 West 20th Street, New York, NY 10011-4211, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
Ruiz de Alarcn 13, 28014 Madrid, Spain
Dock House, The Waterfront, Cape Town 8001, South Africa
http://www.cambridge.org
First published in printed format
ISBN 0-521-80761-1 hardback
ISBN 0-521-01491-3 paperback
ISBN 0-511-04189-6 eBook
Barry J. Blake 2004
2001
(netLibrary)
CONTENTS
List of gures page ix
List of tables x
Preface to the Second Edition xiii
Preface to the First Edition xv
List of abbreviations xviii
1 Overview 1
1.1 Inectional case 1
1.2 Other manifestations 7
1.3 Competing mechanisms 12
2 Problems in describing case systems 18
2.1 The traditional analysis 18
2.2 Distinguishing cases 19
2.3 Meanings and functions 29
3 Modern approaches to case 47
3.1 Introduction 47
3.2 Grammatical relations 48
3.3 Abstract case 57
3.4 Semantic roles and grammatical relations 62
3.5 Hierarchies 86
4 Distribution of case marking 93
4.1 Introduction 93
4.2 Within the clause 93
4.3 Within the noun phrase 96
4.4 Within the word 104
4.5 Within the subordinate clause 109
vii
5 Survey of case marking 118
5.1 Introduction 118
5.2 Organisation of the core 118
5.3 Dative 142
5.4 Genitive 149
5.5 Partitive 151
5.6 Local cases 151
5.7 Other cases 154
5.8 Inectional case hierarchy 155
6 Life cycle of case systems 161
6.1 Origin 161
6.2 Developments within case systems 167
6.3 Loss of case marking 175
6.4 Derived functions of case marking 180
6.5 Finale 183
Notes 184
Guide to terminology 195
Guide to further reading 207
References 208
Author index 219
Language index 222
Subject index 225
viii
FI GURES
3.1 English and Dyirbal contrasted page 57
3.2 Government and Binding clause structure 58
3.3 Nuclear relations in Relational Grammar 80
6.1 Development of adpositions, sufxes and bound pronouns 170
ix
TABLES
1.1 Turkish case system page 2
1.2 Latin case paradigms 4
2.1 The Russian partitive 22
2.2 Latin locative forms 23
2.3 Latin adverbs 24
2.4 Core case marking in Pama-Nyungan 24
2.5 Lak case marking schema 27
2.6 Margany case system (partial) 28
2.7 Types of case 33
2.8 Latin case system (Simon the Dane) 35
2.9 Latin case system (Martin the Dane) 37
2.10 Planudes case system of Ancient Greek 37
2.11 Hjelmslevs case system of Greenlandic Eskimo 38
2.12 Jakobson 1936: Russian case system 40
2.13 Jakobson/Neidle: Russian case system 40
2.14 Kalkatungu cases 42
2.15 Feature analysis of Latin case system 43
3.1 Person/number marking in Latin 51
3.2 Kalkatungu case and clitics 53
3.3 Case and case relations in Kalkatungu 54
3.4 Sanskrit nominal declension as represented by deva- god 64
3.5 Pattern of advancements 88
4.1 German case inection 102
4.2 Nungali class/case prexes 104
4.3 Archi case marking 105
4.4 Yuwaalaray case marking 106
5.1 Konjo clitic pronouns 123
5.2 Case marking in Eastern Pomo 126
5.3 Core case marking in Indo-Aryan 129
5.4 Avar local cases 152
x
5.5 Labels for local cases 153
5.6 Finnish local cases 153
6.1 Djaru case markers 168
6.2 Early Romance noun declensions 176
6.3 Old English case inection 177
6.4 Latin and Old English pronouns 179
6.5 French and Italian singular pronouns 179
6.6 Kala Lagau Ya case and tense/aspect matchings 180
xi
PREFACE TO THE SECOND
EDI TI ON
About seven years have passed since I wrote the rst edition of Case. It is now time
to review what I wrote then and to take account of recent publications in the eld.
This second edition incorporates a number of signicant additions to the data and
some revised interpretations of data. It also incorporates a number of improvements
and expansions to the discussion of important concepts, taking into account current
developments in the eld. As in the rst edition, I have paid particular attention to
traditional and current notions and terminology, not just in case itself, but in the
areas of word class, structure, agreement, roles and grammatical relations.
The most substantial revision has been to section 3.3, which deals with abstract
case in the Chomskian paradigm. I have updated this section, not without some
difculty. The book as a whole is aimed at students and academics in general lin-
guistics or in languages, but the Chomskian paradigm contains numerous concepts
and terms peculiar to itself. Introducing too many of these notions in a short section
can lead to obscurity, but introducing too few runs the risk of distortion. Moreover,
different authors within the paradigm adopt different approaches and the model is
forever changing. Interested readers can follow up the references given in note 4 to
chapter 3.
I would like to thank Carol-El-Chaar for incorporating the new material.
Bundoora
2000
xiii
PREFACE TO THE FI RST
EDI TI ON
I can remember my rst encounter with case quite clearly. It was in 1949. The lan-
guage was Latin and the book was Latin for today. The rst sentence was Discipul,
pict uram spect ate and it came with a translation Pupils, look at the illustration. I
cannot say that I quickly became enamoured of case. There was not much pleasure
to be had in memorising paradigms, but eventually there were rewards: the rolling
hexameters of Virgil, the cleverly contrived odes of Horace and the epigrammatic
prose of Tacitus, all exploiting the genius of a highly inected language, a language
where grammatical functions were expressed in the most highly condensed fashion,
a single short sufx on a noun expressing case, number and sometimes gender, a
single sufx on a verb expressing tense, aspect, mood, voice and the person and
number of the subject.
There were other minor encounters with the language of Beowulf and the lan-
guage of Njal, but my next signicant encounter with case came in 1966. In that
year I took up a fellowship to study Australian Aboriginal languages and I was sent
to western Queensland to record Kalkatungu, or Kalkadoon in the more familiar
spelling, a language which at that time had no more than a dozen uent speak-
ers. Like most Australian languages Kalkatungu had a well-developed case system.
For Kalkatungu there was no available grammar and therefore no paradigms to
learn. The paradigms had to be built up by a mixture of elicitation and recording of
discourse.
These experiences are reected in the present book. A concentration on Latin is
entirely proper in a book on case, since our traditional notions of case and grammat-
ical relations were developed with reference to Ancient Greek and Latin. If we set
out to label cases in a previously undescribed language, it behoves us to apply the
labels as far as possible in a way that is consistent with the traditional description
of Latin. The strain of Australian Aboriginal examples that runs through the book
is fortuitous, but not, I think, unfortunate. Though it reects the background the
author happens to have, it is not inappropriate, given that Australia provides, or at
least did provide, the richest large-scale concentration of inectional case languages
xv
anywhere in the world. These languages obviously developed without any inuence
from Indo-European and they provide an independent perspective from which to
view the case languages of our western tradition.
This book is aimed at two types of reader. Firstly it is written for senior students
and academics in linguistics. Secondly it is written for senior students and academics
whose eld is a particular language or group of languages, students of the classical
languages, for instance, or scholars of Slavonic. For all readers the book will provide
a global perspective against which particular case manifestations can be judged, and
for those not already versed in the literature of cross-language comparison it will
reveal fascinating regularities.
Case has aesthetic properties. To the student of literature this is probably most
evident in text, where an author successfully exploits the succinct means of relating
words that an inectional case system provides and the freedom of word order
usually attendant on the presence of case. But there is also beauty in the system.
This is nowhere more apparent than in Kalkatungu. I like to tell my students that
it was a language more perfect than the Greek, more copious than the Latin.
Not only did it have a system of nine cases, it had a number of valency-changing
derivations that allowed different alignments of semantic role and grammatical
relation. Moreover, it had a separate system of clitic pronouns, a referent-tracking
system based on alternating the transitivity of the verb, and a word order maximally
sensitive to the demands of discourse. Whether there were poets and orators who
availed themselves of this marvellous instrument I do not know. Full many a ower
is born to blush unseen and waste its sweetness on the desert air.
A book cannot be written without contributions from many sources. I would like
to thank rst of all the following who supplied me with information: Keith Allan,
Ketut Artawa, Peter Austin, Greg Bailey, Joan Barclay-Lloyd, Robert Bauer, Edith
Bavin, Byron Bender, David Bradley, Kate Burridge, Mehmet Celik, Wally Chafe,
Hilary Chappell, Bernard Comrie, Grev Corbett, Bob Dixon, Mark Durie, Nick
Evans, Caspar de Groot, Nurcan Hacioglu, Luise Hercus, Greg Horsley, Edrinnie
Kayambazinthu, Miriam Meyerhof, Marianne Mithun, Isabel Moutinho, Johanna
Nichols, William OGrady, John Painter, Jan Rijkhoff, Graham Scott, Anna
Siewierska, Jae Jung Song, Stan Starosta, Sandy Thompson and Nigel Vincent.
I would also like to thank Julie Reid, who read the manuscript from the point
of view of my potential readership, and I would particularly like to thank Rodney
Huddleston, who oversawthe writing on behalf of the publisher and made numerous
helpful suggestions.
Others who facilitated the work include Judith Ayling, who was always available
on the e-mail to help with queries, and the staff of the Borchardt Library at La
Trobe, particularly the inter-library loan staff. My biggest debt is to the secretaries
xvi
of the Department of Linguistics, Dothea Haynes and Barbara Upton, particularly
Barbara who did the nal formatting. . . . hands worked busily a day, and there she
stands. Willt please you sit and look . . . ?
Bundoora
1992
xvii
ABBREVI ATI ONS
A The agent argument of a transitive verb or any argument
that is treated in the same way grammatically.
ab abessive
abl ablative
abs absolutive
acc accusative
agt agent
all allative
ap antipassive
aux auxiliary
ben beneciary, benefactive
caus causative
com comitative
COR correspondent
dat dative
decl declension
DO direct object
ds different subject
du dual
el elative
erg ergative
evid evidential
exp experiencer
f feminine
fa future actor
fem feminine
n nite
fut future
gen genitive
ger gerundive
xviii
ia instrumental advancement
ill illative
imp imperative
impf (a) imperfect
(b) imperfective
in inessive
inc inclusive
inch inchoative
inf innitive
inst instrumental
int interrogative
IO indirect object
loc locative
m masculine
mabl modal ablative
masc masculine
neut neuter
nm nominaliser
nom nominative
NP noun phrase
npst nonpast
O object
obj object
obl oblique
obv obviative
P The patient argument of a transitive verb or any argument that is
treated in the same way grammatically.
par partitive
part participle
pass passive
pat patient
perf (a) perfect
(b) perfective
pl plural
plur plural
poss possessor
ppart past participle
pperf pluperfect
pres present tense
prop proprietive
xix
prpart present participle
purp purpose, purposive
recip recipient
re reexive
rel (a) relative
(b) relative case
S (a) the single argument of a one-place verb
(b) subject (as in SOV subjectobjectverb)
(c) sentence (as in S NP VP)
sg singular
sing singular
ss same subject
subj subject
trans translative
V verb (as in SVO subjectverbobject)
voc vocative
VP verb phrase
1 (a) rst person
(b) subject (in Relational Grammar)
2 (a) second person
(b) direct object (in Relational Grammar)
3 (a) third person
(b) indirect object (in Relational Grammar)
- separates morphs and the corresponding glosses: Spanish virtud-es
(virtue-PL) virtues
. separates multiple glosses of a single morph or word form: German
trank (drink.PAST) drank
= separates a clitic from its host
xx
1
Overview
1.1 Inectional case
Case is a system of marking dependent nouns for the type of relation-
ship they bear to their heads. Traditionally the term refers to inectional marking,
and, typically, case marks the relationship of a noun to a verb at the clause level or of
a noun to a preposition, postposition or another noun at the phrase level. Consider
the following Turkish sentence,
(1) Mehmet adam-a elma-lar- ver-di
Mehmet.NOM man-DAT apple-PL-ACC give-PAST.3SG
Mehmet gave the apples to the man.
In this sentence - indicates that elmalar is the direct object of the verb vermek to
give. The sufx - is said to be an accusative (or objective) case marker and the
word form elmalar is said to be in the accusative case.
1
The sufx - also indicates
that elmalar is specic, since in Turkish only specic direct objects are marked as
accusative. Adam is marked by the sufx -a which indicates that it is the indirect
object. Adama is in the dative case. Mehmet contrasts with elmalar and adama in
that it bears no overt sufx. It is said to be in the nominative case, which in this
sentence indicates the subject.
2
The term case is also used for the phenomenon of having a case system and a
language with such a system is sometimes referred to as a case language.
Our denition of case refers to marking dependent nouns for the type of relation-
shiptheybear totheir heads. This denitionobviouslyembodies certainassumptions
about what is a head and what is a dependent or modier. The verb is taken to be
the head of the clause, since it largely determines what dependents may be present.
Vermek to give, for instance, is a three-place verb that takes three arguments: a
giver (expressed in (1) by the subject in the nominative case), a gift (expressed in
(1) by the direct object in the accusative case) and a recipient (expressed by the in-
direct object in the dative case). A verb may also have other dependents expressing,
1
Overview
Table 1.1 Turkish case system
nominative adam
accusative adam
genitive adamn
dative adama
locative adamda
ablative adamdan
for instance, time or location, which, though not licensed by a particular verb, are
nevertheless modiers of the verb.
Turkish has a system of six cases as in Table 1.1. The locative marks location
as in Istanbul-da in Istanbul, and the ablative indicates from or out of as in
Ankara-dan from Ankara. The genitive is used in phrases like adam-n ev-i the
mans house where n corresponds to s in English. There is a complication. Note
that ev house bears a sufx -i which is a third-person-possessive form translatable
as his, her or its. In Turkish the mans house is literally the mans, his house.
The genitive meets the denition of case on the assumption that ev is the head of
a noun phrase and adam a dependent.
In (1) the cases are determined or governed by the verb. Vermek to give requires
a subject in the nominative, an indirect object in the dative and a direct object in the
accusative (if specic) or nominative (if nonspecic). Cases can also be governed by
prepositions or postpositions. Turkish has postpositions which govern the ablative
like dolay because of : toplant-dan dolay because of the meeting, and sonra
after: tiyatro-dan sonra after the theatre.
3
The word forms displayed in Table 1.1 make up a paradigm, i.e. they constitute
the set of case forms in which the lexeme adam can appear.
4
In Turkish one could
say that there is only one paradigm in that a constant set of endings is found for
all nouns. It is true that noun stems of different shapes take different inectional
sufxes, but all these differences are phonologically conditioned by principles of
vowel harmony and the like. The locative, for instance, has the form -da following
stems with back vowels and -de following stems with front vowels. The d of this
sufx devoices to t following a stem-nal voiceless consonant: kitap-ta on (the)
book.
5
One could refer to -da, -de, -ta and -te as case markers or one could consider
that at a more abstract level there was only one locative case marker. We need to
make a distinction between cases (of which there are six in a systemof oppositions),
and the case markers or case forms through which the cases are realised. A case
marker is an afx and a case form is a complete word. In Turkish the case afxes
can be separated from the stem, so it is possible to talk about case markers. In some
languages, however, it is not possible to isolate a case sufx, so it is necessary to
2
1.1 Inectional case
talk in terms of the various word forms that express the cases of the stem. These
are case forms. (See also Seidel 1988: 36.)
It is also necessary to make a further distinction between the cases and the
case relations or grammatical relations they express. These terms refer to purely
syntactic relations such as subject, direct object and indirect object, each of which
encompasses more than one semantic role, and they also refer directly to semantic
roles such as source and location, where these are not subsumed by a syntactic
relation and where these are separable according to some formal criteria. Of the two
competing terms, case relations and grammatical relations, the latter will be adopted
in the present text as the term for the set of widely accepted relations that includes
subject, object and indirect object and the term case relations will be conned to
the theory-particular relations posited in certain frameworks such as Localist Case
Grammar (section 3.4.4) and Lexicase (section 3.4.5).
Grammatical relations need not be in a one-for-one correspondence with cases.
In Turkish the nominative expresses the subject, but not all noun phrases in the nom-
inative are subject, since, as noted above, the nominative also marks a nonspecic
direct object of a transitive verb (see (1) in chapter 5).
There is a widely held view, explicit, for instance, in Relational Grammar (sec-
tion 3.4.3), that all dependents can be allotted to a particular grammatical relation
whether purely syntactic or semantic. However, in practice it is often unclear how
certain dependents are to be classied. For this reason I will refer, for the most
part, to cases as having functions or meanings. These terms are traditional and they
can be taken to be theory-neutral or perhaps pre-theoretical. The term function
will range over well-dened grammatical relations such as direct object and other
relations such as agent of the passive verb where different theories might ascribe
the function to different relations. The term meaning will cover not only semantic
roles that are demarcated by case marking or some other formal means, but also
semantic roles that are distinguished only on intuitive grounds, roles whose status
remains unclear in the absence of some argumentation.
Turkish is a convenient language to use to illustrate case since it is an aggluti-
native language, i.e. one in which there are afxes that are easily separable from
the stem and from one another. With nouns, the stem, the number marking and the
case marking are all separable (except for some phonological assimilations). This
can be seen in elma-lar- in (1) where -lar is the plural marker and - the accusative
case marker. However, the traditional notion of case was developed on the basis of
Ancient Greek and Latin where there are several complicating factors. In Latin, for
instance, it is not possible to separate number marking from case marking. The two
categories have fused representation throughout the system or cumulative expo-
nence as Matthews calls it (Matthews 1974/1991). This means separate paradigms
for the two number categories, singular and plural. Moreover, there are different
3
Overview
Table 1.2 Latin case paradigms
1 2 3a 3b 4 5
a-stems o-stems cons.stems i-stems u-stems e-stems
feminine masculine neuter
domina dominus bellum c onsul cvis manus di es
mistress master war consul citizen hand day
singular
Nominative domina dominus bellum c onsul cvis manus di es
Vocative domina domine bellum c onsul cvis manus di es
Accusative dominam dominum bellum c onsulem cvem manum diem
Genitive dominae domin bell c onsulis cvis man us di e
Dative dominae domin o bell o c onsul cv manu di e
Ablative domin a domin o bell o c onsule cv, cve man u di e
plural
Nominative dominae domin bella c onsul es cv es man us di es
Vocative dominae domin bella c onsul es cv es man us di es
Accusative domin as domin os bella c onsul es cvs, cv es man us di es
Genitive domin arum domin orum bell orum c onsulum cvium manuum di erum
Dative domins domins bells c onsulibus cvibus manibus di ebus
Ablative domins domins bells c onsulibus cvibus manibus di ebus
case/number markers for different stem classes. Traditionally ve such classes are
recognised, and there are also variations within the classes. The ve classes, or
declensions as they are usually referred to, are illustrated in Table 1.2: the rst de-
clension ( a-stems), second declension (o-stems), third declension (consonant stems
and i-stems), the fourth (u-stems) and fth ( e-stems). The designations a-stems,
o-stems, etc. are not synchronically transparent and reect the product of historical
reconstruction. For practical purposes there are ve arbitrary declensions, though
the term i-stem has some relevance for those members of the third declension that
have -i in the ablative singular, accusative plural and genitive plural.
In Latin there is also a three-way gender distinction: masculine, feminine and
neuter. With a few exceptions male creatures are masculine and females femi-
nine, but inanimates are scattered over all three genders (though almost all neuter
nouns are inanimate). There is a partial association of form and gender in that
a-stems are almost all feminine and o-stems mostly masculine (except for a sub-
class of neuters represented by bellum in Table 1.2). This means that there can be
fusion of gender, number and case. The point is illustrated in Table 1.2 where we
have domina mistress (of a household) illustrating feminine a-stems and dominus
master (of a household), which is based on the same root, representing masculine
o-stems. As can be seen from Table 1.2 the word form domina simultaneously rep-
resents nominative case, feminine gender and singular number, dominumrepresents
4
1.1 Inectional case
accusative case, masculine gender and singular number, and similarly with other
word forms.
In Latin there is concord between a noun and an attributive or predicative adjec-
tive. This concord is sensitive to case and number, and those adjectives that belong
to the rst and second declension are sensitive to gender so we nd domina bona
good mistress and nauta bonus good sailor where nauta is one of the few nouns
of masculine gender in the rst declension. With adjectives of the rst and sec-
ond declensions the inections simultaneously represent case, number and gender
without exception.
As can be seen, six cases are recognised: nominative, vocative, accusative, geni-
tive, dative and ablative; however, no paradigm exhibits six different forms. In the
traditional descriptions a case is established wherever there is a distinction for any
single class of nominals. The vocative, the case used in forms of address, has a
distinctive form only in the singular of the second declension. Elsewhere there is
a common form for the nominative and vocative; however, distinct nominative and
vocative cases are recognised for all paradigms.
Each case has a number of functions, which can be summarised as follows. The
nominative encodes the subject and nouns that stand in a predicative relation to
the subject as in Dominus est c onsul The master is consul. The accusative encodes
the direct object and nouns that stand in a predicative relation to the object as
in F ecerunt dominum c onsulem They made the master consul. It also expresses
destination as in V ad o R omam I am going to Rome and extent as in the following:
(2) R egn avit is pauc os menss
rule.PERF.3SG he.NOM few.PL.ACC month.PL.ACC
He ruled for a few months.
A number of prepositions govern the accusative including all those that indicate
motion towards or extent. In fact a construction like V ad o R omam where the
accusative expresses destination without being governed by a preposition is mainly
conned to the names of towns and small islands; compare V ad o ad urbem I am
going to the city and V ad o in urbem I am going into the city.
The genitive is mainly used to mark noun phrases as dependents of nouns,
i.e. it is primarily an adnominal case. Among its adnominal functions is the en-
coding of possessor: c onsulis equus the consuls horse. The genitive is also used
to mark the complements of certain verbs. For example, with some verbs of remem-
bering and forgetting it marks the entity remembered or forgotten (3); with some
verbs of reminding the person reminded is encoded as an accusative-marked direct
object and the entity to be remembered is put in the genitive (4), and with verbs of
accusing, condemning or acquitting the accused is expressed as a direct object in
5
Overview
the accusative with the fault or crime in the genitive (5):
(3) Di e meminerit c onsul
day.GEN remember.FUT.PERF.3SG consul.NOM
The consul will remember the day.
(4) C onsulem amicitiae commonef ecit
consul.ACC friendship.GEN remind.PERF.3SG
He reminded the consul of friendship.
(5) Parricdi c onsulem inc usat
parricide.GEN consul.ACC accuse.3SG
He accuses the consul of parricide.
The main function of the dative is to mark the indirect object. A few three-place
verbs like d are to give take a direct object in the accusative and an indirect object
in the dative (6). A few score of two-place verbs take only one object, an indirect
object in the dative. These include cr edere to believe, noc ere to be harmful to
and subvenre to help as in (7):
(6) Dominus equum c onsul dedit
master.NOM horse.ACC consul.DAT give.PERF.3SG
The master gave the horse to the consul.
(7) Mihi subv enist
me.DAT help.PERF.2SG
You have helped me.
The ablative in Latin represents the syncretism or merger of three once-distinct
cases: the ablative, the locative and the instrumental. It is not surprising then to
nd that it expresses source, location and instrument. It is also described as having
a number of other functions including expressing the agent of the passive, i.e.
the demoted subject of the corresponding active as in vsus a c onsule seen by the
consul.
Although the ablative alone can express a variety of relations to the verb of
the clause, in most functions it is usually governed by a preposition. Prepositions
governing the ablative include ex out of (ex Itali a from Italy), in in (in Itali a
in Italy) and cum with (cum amcs with friends). One function where it is
normally used without any preposition is the instrumental as in man u by hand.
A handful of verbs take a complement in the ablative case. These include ut to
use and vesc to feed on.
6
1.2 Other manifestations
The denition of case given in section 1.1 above can be regarded as
a central denition. There are also manifestations of case that do not mark the
relationship of dependent nouns to their heads, and others that do not form a system
for marking nouns, at least not in an obvious sense, inasmuch as the exponents are
prepositions or postpositions.
1.2.1 Concordial case
In some languages, including Indo-European case languages like Latin
and Ancient Greek, case marking appears not only on nouns but on certain depen-
dents of the noun such as adjectives and determiners. The following example is
from Plato. Bios is a nominative singular form of a second-declension (o-stem)
masculine noun, the nominative indicating that bios is the subject of the predicate.
The denite article and the adjective are in the nominative singular masculine form,
their concord in case, number and gender indicating that they are dependents of
bios:
6
(8)
Ho aneksetastos bios ou bi otos anthr op o
the.NOM.SG unexamined.NOM.SG life.NOM.SG not livable.NOM.SG man.DAT.SG
The unexamined life is not livable for man.
This example also illustrates concord between a predicative adjective (bi otos) and
the subject (bios). See also section 4.2.
Although the use of the nominative on ho and aneksetastos would appear to meet
the denition of case in that it marks these words as dependents of bios, it does not
mark the type of dependency. We could compare an adnominal genitive construction
such as ho anthr opou bios (the.NOM.SG man.GEN.SG life.NOM.SG) the life of man
where the genitive signals a type of dependency and meets the terms of the central
denition offered in section 1.1.
1.2.2 Case on non-nouns
Case marking is found on pronouns as well as on nouns, but pronouns
and nouns are clearly subclasses of the larger class nominal. Case marking is
also found on certain classes of word that are not obviously nouns. In the previous
subsection it was mentioned that case could extend via concord to determiners and
adjectives. Adjectives in Ancient Greek and Latin decline like nouns and can appear
as the head of a noun phrase as in Greek hoi polloi (the.NOM.PL many.NOM.PL) the
many and to meson (the.NOM.SG middle.NOM.SG) the middle. Adjectives in these
7
Overview
languages are analysable as a subclass of noun, and the Greek grammarians referred
to them as the noun adjective as opposed to the noun substantive, a usage that
remained current until recent times. Determiners in Ancient Greek and Latin decline
like nouns. They can stand as the sole member of a noun phrase, i.e. they function
as pronouns, or they can accompany nouns as with ho in (8). Like adjectives they
should be taken as a subclass of nominal.
Adverbs of place, time and manner play a role analogous to case-marked nouns.
For instance, Latin Unde fugit Whence ees he? can be answered by an ablative-
marked noun expressing source: Corinth o fugit From Corinth he ees. Unde the
interrogative adverb and a noun in the ablative seem to bear the same relation or
function. Adverbs of place, time and manner may bear no case marking, fossilised
case marking, or case marking parallel with that of corresponding nouns. In Latin,
examples of fossilised case marking are common, but there are also examples like
qu abywhat way? ande abythat way where the - awouldappear tobe parallel with
the ablative - a of the rst declension singular. The presence or absence of identiable
case marking would appear to be of little importance; what is signicant is the
parallelism of function between adverbs and case-marked nouns. If grammatical
relations are to be ascribed to nouns, it would seem logical to ascribe such relations
to adverbs of place, time and manner. One can then specify that a complement of a
particular verb must be in, say, the locative grammatical relation. This requirement
can be fullled in a language like Latin by a noun in the ablative case (usually
with an appropriate preposition) or by a locative adverb. See also section 1.3.3 and
Table 2.3.
1.2.3 Vocatives
In the traditional description of Ancient Greek and Latin a vocative
case appears (Table 1.2). The vocative is used as a form of address. In Latin, for
instance, domine is the form used to address ones master as in Qu o v adis, domine?
(whither go.2SG lord.VOC) Where are you going, master?. Vocatives do not appear
as dependents in constructions, but rather they stand outside constructions or are
inserted parenthetically (see (9) in chapter 4).
7
They are unlike other cases in that
they do not mark the relation of dependents to heads. For these reasons vocatives
have not always been considered cases (Hjelmslev 1935: 4). In Ancient Greek and
Latin the vocatives claim to being a case is structural. The vocative is a word-nal
sufx like the recognised case sufxes. However, modied forms of nouns used as
forms of address also occur in languages that do not have case inection. In Yapese
(Austronesian), for instance, there is no morphological case marking on nouns,
but personal names have special forms used for address. There is no reason to
consider that these modications of names constitute a vocative case (Jensen 1991:
229f ).
8
8
1.2.4 Ungoverned case
In case languages one sometimes encounters phrases in an oblique case
used as interjections, i.e. apart from sentence constructions. Melcuk (1986: 46)
gives a Russian example Aristokratov na fonar! Aristocrats on the street-lamps!
where Aristokratov is accusative. One would guess that some expressions of this
type have developed fromgoverned expressions, but that the governor has been lost.
A standard Latin example is m e miserum (1SG.ACC miserable.ACC) Oh, unhappy
me! As the translation illustrates, English uses the oblique form of pronouns in
exclamations, and outside constructions generally.
1.2.5 Analytic case markers
In most languages adpositions (prepositions or postpositions) play at
least some part in marking the relations of dependent nouns to their heads. In
Japanese, for instance, postpositions perform this function to the exclusion of case
afxes. Inthe followingJapanese example gamarks the subject, ni marks the indirect
object and o marks the direct object:
(9) Sensei ga Tasaku ni hon o yat-ta
teacher SUBJ Tasaku IO book DO give-PAST
The teacher gave Tasaku a book.
Adpositions can be considered to be analytic case markers as opposed to synthetic
case markers like the sufxes of Turkish or Latin. The main difference in case
marking between a language like Japanese and a language like Latin is that in
the former there are no case sufxes, just the postpositions, whereas in the latter
there are case sufxes as well as adpositions. In Latin, which is fairly typical of
languages having analytic as well as synthetic case markers, prepositions are like
verbs in that they govern cases, and combinations of preposition and case sufx
can serve to mark the relations of nouns to the verb. In the following examples we
have a transitive verb governing the accusative (10a), a preposition in governing the
accusative (10b), an intransitive verb governing the ablative (10c) and a preposition
in governing the ablative (10d):
(10) accusative
a. Mlit es vident urbem The troops see the city.
b. Mlit es v adunt in urbem The troops go into the city.
ablative
c. Mlit es potiuntur urbe The troops are in control of the city.
d. Mlit es manent in urbe The troops stay in the city.
9
Overview
In (10d) the ablative indicates location (in the context of man ere to remain and
urbs city) and in species inside as opposed to super above, sub under, etc.
Together the preposition and the case sufx indicate the relationship of urbs to the
verb. Note that in can also govern the accusative as in (10b) where the combination
of in + accusative signals into. Most prepositions in Latin govern one particular
case, but some like in can govern the accusative or the ablative. In some languages
all adpositions require the same case, e.g. in Indo-Aryan languages postpositions
with few exceptions require the oblique case (see (11) below) and in English
all prepositions govern the accusative (with me, from her, etc.). In situations like
these it has been argued that the case sufx is redundant and the adposition bears
the sole burden of marking the relation of dependent nouns to their heads as in
Japanese.
In HindiUrdu, as in a number of other Indo-Aryan languages, there are three
layers of case-marking elements: inectional case, primary postpositions and sec-
ondary postpositions. Leaving aside the vocative, the inectional case system dis-
tinguishes two cases, nominative and oblique. The nominative covers both subject
and object and is generally referred to in Indo-Aryan linguistics as the direct case.
The oblique case is used with the primary postpositions such as se instrumen-
tal/ablative, m e locative, ke genitive and ko dative/accusative (it is used with indi-
rect objects and specic, animate direct objects). There is also a third set of local
postpositions that follow ke genitive:
(11) a. l
e
r
.
ka (nominative, alternatively direct)
boy
b. l
e
r
.
ke (oblique)
c. l
e
r
.
ke ko (oblique + dative postposition)
to the boy
d. l
e
r
.
ke ke sath (oblique + genitive postposition + sath with)
with the boy
Where inectional case and adpositions co-occur in a language, the adpositional
system normally exhibits ner distinctions than the inectional system. This is
nowhere better illustrated than in languages like English and Hindi where the case
systemis near-minimal. In Hindi the secondary postpositions, which mostly express
local notions such as between, in front of and behind, make more distinctions
than the primary postpositions.
9
Although one can easily separate different layers of case marking in a particular
language, as in Hindi for instance, it can be difcult to determine whether a single
layer of case marking in a particular language is afxial or adpositional. Where the
10
markers in question gure in concord, they are clearly afxes, but where they occur
only once in a phrase, usually at the end, there can be some doubt about whether
they are inections or free forms.
There are two kinds of evidence that can be sought, phonological and distribu-
tional. If the case marker in question displays a number of variants determined by
the nature of the word or stem to which it is adjacent (excluding the effect of some
kind of pervasive, phonetically motivated rule), then it is an afx. In Korean, for
instance, the subject-marking form (which corresponds to Japanese ga as in (9))
has the shape -ka after stems ending in a vowel and -i after stems ending in a con-
sonant. This would seem to indicate that the subject marker is a sufx. The -ka/-i
alternation is phonologically conditioned and has some phonological motivation in-
sofar as a consonant-initial sufx follows vowels and a vowel-initial sufx follows
consonants, but it is not part of a pervasive rule (like syllable-nal devoicing of
obstruents in German).
In English there is clear distributional evidence to show that prepositions are
words rather than prexes, since they can stand on their own as in Who did she give
it to? Evidence of this type is not forthcoming for most languages. Co-ordination
of nouns would appear to provide a criterion. One might expect that case markers
would have to appear on every co-ordinated noun whereas a preposition or postpo-
sition could appear only before or after the sequence of co-ordinated nouns. It turns
out, however, that markers which are integrated into a stem, i.e. markers that would
appear to be afxes on phonological grounds, are often restricted to one occurrence
in phrase-nal position with co-ordinated nouns. In Korean, for instance, the nom-
inative or subject marker, -ka/-i, referred to above, cannot be used within a phrase
co-ordinated with kwa (OGrady 1991: 7):
10
(12) a. Joe-kwa Mary-ka yenay-lul hanta
Joe-AND Mary-NOM love-ACC do
Joe and Mary are in love.
b. *Joe-ka-kwa Mary-ka yenay-lul hanta
A piece of distributional evidence that might be thought relevant is whether the
case-marking form always appears adjacent to the head. In Japanese and Korean
the order within noun phrases is determined by a modierhead principle so that a
noun is always in phrase-nal position. In the Australian language Diyari, however,
the order is generally determinernounadjective and the case marking is phrase-
nal. However, there is no chance of taking this marking to be postpositional on
the grounds that it can be separated from the head noun by a dependent adjec-
tive, since the markers in question are very much integrated into the nal word of
the phrase whether it be a noun or an adjective. The ergative sufx, for instance,
11
Overview
has phonologically and morphologically determined allomorphs such as -ndu with
female personal names and -yali with singular common nouns of two, four or ve
syllables whose nal vowel is i or u (Austin 1981a: 489). Interestingly, this in-
tegrated marking need be used only on the last of a series of co-ordinated nouns.
Clearly an afx can have a scope beyond the word of which it forms part.
There is one further complication. There are forms that are analysable as separate
elements from the point of view of syntax (i.e. they are words), but which are pro-
nounced as part of an adjacent word. Such forms are called clitics. If they are
pronounced as part of the following word, they are proclitics. If they are pronounced
as part of the preceding word, they are enclitics. The two varieties can be illustrated
from French where the subject pronouns are normally proclitic, but can be enclitic
in certain circumstances, in interrogatives, for instance, in those varieties that use
subjectverb inversion:
(13) O` u descends-tu?
Where do you get off?
A larr et de la rue de Rivoli, je vais faire des emplettes.
At the Rue de Rivoli stop, Im going to do some shopping.
The forms tu and je are enclitic and proclitic respectively in these examples. They
cannot be stressed or pronounced as words separate from the adjacent verb. If stress
is required, special nonclitic forms toi and moi must be used in addition.
Some prepositions and postpositions may be analysable as proclitics and enclitics
respectively. In Turkish, for instance, there is a form -ile/-ila which can express
instrument or accompaniment. In phrases such as tren-le by train and kz-lar-la
with the girls it appears in reduced form and exhibits vowel harmony. It looks like
a case sufx. However, it can be pronounced as a separate word and, like a number
of other postposition-like forms, it governs the genitive with singular pronouns.
This would indicate that it is a postposition that can be cliticised, i.e. that can be
treated as a clitic. The form -ile/-ila is different from the French pronouns in that
the latter are always clitics, whereas -ile/-ila may or may not be used as a clitic.
The view adopted in this work is that phonological integration into a host is the
best guide to the afxial status of a case marker, but the fact remains that there
are many phrase-nal, nonintegrated case markers to be found in languages, and
grammars often describe them as sufxes, postpositions or particles without any
discussion of the basis for the decision.
11
1.3 Competing mechanisms
Case in its most central manifestation is a systemof marking dependent
nouns for the type of relationship they bear to their head. However, it is not the only
12
grammatical mechanism for marking headmodier relations. One important type
of alternative is the principle of marking the head rather than the dependent (cf.
Nichols 1986). Another common option is to use word order rather than head or
dependent marking. Other means include the use of relator nouns and possessive
adjectives. One could take the view that all these means of expressing grammatical
relations are forms of case marking. The point is discussed briey in section 3.3.
1.3.1 Head marking
In many languages, in the majority in fact, there is some kind of
pronominal representation of certain core grammatical relations quite apart from
their representation via noun phrases. This representation is almost always on the
predicate, often on a grammatical (auxiliary) verb. In a few languages it is en-
clitic to the rst constituent of the clause (e.g. in some Pama-Nyungan and in some
Uto-Aztecanlanguages). Inthe followingSwahili example the third-person-singular
subject is represented by the rst-order prex on the verb (a-) and the third-person-
singular object by the third-order prex (-m-) (Hinnebusch 1979: 219). In Swahili,
as in other Bantu languages, nouns fall into gender classes. Each class is marked by
a prex, which also appears via concord on adjectives and determiners. Nouns for
humans take the prex m- in the singular. This can be seen in the phrase mwanamke
mrembo:
(14) Ali a-na-m-penda m-wanamke m-rembo
Ali 3SG.PRES-3SG-love M-woman M-beautiful
Ali loves a beautiful woman.
The grammatical relation most frequently represented in this way is the subject;
however, some languages, including Swahili, also represent the direct object or
other complements in this way. Traditionally the marking on the verb is referred
to as subject agreement, object agreement, etc., but this term is also used for the
type of subjectverb agreement found in French and the Germanic languages as in
English She runs versus They run. There is an important difference. In French and
the Germanic languages the subject noun phrase cannot normally be omitted with
a nite, nonimperative verb, but in the majority type, illustrated by Swahili, the
noun phrases corresponding to the relations represented on the verb can be omitted.
In Swahili one can say anampenda, which means S/he loves him/her. The term
cross-referencing agreement is sometimes used for the Swahili type as opposed
to the Germanic type (Bloomeld 1933: 193ff).
There is a further complication. It has been claimed that in languages with cross-
referencing agreement the pronominal representation on the verb sometimes is
interpretable as a mere agreement marker with the cross-referenced noun phrase
13
Overview
bearing the appropriate grammatical relation, while in other instances the pronomi-
nal marker is the sole exponent of the relevant relation and the cross-referenced noun
phrase a mere adjunct in a kind of apposition with the pronominal element.
12
It is not
appropriate to go into this question here. Sufce it to say that the cross-referencing
pronominal elements serve as an alternative to case in signalling grammatical re-
lations. Where there is more than one set of cross-referencing elements, the sets
may be distinguished by order or by form or both. In Swahili the subject and object
forms are always distinguished by their position and in a majority of instances by
differences of form as in (14) above.
Where differences of form are involved these differences are likely to reect
earlier differences of case in the free forms from which they are derived. In many
of the Northern languages of Australia the bound pronoun for rst-person-singular
subject is nga and the object form is ngan. The -n clearly signals the object relation,
and probably derives from an accusative, but it is a moot point whether one should
refer to marking on pronominal elements in the verb as case marking.
13
Another type of marking on the verb that could be considered head marking is
marking to indicate change of valency. The presence of a passive marker on the verb,
for instance, is a form of head marking that determines how dependent marking is
to be interpreted. In Latin Gall vincunt is The Gauls conquer, but the addition
of the passive marker -ur to the verb yields Gall vincuntur The Gauls are (being)
conquered.
14
In some languages the person and number of a dependent possessor noun is
cross-referenced on the head-possessed noun as in Hungarian az ember haz-a (the
man house-3SG), literally, the man his house, i.e. the mans house. An example
of marking both the head and the dependent in a possessive construction was given
from Turkish in section 1.1.
Another type of head marking in noun phrases indicates the presence of a de-
pendent. In Semitic languages a noun with a noun dependent is in what is called
the construct state. In Arabic, for instance, the book of the king is kit abu lmaliki,
where kit abu book is in the construct state (a nominative form lacking both the
denite article and the -n that marks indeniteness) and lmaliki is genitive (Kaye
1987: 678). In Persian a noun with dependent is sufxed with -e: ket
`
ab-e m an
book of me, my book (Windfuhr 1987: 532).
1.3.2 Word order
Fairly obviously word order is an alternative to case marking in dis-
tinguishing subject from object in languages like English, Thai, Vietnamese and
Indonesian, all of which use the order subjectverbobject as their unmarked op-
tion. In English the word order also distinguishes the patient object from the recip-
ient or beneciary object in double-object constructions where the patient object
14
always follows the other object: She gave me good marks, She cut me a bunch of
dahlias.
It has frequently been observed that there is a correlation between the presence
of case marking on noun phrases for the subjectobject distinction and exible
word order and this would appear to hold true. From the work of Greenberg it
would also appear that there is a tendency for languages that mark the subject
object distinction on noun phrases to have a basic order of subjectobjectverb
(SOV), and conversely a tendency for languages lacking such a distinction to have
the order subjectverbobject (SVO) (Greenberg 1963). The following gures are
based on a sample of 100 languages. They show the relationship between case and
marking for the 85 languages in the sample that exhibit one of the more commonly
attested basic word orders. The notation [+case] in this context means having some
kind of marking, including adpositions, on noun phrases to mark the subjectobject
distinction (Mallinson and Blake 1981: 179).
(15) VSO [+ case] 3 SVO [+ case] 9 SOV [+ case] 34
[ case] 6 [ case] 26 [ case] 7
The SVOcaseless languages are concentrated in western Europe (e.g. English),
southern Africa (e.g. Swahili) and east and southeast Asia (e.g. Chinese and Viet-
namese).
1.3.3 Adverbs and relator nouns
It is common in Australian languages to use location words in con-
junction with the locative case. The locative signals location or proximity in general
and the location word indicates the specic orientation of the located entity with
reference to the location. The following example is from Pitta-Pitta:
(16) Nhangka-ya thithi kunti-ina kuku-ina
sit-PRES old.bro house-LOC back-LOC
Elder brother is sitting at the back of the house.
Kuku-ina can be identied with kuku (anatomical) back, but it is not an ordinary
noun here; it cannot be qualied. In any event the full range of words that can be
used to signal relative orientation cannot be analysed in terms of stems identiable
with nouns and sufxes identiable with case markers. Words like kukuina in (16)
are used as dependents of the verb in parallel with locative-marked nouns. Their
position relative to the locative-marked noun is not xed and the two need not be
contiguous. A good literal translation of (16) would be Elder brother is sitting by
the house, at the back.
15
Overview
In (16) the adverb-like word is parallel with the locative-marked word. Another
possibility is for the relative orientation to be specied by a word that is in series
between the head and the location. This is the situation with words like top as in
It stands on top of the cupboard and front as in She is sitting in front of the house.
Top and front are nouns in that they stand at the head of noun phrases, but they are
different from the common nouns top and front in that they cannot be modied by
an indenite article nor by adjectives. They belong to a small subclass of nouns
that are sometimes referred to as relator nouns.
15
Location words used in parallel
with locative-marked nouns, i.e. words like kukuina in (16), could also be taken to
be relator nouns rather than adverbs.
Relator nouns are a feature of Chinese. In the following example qi antou front
is the head of a noun phrase just as front is in the translation. It is the complement
of zh` an-zai and it has de as its dependent which in turn has d` a-m en as its dependent
(Starosta 1988: 203):
(17) T a zh` an-zai d` a-m en de qi antou
s/he stand-at big-gate of front
S/he is standing in front of the main door.
Chinese also employs combinations of nouns and locational noun roots. The lo-
cational forms do not occur as separate lexical items and in some descriptions they
are treated effectively as postpositions or locative particles (Li and Thompson
1981: 25, 390ff). Since combinations of noun plus locational form can be governed
by prepositions, a better analysis might be to take the combinations to be pseudo-
compounds (pseudo- since the second element does not occur as a free form).
Such an analysis is adopted in Starosta (1988: 206). In the following example m en
qi an is a pseudocompound:
(18) T a zh` an-zai d` a m en qi an
s/he stand-at big gate front
S/he is standing at the main door front.
Under this interpretation the locative relationship is treated derivationally. In
various varieties of colloquial English such compounding occurs to produce locative
adverbs: He went States-side. She went sundown-way (i.e. west). Location is also
treated derivationally in formations such as homewards, skywards or Keats Lethe-
wards. In Bantu languages locative marking on nouns is derivational. As noted in
connection with the Swahili example given above (14), nouns in Bantu languages
fall into a number of classes. Each class is marked by a prex. The classes are
associated with humans, plants and so on, and there are separate classes for the
16
plurals of the various classes. To express location a noun is derived into one of
several locative classes. In Swahili, for instance, there is a pa- class for specic
location, a ku- class for to, from and nonspecic location, and a m(u)- class to
express inside. The word for house is nyumba. It is normally in the n-class. To
express in the house nyumba is sufxed with -ni and put in the m(u)- class. The
n- and m(u)- prexes do not appear directly with nyumba, but they do appear on
dependents via concord and on the verb via cross-referencing agreement. Compare
the following examples,
(19) Nyumba i-na mi-lango mi-wili
house N-has MI-door MI-two
The house has two doors.
(20) Nyumba-ni m-na wa-tu
house-LOC MU-has WA-person
In-the-house has people.
In (19) nyumba is in the n-class which takes i- as its cross-referencing agreement
marker for subject. In (20) nyumba is sufxed with -ni, which looks like an inec-
tional locative case sufx, but nyumba is now in the mu- inside class as indicated
by the subject marker on the verb. The fact that nyumbani is cross-referenced by
a subject marker indicates that it is a subject. Nyumbani is not an inected form
of nyumba, but a form derived into another class. Normally a noun inected for
locative or marked by a locative adposition could not function as subject or object.
The Bantu languages have incorporated both number (singular versus plural) and
location into a noun class or gender system.
1.3.4 Possessive adjectives
A pronominal dependent of a noun can be expressed in the genitive
case, but in some languages possessive adjectives perform a similar function. In
Old English mn is the genitive case form of the rst-person pronoun; mn b at is
my boat. However, genitive forms like mn were reinterpreted as stems and used
as possessive adjectives. Thus He saw my boat was expressed as H e seah mnne
b at where mn takes the masculine, accusative, singular inection in concord with
b at.
In Latin genitive forms of pronouns were generally used as objective genitives
and possessive adjectives were used in a subjective role. For instance, tua memoria
me (your memory me.GEN) means Your recollection of me where tua, the pos-
sessive adjective, corresponds to the subject of the corresponding verb and me, the
possessive pronoun, to its object or genitive complement.
17
2
Problems in describing case systems
2.1 The traditional analysis
The western tradition of describing case systems can be traced back
to the Greeks. Ancient Greek, like the other older Indo-European languages, was
a fusional inecting language in which case marking could not be separated from
number marking, where there was also some fusion of the stem and inection, and
where gender correlated closely with declensional type in the rst (feminine) and
second (masculine) declension. In other words, Greek was structurally parallel with
Latin as illustrated in section 1.1. Given this kind of structure, it is not surprising
that the Greek descriptions of case were based on the word rather than on stems
and sufxes. The modern concept of the morpheme (the smallest meaningful, or
better, smallest grammatical segment) is not very useful in languages where the
expression of several grammatical categories has fused representation across the
whole language, i.e. what Matthews calls cumulative exponence or cumulation
(Matthews 1974/1991).
Cases were described in terms of what are called here case forms, where a case
formis an inected formof a noun. Since the inection of case and number is fused,
it would be more accurate to talk of case/number forms. The termcase is fromLatin
c asus, which is in turn a translation of the Greek pt osis fall. The term originally
referred to verbs as well as nouns and the idea seems to have been of falling away
from an assumed standard form, a notion also reected in the term declension
used with reference to inectional classes. It is from d ecl n ati o, literally a bending
down or aside. With nouns the nominative was taken to be basic, with verbs the
rst person singular of the present indicative. For Aristotle the notion of pt osis
extended to adverbial derivations as well as inections, e.g. dikai os justly from
the adjective dikaios just. With the Stoics (third century BC) the term became
conned to nominal inection (Sittig 1931: 3, Calboli 1972: 87).
The nominative was referred to as the orth e straight, upright or eutheia
straight, direct form and then as the pt osis orth e, pt osis eutheia or later pt osis
onomastik e nominative case. Here pt osis takes on the meaning of case as we know
it, not just of a falling away froma standard. In other words it came to cover all cases
18
2.2 Distinguishing cases
not just the non-nominative cases, which in Ancient Greek were called collectively
pt oseis plagiai slanting or oblique cases and for the early Greek grammarians
comprised genik e genitive, dotik e dative and aitiatik e accusative. The vocative,
which also occurred in Ancient Greek, was not recognised until Dionysius Thrax
(c. 100 BC) admitted it, which is understandable in light of the fact that it does not
mark the relation of a nominal dependent to a head (see section 1.2.3; cf. Hjelmslev
1935: 4, Calboli 1972: 102). The received case labels are the Latin translations of
the Greek ones with the addition of ablative, a case found in Latin but not Greek.
The naming of this case has been attributed to Julius Caesar (Sittig 1931: 1). The
label accusative is a mistranslation of the Greek aitiatik e pt osis which refers to the
patient of an action caused to happen (aitia cause). Varro (116 BC27?BC) is re-
sponsible for the term and he appears to have been inuenced by the other meaning
of aitia, namely accusation (Robins 1967: 35, Calboli 1972: 100).
Case systems of the type represented by Latin and Ancient Greek present two
major problems for description. One is the problem of distinguishing the cases;
the other is the problem of describing their meaning and function. Distinguishing
the cases is a problem, since different classes of stem exhibit a different range of
distinctions, i.e. the paradigms are not isomorphic (see Table 1.2). The traditional
solution is to identify cases across stem classes on the basis of the functions they
have in common. Describing the meaning and function of the cases traditionally
involved nding a principal meaning, which is reected in the label, as well as
listing a range of separate meanings or functions. Theories of how cases are to
be distinguished and of how their meanings and functions are to be described are
discussed in the remainder of this chapter.
2.2.1 The distributional approach
As noted in section 1.1, cases are traditionally recognised on the basis
of a distinction of case formfor any one group of nouns. There is no requirement that
the distinction be made for all classes of noun. In Latin (see Table 1.2), as in other
Indo-European case languages, the nominative and accusative have distinct case
forms for masculine and feminine nouns, but there is neutralisation, or syncretism
as it is often called, of the distinction with neuter nouns. There is also neutralisation
of nominative and accusative with plural nouns of the fourth and fth declensions
and third declension consonant stems. Nevertheless, we recognise the distinction
as applying to all nouns, since it allows us to make exceptionless generalisations
about the forms used for various functions in various syntactic contexts.
Essentially the traditional method involves aligning paradigms on the basis of
functions and then referring to the rows of the sets of paradigms (as in Table 1.2) in
relating expression and function. In Latin, for instance, we want to be able to make
19
statements like the following:
(i) the accusative is used to express the direct object
(ii) the accusative is used to express the object of certain prepositions
(iii) the accusative is used to express duration: xxvii annos for 27 years.
It does not matter, for the purposes of this rule, that the accusative is realised by a
form identical to the nominative in some paradigms.
This is not the only way of dealing with the neutralised distinctions. We could
recognise only the formally distinct forms in each paradigmand relate these directly
to the functions. For the rst declension singular there would be a nominative
vocative, an accusative, a genitivedative and an ablative. For the second declension
singular non-neuter there would be a nominative, vocative, accusative, genitive
and dativeablative, and so on. Rules for encoding syntactic functions would have
to refer to these case forms. For example, rule (i) above, which states that the
direct object is expressed in the accusative case, would have to be amended as
follows:
(iv) The direct object is expressed thus:
With masculine and feminine singular nouns of any declension the
accusative is used.
With plural nouns of the rst and second declensions the accusative is
used.
With plural nouns of the third, fourth and fth declensions and with all
neuters the nominativevocativeaccusative is used.
In the traditional descriptions the indirect object is described as being in the
dative. If we were to describe the case forms in which it appears, the rule would
read something like this:
(v) The indirect object is expressed thus:
With nouns of the rst and fth declensions singular the genitivedative
is used.
With nouns of the third and fourth declensions singular the dative is
used.
With nouns of the second declension singular and with all plurals the
dativeablative is used.
The list of case forms given in (iv) for the direct object would have to be repeated
for all the functions of the traditional accusative, of which there are about four.
Similarly the list of case forms given in (v) for indirect object would have to be
repeated for all the functions of the traditional dative, of which there are six or so.
20
This shows that an attempt to link the case forms directly to functions in Latin is
unwieldy. A consideration of concord shows it is unworkable.
In Latin, attributive adjectives agree with their head noun in case, number and
gender. Adjectives belong either to the rst and second declensions (e.g. bonus
(masc.) bona (fem.) bonum (neuter)) or to the third (e.g. tristis (masc. and fem.)
triste (neuter)). Where an adjective makes more distinctions than the noun it
modies, an insuperable problem arises for a description based directly on dis-
tinct case forms. The word form dominae, for instance, can be genitive or dative. If
it is modied by a third-declension adjective like tristis (which declines like c vis in
Table 1.2), there is no way of determining whether the adjective should be genitive
tristis or dative trist , yet only one of these will be appropriate to the syntactic con-
text. On the other hand, if the traditional method is used, no difculty arises. One
has a simple rule to the effect that an adjective shows concord in case, number and
gender with the noun it modies. Where there is syncretismas with dominae, which
realises genitive and dative singular (as well as nominative and vocative plural),
two homophonous case forms are recognised. Dominae (genitive) takes tristis to
yield dominae tristis of the sad mistress and dominae (dative) takes trist to yield
dominae trist to the sad mistress.
Where a distinctive case form is very restricted in its distribution, it might seem
tempting not to establisha case that applies across the spectrum. In Latin the vocative
is distinct fromthe nominative only with non-neuter nouns of the second declension
singular (cf. domine and dominus in Table 1.2). One might be tempted to subcate-
gorise only non-neuter nouns that were of the second declension as [ vocative].
Under this interpretation, a rst-declension noun like Maria as in Av e Maria Hail,
Mary! would be nominative not vocative though used as a term of address. How-
ever, this interpretation is hardly workable where a masculine noun of the third,
fourth or fth declension singular is modied by an adjective of the rst/second
declension as in O sol laudande Oh sun, worthy of praise where sol is third de-
clension. Here the distinctive vocative case formappears on the modier but not the
head. To operate the concord rule consistently it is necessary to recognise a vocative
case for nouns in general.
However, there is one point overlooked in the traditional description and that is
the pattern of syncretisms. In a language like Latin these may seem rather random
at rst, but two syncretisms predominate. These are the syncretism of nominative
and accusative in all neuters and many plurals, and the syncretism of dative and
ablative in all plurals plus the singular of the second declension and the singular of
the i-stems in the third declension. If we exclude the vocative on the grounds that it
is not strictly a case and if we note that the genitive is almost exclusively an adnom-
inal case, we are left with four adverbal cases which the pattern of syncretisms
divides into two groups: nominative plus accusative, and dative plus ablative.
21
Table 2.1 The Russian partitive
cheese bread
nominative syr xleb
accusative syr xleb
genitive syra xleba
partitive syra, syru xleba
dative syru xlebu
This grouping can be shown to be signicant on other grounds. The nominative
and accusative, for instance, typically encode the complements of the vast majority
of verbs whereas the dative encodes the indirect object of only a small number of
verbs and the ablative encodes the complement of only a handful.
1
The ablative and
the dative, to a great extent, encode adjuncts (see section 2.3.2). Generalisations
across cases, including syncretisms, can be captured in feature notation. This is
taken up in section 2.3.4 below.
2.2.2 Nonautonomous case
In some languages a particular function or meaning is recognised in
the case system not by any distinctive case forms at all but rather by different
syncretisms in different paradigms. In Russian, for instance, the partitive sense,
as with complements of nemnogo a little, is expressed by using genitive forms.
However, with some masculine singular nouns a form coinciding with the dative
is an option. Table 2.1 presents some case forms of syr cheese and xleb bread.
With syr, one can say nemnogo syra or nemnogo syru a little cheese. With xleb one
can say only nemnogo xleba not

nemnogo xlebu. There is no case form peculiar to
the partitive function, but the function is treated differently from both the genitive
functions (possessor, etc.) and dative functions (recipient, etc.), as a comparison of
paradigms reveals. Melcuk calls a case which does not have a case form or case
marker peculiar to itself a nonautonomous case (1986: 66).
A similar situation occurs in Latin where location is normally expressed by the
ablative usually governed by the preposition in. However, with towns and small
islands, where no preposition is used, location is expressed by case forms identical
to the genitive for names belonging to the rst or second declension singular: R omae
at Rome, M l et at Miletus. There are a few third-declension-singular names
with forms the same as the dative: Karth agin in Carthage (although the ablative
Karth agine is an alternative). There is also r ur (third declension) in the country
and dom at home. Domus home has a mixture of second- and fourth-declension
forms. The genitive is dom us or dom .
22
Table 2.2 Latin locative forms
Rome Miletus Carthage
1st decl 2nd decl 3rd decl
genitive R omae M l et Karth aginis
locative R omae M l et Karth agin
dative Romae M l et o Karth agin
ablative R om a M l et o Karth agine
Since the expression of location involves different patterns of syncretism with
different declensions, the locative function is distinguished and we could recognise
a locative case as in Table 2.2. However, if we try to claim that all forms expressing
location are in the locative case, a problem arises. Take the noun urbs city. In the
city is expressed by urbe as in the following example where urbe is in apposition
with the locative Antiochae.
(1) Archias poeta Antiochiae natus est,
Archias poet Antioch.LOC born is.3SG
celebr quondam urbe et copios a
thronging.ABL once city.ABL and wealthy.ABL
The poet Archias was born in the once populous and wealthy city
of Antioch. (Cicero, pro Archia poeta 4:53)
This might at rst appear to conrm the idea that urbe and all ablative forms
expressing location are locative forms homophonous with the ablative; after all,
apposition is normally considered to involve case concord. But note the adjective
copios a which agrees with urbe. If a locative noun like Antiochae were to control
concord, it would require copiosae.
2
Clearly there is a paradigmatic opposition be-
tween forms in -ae (locative) and forms in - a (ablative). We can make the opposition
clearer by considering the apposition of phrases expressing at my home and at
my villa, namely dom meae, vill a me a. The case form dom is traditionally called
locative (it is homophonous with the genitive). The case form vill a is ablative. If
we claim that it is a locative homophonous with the ablative, we are confronted
with having to explain an opposition between meae, which is required by dom, and
me a, which is required by vill a, both these forms belonging to the same paradigm,
namely rst declension singular. The conclusion must be that there is a locative
case, but that it is conned to certain nouns such as domus and to adjectives in
concord with them. One would also have to allow that in the situation illustrated in
(1), and perhaps more generally, apposition operates on the basis of a function or
23
Table 2.3 Latin adverbs
to at from
qu o whither ubi where unde whence
h uc hither h c here hinc hence
ill uc thither ibi, ill c there illinc thence
relation rather than case, in this instance on the basis of the locative grammatical
relation.
Quite apart from these locative forms there are locative adverbs. In fact Latin
makes a three-way distinction between to, at and from forms (Table 2.3).
The occurrence of a set of distinct forms expressing location adds to the evi-
dence afforded by the vestigial locative case that there is a locative grammatical
relation.
2.2.3 The formal approach
In section 2.2.1 it was argued that it would be unwieldy to link case
forms directly with functions and impossible where there was concord involving
a modier that belonged to a paradigm more differentiated than the paradigm of
the head (recall the distinction between dominae tristis of the sad mistress and
dominae trist to the sad mistress). Nevertheless, there are descriptions of case
that deal either wholly or at least partly in case forms or case markers rather than
traditional cases, although such descriptions do not involve the worst-case scenario
outlined with reference to Latin in section 2.2.1.
Direct reference to case markers and case forms is common practice among Aus-
tralianists describing the core case system of Pama-Nyungan languages. Among
these languages it is common to nd the case-marking schema shown in
Table 2.4.
Table 2.4 Core case marking in Pama-Nyungan
function noun pronoun
P - -nya or -nha
S - -
A -lu, -ngku, -Tu -
Instrumental -lu, -ngku, -Tu -
T stands for an alveolar stop which may assimilate to
match a dental, retroex or palatal stem-nal consonant.
24
At this point the pre-theoretical syntactic functions S, Aand P will be introduced.
S is the single argument of an intransitive predicate; A is the agent argument of a
transitive verb, and P is the patient argument of a two-place transitive verb. The
terms A and P extend beyond agent and patient to other roles that are treated
grammatically in the same way as agent and patient respectively, so not only would
Martha hit Ruben be described in terms of A and P, but so would Martha saw
Ruben. The denition A is the agent argument of a transitive verb is meant to
exclude the agent adjunct of the passive, since this agent is not treated the same way
grammatically as the agent argument of a transitive verb.
3
As can be seen from Table 2.4, nouns in Pama-Nyungan languages have case
marking for Afunction opposed to an absence of case marking for Sand Pfunctions.
The marking for A function is generally known as ergative case marking and a
system of case marking that makes an SP-versus-A distinction is referred to as an
ergative system, an ergativenominative system or an ergativeabsolutive system.
The term absolutive is used in some circles for a case or case form covering S+P
functions.
4
Normally such forms are unmarked. With pronouns, or at least with rst-
and second-person pronouns, most Pama-Nyungan languages have case marking
for P, i.e. accusative case marking, opposed to a common marking for S+A, which is
usually zero, at least for nonsingular pronouns. Asystemthat makes an SA-versus-P
distinction is called an accusative system.
If one were to recognise cases on the basis of distinctive marking in any one
paradigm or distinctive treatment across paradigms as in the classical tradition,
then one would need to recognise a nominative for S function, an accusative for
P function and an ergative for A function. These cases would not be exclusive to
S, P and A since they encode nominals in a predicative relation to S and to P (Mick
is a man, They made him a man), both objects in double-object constructions (She
gave the man bread ) and nominals in apposition to S, P and A. Also, in a majority of
Pama-Nyungan languages the ergative expresses the instrumental function as well
as A. In a traditional-type analysis the accusative would be said to be expressed by
-nha (or in some languages -nya) with pronouns and by - with nouns. The ergative
would be described as being expressed by case markers such as -lu, -ngku, etc. with
nouns and with pronouns. However, in Australianist practice nouns are said to op-
erate in an ergativeabsolutive paradigm and pronouns in a nominativeaccusative
paradigm. A noun marked by -lu or -ngku etc. is said to be in the ergative case, but
an unmarked pronoun in A function is described as being in the nominative case.
Conversely a pronoun marked by -nha or -nya is said to be in the accusative case,
but an unmarked noun in P function is said to be in the absolutive case.
Since one of the paradigms involved is a noun paradigm and the other a pronoun
paradigm, problems of concord do not loom large, but they do occur.
5
25
Consider the following sentence from Pitjantjatjara (Bowe 1990: 49):
(2) Paluru=rni tjitji tjukutjuku-ngku nya-ngu
3SG.ERG=lSG.O child small-ERG see-PAST
He saw me when he was a small child.
Paluru is a pronoun in A function. The form is the same as would be used for S
function; there is no specically ergative marking, and the ergative gloss given here
is in accordance with the traditional interpretation of case. The bound pronoun -rni,
which is enclitic to the rst constituent, encodes the direct object (glossed as O).
The phrase tjitji tjukutjuku-ngku bears the ergative case marker indicating that it is
to be construed with paluru rather than -rni. The simplest way to account for this
ergative marking is via a general rule for predicative nominals to the effect that they
agree in case with their controller, but such a rule cannot be stated if it is dependent
on whether the controller bears particular case marking.
Another problem arises with wholepart constructions where, as in Australian
languages generally, the whole and the part appear in parallel with the case appro-
priate to their function in the clause. In Pitjantjatjara a pronoun whole and a noun
part encoding the patient of a transitive verb will not match in case marking, since
the pronoun will have accusative marking but not the noun (Bowe 1990: 54):
6
(3) Tjilka-ngku ngayu-nya tjina waka-rnu
prickle-ERG 1SG-ACC foot.ACC pierce-PAST
A prickle pierced my foot.
If we adopt the traditional approach (as anticipated in the glossing), then an example
like this will simply fall under the general rule of case matching and will not require
specication of case marking for nouns in P function as opposed to pronouns in P
function.
7
Another circumstance where a noun and pronoun are likely to be used in parallel
is where a noun is used to ll in the reference of a nonsingular pronoun. For instance,
the white man and I is typically something like ngali walpala (we.two white.fella)
in Pama-Nyungan languages. If such a phrase is used to encode the patient of
a transitive verb, there would be accusative marking for the pronoun but not the
noun, and therefore another exception, under Australianist practice, to an otherwise
general principle of concord.
The use of case forms rather than cases in the syntax would have been impossible
for a language like Latin or Ancient Greek for the reasons given earlier. However,
the practice manages to survive in Australia for a number of reasons:
(a) The case-marking asymmetry is largely conned to the core cases, i.e.
those expressing S, A and P (see section 2.3.2).
26
Table 2.5 Lak case-marking schema
function noun pronoun
P, S nom nom
A gen nom
Genitive gen gen
(b) The number of separate paradigms involved is fewer.
(c) The fact that the differences are between a noun paradigm and a pro-
noun paradigm reduces the concord problem.
(d) The distribution of the morphologically unmarked forms is probably
seen as signicant. In a traditional analysis of case in terms of distribu-
tion, this would be obscured. However, if cases are described in terms
of distinctive features, shared characteristics can be captured. This is
discussed in section 2.3.4.
In Lak, a Daghestan language of the Caucasus, the A of a transitive verb is
expressed in the genitive except with rst- and second-person pronouns where it
appears in the unmarked nominative form (Table 2.5). This distribution raises the
possibility of recognising a nonautonomous ergative case that is homophonous
with the genitive in the noun paradigm and with the nominative in the rst- and
second-person pronoun paradigms. Melcuk (1986: 678), however, argues against
establishing a nonautonomous ergative on the grounds that no particular lexical
items need be referred to.
The partitive in Russian is also nonautonomous (see section 2.2.2) in that it always
coincides with the genitive or dative. However, it is peculiar to certain lexical items,
and Melcuk suggests that it would be more complex to specify that one should
use the dative in certain contexts if the noun concerned were x, y, z, etc., than
to specify that the partitive should be used. Using the partitive enables us to state
the expression of dependents in three or so contexts in an exceptionless way, and the
specication of the words that take the dative-type form need be given only in the
lexicon.
8
In grammars of Turkish one nds reference to two cases occurring in one syntactic
environment (see, for instance, Lewis 1967). A number of postpositions are said
to govern the genitive of singular pronouns and the nominative of other nominals.
One such postposition is i cin for.
9
(4) a. siz-in i cin (you-GEN for) for you (sing.)
b. siz-ler i cin (you-PL.NOM for) for you (plur.)
c. bu adam i cin (this man.NOM for) for this man
27
In Turkish there is no concord and one of the stemclasses involved is pronominal
anyway. There would be no advantage in establishing a nonautonomous case just
for the object of one group of prepositions.
A similar situation occurs in Hindi. The ergative postposition ne, like other post-
positions, governs the oblique case.
10
However, with rst- and second-person pro-
nouns, it governs the nominative. There would be no advantage in establishing a
third case which coincided with the nominative for rst- and second-person pro-
nouns and the oblique elsewhere, particularly as such a nonautonomous case would
be peculiar to a single syntactic environment, namely complement of ne.
One nal point. In a majority of Pama-Nyungan languages the same case markers
express the A of a transitive verb and the instrumental. The two functions can be
distinguished on syntactic grounds (see section 3.2.1), but there is no reason to
recognise two cases. Typically there is a different distribution of case markers with
respect to these two functions since the instrumental does not normally occur with
pronouns. Instrumentals are virtually always inanimate and inanimates tend not to
be represented by pronouns in Australian languages. In a few languages, however,
the instrumental expresses the indirect cause (because of ) or aversive function
(what is to be avoided). This is the situation in Margany, for instance (Breen
1981: 3038). In such a language the instrumental occurs with pronouns and there
is an opposition in the pronoun paradigm between - case marking for A function
and instrumental markers like -tu for instrumental function. In this situation two
separate cases need to be recognised, ergative and instrumental as displayed in
Table 2.6.
2.2.4 Summary
The traditional model for describing case systems is based on Ancient
Greek and Latin where three factors are prominent:
(a) There are numerous paradigms (when number is taken into account)
with different patterns of syncretism (i.e. the paradigms are not iso-
morphic).
Table 2.6 Margany case system (partial)
function case noun pronoun
P accusative - -nha
S nominative - -
A ergative -ngku, -tu -
Inst instrumental -ngku, -tu -tu
28
2.3 Meanings and functions
(b) For each case there are a number of functions.
(c) There is concord between the head of a noun phrase and its dependents.
Given these factors it is difcult to see how one could arrive at a simpler analysis
than the traditional one. The only disadvantage is that there is no reference in a
traditional description to the pattern of syncretisms in particular paradigms. All
such syncretism is treated as nonsignicant.
Where there are only a small number of paradigms, say two, where the number of
functions are small, and where there is little or no concord, it is not impossible to deal
in case forms (or case markers). This is the current practice among Australianists
where nouns have ergative marking for A function and zero for S and P, while
pronouns have accusative marking for P and zero for S and A. Such a model of
description picks out the pattern of what are, from the traditional point of view,
syncretisms.
Where a function or functions can be expressed by one case form with some
nouns and another case form with others, it is possible to establish a nonau-
tonomous case with no forms peculiar to itself. This is worthwhile in some in-
stances as with the Russian partitive where the syncretisms are irregular (lexical)
and there are a number of functions involved. Where a single function or syntac-
tic environment is concerned and where the case form alternation is simple, as in
Turkish where certain prepositions govern the genitive of singular pronouns and
the nominative otherwise, it is simpler to treat the alternation as an alternation in
case.
Traditionally cases are described as having a number of functions or
meanings. The accusative in Ancient Greek, for instance, is described as denoting
the direct object and the two objects in a double-object construction.
(5) Edidaksan ton paida t en mousik en
taught.3PL the.ACC boy.ACC the.ACC music.ACC
They taught the boy music.
There is also an accusative of respect or reference which typically marks the
complement of adjectives and indicates in respect of which eld the adjectival
property holds (Smith 1888: 150).
(6) Oudeis anthr opos autos panta sophos
no man.NOM self.NOM all.ACC wise.NOM
No man is himself wise in everything.
29
Lastly there is the adverbial accusative which expresses extent in distance or time.
(7) Treis m enas emeinen
three.ACC months.ACC stayed.3SG
He stayed three months.
Typically the rationale for separating meanings or functions is not made explicit.
Some distinctions appear to have a syntactic basis, but others are semantically based
and in the absence of explicit criteria there is the possibility that the distinctions
are based on the intuition of the grammarian. In fact this is more than a possibility
given that different writers come up with different classications of the meanings
and functions of a particular system. However, with the examples of the Greek
accusative quoted above, there would appear to be a sound syntactic basis for the
distinctions. Objects are governed by a specic set of verbs, namely transitive verbs;
the accusative of respect is a complement of adjectives and participles, and adverbial
accusatives are adjuncts that can be modiers of any predicate and can co-occur
with objects and respect complements.
In the description of case meanings and functions a number of general distinc-
tions emerge. The rst is between the nominative and the other cases; the second is
between the grammatical and the semantic cases, and the third is between comple-
ments and adjuncts. These distinctions are described in the following subsections.
2.3.1 Nominative versus oblique
One of the distinctions that goes back to the Greeks is that between
the nominative and the other cases, collectively the oblique cases (cf. section 2.1).
In some languages there is a formal distinction between the nominative and the
oblique cases inasmuch as there is a special stem for the obliques. This is the
situation in many Dravidian and Daghestan languages, for instance (see examples in
section 4.4.2). The term nominative (Greek onomastik e, Latin n ominat vus) means
naming; the nominative is the case used outside constructions, the case used in
isolation, the case used in naming. In most languages the nominative bears no
marking, but consists of the bare stem; it owes its status as nominative to the
existence of marked cases. Indo-European case languages are unusual in having a
marked nominative in most paradigms. In chapter 2 of Peri hermeneias Aristotle
declared that only the nominative is a noun, the other forms being cases of nouns.
The distinguishing feature of the nominative for the Greeks was that it was the only
form that could encode the subject of an existential predicate. One modern view
is that the nominative represents the noun as a concept pure and simple (Juret
1926: 16), that the nominative form is the case of pure reference (de Groot 1956:
189).
11
Hjelmslev repudiated the nominativeoblique distinction, but described the
30
nominative as a form that could only be dened negatively (Hjelmslev 1935: 2, 96).
Behind all these views lies the notion that the nominative simply denotes an entity
not a relation between an entity and a predicate.
Jakobson saw the leading role of the nominative to be the expression of the topic,
another notion that goes back to the Greeks and the division of the clause into
onoma and rh ema. Onoma, originally name, later noun, is effectively subject
when applied to the structure of the clause. Rh ema saying, proverb, later verb,
is effectively predicate phrase (cf. Robins 1967: 267, Calboli 1972: 89). The nom-
inative is compatible with carrying the topic function, since the nominative has no
clear relational content. People frequently want to talk about entities designated by
nouns, but not so often about relations to entities. For this reason, locatives, abla-
tives, etc. are not normally topics. In fact the locative subjects of Bantu languages,
which can be topics (see (20) in chapter 1), are unusual.
As noted above, the nominative is generally thought of as the case used outside
syntax, the case used in naming, the case used in talking about a lexeme, but Rubio
argues that in Latin the accusative as well as the nominative is used in isolation and
metalinguistically. He sees both of them as cases of pure denotation (1966: 957).
We are reminded of the use of the oblique forms of pronouns in English: Who wants
it? Me. Me, Ill get it. De Carvalho also notes that the accusative in Latin is used
out of context (1982: 257ff, 1985). He sees the nominative in more positive terms
as the case in which one expresses the protagonist (1982: 248, 263).
A curious fact about the nominative is its neglect in grammars of the classical
languages. It is true that it has fewer functions and is used in a smaller range
of syntactic contexts than any of the oblique cases, but it tends to be dismissed
cursorily. Woodcock 1959, in a fairly detailed treatment of the Latin case system,
omits it altogether.
In sum, the nominative is the case used in isolation and is usually morphologically
unmarked. It is the case in which the subject is normally encoded (though not in
ergative or active languages; see sections 3.2, 5.2.2 and 5.2.3) and since the subject
is characteristically associated with the topic of a sentence, there is an association
of nominative and topic.
2.3.2 Grammatical and semantic cases
Another distinction that is often made is between grammatical (or syn-
tactic) cases and semantic (or concrete) cases. The grammatical cases are tradi-
tionally taken to include the nominative and accusative and often the genitive
(Kurylowicz, for instance, includes the genitive (1949, 1964: 188)), and should
include the dative and the ergative. The basis for the distinction is often not made
clear, but if the criterion for classifying a case as grammatical or syntactic is that
it expresses a purely syntactic as opposed to semantic relation, then the dative and
31
ergative should certainly be included since the dative encodes the indirect object
and the ergative encodes A, or what is often referred to as transitive subject. In
this text the transitive subject will be referred to as the ergative relation. The dis-
tinction between A and the ergative relation is that A is a pre-theoretical term of
universal applicability, whereas the term ergative relation should only apply where
A is demarcated by some formal means such as ergative case marking.
If the distinction between grammatical and semantic cases were to be clearcut, the
grammatical cases would encode only purely syntactic relations and the semantic
cases wouldencode onlyhomogeneous semantic relations suchas locationor source.
However, it is common for a syntactic case to encode a semantic relation or role
that lies outside of whatever syntactic relation it expresses. In Latin, for instance,
the accusative not only expresses the direct object, it also expresses the semantic
role of destination, which can be shown not to be subsumed under direct object.
On the other hand, there are situations where the so-called semantic cases encode a
purely syntactic relation. This commonly occurs with the passive where the demoted
subject is often expressed via a semantic case. In Latin the agent of the passive is
normally expressed in the ablative governed by a/ab. This agent is not a semantic
agent, but may cover a range of roles that can be expressed as subject including true
agents (occ sus a c onsule killed by the consul), experiencers (am ata a c onsule
loved by the consul) and perceivers (v sus a c onsule seen by the consul).
The Latin ablative clouds the distinction between the grammatical and semantic
cases in another way inasmuch as the ablative is semantically diffuse. As pointed
out in section 1.1, the ablative expresses the semantic roles of source, location and
instrument. Latin has no semantic case where there is a one-for-one relationship
betweeen case and semantic role, but in languages with larger case systems there
are frequently cases that at least approximate to being semantically homogeneous
(cf. Agud 1980: 454).
Perhaps one could argue that a case is syntactic if its primary function is to
encode a purely syntactic relation. Kurylowicz (1964: 1813) argues that expressing
the direct object is the primary function of the accusative since all the adverbial
functions are determined by context, in particular by the choice of verb or the choice
of accusative noun. An accusative of destination is found only with verbs of motion
(Latin: re R omam to go to Rome) and an accusative of extent is used only with
nouns referring to measures of distance or periods of time (Latin: r egn are septem
ann os to reign seven years, ambul are m lia tria to walk three miles).
It is also convenient for the purposes of making certain generalisations (see,
for instance, section 5.2) to distinguish the cases that encode the complements
of typical one-place and two-place transitive verbs. These comprise nominative,
ergative and accusative. I shall refer to these as the core cases and to the others as
peripheral cases. There is another set of terms in use, namely direct and oblique. But
32
Table 2.7 Types of case
nominative
core accusative
grammatical ergative
genitive
dative
locative
local
ablative
semantic allative
perlative
instrumental
comitative
etc.
Other local cases are possible, but they all involve the
combining of location, source, destination and path
with other notions.
this use of oblique is confusing, since the nominativeoblique distinction is better
established.
In some treatments the termlocal cases is used to cover cases designating notions
of place such as location, source, destination or path.
These distinctions are summarised in Table 2.7.
2.3.3 Complements and adjuncts
A number of writers including Kurylowicz (1949, 1964) and Touratier
(1978) distinguish between instances where cases are governed by verbs or prepo-
sitions and instances where they express adjuncts. What they appear to have in
mind is the redundancy involved in marking a dependent whose role is already
implicit in the choice of verb or preposition. In Latin most two-place verbs take an
accusative complement, but, as illustrated in section 1.1, some take complements in
the genitive, dative or ablative.

Ut to use, for example, governs the ablative as in
gladi o utor I am using a sword. We might compare the use of the ablative to
express an adjunct with instrumental function as in gladi o fer re to wound with
a sword. Kurylowicz takes governed, peripheral case marking with verbs (as in
gladi o utor) to be voided of its semantic contents and to have become an allo-
morph of the ending of the accusative (of direct object), a simple sign of syntac-
tical subordination (1974:193). In similar vein Nichols considers that in Russian
there is no difference in syntactic relations between the accusative complement of
ljubit to love, the instrumental complement of interesovatsja to be interested
33
in, the dative complement of udivljatsja to be surprised at and the prepositional
complement of serditsja to get angry with, which takes a preposition na on(to),
which in turn governs the accusative. She describes all of these as the rst object
(Nichols 1983: 171).
However, there are problems with these views. Complements with indirect or
peripheral case marking can usually be shown to be syntactically distinct from
accusative-marked direct objects in that while direct objects match the subject of a
corresponding passive, for two-place verbs with a peripheral-marked complement
there is no corresponding passive. Semantically there are problems too. If one takes
the governed ablative of a Latin verb like ut to use to be voided of its semantic
contents, this implies, as Touratier points out, that the ablative in an adjunct phrase
as in gladi o fer re to wound with a sword realises a different morpheme (Touratier
1978:102). However, the choice of governed case is relatable here and elsewhere
to the use of cases in adjunct functions. The ablative in Latin is the case used
to express the instrumental function and this would appear to be appropriate for
ut to use. Worse still for the viewthat governed cases are not signicant is the fact
that some alternations occur and these tend to be meaningful. In Latin moder ar to
govern takes the dative or the accusative. The dative tends to be used with reference
to governing ones tongue moder ar linguae or ones anger moder ar rae, i.e. to
exercising self-restraint, whereas the accusative tends to be used for governing or
controlling something external whether it be horses or territory.
12
It is true that
case marking is partly arbitrary, but it is not random (cf. Nichols 1983: 190). One
broad generalisation that holds across languages is that verbs taking peripheral
complements are not activity verbs involving impingement on a patient, though this
is not to imply that all verbs with accusative-marked direct objects are such verbs.
Writers like Kurylowicz and Touratier are right to point out the redundancies in
case use. A clear instance can be found in English where verbs and prepositions
always govern the accusative, which therefore contributes practically nothing to
the meaning. However, where verbs and prepositions can govern various cases,
the choice is likely to be signicant. Even where a case makes no independent
contribution to the meaning because the role is determined by the verb or adposition,
the case can still have a meaning, even if there is redundancy (as with the Latin verb
ut to use governing the instrumental ablative). Moreover, the choice of case may
have syntactic signicance, e.g. marking a direct object as opposed to a peripheral
complement, which in turn can have discourse-pragmatic signicance.
13
2.3.4 Case features
As noted in section 2.3, cases are traditionally described as having a
number of functions. Atypical Latin grammar will list the principal functions of the
ablative as separation, instrumental and locative. Within separation there will be
34
subtypes such as separation or source (Ath en s rede o I return fromAthens), origin
(optim s parentibus n atus born of excellent parentage) and comparison (l uce
cl arior clearer than daylight). In using a single expression unit such as the ablative
case to express a number of different functions, the language is exploiting the
notion of complementary distribution. The functions can usually be distinguished by
the choice of governor or the choice of lexical item bearing the case. In the phrase
Ath en s rede o I returnfromAthens the separationor source sense is deducible from
the meaning of red re to return; in Ath en s habit o I live in Athens the locative
sense is deducible from the meaning of habit are to live in; in the sentence Tot a
Itali a erat bellum (all.ABL Italy.ABL was war) There was war throughout Italy the
locative sense is deducible from the fact that Italia refers to a place (cf. Rubio
1966:15587). This complementarity raises the possibility of arriving at a single
generalised meaning of which the particular meanings are predictable variants. The
notion of a generalised meaning, for which the German term Gesamtbedeutung
(whole/aggregate meaning) is sometimes used, can also be arrived at via a different
path. Instead of considering the particular meanings or functions of a case, one
can look at cases solely from the point of view of their position in the system of
cases, in the system of oppositions. One can then allot them a characterisation that
is sufcient only to distinguish each case from the others in the same system.
The origin of the generalised meaning goes back to the classical period, but it
becomes prominent in the thirteenth century, both in the work of the Byzantine
grammarian Maximus Planudes (1260c.1310) and in the work of the scholastic
grammarians (Hjelmslev 1935: 1lf, Serbat 1981a: 246), where generalised char-
acterisations were exploited in presenting case inventories as systems. Simon
the Dane, for instance, characterised the Latin genitive and ablative as express-
ing origin as opposed to the dative and accusative which express what I will call
destination. A cross-cutting distinction is made between relations of substance to
substance, i.e. adnominal relations, and relations of substance to action, i.e. adverbal
relations (Serbat 1981a: 246). These binary distinctions can be presented in terms
of distinctive features as shown in Table 2.8.
The notion of markedness was not a feature of pre-twentieth-century linguis-
tics. Obviously there is some redundancy in having values for both substance-to-
Table 2.8 Latin case system (Simon the Dane)
nom acc gen dat abl
origin + +
destination + +
substance-to-substance + +
substance-to-action + +
35
substance and for substance-to-action. Simon includes a further distinction between
intransitive cases (nominative and vocative) and transitive ones. Nominative is char-
acterised as expressing the suppositum actuale and the vocative as expressing the
suppositum virtuale. The suppositum is what is presupposed or given and the sup-
positum actuale relates to the discourse function of the subject. Communication
presupposes a speaker (or writer or signer) and an addressee. The addressee is
the suppositum virtuale. Simons treatment of the nominative/vocative opposition
does not bring out clearly the unmarked character of the nominative vis-` a-vis the
vocative, although actuale as opposed to virtuale could be interpreted as indicat-
ing this. Establishing these generalised characterisations can involve overlooking
or minimising part of the evidence. Simon takes the dative to express the rela-
tions of substance-to-substance (adnominal), but the dative has adverbal as well as
adnominal functions.
A contemporary, Martin the Dane, produced a system which recognised that
the dative and ablative cases had both adnominal and adverbal functions, though
he seems to have overlooked the fact that the genitive in Latin has some limited
adverbal functions. As noted in section 1.1, a few verbs such as oblivisc to forget
and miser er to pity take genitive complements. His system has the virtue of
picking out the cases that can be adnominal (genitive, dative, ablative) from those
that cannot (see (10) below). Anotable feature of Martins systemis that it is localist.
In a localist theory not only are the local cases characterised by local notions such
as source and destination, but the other cases are also interpreted as more abstract
instantiations of these notions. In Martins system the nominative is seen as a case
of origin since the action typically proceeds from the subject, and the vocative is
characterised in terms of the feature towards since communication is directed
towards an addressee (Serbat 1981a: 267). A point about this system that may
seem curious is that Martin characterises the dative as a to case and the accusative
as a towards case. Even those unsympathetic to abstract characterisations of cases
would probably be happy to see a common feature such as to used to capture the
notion of destination common to the accusative and dative. Martin may have been
motivated to nd a way of distinguishing the [+ terminus] cases without resorting
to the adnominal/adverbal distinction. His system is displayed in terms of features
in Table 2.9.
The interpretation of the case system of Ancient Greek developed by the Byzan-
tine scholar Maximus Planudes, as presented in Hjelmslev (1935: 1315), is dis-
playedinTable 2.10. InGreekthere were ve cases: nominative, accusative, genitive,
dative and vocative (though the vocative is excluded in the scheme of Planudes).
The genitive covers the notion of from as well as possessor and the dative includes
locative and instrumental functions. According to Hjelmslev, Planudes was the rst
to see the cases fromthe point of viewof the object rather than the subject. Note that
36
Table 2.9 Latin case system (Martin the Dane)
nom acc gen dat abl voc
origin + +
terminus + + +
substance-to-substance + + +
substance-to-action + + + +
to +
from +
towards + +
the dative holds the central, neutral, pivotal position on a scale running from what
Hjelmslev calls rapprochement (bringing nearer) to eloignement (taking away). The
nominative is also neutral with respect to those two poles, but it is independent rather
than dependent.
The dominant model for describing case systems in the Renaissance and right
down to the present day has been the traditional method of listing a number of
meanings and functions for each case as illustrated in section 2.3 above. However,
in the early nineteenth century, a period famous for the development of comparative-
historical linguistics, the notion of cases having a single, abstract meaning can be
found in the work of linguists such as Rask, Bopp and W ullner. Not surprisingly,
the notion of Gesamtbedeutung becomes prominent in the structuralist period of
the twentieth century. The leading proponents are Hjelmslev and Jakobson, both of
whom published important, inuential works in the mid thirties.
Hjelmslev
For Hjelmslev, a case, like linguistic forms in general, does not signify
several different things. It signies a single abstract notion from which one can
deduce the concrete uses (Hjelmslev 1935: 85). The meaning of a case cannot be
determined in isolation, but only from a consideration of the oppositions within
the case system. Hjelmslev was a localist, in fact it was he who rescued Maximus
Planudes from oblivion (see above), and he captured the oppositions within a case
system in localist terms. The rst dimension in a case system is direction, of which
the positive termis rapprochement (bringing nearer) and the negative eloignement
Table 2.10 Planudes case system of Ancient Greek
(to) + 0 (from)
dependent acc dat gen
independent nom
37
(taking away) (1935: 128). To see how this works, let us look at Hjelmslevs inter-
pretation of Greenlandic Eskimo (Hjelmslev 1937: 6575). In this language there
are the following cases: nominative, ergative-genitive, equative (or predicative), in-
strumental, ablative, allative, locative and a through case which is called perlative
elsewhere in this text (Hjelmslev calls it prosecutive). The last four of these cases
are clearly local, and Hjelmslev captures them in terms of rapprochement to and
eloignement from as follows:
ablative from
allative to
locative neither from nor to
perlative both from and to
The second dimension in Hjelmslevs system is degree of intimacy: coherence
versus incoherence (1935: 128). This is another local notion and Hjelmslev
illustrates the distinction with prepositions. Coherence involves contact or pene-
tration whereas incoherence merely involves proximity. She went into the building
(coherent); she went to the building (incoherent); she went between the buildings
(indifferent to coherence). With Eskimo Hjelmslev uses this dimension to capture
the distinction between the local cases and the syntactic cases. The local cases
were taken to be indifferent to the distinction and the syntactic cases to be inco-
herent. However, in Table 2.11 the local cases are shown as [+ coherent] since, for
Hjelmslev, they represent the positive termof the opposition. He aligns the ergative-
genitive with the ablative (the action proceeds fromthe agent) and the equative with
the allative (A is equal to B). The nominative which is pivotal and which does not
involve either the notion of to or of from is matched with the locative, and the
instrumental with the perlative, an analogy which can be rationalised in terms of
Table 2.11 Hjelmslevs case system of
Greenlandic Eskimo
[coherent] [+coherent]
+from
to
ergative ablative
+to
from
equative allative
+from
+to
instrumental perlative
from
to
nominative locative
38
the action passing through the instrument to the patient. The tabloid summary is
presented in Table 2.11 (after Hjelmslev 1937: 74).
Jakobson
Like Hjelmslev, Jakobson broke away sharply fromthe traditional prac-
tice of merely listing case uses or functions. As van Schooneveld describes the
situation (1986: 374):
The impact of such listings was that the majority of linguists considered the
extraction, in a given language, of a semantic invariant of each case beyond
reach. And suddenly, in the same decade [the thirties], appears this native
speaker of Latin (because as far as case is concerned, Latin and Greek are
similar to Russian) who states that to him, there is in his native language an
invariant meaning for each case. Not only do these semantic characteristics
of each individual case have elements in common but they constitute also a
paradigmatic structure.
Jakobson distinguished between the invariant, intensional meaning of a case
and its syntactically and/or lexically conditioned variants which make up the exten-
sion of the case (cf. (5), (6) and (7)). The Gesamtbedeutung of a case is indepen-
dent of the environment and cannot be determined from the individual meanings
(Sonderbedeutungen) nor from the principal meaning (Hauptbedeutung). Cases are
correlative and take their value from their relation to other cases in a system of
oppositions (Jakobson 1936/1971: 356).
Jakobson applied his theory to the case systemof Russian. The cases are: nomina-
tive, accusative, dative, instrumental, genitive and locative, plus two rather restricted
cases: genitive II or partitive (see section 2.2.2) and locative II. The locatives are
also called prepositional. Jakobson takes the nominative to be unmarked. Opposed
to it is the accusative, which is always subordinated to it and which signals di-
rection or goal. The instrumental and dative are opposed to the nominative and
accusative as marginal or peripheral cases (Randkasus) opposed to direct or central
cases (Vollkasus).
(8) central nom acc
marginal inst dat
The dative is aligned with the accusative in that both express the goal of an event.
The instrumental is taken to be the unmarked marginal case just as the nominative
is the unmarked central or direct case. In what most would take to be its core
meaning, namely that of expressing an instrument in a sentence like Petr rezal mjaso
no zom (Peter-NOM cut meat.ACC knife.INSTR) Peter was cutting the meat with
a knife, the instrument is backgrounded relative to the subject and direct object,
39
Table 2.12 Jakobson 1936: Russian case system
central (nom acc) (gen I gen II)

marginal (inst dat) (loc I loc II)
a backgrounding iconically reected in its adjunct and marginal syntactic status
(cf. Wierzbicka 1980). The instrumental case is also used to express backgrounded
agents in the passive. The genitive and locative focus upon the extent to which
the entity takes part in the message (Jakobson 1958/1971: 179), the locative also
being considered marginal. The genitive II (partitive) and locative II are taken to
be marked with respect to genitive I and locative I respectively. Table 2.12 (from
Jakobson 1936/1971: 65) shows the complete system. Within each opposition the
marked case is to the right or below.
In his 1958 paper Jakobson presented the systemin terms of a cube, with what are
features serving as the dimensions: [ direction], [ marginal] (as in Table 2.12)
and [quantication] (partial involvement). The feature [directional] is extended
from the accusative and dative to cover genitive I and locative I. It is also referred to
as [ascriptive] since directional is not so appropriate when it is extended to cover
genitive and locative. I think it would be fair to say that Jakobsons characterisation
of the Russian cases in terms of features is less than perspicuous and there is less
than adequate demonstration of how the feature analysis can be exploited (see the
discussion following Table 2.15). Neidles representation of the system in terms of
a matrix of distinctive features is shown in Table 2.13 (Neidle 1982: 397).
The notion that cases have a single, abstract meaning still has supporters such as
Rubio (1966) and de Carvalho (1980). However, a description of meaning purely
in terms of Gesamtbedeutungen has limitations. As Wierzbicka points out, the
meanings Jakobson ascribes to the Russian cases are too broad to be predictive
Table 2.13 Jakobson/Neidle: Russian case system
Marginal Quantifying Ascriptive
Nominative
Accusative +
Genitive I + +
Genitive II +
Locative II + +
Locative I + + +
Dative + +
Instrumental +
40
and one cannot learn to use the Russian cases from his formulas (1980: xv). The
extension of a case is not always derivable fromthe generalised meaning. However, it
does not followfromthis observation that one should abandon generalised meanings
or at least generalised characterisations. In section 2.2.1 the utility of the traditional
cases was demonstrated. However, it was also pointed out that a description entirely
in terms of distributional cases (the rows in the paradigms) neglected syncretisms
(the vertical dimension). One thing generalised meanings or at least generalised
characterisations are useful for is to serve as the basis for grouping cases into a
hierarchy of sets, each set bearing one or more features in common. A syncretism
between two or more cases is likely to be signicant, especially if it recurs in a
number of paradigms. In the Indo-European case languages there is syncretism
of nominative and accusative (and usually vocatives) for neuters no matter which
declensionor number theybelongto. InRussianthere is syncretismof accusative and
genitive. This covers animate nominals of the rst declension masculine singular or
any animate plural or any personal pronoun. Signicantly the syncretism covers all
the different forms that occur in these paradigms, which would rule out phonological
motivation: slon a (ACC/GEN) elephant, slon ov (ACC/GEN.PL) elephants, etc.
Such syncretisms can be captured in terms of features of the type illustrated from a
variety of linguists from Simon the Dane (Table 2.8) to Jakobson (Table 2.13). The
focus in such an analysis is more on shared formal properties than on meanings,
but the idea of a generalised characterisation remains the same.
As in phonology, where the use of features rather than atomic phonemes becomes
more attractive in proportion to the number of generalisations that can be captured,
so in case the feature analysis becomes more convincing as cases are seen to exhibit
shared behaviour across a range of areas. Here are some kinds of behaviour that
cases share:
(a) Syncretism
Syncretism in particular paradigms is the most readily observable and
the most widespread example of shared properties between cases.
(b) Compound case
In many languages semantic features shared between cases are reected
in the use of separate markers for the features, yielding what Melcuk
calls compound cases (see section 4.4.3). Consider the case system
of Kalkatungu (Table 2.14) which contains an unmarked nominative,
plus four simple cases and four compound cases (only one marker for
each of ergative, dative and locative I is shown). The distribution of
the marker -thi suggests that the locative and ablative have something
in common and a comparison of -thi and -thingu identies -ngu with
the notion from. The aversive is used to express what is to be avoided
41
Table 2.14 Kalkatungu cases
nominative
ergative -thu aversive -thungu
dative -ku allative I -kunha
locative I -thi ablative -thingu
locative II facing -ngii allative II towards -ngiinha
(Keep away from the re!) or the indirect cause (sick from (eating)
bad meat). The marker -thungu is based on the ergative, which ex-
presses agents and instruments, combined with -ngu expressing from.
Both components of the marking are clearly motivated. Similarly one
can identify -nha as to and see motivation for its distribution.
Naturally there can be shared features that are not revealed via com-
pound marking, but the distribution of recurring markers would seem
to be signicant and worth pointing out.
(c) Type of stem
In some languages cases can be distinguished according to the type of
stem they attach to. In Tamil, for instance, the oblique cases of some
singular nouns are sufxed to an oblique stem. The word for tree has
a nominative maram, but the other cases are sufxed to a stem maratt-.
The accusative is maratt-ai, the dative maratt-ukku, and so on. This
distinction in the stem serves to set the nominative off from the other
cases (cf. section 4.4.2).
In some languages relator nouns with meanings like top, behind
and bottom can appear only in the local cases.
(d) Complement v. adjunct
Cases may group according to whether they express complements or
not. Insome languages there is a clear distinctionbetweencomplement-
marking cases and cases that mark adjuncts exclusively. In Pitta-Pitta,
for instance, only nominative, ergative, accusative and dative mark
complements.
(e) Adverbal v. adnominal
Core cases such as nominative and accusative are not normally able
to be used adnominally. In Latin the peripheral cases (genitive, dative
and ablative) can be used adnominally, but the nominative, vocative
and accusative cannot.
In some languages one case may be substituted for another. Such a substitution is a
sign of a relatively close relationship and an indication of a markedness relationship
42
between the two. In Latin, the nominative may be substituted for the vocative as
in the following example from Livy where the vocative would have been popule
Alb ane
(9) Aud , t u, populus Alb anus
hear.IMP thou people.NOM Alban.NOM
Hear, you people of Alba.
This conrms what we know from syncretism, namely that the vocative and
nominative are treated similarly. The fact that the nominative may be sub-
stituted for the vocative suggests the nominative is the unmarked member of the
pair.
There can also be evidence for markedness with respect to particular functions.
Ramarao (1976), for instance, notes that in Telugu a verb cannot take a dative
or instrumental unless it has an accusative. He argues that the accusative is un-
marked with respect to the dative and the instrumental as a means of encoding
complements.
Across languages the distribution of syncretism, stem-forming sufxes, case-
marking formatives and a consideration of the syntactic distribution of cases re-
veals that the nominative/oblique, core/peripheral and nonlocal/local groupings are
widespread. In Table 2.15 these are applied to Latin with oblique, peripheral and
local being taken as positive. The vocative is set off from all the other cases by the
feature [+ addressee] and the genitive is distinguished from the other peripheral
cases by the feature [+ possessor].
It is interesting to note that this system of features, which is largely derived from
data in a large number of languages, facilitates not only the capturing of pervasive
syncretisms in Latin, but also a number of other generalisations about syntactic
distribution.
The feature specication [addressee] captures the set of adverbal cases, i.e. all
the cases that can mark dependents of the verb.
Table 2.15 Feature analysis of Latin case system
voc nom acc gen dat abl
addressee +
oblique + + + +
peripheral + + +
local + +
possessor +
43
The feature [oblique] captures nominativevocative syncretismand[+oblique]
captures the set of cases that can be adjuncts to verbs and which can be dependents
of participles and adjectives. In Latin a genitive, dative or ablative phrase can be
adnominal (see (10) below) but with adjectives and participles there is an extra
possibility. There are phrases such as n udae lacert os (nude.NOM.PL arm.ACC.PL)
bare as to the arms where an accusative phrase is dependent on an adjective. This
accusative of respect is in imitation of the Greek construction illustrated in (6)
above. The feature [+ oblique] is also relevant to the gerund or verbal noun. The
gerund can only be [+ oblique].
14
The feature specication [ peripheral] captures the syncretism of nominative,
vocative and accusative found in all neuters and in the plural of the third, fourth
and fth declensions. The specication [+ peripheral] captures the cases that can
be adnominal:
(10) a. genitive
vir magn ingeni
man great.GEN talent.GEN
a man of great talent
b. dative
locus r egn o
place kingdom.DAT
a place for a kingdom
c. ablative
vir summ a prudenti a
man highest.ABL prudence.ABL
a man of the greatest prudence
The syncretism of the dative and ablative is captured with two features [+ pe-
ripheral] [ possessor]. This syncretism is found in all plurals and in the singular
of the second declension and third-declension i-stems.
The genitivedative syncretismcan be captured with the specication [+oblique]
[local]. It occurs only in the singular of the rst and fth declensions and is there-
fore somewhat marginal, i.e. closer to being an accidental rather than systematic
syncretism. Nevertheless genitivedative syncretism is not uncommon across lan-
guages and since the proposed system is based on widely distributed properties of
language it can capture this syncretism fairly economically.
In the third-declension-singular i-stems there is nominativegenitive syncretism.
This is isolated and therefore regarded as accidental. The feature system is not
designed to capture this syncretism which is not common across languages.
44
The feature [+ local] covers accusative (which expresses to and through)
and the ablative (which expresses in/at and from).
15
These are the two cases
that can be governed by prepositions, and which can complete the valency of a
verb like ponere to put down. They are also the only cases in which the supine
can appear. The supine is a nominal form derived from a verb. In the following
sentences admonitum and dict u are supines.
(11) a. V enimus t e admonitum
come.PERF.1PL you.ACC remind.ACC
We have come to remind you.
b. Facilis dict u
easy say.ABL
easy to say
It is clear that an analysis of cases in terms of features can facilitate generalisations
covering classes of cases. The value of a feature analysis will depend, in part, on the
overall grammatical framework adopted for description. Here is an example of how
feature notation can be exploited in a LexicalFunctional Grammar framework. The
example is from Neidles description of case in Russian (1988).
In this language the object of the verb may be in the accusative or the genitive.
(This is quite a separate matter from the accusative/genitive syncretism referred to
above.) Neidle treats object case marking as something that is assigned structurally.
The rule expanding the verb phrase is annotated to show case assignment on the
object:
(12) VP V NP
(CASE = [ , (), +])
(12) says that a verb phrase (VP) consists of a verb (V) and a noun phrase (NP).
This noun phrase has a case specied in terms of three features. These three fea-
ture specications refer to [ marginal], [ quantifying] and [ ascriptive] as in
Table 2.13. Actually Neidle uses a different set of features in her 1988 description,
but this does not affect the issue in hand. The notation in (12) says that the noun
phrase in the verb phrase has case with a minus value for marginal and a plus
value for ascriptive. The value for quantifying, which will distinguish accusative
and genitive, is left unspecied (hence the round brackets), but is given a default
value negative which characterises the accusative. This notation neatly captures
the pair accusative and genitive and the fact that the accusative is the unmarked
option. Oversimplifying the facts we can say that the genitive is used for a non-
specic object and generally for the object in a negative clause. Interestingly the
45
accusative/genitive case alternationalsoshows upontime expressions (Neidle 1988:
10, 167).
(13) a. On ne spal odnu minutu
he not slept one.ACC minute.ACC
He did not sleep (= was awake) for one minute.
b. On ne spal odnoj minuty
he not slept one.GEN minute.GEN
He did not sleep (even) for a minute.
Neidle suggests that the case assignment is structural and applies to postverbal noun
phrases irrespective of whether they are objects. However, word order in Russian
is grammatically fairly free, so the notion of structurally assigned case involves
a basic subjectpredicate structure which can be captured with rules of the type
Sentence NP VP, VP V NP with subsequent scrambling rules. The notion
of assigning case structurally comes up again in section 3.3.
2.3.5 Summary
Traditionally cases are described as having a number of functions or
meanings. The Greek or Latin accusative, for instance, would be said to express the
direct object and to express the meaning of extent in place or time. There is another
approach in which cases are seen as a system, each one having a single, general
meaning. These general meanings are not self-sufcient; one cannot predict from
the generalised meaning to the set of contexts in which a case can be used. However,
generalised meanings, or at least generalised characterisations, can form the basis
for a componential analysis of case whichenables one tocapture similarities between
sets of cases.
46
3
Modern approaches to case
3.1 Introduction
This chapter deals with themes that have come to the fore over the
course of the last forty years. Since the early sixties linguistics has been dominated
by the theories of Chomsky. His inuence is evident not only in works couched
in a Chomskian framework, but in the Case Grammar of Fillmore, the Relational
Grammar of Perlmutter and Postal and the LexicalFunctional Grammar of Bresnan.
These theories and indeed most of the thirty-odd theories that have been advanced
over the last few decades have been provoked by and are a reaction to some facet
of Chomskys theoretical approach.
At the beginning of chapter 1 case was described as essentially a system for
marking dependent nouns for the type of relationship they bear to their heads.
It was pointed out that the term case traditionally referred to inectional mark-
ing (section 1.1), but could be extended to cover prepositions and postpositions
(section 1.2). Other means of signalling the type of relationship dependent nouns
bear to their heads, such as word order, were referred to as competing mechanisms
(section 1.3). In recent theories the view has emerged, foreshadowed in Hjelmslev
(1935: 21), that all these mechanisms may be used to signal case, that case is abstract
existing independently of the means of expression, that it is universal. This view is
described in section 3.3 below.
While attention has been focussed on what constitutes the expression side of case,
even more attention has been paid to the question of determining what the relations
are that cases express. A prominent theme has been the view that there is a small
list of universal semantic roles such as agent, experiencer, patient and instrument
(see section 3.4).
A third theme of the recent literature has been the notion of the hierarchy.
Case marking, cases, semantic roles and grammatical relations can all be arranged
hierarchically. This notion is described in the nal section of the chapter
(section 3.5).
As a preliminary to this review of recent approaches to case some data from
Kalkatungu is presented (section 3.2) since this language affords a neat example of
47
howgrammatical relations can be distinguished fromcases, howconicting systems
of organising the core can coexist, and the issue of whether SP rather than SA could
be described as subject.
3.2 Grammatical relations
In the previous chapter cases were described as having functions
(e.g., object) or meaning (e.g., source). In the recent literature it has become com-
mon to talk of morphosyntactically distinct grammatical relations. Grammatical
relations may be semantically heterogeneous (the subject, for instance, typically
expresses a variety of semantic roles including patient (He got run over), agent
(She did it) etc.) or they may be homogeneous. Peripheral grammatical relations
tend to be semantically homogeneous; the locative case relation, for instance, may
encode just locations.
The notion of grammatical relations that are not in a one-for-one correspondence
with morphological cases is not new. Among the grammarians of the Roman em-
pire the question of recognising different semantic roles (which they called cases)
within the Latin ablative was raised, but the criteria used tended to be intuitive or
based on historical or comparative grounds (Calboli 1972: 106ff ). Quintilian, for
instance, distinguished the source role of the ablative from the location and instru-
ment roles, because in Greek, which had lost its ablative, source was expressed by
the genitive, and location and instrument were expressed by the dative (Institutio
oratoria: 1.4.26). Using the insights of structuralist and more recent approaches
to morphosyntax, it is now possible to demarcate grammatical relations on more
objective grounds.
3.2.1 Distinct grammatical relations within one case
Kalkatungu affords some good examples of how grammatical rela-
tions can be distinguished from cases. In this language there are two core cases,
the ergative and the nominative. Intuitively the ergative would appear to encode the
A of a transitive verb and the instrumental function. By the use of paradigmatic
oppositions, one can show that A and instrumental are in fact distinct grammat-
ical relations. In Kalkatungu there is an antipassive construction, which is used
instead of the transitive construction to express action that is ongoing, habitual or
characteristic, or activity directed towards a nonspecic or generic patient.
1
In this
derived construction a nominative-marked S corresponds to A and a dative to P.
A transitive clause with two tokens of the ergative is illustrated in (1a). (1b) is a
corresponding antipassive. Note how this derivation differentiates A and the in-
strumental. The ergative phrase encoding A in (1a) corresponds to the nominative
48
in (1b), but the ergative phrase encoding the instrumental is unaffected by the
derivation.
(1) a. Marapai-thu rumpa-mi ithirr matyamirla-thu
woman-ERG grind-FUT seed.NOM grindstone-ERG
The woman will grind the seed with the grindstone.
b. Marapai rumpa-yi-mi ithirr-ku matyamirla-thu
woman.NOM grind-AP-FUT seed-DAT grindstone-ERG
The woman will grind seed with the grindstone.
The reexive in Kalkatungu is also a derived intransitive. Here the A of the
transitive verb is re-expressed as S and P is deleted. An instrumental is
unaffected.
(2) a. Marapai-thu karri-mi pirlapirla thupu-ngku
woman-ERG wash-FUT child.NOM soap-ERG
The woman will wash the child with soap.
b. Marapai karri-ti-mi thupu-ngku
woman.NOM wash-REFL.FUT soap-ERG
The woman will wash with soap.
An instrumental can appear with underived intransitives, but in the nature of
things there are only a few possibilities such as tyanparra-thu ingka to walk with
a stick.
There is a kind of converse to the antipassive and reexive in which the instru-
mental phrase rather than A can be re-expressed in the nominative, though as P
rather than as S. This is illustrated in (3), which should be compared with (1a).
The verb is marked by -nti-, which signals the advancement of instrumentals and
locatives to the nominative. Note that the ergative phrase for A is unaffected. Note
too that the P of the transitive verb rumpa is demoted to the dative.
(3) Marapai-thu rumpa-nti-mi matyamirla ithirr-ku
woman-ERG grind-IA-FUT grindstone.NOM seed-DAT
The woman will use the grindstone for [to grind] the seeds.
There is another means of distinguishing the two types of ergative phrase in
Kalkatungu. The number and person of an A phrase (or S phrase) but not an instru-
mental phrase can be represented on the verb. The third-person-plural formfor Sand
A is -na; the third-person-singular form for S and A is -. In (4a) the instrumental
49
phrase but not the A phrase is plural, so the verb is not overtly marked. In (4b) the
A phrase as well as the instrumental phrase is plural and the third-person-plural A
is marked on the verb.
(4) a. Papi-yi kati-mba pirlapirla malhtha-yi kulapuru-thu
fas.mo-ERG cover-PERF baby.NOM many-ERG blanket-ERG
Granny covered the baby with blankets.
b. Papi-mia-thu kati-mba-na pirlapirla malhtha-yi
fas.mo-PL.ERG cover-PERF-3PL baby.NOM many-ERG
kulapuru-thu
blanket-ERG
The grannies covered the baby with blankets.
The above illustration involves isolating a case morphologically and then demon-
strating that there are syntactic grounds for saying that it expresses more than one
relation. This raises the question of whether, in the absence of such evidence, a
case expresses a single relation, or, in other words, whether every morphological
case expresses at least one case relation. Typically with peripheral cases there is
no syntactic evidence beyond the morphology. Peripherally marked noun phrases
do not normally control agreement or missing subjects of innitives, for instance.
The view taken here is that each morphological case expresses at least one gram-
matical relation except where a number of cases are co-hyponyms or where the
meaning of one is included in the meaning of another. In Kalkatungu, there is
a to-case (allative) and a towards-case (allative II) (see Table 2.14). These both
express destination, but differ with respect to whether the journey is seen as com-
pleted or not. One would not want to say they represented separate grammatical
relations.
3.2.2 Conicting systems of organising the core
Interestingly the systemof bound pronouns in Kalkatungu is in conict
with the case system. While the core case system is organised on a nominative
ergative basis, the bound pronoun system operates on an SA/P basis. This is
illustrated in the following set of examples. Note rst the inectional case sys-
tem on noun phrases. The second person singular is represented by nyini in S
function (5) and in P function (7). It is represented by nyinti (< *nyin-tu) in A
function (6). Now note the forms of the bound pronouns. We nd -ni in S func-
tion (5) and A function (6), and -kin in P function (7). In these examples the
bound pronouns are enclitic to an auxiliary particle (or auxiliary verb perhaps)
which marks the purposive/future aspect. The SA form is glossed as the subject
50
form and the P form as the object form (see also Table 3.2 for the system of
marking).
(5) Nyini a=ni ingka?
you.NOM(S) PURP=2SG.SUBJ go
Are you going?
(6) Nyin-ti a=ni nuwa?
you-ERG(A) PURP=2SG.SUBJ see
Do you want to see him/her/it?
(7) Nyini a=kin nuwa
you.NOM(P) PURP=2SG.OBJ see
S/he wants to see you.
A number of languages, probably the majority, are like Kalkatungu in having
some kind of bound pronominal representation, usually on the lexical verb or on a
grammatical (auxiliary) verb. I refer to these markers as bound pronouns. In some
languages they can be analysed as clitics, i.e. as elements that are words from
the syntactic point of view but which attach to a host phonologically (see (13) in
chapter 1). In other languages they are better analysed as inection. This is the
traditional analysis of the representation of the person and number of the subject
found in most Indo-European languages. A sample of the person/number marking
in Latin is presented in Table 3.1.
An important question that arises with bound pronouns is whether they alone
represent the core grammatical relations or whether both they and the noun phrases
they coreference represent them jointly. In the traditional interpretation of Latin
and Ancient Greek the representation of the subject on the verb was referred to as
agreement, even though this marking could occur without there being any corefer-
ent noun phrase.
2
In the Indian interpretation of Sanskrit, however, the marking on
the verb was taken to represent the subject, and the nominative noun phrase was
taken to be an adjunct in a kind of apposition with it (see section 3.4.1 below). It
seems to be the agreed interpretation of many languages with pronominal repre-
sentation of all core relations on the verb that any noun phrases expressing core
arguments are adjuncts in a kind of apposition with the bound pronouns and that
Table 3.1 Person/number marking in Latin
singular plural
1st person habit o habit amus
2nd person habit as habit atis
3rd person habitat habitant
51
the bound pronouns bear the grammatical relations. In these languages the noun
phrases expressing the core arguments are generally unmarked. However, this does
not seem to be an appropriate analysis for Kalkatungu where nouns in A function
bear ergative case and where the nominativeergative case distinction rather than
the subjectobject bound pronouns can be shown to be relevant to a number of syn-
tactic rules. Moreover, bound pronouns are not used at all in some clauses. In these
clauses the core arguments are represented solely by noun phrases, yet there is no
reason to distinguish these clauses syntactically from those with bound pronouns
for the core arguments. What we appear to have is dual representation of the core re-
lations in two different systems, an SA/P system with bound pronouns and an SP/A
system with nouns. One could try to avoid this conclusion by taking the clitics to
represent a nominal paradigm, then one would have two nonisomorphic paradigms
and an alignment of the two on the basis of function would yield a system of three
core cases: nominative, ergative (distinguished in the free nominal paradigm) and
accusative (distinguished in the clitic paradigm). This analysis would be the same as
that obtained by applying the traditional analysis to a language with ergative mark-
ing on nouns and accusative marking on free pronouns (see Table 2.4). However,
there are certain disadvantages with such an analysis applied to nominals and clitics.
Let us consider rst of all wholepart constructions, in particular, constructions
involving a person and their body parts. In Kalkatungu, as in Australian languages
generally, the possessor and the body part appear in parallel, both marked with the
case appropriate to their semantic role in the clause. The English sentence The snake
is crawling towards your leg would be translated literally as The snake is crawling
towards you, towards leg, and Ants are crawling on your leg would be Ants are
crawling on you, on leg. If a possessor and their body part are the patient of a transi-
tive verb, then the possessor may be represented by an object clitic. In the following
example the possessor is encoded both as an object clitic and by a noun phrase.
(8) Ngayi a=ngi-(i)na itya milthi milnga-ngku
me.NOM PURP=1SG.OBJ.3PL.SUBJ bite eye.NOM y-ERG
Flies are biting me [in the] eye(s).
If we treat the clitic pronouns as a nominal paradigm and recognise an accusative
case on the basis of the bound form for the object, then ngayi in (8) would be
accusative but not milthi. To make this clear some free and bound forms have been
aligned on the basis of function in Table 3.2. Only one ergative marker is shown
and only two clitic paradigms. The forms ngayi and ngi in (8) correspond with the
row labelled P and the form milthi corresponds with the row labelled other.
This approach has the disadvantage of throwing up an apparent exception to the
general rule that possessors and their body parts match in case, and has the further
disadvantage of requiring the recognition of an extra case, an unmarked case for
52
Table 3.2 Kalkatungu case and clitics
clitics
noun, pronoun
function marking 1SG 2SG
S - lhaa n(i)
A -ngku lhaa n(i)
P - ngi kin
other* -
See (8), (9) and (10)

the body part (see the row labelled other in Table 3.2). Note that this example
is not parallel to the Pitjantjatjara example presented as (3) in chapter 2 where it
was argued that an accusative should be recognised for all nominals if accusative
marking appeared on some.
A similar problem arises with the double-object construction. Kalkatungu is like
English in having two constructions with the verb anyi to give, one in which the
recipient is expressed as a destination (9a), and the other in which both the recipient
and the gift are expressed in the case appropriate for P (9b):
(9) a. Yanyi-ngku a=ina awa ntia
white.man-ERG PURP=3PL.SUBJ give money.NOM
nga-tyinha
me-ALL
The white men are going to give money to me.
b. Yanyi-ngku a=ngi=(i)na ngai
white.man-ERG PURP=1SG.OBJ=3PL.SUBJ me.NOM
awa ntia
give money.NOM
The white men are going to give me money.
Since the recipient is cross-referenced by an object clitic, I take it to be the direct
object and the patient to be a secondary object. As with constructions involving
possessors and body parts, taking the clitics to be a nominal paradigm would mean
taking the recipient object to be accusative and the patient object to be nominative.
This would hold not only in clauses where there was an object clitic, but even in
clauses where the objects are represented only by noun phrases.
At this point it might be worth considering again just what a case is. Cases are
categories into which the relations nouns bear to their heads are classied. Cases are
not always in a one-for-one correspondence with grammatical relations. If that were
the situation, we would deal only in relations and the word forms or markers that
53
express these relations. If, for example, the nominative forms in Latin expressed
only subject, the accusative only direct object, the dative only indirect object, and
so on through the whole system, then we would talk of subject forms, direct object
forms and indirect object forms. There would be no need for the notion of case,
just as there is no need for notional categories between tense markers and tense
categories or aspect markers and aspect categories.
3
The problem with integrating a clitic system into a case-marking system is that
the clitic system picks out particular grammatical relations. If syntactic data can be
imported into morphological paradigms, adpositional systems or systems involv-
ing both inectional case and adpositions, then we are confusing two classica-
tions and destroying the opportunity to explicate the relationship between the two
classications.
The effect of the cross-cutting of the two systems of core grammar is illustrated
in Table 3.3. As can be seen, S is expressed by a nominative noun or a subject-bound
pronoun, A by an ergative noun or a subject-bound pronoun. As illustrated in (4),
an instrumental is expressed in the ergative, but cannot be expressed by a bound
pronoun. P is expressed by a nominative noun and an object-bound pronoun. Note
the top right-hand box labelled adjuncts and complements in the nominative case.
Here are some of the possibilities it covers:
(a) the secondary ( patient) object in the double object construction as
in (9)
(b) the body part adjunct to P in transitive clauses like (8)
(c) the body part adjunct to S in clauses like (10)
Table 3.3 Case and case relations in Kalkatungu
54
(10) Nyini arnka unu=n putu
you.NOM ail LEST=2SG stomach.NOM
You might get sick [in] the stomach.
As we shall see in section 3.4.3, Relational Grammar treats these complements and
adjuncts as having been displaced from the absolutive relation.
Although a comparison of the case system and the bound pronominal system
isolates S, A and P, it is not necessary to refer to S in the grammar. Most of the
syntactic rules refer to the absolutive grammatical relation (SP) (see section 3.2.3).
The subject relation and object relations are manifest only in the bound pronoun
system.
3.2.3 The status of SP as a grammatical relation
In an accusative language S and Aare identied by such characteristics
as case, control of agreement on the verb, word order, and the fact that they are not
normally expressed with nonnite verb forms. The SA combination is described as
bearing the subject relation. In a typical ergative language S and P are identied
via case and the question arises of whether they should be described as bearing
the subject relation. Although a number of suggestions to this effect have been
put forward over the last hundred years or so, they have not been numerous over
the last generation. Some eschew the idea of calling SP subject because it would
be confusing. Others point out that in most languages with a nominativeergative
case system, there are identications of S and A in other parts of the grammar, in
the case marking of pronouns, for instance, or in the bound pronoun system (as we
have just seen in the previous section). They prefer to use the term subject for an
alignment of S and A wherever it is found.
Kalkatungu, fromwhat we have seen, has a subjectobject-bound pronoun system
along with a nominativeergative case system. However, it should be noted that the
SA alignment shows up only in the bound pronoun system; elsewhere it is the
relation we are calling absolutive (i.e., SP) that prevails:
(a) A purpose clause must be detransitivised if its A is coreferent with
the absolutive of the governing clause (illustrated in (50) and (51) of
chapter 4).
(b) In an innitive clause it is the absolutive that is unexpressed (illustrated
in (23) in chapter 6).
(c) Only the absolutive can be relativised.
This last point is illustrated in (11), (12) and (13), at least as far as S, A and P are
concerned. Each example contains a subordinate clause with a participial form of
55
the verb. In (11) the relativised function in the subordinate clause is S which remains
covert and is represented here by [ ],
(11) Ngulurrrnayi-nha nga-thu yurru [ ] ngartathati-nyin
grab-PAST I-ERG man.NOM sit-PARTICIPLE
I grabbed the man [as he was/who was] sitting down.
In (12) the relativised function is P so P remains covert,
(12) Ngulurrmayi-nha nga-thu yurru [ ] thuku-yu itya-nyin
grab-PAST I-ERG man.NOM dog-ERG bite-PART
I grabbed the man [as he was/who was] being bitten by a dog.
However, if any role that is not normally encoded as absolutive is to be relativised,
then a derived verb must be used. In order to say I grabbed the man as he was
hitting the dog, where the relativised function is an agent which would normally
be expressed as A, an antipassive derivation must be used. In this construction
S corresponds to the A of a transitive verb and a dative to P (see (1b) above). S in
the following example is, of course, covert.
(13) Ngulurrmayi-nha nga-thu yurru [ ] thuku-u lha-yi-nyin
grab-PAST I-ERG man.NOM dog-DAT hit-AP-APART
I grabbed the man [as he was/who was] hitting the dog.
Clearly the absolutive (SP) plays a part analogous to that played by SA in an
accusative language. Clearly SP is the primary grammatical relation, the privileged
relation in that it is the only one that can be relativised. This being so, the antipassive
is a derivation that allows A access to the privileged relation and indeed it was this
that inspired Silverstein to choose the term antipassive; it is the analogue of the
passive in an ergative system (Silverstein 1976).
Since ergative languages often exhibit SA alignments in parts of their grammar
and since many ergative languages seem to be only supercially ergative in that
they exhibit an SP/A alignment only in case marking, I do not use the term subject
for the grammatical relation embracing S and P. I follow what is fairly widespread
practice in calling it the absolutive relation and I recognise that in some ergative
languages the absolutive is the primary grammatical relation.
Although it has not been popular to use the term subject for SP even in languages
where SP is not only in the unmarked case but clearly the privileged grammatical
relation, there are formal analyses that effectively take SP to be subject. The rst was
Dixons analysis of another Australian language, Dyirbal (Dixon 1972). Casting the
56
3.3 Abstract case
Figure 3.1English and Dyirbal contrasted
language into Chomskys Aspects model (Chomsky 1965), in which the sentence
is rst divided into NP (functioning as subject) and VP (functioning as predicate),
Dixon took the NP immediately dominated by the sentence node in a transitive
clause to be P and the NP immediately dominated by the VP to be A (Figure 3.1).
The analysis offered for Dyirbal would be applied to any language in which S
and P were identied syntactically, i.e., to any language which has ergative syntax.
Analyses along similar lines have also been offered by Marantz (1984) and Levin
(1987).
One last point about grammatical relations. I believe that the term double-object
construction can be applied in Kalkatungu, since a direct-object can be picked out by
the object series of clitics. I take the patient in examples like (9b) to be a secondary
object, since it is marked like an object and it is a complement (as opposed to the
similarly marked body-part noun phrases as in examples like (8)). For the purposes
of most syntactic rules the direct object is subsumed under the larger relation of
absolutive. In a language with a nominativeergative case system, no object-bound
pronouns, and syntactic rules based on the absolutive, the term object and more
so the term double object is difcult to justify on internal grounds. Yalarnnga,
Kalkatungus southern neighbour, appears to have been such a language, although
it must be admitted that the evidence for syntactic ergativity is meagre. It has a
construction with nguny- to give that is analogous to the Kalkatungu construction
illustrated in (9b) but without the auxiliary.
3.3 Abstract case
Even in languages with inectional case systems, case is quite often
abstract in the sense that it is not always realised by a distinctive inectional formor
marker. In English there is an inectional case system, which is conned to personal
pronouns (I/me, he/him, etc.) and who/whomfor some speakers. One could take the
57
view that all nominals in English take case, but that it is realised morphologically
only on personal pronouns. Under this viewthe abstract nature of case becomes quite
apparent, because only a subclass of nominals show any marking. Some linguists
take the view that case is universal. This involves taking all means of indicating
the relationship of dependent nouns to their heads to be potential markers of case.
However, as noted in section 3.2.2, one such mechanism, namely the use of bound
pronouns on the head of a construction to mark grammatical relations, tends to
involve systems in which there is a one-for-one correspondence between marking
and relation. The notion of case is useful only where cases can express more than
one relation. There is a danger in the notion of universal, abstract case of confusing
case with grammatical relations.
Abstract case is found in a number of other recent theories. In Chomskys
Government and Binding model case theory, which is largely about abstract rather
than morphological case, is an important module or component of the grammar. In
this theory a distinction is made between structural case and inherent case.
4
Struc-
tural case (conventionally written as Case with a capital C) is assigned to noun
phrases according to their position in a structural conguration.
5
As we have seen
in the previous section, Chomsky originally took the basic clause structure of the
sentence (S) to consist of a noun phrase (NP) functioning as subject and a verb
phrase (VP) functioning as predicate.
6
The VP consists minimally of a verb (V); in
a transitive clause the object NP is a sister to the verb. In the Government and Bind-
ing model the structure of a clause is considerably more elaborated. An Inectional
Phrase is introduced. The head of this phrase hosts tense and modality. It is referred
to as In. The verb phrase is its complement or inner modier and the subject its
specier or outer modier. This is illustrated in Figure 3.2 with the sentence They
IP
They I'
I
(will)
VP
eat worms
Figure 3.2Government and Binding clause structure
58
3.3 Abstract case
eat worms. Note that I (In) and its complement (VP) form a constituent labelled I
within an IP (In Phrase). In a sentence such as They will eat worms the auxiliary
will would occupy the In position.
The verb assigns accusative case to the NP it governs, namely the object NP.
This assignment of structural case is manifest in morphological accusative marking
in some languages but not all languages. In, if nite, assigns structural nomi-
native case to the subject NP. In most languages the subject will be unmarked
or bear the nominative morphological case, but in a morphologically ergative lan-
guage like Warlpiri the subject of a transitive verb will carry morphological ergative
case.
7
Where these discrepancies between abstract structural case and morpholog-
ical case arise it is clear that abstract structural case approximates grammatical
relations.
Where there are case assignments peculiar to particular verbs or sets of verbs,
these assignments are held to be inherent. In German, for instance, most two-place
verbs take an accusative complement, but a few take a dative complement. The
default assignment is accusative and this can be handled structurally, but the as-
signment of dative needs to be specied in the lexicon for particular verbs and
therefore is taken to be inherent.
The contrast between structural and inherent case shows up in passivisation. In
Government and Binding Theory all noun phrases must be assigned case. Apassive
verb is considered incapable of assigning case and this motivates the movement
of the object to the subject position where it receives nominative case from In
(if nite). Naturally the abstract accusative case associated with the direct-object
relation and the accusative morphological case that marks this abstract case/
grammatical relation do not follow the NP in its movement to subject position.
However, where a verb assigns inherent case this is retained. The contrast is illus-
trated in (14) and (15). In (14) the accusative-marked object of (14a) shows up as
a nominative subject in (14b). In (15) the dative-marked complement of helfen in
(15a) shows up as a dative subject in (15b) (Haegeman 1994: 186f ).
(14) a. Sie sieht ihn b. Er wird gesehen
she.NOM sees him.ACC he.NOM is seen
She sees him. He is seen.
(15) a. Sie hilft ihm b. Ihm wird geholfen
she.NOM helps him.DAT him.DAT is helped
She helps him. He is helped.
Clearly there is a contrast, but in many analyses (15b) would be taken to be an
impersonal passive with no overt subject and ihm in (15b) would be taken to hold
the same grammatical relation as in the active (15a).
59
Verbs and adpositions regularly assign case, but rarely receive it.
8
Adjectives and
nouns regularly receive case, but do not assign it in all languages. Van Riemsdijk
has suggested that there is a hierarchy of case assigners (1983: 249),
(16) verb > preposition > adjectives > nouns
He also suggests that case assigners lower on the hierarchy are more likely to assign
a case nearer to the oblique end of the hierarchy of cases. Where adjectives and
nouns assign case, it is typically the dative or genitive rather than the accusative.
Nouns and adjectives are not considered to be structural case assigners normally.
A phrase such as proud John as in *Dad is proud John is considered ungrammatical
because John has not been assigned case. All noun phrases must be assigned case.
However, the phrase in question can be saved by the insertion of of to yield Dad
is proud of John. Of is regarded as a meaningless case marker for the abstract
genitive case. Note, however, that of, like all other prepositions in English, governs
the accusative morphological case: Dad is proud of him.
Genitive case is held to be inherent rather than structural. Consider the following
example:
(17) *Dad is proud [John to be the winner]
This cannot be turned into a grammatical English sentence by the insertion of of.
Intuitively the reason is clear. Dad is not said to be proud of John, but of the content
of the proposition [John to be the winner]. The adjective is taken to be an inherent
case assigner rather than a structural case assigner. A structural case assigner is
sensitive only to structure, whereas an inherent case assigner is sensitive also to
semantic roles (called theta roles in Government and Binding Theory). An inherent
case assigner can assign case only where it assigns a semantic role. In (17) the role
for the complement of proud applies to the clause shown in brackets, not just to John.
The principles of case assignment in Government and Binding assume a VP. In
some languages the verb and its object are not adjacent. This is true of the Celtic
languages, for instance, where, for the most part, the order is verbsubjectobject.
Such an order is held to be the result of movement, one possibility being that
the verb moves to the left of the subject. In some languages word order seems to
be particularly responsive to the demands of discourse and there is little evidence
for a VP constituent. In other languages of the polysynthetic type arguments are
frequently represented in the verb; and where there are noun phrases for arguments
there is the possibilityinsome instances that these are adjuncts ina kindof apposition
with the pronominal representation in the verb. In such languages too there is little
60
3.3 Abstract case
evidence for a VP. However, in all these cases the Chomskian tradition assumes an
underlying conguration with a VP.
In Government and Binding Theory an innitive is considered unable to as-
sign case to its subject, so a sentence like (18a) is held to be ungrammatical
since have cannot assign case to her. (18a) can be rescued, however, by the ad-
dition of a preposition as in (18b) where for is considered to assign accusative case
to her:
(18) a. *[[Her to have to chop the wood ] was just too much]
b. [[For [her to have to chop the wood ]] was just too much]
Where an innitival clause functions as a complement, a noun phrase like him in
(19) is taken to be the subject of the innitive and to receive its accusative case from
the nite governing verb.
9
(19) We believed [him to be the culprit ]
(18b) and (19) are examples of exceptional case marking in that case is assigned
outside the normal scope of government or, to put it another way, case is assigned
across what is normally a barrier, in these instances the boundary of an inni-
tival clause. The assumed boundaries are shown by bracketing in (18a), (18b)
and (19).
Latin provides a useful example of a verb assigning both inherent and structural
case and making an exceptional case assignment. The verb cr edere to believe,
to trust takes a dative complement for the entity believed (in) or trusted and less
frequently an accusative complement, i.e., a direct object, for what is believed or
admitted. It also takes an accusative and innitive construction as a complement
as in (20):
(20) Cr edidimus Iovem regn are
believe.PERF.1PL Jove.ACC reign.INF
We believed Jupiter to reign.
Cr edere would need to be allotted three lexical entries. In one a dative complement
would be specied. This would be an instance of inherent case assignment. In a
second an NP complement would be specied and in a third an innitival clause
complement. No case would be assigned in the second and third entries. The ac-
cusative would be assigned structurally as a default and where an innitival clause
complement occurred as in (20) the exceptional case-marking principle would apply
61
to assign the accusative to the noun phrase functioning as the subject of the inni-
tive. Since Latin is a language where word order is determined largely by pragmatic
considerations, this would be another example of a language where structural case
assignment would assume an underlying VP structure that is frequently disguised
by the operation of various movement rules.
In more recent work within Chomskys Minimalist Program (see, for instance,
Chomsky1993) Inis split intoa number of different nodes includinga T(ense) node
and it is widely assumed that in underlying structure the subject is a specier within
the VP (the VP-Internal Subject Hypothesis). The motivation for this can be seen
if we note that a subject is an argument of a lexical verb, not of tense or modality.
These are higher predicates and the subject raises from its specier position within
VP to become the subject of whatever predicates of tense or modality are present.
The split In also contains subject and object nodes each projecting a phrase with
a specier position. Subjects and objects raise, either overtly or covertly, to have
their nominative or accusative case features checked in these specier positions.
This procedure ts into a larger framework of checking theory in which lexical
items are licensed by functional heads. It would take us too far aeld to describe
this system. Sufce it to say that case theory may not continue as a recognisably
separate module of the theory.
3.4 Semantic roles and grammatical relations
Since the late 1960s a number of theories have been put forward claim-
ing that the semantic relationships borne by nominal dependents to their governors
make up a small, universal set. Since obviously there is a great deal of variation
between languages as to how many cases and adpositions they have, the semantic
relationships that are posited are not always reected directly in the morphosyntax.
Theories positing a universal set of semantic relations include Fillmores proposal
for Case Grammar (1968, 1971), John Andersons Localist Case Grammar (1971,
1977, 1997), Starostas Lexicase (1971, 1988) and Diks Functional Grammar
(1978).
All modern theories allowfor some kind of semantic relations that are not always
reected directly in the morphosyntax, but they differ in the extent to which they
use syntactic rather than semantic evidence to isolate the semantic relations. There
is also a lot of confusing variation in the terminology. Fillmore began by positing
a universal set of relations with traditional case-like labels (agentive, instrumental,
dative, factitive, locative, objective) (1968: 245), but later switched to agent, expe-
riencer, instrument, object, source, goal, place and time, which, except for object,
are more semantically transparent and less confusible with traditional case labels
(1971). He called these syntacticsemantic relations cases and his conception of
grammar and its congeners came to be referred to as case grammar (1968: 19). It
62
has become common over the last generation to refer to Fillmorean-type cases as
deep cases and traditional cases as surface cases. The most widespread terms for
purely semantic relations are semantic roles, case roles, thematic roles (Lexical
Functional Grammar) and theta roles (as in Government and Binding).
10
I will use
semantic role or simply role for a semantic relation.
These semantic roles need to be distinguished from grammatical relations
(see section 1.1). On the basis of formal distinctions made in case, agreement,
word order, adpositions and the like we can isolate a number of grammatical rela-
tions. Some of these relations are purely syntactic: subject, object, indirect object,
ergative and absolutive, each of which subsumes a number of roles. Other relations
are semantically homogeneous. It is not uncommon to nd, for instance, that loca-
tive and instrumental are demarcated by case. In this situation grammatical relation
and semantic role coincide. Some roles such as agent, experiencer and patient are
normally subsumed under purely syntactic relations and can usually only be isolated
by semantic tests. Some roles such as instrument and location, which are typically
expressed outside the purely syntactic relations, can sometimes be expressed via a
purely syntactic relation. An example of the instrumental being expressed via the
absolutive relation in Kalkatungu was given in (3) above.
As noted in section 1.1, the term case relation also appears in the literature.
For some it is synonymous with grammatical relation as described above (e.g., in
Silverstein 1976), but it is perhaps more frequently used with reference to the
syntacticsemantic relations positedincertaintheories. Inthis text it will be conned
to the relations of Localist Case Grammar (section 3.4.4) and Lexicase (section
3.4.5).
The notion of a set of syntacticsemantic relations that have some independence
from noun-phrase marking is not entirely new. In fact relations of this type are to
be found in the work of the ancient Indian grammarian P an
.
ini, so it is appropriate
to discuss these briey before discussing the work of recent decades.
3.4.1 P an
.
ini and k araka theory
P an
.
inis grammar of Sanskrit is known as the As
.
t
.
adhy ay (eight
books). Its date is uncertain, with estimates ranging from 600 BC to 300 BC.
It is a work of great sophistication clearly deriving from a lengthy tradition and
it has provided a model for the study of language in greater India down to the
present day. Bloomeld described P an
.
inis grammar as one of the greatest mon-
uments of human intelligence (1933: 11). As is well known, the discovery of
Sanskrit by European scholars in the late eighteenth century and the realisation that
it is genetically related to Greek and Latin provided great stimulus to comparative-
historical linguistics. It has also been claimed that familiarity with P an
.
ini and
the ancient Indian tradition helped raise the level of sophistication in descriptive
63
linguistics; however, the aspect of P an
.
inis theory that we are concerned with here,
namely k araka theory, has not been inuential though discussed by various scholars
(Cardona 1976a: 215ff ).
K arakas are semantic relations holding between nouns and verb. The verb is held
to be the head of the clause and each nominal dependent is assigned to one of
six k arakas. These k arakas are listed in (21) with their common translations. The
term object is unfortunate since it suggests a purely syntactic relation. A term like
patient would have been more appropriate.
(21) kartr
.
agent
karman object
karan
.
a instrument
sam
.
prad ana destination
ap ad ana source
adhikaran
.
a locus
To appreciate the signicance of k arakas it is useful to compare them with the
case system of Sanskrit, since the cases are obviously the main means of express-
ing k arakas. In Sanskrit there are eight cases, all of which are distinguished in
the singular of masculine a-stems. In P an
.
inis grammar and in the Indian tradi-
tion generally they are numbered as in Table 3.4 and often referred to by their
numbers.
The vocative does not mark a dependent of the verb and therefore a vocative-
marked noun bears no k araka. The genitive is held to be adnominal and so no
k araka is assigned to genitive-marked nouns. However, the genitive does have one
adverbal function, namely to mark the complement of a small number of verbs as
in: m atuh
.
smara-ti (mother.GEN remember-3SG) S/he remembers his/her mother
(Filliozat 1988: 86). Withthe other sixcases there are regular associations of case and
Table 3.4 Sanskrit nominal declension as
represented by deva- god
1 devas nominative
2 devam accusative
3 devena instrumental
4 dev aya dative
5 dev at ablative
6 devasya genitive
7 deve locative
8 deva vocative
64
k araka as follows:
(22) nominative kartr
.
agent
accusative karman object
instrumental karan
.
a instrument
dative sam
.
prad ana destination
ablative ap ad ana source
locative adhikaran
.
a locus
This list is, however, misleading insofar as it suggests a one-for-one correspondence
between case and k araka and in that it suggests the nominative expresses a k araka.
A consideration of an active sentence and its passive counterpart demonstrates how
the one-for-one association needs qualication.
(23) Devadatta odana-m paca-ti
Devadatta.NOM rice-ACC cook-3SG
Devadatta cooks the rice.
(24) Devadatt-ena pac-ya-te odana-h
.
Devadatta-INST cook-PASS-3SG rice-NOM
The rice has been cooked by Devadatta.
The person marking on a nite verb is held to encode a k araka, in the active the
kartr
.
. The nominative is held to encode only the denotation of the person marker
on the verb. Since the nominative-marked noun refers to the same denotatum as the
person marker in the verb (-ti in (23)), it names the referent holding the k araka of
kartr
.
(agent). In the active the accusative encodes the karman. In the passive the
person marker on the verb is interpreted as encoding the karman. The nominative
in cross-reference with this person marker (-te in (24)) names the karman. The
instrumental encodes the kartr
.
.
A consideration of the passive then demonstrates that there is no one-for-one
correspondence between case and k araka and that the nominative does not encode
a k araka. However, P an
.
inis treatment of the passive can give the misleading im-
pression that k arakas are like the deep cases or semantic roles of modern theories,
which tend to remain constant under paraphrase. Consider now the following pair
and their interpretation (Cardona 1976b: 224):
(25) Bh um av as-te Devadatta-h
.
ground.LOC sit-3SG Devadatta-NOM
Devadatta is sitting on the ground.
(26) Bh umi-m adhy- as-te Devadatta-h
.
ground-ACC on-sit-3SG Devadatta-NOM
Devadatta is sitting on the ground.
65
Both these sentences can refer to the same objective reality in which Devadatta
refers to the person sitting and bh umi to the location. In (25) bh umi is encoded in
the locative and interpreted as expressing adhikaran
.
a (locus). In (26) the transitive
verb adhy as- (adhi on + as sit) is used. Bh umi is in the accusative and interpreted
as expressing karman. There are a number of examples like this available that
show P an
.
ini distinguished the linguistic encoding of events from objective reality,
that is, he allowed for more than one semantic interpretation of an event. I have
not been able to ascertain exactly what the difference in semantic interpretation
is between (25) and (26), but there are similar pairs to be found in a variety of
languages including Latin, English, Indonesian and Kalkatungu where the effect of
advancing a locative to direct object is to add a sense of affecting an entity. English
examples are given in (34) and (51) below. A Latin example is given in note 5 to
chapter 6.
3.4.2 Fillmore
Fillmore is responsible for bringing to the fore the notion that there
is a universal set of atomic semantic roles. In his seminal paper The Case for
Case, published in 1968, he proposed a set of six cases, which he later revised
and extended to eight (see section 3.4). These cases were deep-structure cases,
described as being underlying syntacticsemantic relationships. They were to be
distinguished from case forms, which comprise the means of expressing cases:
sufxes, suppletion, adpositions, etc. (Fillmore 1968: 21ff ). Sets of roles similar to
Fillmores can be found in the writings of other linguists including Halliday 19678,
Chafe 1970, Longacre 1976, Dik 1978 and Cook 1979.
To establish a universal set of semantic roles is a formidable task. Although some
roles are demarcated by case or by adpositions in some languages, in many instances
they have to be isolated by semantic tests. There are no agreed criteria and there is
certainly no consensus on the universal inventory. To a great extent establishing roles
and ascribing particular arguments to roles involves an extra-linguistic classication
of relationships between entities in the world. There tends to be agreement on salient
manifestations of roles like agent, patient, source and instrument, but problems arise
with the classication of relationships that fall between the salient ones. There are
also problems with determining how ne the classication should be. Consider, for
instance, an entity that is presented as the material fromwhich something is made, as
in She made the bowl from clay. The notion is conceptually distinct, but there is not
normally any marking specic to this notion. On the other hand, since it is encoded
differently in different languages, for instance, in the ablative or the instrumental,
then it needs to be recognised in a cross-language comparison. It is not proposed to
deal with these problems here, but the following list of roles is offered as a checklist
66
of roles that have been frequently distinguished in the literature. It also represents
a list of the roles mentioned in the present text, and on a few points it reects my
own preferences.
patient
Almost all inventories of semantic relations include a role that covers the following:
(i) an entity viewed as existing in a state or undergoing change:
The sky is blue
The ame grew bright
(ii) an entity viewed as located or moving:
The lion is in the cave
The stone moved
He moved the stone
(iii) an entity viewed as affected or effected by an entity:
The bird ate the worm
The bird sang a song
Fillmore called this role object, later objective; Gruber called it theme; others have
called it goal and patient.
11
The labels objective and object are unsatisfactory
in that they are confusible with a case label and a label for a grammatical rela-
tion respectively. The label theme, which derives from Grubers 1965 dissertation
(see also Gruber 1976), is also unsatisfactory since the term had been established
earlier in Prague School linguistics for a discourse-pragmatic function. The label
goal is sometimes used for the role that I call destination, so it too is less than ideal.
The label patient is the most widely used of the various alternatives, even if it is
not appropriate for all the examples to which it is applied (like those in (i) and the
rst two in (ii)). A number of linguists in fact make a distinction between theme
and patient using theme for (i) and (ii) and patient for (iii). Others combine them
under the label patient / theme. In this text I use the term patient, qualifying it where
necessary as affected patient, effected patient or neutral patient. It could be said
that such a patient is more like a syntactic entity than a semantic one, but I doubt
whether one needs to subdivide it to cater for a contrast within the clause (see also
the discussion of (27) below). The patient, in the wider sense, is a fundamental
role in that it has the closest semantic relationship with the predicate. A number
of linguists including Gruber (1976: 38) have claimed that it is an obligatory role,
though that would not appear to be true for a small number of one-place predi-
cates such as SHOUT and URINATE. In the case of predicates of bodily emanation
there is a patient (the emanated substance), but it is not expressed as an argument
(Blake 1982).
67
agent
The entity that performs an activity or brings about a change of state.
The robots assembled the car
The crowd applauded
The sun melted the ice
The race was won by the favourite
instrument
The means by which an activity or change of state is carried out.
She squashed the spider with a slipper
experiencer
The creature experiencing an emotion or perception.
They love music
They see everything
Some writers distinguish the perceiver or cogniser of verbs like see or hear fromthe
experiencer of verbs like love. The perceiver is almost always aligned syntactically
with the agent, whereas the experiencer is often treated differently (see the brief
discussion of inversion in section 3.4.3). In some Northeast Caucasian languages
the perceiver is treated differently from both the agent and the experiencer. See (6),
(7) and (8) in chapter 5.
location
The position of an entity. The view taken here is that location and the other local
roles can refer to time as well as place. Some linguists, however, including Fillmore
and Dik, distinguish temporal and spatial roles.
The vase is on/under/near the table
Australia Day fell on a Tuesday
source
The point from which an entity moves or derives.
They got news from home
Since June everything has been all right
path
The course over which an entity moves.
The dog chased the cat along the path and through the conservatory
They managed to survive through the drought
68
destination
The point to or towards which an entity moves or is oriented.
He turned to the altar and walked towards it
She slept till dawn
The terms direction and goal are alternatives, but the meaning of the former is not
transparent and the latter has also been used for patient / theme and for recipient.
recipient
A sentient destination.
She gave her spare change to the collectors
purpose
The purpose of an activity.
He went to the Red Rooster for some take-away
beneciary
The animate entity on whose behalf an activity is carried out.
She did the shopping for her mother
manner
The way in which an activity is done or the way in which a change of state takes
place.
He did it with great skill
extent
The distance, area or time over which an activity is carried out or over which a state
holds.
It lasted the winter
He ran ( for) three miles
possessor
The entity that possesses another entity.
I saw Johns golf clubs
Some writers refer to the possessed item as the possessum. It should be equated
with the neutral patient (theme) and in some languages there is formal evidence for
such equating. See, for instance, the Abaza and Yupik examples in section 5.4.
69
Although there is no consensus on the universal inventory of roles, linguists tend
to adhere to a common set of practices in ascribing roles:
(a) the inventory is kept small
(b) a role can be assigned only once in a clause
(c) no dependent can bear more than one role
(d) roles remain constant under paraphrase.
Keeping the inventory small
All the inventories of roles that have been proposed are fairly small,
usually about the same as the number of cases found in a typical case language, i.e.,
between six and ten or so. Where complements are concerned there is an obvious
relationship between the meaning of the verb and the role of its arguments. The
verb hit implies a hitter and a hitee, the verb scrape implies a scraper and a scrapee,
and so on. No one, of course, suggests treating hitter and scraper as separate roles,
rather one abstracts the notion of agent. Similarly no one suggests taking hitee
and scrapee to be separate roles; one abstracts the notion of patient. The problem,
however, is to determine howbroad the roles are to be. Consider the verb watch as in
The pelican watched the sh. The pelican can be described as an agent, but the sh
is not a patient in the sense of an entity that is affected by an activity (at least it is
not necessarily affected), and some would describe it as bearing the role of theme
(see above). However, the affected patient of hit and the unaffected or neutral patient
of watch do not contrast syntagmatically nor are they opposed paradigmatically, so
one can treat them as sharing the same role. If one does this, one is exploiting a
principle of complementary distribution, which is just what languages do. It is also
efcient to take this step, since the only function the roles of complements can play
in the grammar is to indicate the orientation of a verb.
12
Consider, for instance, the
following sentence and its French translation:
(27) a. He misses you very much
experiencer patient
b. Tu lui manques beaucoup
pat exp
With the English verb miss the subject expresses the experiencer of the feeling
of loss and the direct object expresses the neutral patient. With the French verb
manquer, when used in the corresponding sense, the subject encodes the neutral
patient and the indirect object (lui) encodes the experiencer. Obviously grammars
need to indicate which grammatical relation encodes which role, but to do that
70
they do not need certain distinctions such as the one between affected patient and
unaffected patient or theme. Of course if one exploits the complementary dis-
tribution fully, one tends to wind up with syntactic entities rather than semantic
ones.
While an exploration of the syntactic relations raises questions of how many
distinctions are needed, an examination of adpositions raises the opposite issue
of how few are required. The set of adpositions found in many languages runs to
forty or so members and in some Fennic and Northeast Caucasian languages the
case systems approach this gure (but see section 5.6). Most of the forms making
up these sets are local, expressing notions such as above, below, near, on
etc. If one were to take these as expressing separate roles, the inventory would
be much larger than it typically is. It is clear that one cannot take all these local
forms to represent separate roles and maintain the notion of atomic roles. These
local forms can be analysed in terms of the notions of source, location, path and
destinationplus markers of relative orientationsuchas above, below, etc. See also
Table 5.4.
Once per clause
The notion that each role can be assigned only once per clause is
generally agreed. However, it needs to be interpreted as allowing co-ordination
(Jack and Jill went up the hill ) and multiple specication of a particular location
(It is on the shelf, to the left, behind the door). Where nominals stand in apposition
they may share a role, but it could be argued that only one of the apposed nominals
is in the clause proper. In any event apposed nominals are coreferent, so only one
referent is understood as bearing the role assigned.
13
Multiple assignment of a role
is, however, a feature of John Andersons Localist Case Grammar. Examples are
given in section 3.4.4.
No dependent with more than one role
Most linguists adhere to the principle that each dependent bears but
a single role to its governor, but certain verbs have been discussed as providing a
challenge. These include the pair buy and sell. Consider the following assignment
of roles:
(28) Fred bought the book from John
agent patient source
(29) John sold the book to Fred
agent patient destination
71
In (28) Fred is an agent in that he initiates an activity; the book is a patient in that it
passes from the possession of John to Fred, and John is marked as the source from
whom the book passes. In (29) John is an agent in that he initiates an activity; the
book is a patient, and Fred is marked as a destination. However, some have pointed
out that Fred is a destination in (28) in that he receives the book and John is a source
in (29) in that the book passes from him. But if one takes this view one would
also have to consider the passage of the money in the opposite direction from the
book.
14
It would seem superuous to extract all these roles which are entailed in
the denitions of the two verbs. The ascription of the role of agent to the subject
and patient to the object, however, links buy and sell to the large class of activity
verbs, but if these verbs are marked [+ activity] even the specication of agent and
patients is predictable as the default role for the subject and object respectively of
an activity verb.
In Foley and Van Valin (1984) and in later work by Van Valin in the framework
of Role and Reference Grammar ( Van Valin 1993) two macroroles are introduced,
namely Actor and Undergoer, rather than roles such as agent, patient, experiencer,
etc. Actor and Undergoer are held to be sufcient to describe the orientation of a
predicate (as illustrated with (27) above) with more particular roles such as agent
or experiencer being predictable from the semantics of the verb.
Roles constant under paraphrase
In the generally accepted tradition roles may be identied across para-
phrases and across translational equivalents; they are independent of expression
(cf. Agud 1980: 456). Fillmore would identify an agent expressed as subject with
a dative of agent or an ablative of agent (1968: 19). All these possibilities can be
found in Latin where the dative is used with the gerundive
15
and the ablative with
the passive (though the ablative can be used with the gerundive where ambiguity
threatens and the dative is a possibility with the passive).
agent as subject
(30) Mlt es hanc provinciam d efend erunt
troops.NOM this.ACC province.ACC defend.PERF.3PL
The troops defended this province.
passive with ablative of agent
(31) Haec provincia a m litibus d efensa est
this.NOM province.NOM by troops.ABL defended.NOM is
This province has been defended by troops.
72
gerundive with dative of agent
(32) Haec mihi provincia est d efendenda
this.NOM 1SG.DAT province.NOM is defend.GERUNDIVE.NOM
This province is to be defended by me.
An active sentence and its passive counterpart are normally taken to be syn-
onymous in terms of propositional content though different in terms of discourse-
pragmatic structure. Mlt es would be taken to be agent in both (30) and (31), but in
the active Mlt es has no marking that is peculiar to a role; the nominative marks the
subject grammatical relation encoding several roles such as agent, experiencer and
patient. In the passive, however, Mlt es appears in the ablative governed by a/ab
(the latter variant being obligatory before vowel-initial words) which means from:
ab e o loc o from that place. In the normal practice of modern grammarians of vir-
tually all schools the choice of case and/or adposition is disregarded in a situation
like this. The preposition a/ab is not interpreted as having its normal meaning, but
is regarded as a grammatical marker.
The use of a/ab plus the ablative to express the agent of the passive makes
an interesting comparison with the following example where the clause with the
ablative of agent is not passive (De Bello Gallico 1:20, 4 via Rubio 1966: 89).
(33) S quid e a Caes are gravius
if anything 3SG.DAT from Caesar.ABL more.serious.NEUT
accidisset, . . .
happen.PPERF.SUBJ.3SG
If anything drastic befalls him from/on Caesars part, . . .
Here where there is no passive and therefore no possible active counterpart, it
is difcult to avoid taking a Caes are at face value as a from-phrase. No agency
is expressed, but an interpretation of what is meant entails agency on Caesars
part.
The situation with the dative of agent is rather similar. Although (32) refers to
a situation in which ego will have to be an agent, (32) does not present ego as an
agent. (32) could be translated as It is up to me to defend this province or This
province is for me to defend. As Rubio points out (1966: 150), the term dative
of agent is inappropriate. The dative in a sentence like (32) can be considered
a dative of the person indirectly involved, i.e., a typical indirect object, with the
sense of agency merely being entailed by the context of the gerundive. The dative
in (32) is rather like the ablative in (33). In both instances agency is not expressed
directly.
73
There is a range of examples that presents problems for Case Grammar. One
type, recognised by Fillmore as problematical (1968: 489), involves valency al-
ternations with different choices of object. Suppose we are describing a situation
in which an agent (John) moves an object (a smoking pipe) and causes it to come
into light contact with another object (a wall). In English we could say either (34a)
or (34b):
(34) a. John tapped the wall with his pipe
b. John tapped his pipe on the wall
In Case Grammar deep cases remain constant under paraphrase. Either we consider
that the wall in (34a) is a patient and the pipe an instrument and transfer these roles
to (34b), or we take the pipe in (34b) to be patient and the wall to be locative and
transfer these roles to (34a). Obviously we cannot do both and maintain that roles
remain constant under paraphrase. If one recognises that (34a) and (34b) represent
different encodings of the same physical event, the problem disappears. (34a) is
likely to be our choice if, for instance, John is seen to be tapping the wall with
his pipe to see if the wall is hollow. The wall is seen as a patient and his pipe an
instrument. (34b) is likely to be chosen if John is seen to be tapping his pipe on the
wall to dislodge some wet tobacco from the pipe. His pipe is seen as a patient and
the wall as a location.
16
Encoding an entity as an object rather than as a noncore dependent often adds
a sense of affectedness. This can be seen in (34a) and (34b). In both sentences the
tapping is aimed at affecting the object. There are numerous examples of this in the
literature, including some paralleling P an
.
inis pair (25) and (26). I do not know if
there is such a difference between the Sanskrit sentences, but such a distinction can
easily be captured. The sense of affectedness can be related to the fact that bh umim
is the direct object of an activity verb. The sense of location is attributed to the
locative or stance feature of the verb.
These examples are not meant to imply that roles can never be tracked across
paraphrases. Obviously one can identify a patient and a recipient in both Mary gave
a book to Martha and the paraphrase Mary gave Martha a book, but allowance
needs to be made for more than one perception of a situation (the wall tapping
examples) and roles that are implied but not encoded (the Latin dative-of-agent
example).
Fillmores case grammar and similar attempts by others to establish a small list
of universal roles have fallen somewhat into disrepute largely because no one has
been able to produce a denitive list. However, a number of major theories such as
Government and Binding and LexicalFunctional Grammar embrace the notion of
semantic roles, but they remain uncommitted about the universal inventory.
74
3.4.3 Relational Grammar
Relational Grammar is a theory originally developed by Perlmutter and
Postal in the early 1970s.
17
It is of particular relevance to the study of case since it
has concerned itself almost entirely with grammatical relations. In this theory gram-
matical relations are taken as undened primitives. A distinction is made between
the grammatical relations subject, direct object and indirect object, collectively
known as terms, and the obliques such as locative, benefactive and instrumen-
tal. The terms are pure syntactic relations whereas the obliques are semantic. The
grammatical relations form a hierarchy as follows:
(35) subject direct object indirect object obliques
1 2 3
The theory is multistratal, i.e., the dependents of the verb may bear different
relations in different strata. The relations held at the initial stratum are determined
semantically. An agent or experiencer will be a subject, a patient a direct object, a
recipient an indirect object, and other roles such as locative will be encoded in the
appropriate oblique relation. The following sentences reect initial stratumrelations
directly. From here on the Relational Grammar convention of referring to relations
by their position on the hierarchy will be adopted (1 = subject, 2 = direct object,
3 = indirect object):
(36) Abel slew Cain
1 2
(37) Eve gave the apple to Adam
1 2 3
A passive sentence or a double-object construction, however, will be interpreted as
involving two strata. In the passive Abel was slain by Cain the argument Abel is said
to be advanced from 2 to 1. The argument Cain is said to have become a ch omeur.
The French word ch omeur means unemployed person and is part of an extended
metaphor relating to work that runs through a good deal of Relational Grammar
terminology.
(38) Abel was slain by Cain
2 1 initial stratum
1 ch omeur nal stratum
Ch omeurs are not found in the initial stratum, but arise in multistratal clauses
where a dependent is displaced from term status by a revaluation, as the initial 1 is
75
displaced by the advancee from 2 in (38). A ch omeur lacks some of the properties
of the corresponding termand cannot be advanced back to termstatus. In Relational
Grammar there are no revaluations to oblique status, i.e. to locative, instrumental,
etc. The 1-ch omeur in the passive may come to be expressed by a particular prepo-
sition, postposition and/or case, but this marking is never interpreted as indicating
oblique status.
The following sentence illustrates the interpretation of a double-object construc-
tion. The recipient is taken to be an initial 3 which advances to 2. As a result of this
advancement the initial 2 (the patient) is demoted to ch omeur status.
(39) Eve gave Adam the apple
1 3 2 initial stratum
1 2 ch omeur nal stratum
The principal piece of evidence for taking the recipient object to be the direct object
is that this object can be advanced to subject by the passive:
(40) Adam was given the apple by Eve
3 2 1
2 cho 1
1 cho cho
In languages with bound pronouns for direct object, the recipient object in the
double-object construction is normally represented by the direct object series which
suggests that the recipient object is in fact the direct object. The point was illustrated
fromKalkatungu in (9) above. In English, however, the analysis of the double-object
construction is controversial. The recipient object has claims to being direct object
in that it comes immediately after the verb and can be advanced to subject via
the passive as shown in (40). However, where the rst object in the double-object
construction is a beneciary (I made her a cake), it cannot be advanced to subject.
Moreover, the rst object, whether recipient or beneciary, cannot be relativised
freely (?I saw the man who you gave/made a cake) as the direct object can (I saw
the man who you admire). Hudson (1992) argues that the second object in the
double-object construction in English should be taken to be the direct object, but
in the present work the recipient or beneciary object will be taken to be the direct
object since this analysis seems to hold across languages.
There is one other Relational Grammar analysis that is worth mentioning and that
is possessor ascension, since it is relevant to the Kalkatungu examples involving
body parts presented as (8) and (10) above. We will consider a sentence similar to
(8), but cast in the future tense rather than the purposive modality, since this enables
76
us to dispense with the complication of bound pronouns. It will be recalled that in
(8) the possessor of a body part and the possessed body part appeared in parallel.
(41b) presents the propositional content of (8) and it is paired with (41a), which
shows a possessor expressed in the dative (the case that covers possessive function).
In Kalkatungu the dative is not normally used to refer to the possession of ones
own body parts, so (41a) would normally be taken to refer to the possession of an
eye other than ones own.
(41) a. Milnga-ngku itya-mi ngatyi milthi
y-ERG bite-FUT 1SG.DAT eye.NOM
The y/ies will bite my eye(s).
b. Milnga-ngku itya-mi ngayi milthi
y-ERG bite-FUT 1SG.NOM eye.NOM
The y/ies will bite me [in the] eye(s).
In Relational Grammar these two sentences would be based on the same initial
stratum, one directly underlying (41a), i.e., one in which the rst-person possessor
is taken to be a POSSessor dependent within a possessor phrase with the possessed
body part (milthi) as head. (41b) would be held to reect a revaluation, namely the
ascension of the possessor out of the possessor phrase to become a dependent of the
verb. In the revaluation the possessor assumes the relation borne by the possessed
in the initial stratum (namely 2) and pushes the possessed into ch omage.
(42) Milnga itya-mi ngayi milthi
[1 P [POSS 2]]
[1 P 2 cho]
An analogous analysis would be provided for a sentence like (10) where the pos-
sessor of the body part is in S function.
If we consider (8), (9) and (10) together ((9b) is an example of the double-object
construction), we see that in each clause with two nominative-marked noun phrases
one is interpreted as absolutive (S or P) and the other as a ch omeur.
Relational Grammarians have not attempted to establish a universal set of se-
mantic roles and their conception of terms (1, 2 and 3), which are semantically
heterogeneous, and obliques (locative, instrumental, etc.), which are simply seman-
tic roles, is a conventional one. Their main interest in the present context lies in the
following:
(a) the Universal Alignment Hypothesis
(b) the treatment of the absolutive and ergative relations
(c) the concept of ch omage
77
Universal alignment hypothesis
As mentioned above, initial stratum grammatical relations are deter-
mined semantically. Perlmutter and Postal proposed a Universal Alignment Hypo-
thesis (1984: 97; orig. 1978). According to this theory agents, experiencers and
cognisers are encoded as 1, patients as 2, and recipients as 3. The hypothesis is not
entirely unproblematic. Consider, for example, the semantic role of the object in
She speaks French. A survey of a number of languages reveals that the translational
equivalent can appear in a great variety of cases and with a great variety of apposi-
tions. In Australian languages alone the locative, instrumental, ablative, perlative,
dative, accusative and nominative (in an intransitive construction) are used. This is
hardly surprising. The relationship between the name of a language, dialect, etc. and
a predicate meaning speak language X falls between the salient roles of patient,
location, instrument, etc. Languages encode it in different ways, some allowing
alternatives as English does (She speaks French, She spoke in French) (Blake 1987:
27, Blake 1990: 24f ). The Universal Alignment Hypothesis is recognised as not
being tenable in its strict form (Rosen 1984), though it is unproblematic in most
instances. Its main effect is to force a multistratal interpretation of clauses that do
not exhibit the universal alignment. Consider, for instance, the following English
examples:
(43) a. That girl speaks to me
agent recipient
b. That girl matters to me
patient experiencer
(44) a. The prisoner escaped connement
agent source
b. The reason escapes me
patient experiencer
In the (a) examples the default alignment of role and relation occurs, but in the (b)
examples there is a marked alignment. In the Relational Grammar analysis of (43b)
me is an initial 1, because it is an experiencer which demotes to 3. The girl is an
initial 2, since it is a patient, which advances to 1 to provide a subject. In Relational
Grammar the nal stratum must contain a subject.
(45) That girl matters to me
2 1
2 3
1 3
78
A similar analysis applies to (44b) except that the initial 1 demotes to 3 and then
subsequently advances to 2 (Perlmutter and Postal 1984, orig. 1978):
(46) The reason escapes me
2 1
2 3
1 3
1 2
The (b) sentences cannot be passivised. Perlmutter and Postal attribute this to a
principle to the effect that there can only be one advancement to 1 in a clause. These
clauses already have an advancement to 1.
Predicates that encode a role normally associated with subject as an indirect
object, or occasionally a direct object, are known in the Relational Grammar litera-
ture and elsewhere as inversion predicates. Inversion is common with the meaning
to please, witness German gefallen, Latin plac ere, Italian piacere a, Spanish
gustar a and French pla re ` a where the neutral patient is encoded as subject and the
experiencer as indirect object. For examples see (48) and (55) in chapter 5.
Absolutive and ergative
As noted above, Relational Grammar recognises a set of primitive
(undened) grammatical relations comprising subject, direct object, indirect object
and a yet-to-be-determined number of obliques. It also admits absolutive and erga-
tive as derived (dened) relations. The absolutive covers the 2 of a transitive verb
and the 1 of an intransitive verb ( but see below). The ergative is the 1 of a transitive
verb.
The derived relations of absolutive and ergative provide a cross-cutting clas-
sication of the primitive relations of subject and object. The two classications
between them provide for any mixture of SA/P and SP/A alignments such as the
one illustrated from Kalkatungu in section 3.2.2.
The denition of absolutive given above is adequate in a system that recognises
only surface structure. However, Relational Grammar recognises underlying strata
and it analyses some intransitive predicates as having an initial stratum 2, mostly
predicates that take a patient argument such as fall, die, be small. The denitions
of the derived relations are meant to be applicable to all strata, so the absolutive
should be dened more strictly as covering the 2 of a transitive verb and the nuclear
term (1 or 2) of an intransitive predicate. (See Figure 3.3.)
It should be noted in passing that most of the revaluations posited in Relational
Grammar must be recognised in any theory. Although there are revaluations like
those illustrated with matter to (43) and escape (44), which have no direct surface
79
Figure 3.3Nuclear relations in Relational Grammar
manifestation, most revaluations concern valency-changing derivation: passive, an-
tipassive, reexive verb formation (see (2b) above), advancement to direct object
(see (3) above), causativisation, etc.
Ch omage
As noted earlier in the discussion of (38) and (39), where a depen-
dent is displaced from term status by a revaluation (usually the advancement of
another dependent) it becomes a ch omeur. In some instances the ch omeur retains
its marking as in the double-object construction (39); in others it assumes oblique
marking as in the passive (38). It is the assumption of oblique marking that is
relevant to case. In the Latin passive quoted above in (31) the displaced subject
is marked by the preposition a/ab meaning from plus the ablative, one of the
functions of which is to express from. In the Relational Grammar interpreta-
tion the displaced subject takes on no more than the marking a/ab plus ablative.
It is not interpreted as being recategorised as the grammatical relation otherwise
expressed by a/ab plus the ablative and certainly not as being semantically recat-
egorised. As noted in connection with (38) there are no revaluations to oblique
relations.
3.4.4 Localist Case Grammar
In his Localist Case Grammar John Anderson (1977, 1997) establishes
a set of universal case relations comprising only four members:
(47) absolutive
locative
ergative
ablative
It is immediately obvious that these entities are not the generally accepted gram-
matical relations. The question arises of how the larger number of semantic roles
recognised by scholars like Fillmore, Chafe and Cook can be accommodated.
80
There are three characteristics of Localist Case Grammar that account for this:
(a) localism
(b) case relations interpreted according to syntagmatic contrast
(c) multiple assignment of case relations
(a) Localism
As noted in section 2.3.4 a localist interpretation of case involves taking
syntactic case categories to be extensions of local case categories into an abstract
domain. For example, in a localistic interpretation the subject of a verb like have as
in She has it in English will be [locative]. This can be rationalised by reference to the
fact that the object is located in the possession of the subject. If this type of interpre-
tation is exploited liberally, obviously the inventory of case relations can be reduced.
(b) Syntagmatic contrast
The absolutive and the locative are interpreted as goal (destination)
in the context of ergative and ablative respectively. For instance, the noun phrase
the stick will be [abs] both in The stick broke and in The teacher broke the stick,
but will be interpreted as a goal of action ( patient) only in the transitive clause.
Similarly [loc] will be interpreted as a spatial goal (destination) in (48a) but not
in (48b) (cf. Anderson 1977: 115):
(48) a. They rode from Ghent to Aix
[abl] [loc]
b. They remained in Aix
[loc]
(c) Multiple assignment
In Localist Case Grammar an argument can bear more than one case
relation. Combinations of case relations obviously have the effect of extending the
inventory and the case relations effectively become features.
Most systems of semantic roles would distinguish an agent subject in She bit it
from an experiencer subject in She knew it. Anderson distinguishes them as follows
(Anderson 1977: 45, 87, 145):
(49) a. She bit it
[erg] [abs]
b. She knew it
[erg] [abs]
[loc]
81
The case relation [loc] is motivated by a comparison with sentences such as the
following where the knower is [loc]:
(50) He taught English to migrants
[erg] [abs] [loc]
[abl]
Clearly Anderson is motivated by a desire to maintain a xed mapping of ex-
tralinguistic roles to case relations. The localist nature of his interpretations also
comes out in his interpretation of the teacher as [abl] since the teacher is the
abstract source of the knowledge. Note that it is this [abl] that forces the spatial goal
interpretation of the [loc] case relation in the same clause.
Anderson uses multiple case assignment to handle pairs of sentences like the
following:
(51) a. He sprayed the paint on the wall
[erg] [abs] [loc]
b. He sprayed the wall with paint
[erg] [loc] [abs]
[abs]
As noted in section 3.3.2 (see (34)), pairs of sentences like these were discussed
by Fillmore as being problematic for a theory that holds roles constant under para-
phrase. In section 3.3.2 I suggested that the principle be abandoned and more than
one conceptualisation recognised, but Anderson has a way of capturing the objec-
tive properties of the arguments and at the same time recognising the difference in
perspective. The sentence in (5l a) is allotted the typical roles for transitive subject,
direct object and local phrase. The sentence in (51b) is interpreted as involving
a marked relationship between semantic role and grammatical relation. The wall
is now marked [abs] as well as [loc] and the paint retains its [abs] relation from
(51a). Sentence (51b) has as its unmarked interpretation a holistic reading, i.e., the
sentence would be used where the wall was wholly covered or at least substantially
affected by paint. Anderson captures this by tying the holistic reading to the case
relation [abs]. Every clause contains at least one exponent of [abs] and S and P will
always be [abs], no matter what other relation either may hold. Note that Anderson
effectively treats the preposition with as marking a demoted [abs] rather than an
instrument. This is essentially the same treatment that would be found in Relational
Grammar where with would be interpreted as marking a direct-object ch omeur
(cf. the treatment of the agent in the passive discussed in conjunction with (38) in
the previous section).
82
Undoubtedly the most interesting aspect of Andersons approach is his belief that
the core grammar of all languages is organised on an absolutiveergative basis at
an underlying level, even a language like English where the grammatical relation of
subject is so prominent. Consider the assignment of case relations in the following
(Anderson 1977: 82):
(52) a. The stone moved
[abs]
b. John moved
[abs]
[erg]
c. John moved the stone
[erg] [abs]
The semantic motivation is transparent. An agent is [erg] and a patient is [abs].
In a typical reading of John moved the subject John is agent (John instigates the
movement) and patient (he is displaced).
18
Anderson believes that all languages
have this assignment of case relations at an underlying level, but that many lan-
guages exhibit the phenomenon of subject formation which obscures the absolutive
and ergative case relations. Languages of the type that Klimov (1973) calls active
(see section 5.2.3) reect the case relations directly in their morphological mark-
ing. Ergative languages collapse [abs, erg] with [abs], but otherwise reect the case
relations shown in (52). In a language like English it is possible to distinguish rules
that are sensitive to the putative case relations from those that are based on the
subject-grammatical relation. The host of raising, for instance, is [abs], i.e. there is
raising to S and to P but not to A:
19
(53) The patient seems [ ] better
(54) The doctor expected the patient to [ ] recover
On the other hand, rules of case assignment (he v. him etc.), word order and
verb agreement are sensitive to subject as are other rules such as the requirement
that the subject remain unexpressed, in almost all circumstances, with a nonnite
verb.
3.4.5 Starostas Lexicase
Lexicase, a model of grammar developed by Starosta (1971, 1988),
is similar to Localist Case Grammar in that it takes the core grammar of all lan-
guages to be organised on an absolutiveergative basis. There are ve atomic,
83
syntacticsemantic case relations:
(55) PATIENT
AGENT
CORRESPONDENT
LOCUS
MEANS
PATIENT covers Splus P. i.e., what we are calling the absolutive relation in this text.
AGENT corresponds to A, i.e., the ergative relation. The other three relations have
inner and outer subtypes. An inner subtype has the PATIENT as its scope; the outer
subtype has the clause as its scope. A second ( patient) object in a double-object
construction would be an inner CORRESPONDENT, whereas an outer CORRESPONDENT
would be marked by an adposition and/or oblique case. An outer CORRESPONDENT
covers roles like purpose and beneciary. LOCUS covers all local relations, dif-
ferences between source, path, location and destination being attributable to case
marking (inection, adpositions, etc.) and choice of verb (leave, cross, inhabit, ap-
proach). MEANS covers roles like instrument and manner. Some illustrations follow:
(56) In the springtime birds nest
[+LOC] [+PAT]
[+outer]
(57) The diner gave the tip to the waiter
[+AGT] [+PAT] [+LOC]
[+inner]
(58) The diner gave the waiter a tip
[+AGT] [+PAT] [+COR]
[+inner]
(59) Lassie caught a rabbit for us
[+AGT] [+PAT] [+COR]
[+outer]
(60) The kids smeared paint on the nursery wall
[+AGT] [+PAT] [+LOC]
[+inner]
(61)
The kids smeared the nursery wall with paint with their ngers
[+AGT] [+PAT] [+MEANS] [+MEANS]
[+inner] [+outer]
84
As can easily be seen, there is no attempt to keep case relations constant un-
der paraphrase. Semantic roles may well remain constant under paraphrase, but
these play no part in the theory except for the interpretation placed on the case
relations according to marking and choice of verb. An AGENT, for instance, will
be interpreted as an agent with activity verbs and as an experiencer with verbs of
emotion.
Clearly the case relations are primarily syntactic rather than semantic, and the
same question arises with Lexicase as with Localist Case Grammar, namely how
can AGENT and PATIENT be justied in languages which obviously have a subject?
In fact the PATIENT is exploited in much the same way as [abs] is in Localist Case
Grammar. The PATIENT is the scope of inner case relations. In (62) and (63) the nal
local phrase is inner and refers to the location of the PATIENT:
(62) In the summer they sit in the shade
[+LOC] [+PAT] [+LOC]
[+outer] [+inner]
(63) In the summer they put their meat in the ice chest
[+LOC] [+AGT] [+PAT] [+LOC]
[+outer] [+inner]
The inner/outer distinction applies to innitival dependents. The PATIENT is the
controller of the missing subject of an innitival complement (Starosta 1988:
1337). In (64) the missing subject of swim is interpreted as the PATIENT of the
intransitive verb, namely Mary. In (65) the missing subject of go is interpreted as
coreferential with the PATIENT of the transitive verb tell, again Mary.
20
(64) Mary tried to [ ] swim the river to [ ] impress John
[+PAT] [n] [n]
[+inner] [+outer]
(65) John told Mary to [ ] go home to [ ] please his father
[+AGT] [+PAT] [n] [n]
[+inner] [+outer]
The target for raising can be captured in terms of [+PATIENT]. A raising analysis
is not used in Lexicase, but raising verbs such as seem (66) and expect as in (67)
have the property that they do not impose selection restrictions on their PATIENT.
The PATIENT will control the missing subject of an inner innitival complement and
85
this subject will fulll the selectional requirements of the innitive (cf. Starosta
1988: 158f ):
(66) Cindy seems to [ ] speak French
[+PAT] [+PAT]
(67) I expected Cindy to [ ] speak French
[+AGT] [+PAT] [+PAT]
Rules of word order and verb agreement are based on case. In English, for in-
stance, the verb agreement is controlled by the nominative-marked noun phrase.
One other entity is needed. In some languages with a nominativeergative system
of case marking S and A are treated alike in terms of certain syntactic rules even
though they are not both in the nominative. An example from Kalkatungu was
given in section 3.2.2 where the case system was nominativeergative, but where
the bound pronominal elements operated in terms of SA versus P. To capture SA
alignment Lexicase invokes the notion of [+actor] which is a feature independent
of the system of case relations.
3.5 Hierarchies
Grammatical relations, cases and case markers do not form sets of
equipollent entities. As we saw in the previous chapter, sets of cases share prop-
erties, and markedness relations can be found between cases. Shared properties
and markedness can be captured in terms of a feature analysis (section 2.3.4); to
some extent markedness can also be captured by means of hierarchical ordering.
It is a moot point whether grammatical relations should be given a componential
analysis and it partly depends on which set of relations one recognises, but there
is no doubt grammatical relations can be ordered hierarchically. Case marking is
an abstraction covering different types of exponent, e.g., adpositions and afxes.
The various types of case marker can be ordered in terms of their formal properties
(e.g., free vs. bound), and these formal properties can be shown to covary with the
hierarchical ordering of cases and grammatical relations.
Semantic roles can be ordered hierarchically with respect to grammatical rela-
tions. Under some views this order is also relevant to the acceptability of certain
constructions such as reexives and passives.
3.5.1 Grammatical relations
As we saw in section 3.4.3 the following hierarchy of grammatical
relations is posited in Relational Grammar:
86
3.5 Hierarchies
(68) 1. subject
2. direct object
3. indirect object
4. obliques (locative, instrumental, etc.)
The hierarchy manifests itself in various ways. In some languages the unmarked
word order follows the hierarchy. This is true of English, French (clitic pronouns
apart), Thai, Indonesian and Swahili, for instance. In some the hierarchy shows up
in causativisation. If one causativises a verb, a new agent is added to the valency
of the verb. This agent is encoded as subject and it displaces the old subject. In
languages such as Turkish, French and Italian, the ousted subject takes up the
rst available relation on the hierarchy: direct object, indirect object, oblique. The
following examples are fromTurkish. In (69a) we have an intransitive clause. When
the new agent is added as subject in (69b), the old subject is expressed as the direct
object. In (70a) we have a transitive clause with a direct object. When the new
subject is introduced in (70b), the old subject is expressed as an indirect object. In
(71a) we have a ditransitive verb with a subject, direct object and indirect object.
When the newsubject is introduced in (71b), the old subject is expressed in a phrase
governed by the postposition tarafndan (Comrie 1976, 1989: 1756).
(69) a. Hasan ol-d u
Hasan.NOM die-PAST
Hasan died.
b. Ali Hasan- ol-d ur-d u
Ali.NOM Hasan-ACC die-CAUS-PAST
Ali killed Hasan.
(70) a. M ud ur mektub-u imzala-d
director.NOM letter-ACC sign-PAST
The director signed the letter.
b. Ali mektub-u m ud ur-e imzala-t-t
Ali.NOM letter-ACC director-DAT sign-CAUS-PAST
Ali got the director to sign the letter.
(71) a. M ud ur Hasan-a mektub-u g oster-di
director.NOM Hasan-DAT letter-ACC show-PAST
The director showed the letter to Hasan.
b. Ali Hasan-a mektub-u m ud ur tarafndan
Ali.NOM Hasan-DAT letter-ACC director by
g oster-t-ti
show-CAUS-PAST
Ali got the director to show the letter to Hasan.
87
The hierarchy also shows up in access to relativisation. English can relativise
subjects (the man who left), objects (the man I saw) and oblique relations (the gun
with which I shot the intruder). However, Keenan and Comrie (1977) have shown
that some languages can relativise only subjects, some only subjects and direct
objects, some only subjects, direct objects and indirect objects, and so on.
21
Most languages allow for some verbal derivations that change the valency of the
verb. The passive is a widespread example. On the basis of the hierarchy of gram-
matical relations, such verbal derivations can be described in terms of advancement
(or promotion) and demotion. The passive involves the advancement of the direct
object to subject and the demotion of the subject to a peripheral relation.
Except for the passive there are not many general valency-changing derivations
in Indo-European languages, though there are some possibilities available through
alternative lexical choice as with English steal and rob. However, in some languages,
especially the Bantu languages, the Austronesian languages and some Australian
languages, there are extensive verbal derivations available that allow a grammatical
relation (and its associated role or roles) to be advanced up the hierarchy. There
are two patterns of advancement. In accusative languages where the core grammar
is organised on a subject /object basis, advancements to object feed the passive
enabling an indirect object or oblique relation to be advanced to subject. In ergative
languages like Dyirbal and Kalkatungu, where the core grammar is organised on
an absolutive/ergative basis, advancements are to the absolutive, which is at the
top of the relational hierarchy. There is no passive, but there is an antipassive that
enables an ergative to be advanced to absolutive (Table 3.5). An example of the
Table 3.5 Pattern of advancements
accusative languages
Subject Object
agt pat
recip
advancement to Object loc
ben
inst, etc
passive
ergative languages
Absolutive Ergative
pat agt
recip
advancement to Absolutive loc
ben
inst, etc
antipassive
88
3.5 Hierarchies
advancement of a locative to direct object was given in (25) and (26) from Sanskrit.
An example of the advancement of an instrumental to absolutive was given in (3)
from Kalkatungu.
In some Western Malayo-Polynesian languages (Austronesian) such as Tagalog
and Balinese there are alternative alignments of semantic role and grammatical
relation. InBalinese, for instance, the patient is the subject inthe unmarkedtransitive
construction and the agent in the marked transitive construction in which the verb
bears a nasal prex. The effect of advancement of an oblique is to displace the
patient from subject (unmarked construction) or object (marked construction). This
is akin to having the accusative pattern of advancement and the ergative pattern in
the one language (Artawa and Blake 1997).
Irrespective of the organisation of the core grammar the advancement displaces
the patient/theme argument. In the recent literature constructions reecting such
advancements have come to be called applicative.
3.5.2 Case
At the end of chapter 5 (section 5.8) it is argued that inectional case
systems tend to be built up in the following order,
(72) nom acc/erg gen dat loc abl/inst others
The hierarchy is to be interpreted as follows. If a language has a certain case, say
locative, it will usually have a nominative, at least one other core case, a genitive
and a dative. If it has an instrumental or ablative, it will usually have a nominative,
accusative or ergative or both, genitive, dative and locative, and so on. Where there
are gaps in the hierarchy it is usually because the higher grammatical relations
are marked by bound pronouns or differentiated by word order. The hierarchy that
emerges is a hierarchy of grammatical relations, an elaboration of the Relational
Grammar hierarchy given in (68).
Pinkster has demonstrated that cases can be arranged hierarchically according to
the extent to which they encode arguments of a predicate rather than adjuncts. In
Latin the hierarchy is as follows (after Pinkster 1985: 167):
(73) nom > acc > dat > abl > gen
The order nom > acc > dat > other (or nom > erg > dat > other in ergative
languages) would appear to hold generally, but it must be remembered that the label
dative is often allocated to a case because it is the favoured choice for complements
after the core cases (see section 5.3).
89
Ramarao (1976) shows that in some languages, such as the Dravidian language
Telugu, the priority of the accusative over the noncore cases can be established from
the fact that the noncore cases cannot encode a complement unless an accusative-
marked complement is present.
3.5.3 Marking
There are generalisations to be made about the distribution of morpho-
logically unmarked forms, about synthetic versus analytic marking, and about head
versus dependent marking.
If there is an unmarked (zero) case, and there usually is, this will normally be
the nominative encoding SA (subject) in accusative languages and SP (absolutive)
in ergative languages. Indo-European was unusual in having a positive marker
*-s for nominative. Some nouns, however, had a zero nominative, namely those of
the a declension: Greek ch ora land, Latin m ensa table. In Modern Greek some
accusative forms are unmarked: logos word (nom), loge (voc), logo (acc), logou
(dat). There are a few languages in which the accusative rather than the nominative
is unmarked for all nominals. Wappo (Penutian) is an example (Li and Thompson
n.d.).
There is a clear correlation between the type of marking employed for a grammat-
ical relation and the position of that relation on the hierarchy. Analytic case markers
(adpositions) are more likely for the noncore or peripheral relations, especially
the semantic relations (see Table 2.7 for these distinctions). Conversely, synthetic
markers (inection) are more likely than analytic ones for the purely grammatical
relations, especially the core ones. English is an example to the extent that it has
suppletive accusative inection with pronouns and a genitive marker s with nouns,
but prepositions for most other functions. The status of the genitive s is unusual in
that its scope is phrasal: the man over theres dog.
22
Head marking, i.e., marking the head of a construction rather than the dependents
to show grammatical relations (section 1.3.1), is practically conned to the purely
grammatical relations. Some languages have cross-referencing bound pronominal
forms, usually on the verb, for the subject only. This is the traditional Indo-European
type exemplied by Latin (Table 3.1). Other languages have cross-referencing for
the object as well (Swahili) and a smaller number cross-reference the indirect object
too (Georgian). It is very common to cross-reference an adnominal noun phrase in
possessor function on the head of the construction. An example was given from
Turkish in section 1.1: adam-n ev-i (man-GEN house-3SG.POSS) the mans house.
23
The distribution of inection versus adpositions and the distribution of head
versus dependent marking has implications for the interpretation of the historical
development of case systems. These developments are described in chapter 6.
90
3.5 Hierarchies
In some languages there is no marking for any core function, either on the noun
phrases or on the verb, though there is marking for all the peripheral functions. In
such languages a subject/object distinction is made via word order, usually subject
verbobject order as in Thai or Cambodian.
Case marking, particularly core case marking, may be sensitive to the category
of the stem (common noun vs. proper noun, noun vs. pronoun, etc.) including other
categories which may be marked on the stem such as number. The sensitivity of
case marking to the category of the stem is discussed in section 5.2.6.
3.5.4 Roles
When Fillmore proposed the notion of semantic roles independent
of the morphosyntax, he also introduced the idea of a subject-choice hierarchy
(1968: 33).
Most modern theories incorporate such a hierarchy, some adding an object-choice
hierarchy. The following example is to be found in Diks Functional Grammar (with
patient substituted for goal) (Dik 1978: 74f ).
(74) agent > patient > recipient > beneciary > instrument > location >
temporal
This represents a hierarchy of accessibility to subject. Agent is most accessible to
expression as a subject, patient next most accessible and so on. (74) also represents
a hierarchy of accessibility to direct object except that the agent cannot be expressed
as object. In Diks view a continuous segment of the hierarchy has access to subject
and another continuous segment has accessibility to direct object. Assignments of
lower roles to subject and object are less frequent within languages and across
languages; they are usually subject to lexical constraints and they typically involve
markedconstructions. InEnglish, agent is the rst choice for subject inthe unmarked
(active) construction (I shouted, I ate it), patient is the next choice (I fell ), but a
marked construction (the passive) must be used if it is to displace an agent argument
(He got shot by a madman).
A recipient of the verb give may displace a patient in the choice for object (I gave
the book to him, I gave him the book), but the double-object construction would
have to be considered marked with respect to the construction with a prepositional
phrase which is obviously much more common across verbs in general. The recip-
ient can be promoted to subject via the passive (He was given the book). However,
the advancement of a recipient to direct object is not possible with all three-place
verbs (cf. donate). A beneciary can be advanced to direct object with some verbs
(He built me a carport ), but not others (e.g., construct ) and it is difcult to nd
felicitous examples of passives with beneciaries as subject.
24
91
English is highly unusual in allowing roles low on the hierarchy to be encoded
as subject, at least with a few verbs. In The rst chapter contains the denitions
the subject would appear to be locative and in The war years saw a reduction in
the crime rate the subject is arguably temporal, although one could take it to be a
metaphorical perceiver.
Across languages the passive can be seen as a means of making a marked choice
for subject. Similarly derived verbs of the type illustrated in (3) above allow a
marked choice for object.
Dik does not recognise an experiencer role. If an experiencer were included, it
would presumably outrank the patient, since it is more often expressed as subject
than the neutral patient is (I like/hate/fear him). In a number of role hierarchies
not only is the experiencer placed above the patient, but several other roles are
as well (see, for instance, Jackendoff 1972: 43, Foley and Van Valin 1984: 59
and Bresnan and Kanerva 1989: 23). To some extent these hierarchies in which
the patient is ranked low are based on different premisses and used for different
purposes. Jackendoff, for instance, introduces the hierarchy in (75) and uses it as a
basis for the Thematic Hierarchy Condition on passives and reexives. In a passive
the agent phrase is supposed to be higher than the derived subject and in reexives
the reexivised argument may not be higher than its antecedent (Jackendoff 1972:
43, 148).
(75) 1. Agent
2. Location, Source, Goal
3. Theme.
The Thematic Hierarchy Condition could be invoked to show why the following
sentences do not have passives. In both, the subject is a neutral patient or theme and
the other argument an experiencer which Jackendoff would take to be an abstract
goal.
(76) That girl matters to me
(77) The reason escapes me
These sentences were previously presented in the discussion of universal align-
ment of roles and relations in Relational Grammar (see (43) and (44)).
25
92
4
Distribution of case marking
4.1 Introduction
This chapter describes the distribution of case marking within the sen-
tence. The order of presentation is clause level (4.2), phrase level (4.3) and then
word level (4.4). The distribution of case marking in subordinate clauses is treated
in the nal section (4.5).
4.2 Within the clause
In the clause case marks the relationship of various complements and
adjuncts to the predicate. These complements and adjuncts are usually realised by
noun phrases or adverb phrases, and the predicate is usually a verb. This function of
case is the basis for the central denition of case given at the beginning of chapter 1,
and numerous examples of this function are to be found throughout the text and
do not therefore need further illustration here. However, as noted in section 1.2.1,
case marking may also be found on dependents within a noun phrase or on words,
mostly nouns and adjectives, in predicative function.
In a construction such as Virgils v enit summa di es The last day came the
nominative case on di es day indicates that it bears the subject relation to the verb
v enit and this usage ts our central denition. However, the nominative on the
adjective summa highest does not indicate that summa is subject, but that it is a
modier of the subject. This usage is somewhat marginal to our central denition,
since the nominative here is merely specifying that summa is a dependent of di es
without specifying a type of relation between summa and di es (see also section 4.3).
Where there is concord between a predicative nominal and a subject, the case
marking is even further from our central denition. In many languages a nominal
mayinstantiate the predicate without there beinganygrammatical verb, at least inthe
unmarked tense and mood.
1
In Latin, for instance, a noun or adjective can instantiate
the entire predicate as inNihil bonumNothing[is] good, where the adjective bonum
is the predicate and it shows concord in case (nominative), number (singular) and
gender (neuter) with the subject nihil nothing. If one takes the predicate to be
the head of the clause, even when realised by a nominal, the concord is of a head
93
with a dependent.
2
However, this is not the only possible interpretation. One could
take the view that both the subject and the nonverbal predicate are dependents of
an abstract predicate or that the construction is exocentric and lacks a head. Under
either of these views any case concord will be between two dependents.
If a verb is used with a predicative noun or adjective, then the verb will be the
grammatical head of the clause and the concord will be between two dependents.
In the following example from Pliny the verb esse to be is used:
(1)
Homin d elissim sunt equus et canis
man.DAT most.faithful.NOM.PL be.3PL horse.NOM and dog.NOM
To a human the horse and the dog are the most faithful.
Here the concord is between the subject equus et canis and the predicative adjective
d elissimi. Fid elissimi is the predicate word that determines the valency (the subject
and the dative complement) but sunt, the nite verb, is the head of the predicate
phrase.
Anoun or adjective can be predicative to nouns in cases other than the nominative.
In Latin predicates controlled by an accusative object were common:
(2) Cicer onem c onsulem cre av erunt
Cicero.ACC consul.ACC made.3PL
They made Cicero consul.
Here the noun c onsulem is a predicate exhibiting accusative concord with the di-
rect object. Under the traditional analysis it would be referred to as the object
complement and considered part of the clause containing the object. In one modern
analysis Cicer onemc onsulemwouldbe takentobe a verbless clause servingas a con-
stituent of the larger clause: [[Cicer onemc onsulem] cre av erunt]. In another analysis
Cicer onem would be taken to have been raised from the clause Cicer o c onsul to
become a constituent of the main clause, leaving the nominal predicate c onsul as
the sole constituent of a nonnite subordinate clause. The missing subject of this
nonnite clause would be identied with the object Cicer onem of the main clause.
3
Besides nominals in predicate function exhibiting concordial case there are in any
case language also phrases in predicate function bearing relational case. Consider
the following Latin examples:
(3)
Cicer o, vir capite candid o, dixit. . .
Cicero.NOM man.NOM head.ABL white.ABL speak.PERF.3SG
Cicero, a man with white hair, said. . . .
94
4.2 Within the clause
(4) Vir est capite candid o
man.NOM is.3SG head.ABL white.ABL
The man is white haired.
In (3) the ablative in capite (which also shows up in candid o via concord) expresses
the relation of the modier caput to the head vir and thus ts our central denition. In
(4) the ablative in capite expresses the same relation to vir, but caput is not a depen-
dent of vir. It seems that we need to extend our denition of case to cover the marking
of nominals in predicative function expressing their relation to their controller.
Now to return to concord. While concord of case, number and gender of pred-
icative adjectives is the norm in Latin and Ancient Greek, in many other languages
predicative adjectives display no concord. This is the situation in German where the
predicative adjective consists of the bare stem and is distinct from the various case
forms. The following sentences illustrate the difference between the attributive ad-
jective in nominative forms and the predicative adjective. In the rst three examples
below the attributive adjective is in the weak form following the denite article.
In (8) it is the strong form, since there is no denite article.
(5) Der alte Tisch ist sehr gut
The old table (masc.) is very good.
(6) Das gute Kind ist sehr nett
The good child (neut.) is very nice.
(7) Die nette Frau ist sehr alt
The nice lady (fem.) is very old.
(8) Alter Wein ist sehr nett
Old wine (masc.) is very nice.
Hjelmslev (1935: 120) took the predicative to be a fth case separate from the four
recognised cases (nominative, accusative, genitive and dative). The justication for
this analysis is not strong, since the sole function of the predicative lies outside
the central range of case functions. A better analysis would be to say that German
predicate adjectives lack case altogether.
Hjelmslev also notes that in Tabassaran, a Daghestan language of the Caucasus,
a language with a very large case system (see section 5.6), there is a predicative
case marked by -u-, - a- or - o-. The markers precede the markers for number and
gender, whereas markers for other cases are word-nal as in mu er-ar sib-u-r-ar-u:
These pears are green where sib- green, not ripe is sufxed by -u-, Hjelmslevs
predicative case, followed by -r-, the gender marker, and -ar- the plural marker
(Hjelmslev does not gloss the nal u:). Hjelmslev is not deterred by the position of
95
the markers for predicative case, since he sees the function as all important (1937:
139). However, as stated above, the function is not a typical case function. What
Hjelmslev analyses as predicative case markers would appear to be derivational
afxes. The predicative case markers in Tabassaran are matched by an attributive
case marker, which is discussed in section 4.3.2 below.
Besides concord within the noun phrase and concord exhibited by predicative
nominals and adjectives, there is also apparent concord between what look like
separated parts of a noun phrase. In Latin, it is possible to take a word that would
appear to modify a noun and express it in a phrase separate from the noun. The
following example is from Virgil (Aeneid II:3),
(9) Inf andum, r egna, iub es renov are dol orem
unspeakable.ACC queen.VOC order.2SG renew.INF sorrow.ACC
Unspeakable, [O] queen, [is] the sorrow you order [me] to rekindle.
Here the gerundive adjective Inf andum is displaced from the word it might be
thought to modify, namely dol orem. In Australian languages the noncontiguous ex-
pression of words that would normally appear within a single noun phrase in English
is commonplace. The translational equivalents of English adjectives frequently ap-
pear late in the clause separated from their apparent heads. The following examples
are from Warlpiri (Pama-Nyungan). In this language case marking is phrase-nal.
In (10) the ergative sufx appears on the nal word of the phrase tyarntu wiri (dog
big). In (11) tyarntu and wiri are separate phrases, and each takes the phrase-nal
case marking. The form =tyu is an object clitic attached to the rst constituent of
the clause (Hale 1973b: 314).
(10) Tyarntu wiri-ngki = tyu yarlki-rnu
dog big-ERG = 1SG.OBJ bite-PAST
The big dog bit me.
(11) Tyarntu-ngku = tyu yarlki-rnu wiri-ngki
dog-ERG = 1SG.OBJ bite-PAST big-ERG
The big dog bit me.
The dog bit me, big (one).
4.3 Within the noun phrase
4.3.1 Internal relations
The relationship of dependent nouns to the head noun in a noun phrase
can normally be expressed by a variety of noncore or peripheral cases. In English
96
any preposition can be used (the house on the corner, the strike over leave loadings,
etc.); in Latin any peripheral case:
(12) a. genitive: c onsulis equus the consuls horse
b. dative: obtemper ati o legibus obedience to laws
c. ablative: vir praest ant prudenti a a man of outstanding wisdom
Typically one case emerges as a specialised adnominal case with a variety of
functions. In Latin the genitive is almost exclusively an adnominal case; it marks
the complement of only a handful of verbs. One of its important functions is to
provide a means of expressing the adnominal equivalent of an adverbal complement.
Consider the following pairs:
(13)
a. Catilna coni urat adversus rem p ublicam
Catiline.NOM conspire.3SG against matter.ACC public.ACC
Catiline conspires against the state.
b. coni ur ati o Catilnae
conspiracy.NOM Catiline.GEN
Catilines conspiracy
(14) a. Metuit calamit atem
fear.3SG calamity.ACC
S/he fears calamity.
b. metus calamit atis
fear.NOM calamity.GEN
fear of calamity
In (13b) the genitive corresponds to the subject of the corresponding verb and
is an example of the subjective genitive. In (14b) the genitive corresponds with
the object of the corresponding verb and is an example of the objective genitive.
The genitive could be used to express the adnominal equivalent of a complement
in the dative, ablative or even genitive case, and it could also be used with nouns
that are not derived from verbs. In the following example there are two genitives
dependent on the noun iniuria injury which does not have a verbal analogue (cf.
English injury/injure). One genitive must be interpreted as a subjective genitive and
the other as an objective genitive, but this can only be done on the basis of extra-
linguistic knowledge, although where both genitives have human referents there
is a preference for a prenominal subjective genitive and a postnominal objective
97
genitive (Torrego 1991: 284).
4
The following example is a standard one. It comes
from Caesar (de Bello Gallico 1, 30, 2).
(15)
pr o veter-ibus Helv eti- orum ini uri-s popul- Rom an-
for old-ABL.PL Helvetii-GEN injury-ABL.PL people-GEN Roman-GEN
. . . for the old injuries inicted by the Helvetii on the Roman people.
The emergence of a specialised adnominal case is common across languages and
such a case plays an important part in the nominalisation of clauses. For examples,
see section 4.5.1.
4.3.2 External relations
The case relation of a noun phrase is typically marked by an afx, an
adposition or both. The position of the adposition, i.e. whether it is a preposition
or a postposition, is related to whether the language has the order verbobject
(VO) or objectverb (OV). Part of the reason for this is that adpositions frequently
derive from verbs and naturally tend to retain the same position with respect to
their complement. In English, for instance, the participle concerning has become
an adposition as in concerning the students; it has become a preposition rather than a
postposition for the obvious reason that concern, like all other verbs in English, takes
a following object. Conversely, the corresponding German participle betreffend has
become a postposition since participles in German, like most dependent verbs, take
a preceding object (Vennemann 1973: 31). If similar developments have occurred
generally in verb-nal languages (German is verb-nal only in subordinate clauses),
then this will mean an association of postpositions and verb-nal syntax. It may
be true, as is commonly believed, that languages tend to maintain a consistent
headmodier order at clause and phrase level. Since the verb is the head of the
clause and the adposition the head of the adpositional phrase, this would mean
an association of prepositions with verb-initial languages and of postpositions with
verb-nal languages. Whatever the reason, there is such an association plus a strong
association of prepositions with subjectverbobject (SVO) order as in English.
It has been argued that afxes rather than stems are the heads of words. This
is most obvious in derivation where the afx determines the part-of-speech. The
addition of -ish to fool in English produces the adjective foolish; the addition of
-ness to the stem foolish produces the noun foolishness. With inection the principle
is not so clear, but it can be said that the part-of-speech category can be read from
the inection. If we accept this idea, then we could ask whether the consistent
ordering of heads and modiers extends to word structure. If it does, we would
98
expect the following correlations:
(16) headmodier modierhead
verb-initial verb-nal
preposition postposition
prex sufx
Even if we reject the notion of afxes being heads, for after all they are semantically
less important than stems, or if we reject the notion of there being a tendency for
heads and modiers to order themselves consistently, we should still expect the
correlations in (16) to hold with respect to case since there is good evidence that
many case afxes derive from adpositions (see also section 6.1). In fact it turns out
that in verb-nal languages (almost all SOV) afxes are indeed almost all sufxes,
but in verb-initial languages (predominantly VSO) and in SVO languages there
are not the expected prexes. These languages tend to have sufxes. In a study of
the distribution of afxes Cutler, Hawkins and Gilligan 1985 found no language
with case prexes (but see section 4.4.1 below). They posit a general tendency for
languages to prefer sufxes and they advance a psycholinguistic explanation. It can
be established independently of the data on afxation that the initial part of a word
is the most salient. Cutler, Hawkins and Gilligan suggest that stems favour this
salient initial position, because stems are processed before afxes. The contribution
of afxes can only be determined after stems have been identied.
There are two common distributions for the case sufxes. In one type case ap-
pears not only on the head nominal in the noun phrase but also on the determiner
and usually the adjective (word-marking languages). This system is familiar from
Indo-European languages and can be illustrated from Ancient Greek. In (17) the
determiner and attributive adjective showconcord for case, number and gender with
bios which is masculine.
(17) ho aneksetastos bios
the.M.NOM.SG unexamined.M.NOM.SG life.M.NOM.SG
the unexamined life
Case concord is also found in Balto-Finnic, Semitic, in some Pama-Nyungan lan-
guages and elsewhere, but it is a minority system.
In the other type, case marking appears on the nal word in the noun phrase
(phrase-marking languages). We can distinguish two subtypes. In the rst the nal
word is the head noun in the noun phrase. This type is widespread. It is found, for
instance, in Quechua. It is common among the Papuan languages and there is a con-
centration of the type in Asia including the Turkic, Mongolian and Tungusic families
99
north of the Himalayas as well as the languages of the subcontinent whether they
be Dravidian, Munda or Indo-Aryan (though a number of Indo-Aryan languages
have vestigial concord). In this area most languages are consistent modierhead
languages with SOV order at clause level and determinernoun, adjectivenoun
order at phrase level. The following example is from the Dravidian language
Kannada (Giridhar 1987: 33):
(18) Naanu ellaa maanava janaangavannu priitisutteene
I.NOM all human community.ACC love.1SG
I love all mankind.
In the other subtype the nal word in the noun phrase is not always the head.
This is the situation in various Australian and Amazonian languages (Derbyshire
and Pullum 1986, 1989, 1991). The phenomenon also occurs in Basque (Saltarelli
et al. 1988: 77):
(19) etxe zaharr-etan
house old-PL.LOC
in old houses
Besides these two common distributions, there are a number of minority pat-
terns. Among Australian languages there are some where only one constituent of
the noun phrase needs marking, not necessarily the head and not necessarily the
last word in the phrase. In Nyigina and Gooniyandi
5
discourse principles play some
part. In Nyigina, for instance, the marking is often on the rst word: gudyarra-ni
wamba (two-ERG man) two men, or frequently on the word that is most signicant:
ginya marninga Wurrawurra-ni (that woman Wurrawurra-ERG) that Wurrawurra
woman (Stokes 1982: 59f). In another Australian language, Uradhi, the case mark-
ing applies obligatorily to the head, but only optionally to the dependents (Crowley
1983: 371f ).
(20) Utagha-mpu amanyma(-mpu) udhumpuyn ighanhanga-n
dog-ERG big(-ERG) back.ACC break-PAST
The big dog broke [the other dogs] back.
In some languages case is marked on the determiner and the noun but not on
the adjective. Traditionally the determiner is considered to be a dependent of the
noun, but this should not be taken for granted. In many languages the determiner is
morphologically the same as or similar to third-person pronouns or demonstrative
pronouns, that is, certain forms can appear as the head of a noun phrase (pronoun
100
usage) or they can appear with nouns in a noun phrase (determiner usage). Where
these pronominal forms are used as determiners, it is not certain that they are
dependents; they may be heads. One such language is Diyari (Pama-Nyungan)
where the determiner, which comes rst in the noun phrase, is inected for case and
so is the nal word of the rest of the phrase (Austin 1981a).
6
(21) Nhu-lu karna pirna-li wama thayi-yi
he-ERG man big-ERG snake.ACC eat-PRES
The big man eats snake.
Yaqui (Uto-Aztecan) provides a straightforward example of a language with case
marking on determiners and nouns but not adjectives (Lindenfeld 1973: 60).
(22) Ini-e tu?i usi-ta=u noka-?e
this-OBL good child-OBL=to speak-IMP
Speak to this good child.
Another language in which only the determiner and noun take case marking is
Hungarian (Abondolo 1987: 591). In this language there is an article separate from
the determiner. It is not marked for case.
(23) ez-ek-ben a nagy g or og l ad a-k-ban
this-PL-in the large Greek crate-PL-in
in these large Greek crates
A similar situation is found in Tabassaran where adjectives all bear a marker -i
rather than the case marking found on the head noun: sib-i ve c green apple. As
noted in section 4.2, predicative adjectives in Tabassaran bear a predicative marker,
which Hjelmslev considered to be a marker of predicative case. He also took the
attributive marker to be a marker of attributive case (Hjelmslev 1937: 13940),
but the two markers could be considered derivational markers of their respective
classes. This analysis would seem compelling in that there is a third marker in the
set, namely -d-, which marks adjectives serving as heads of phrases. This -d- would
appear to mark the derivation of nouns from adjectives. It is followed by case suf-
xes: bicur-d-i qapur the little one said where -d- is followed by the ergative-
instrumental -i.
In the Indo-Aryan languages adjectives exhibit fewer case distinctions than nouns
and in some, including Bengali (Klaiman 1987: 499), they exhibit no concord at
all. In Germanic languages that exhibit a morphological case system it is noticeable
that in noun phrases with a determiner, an adjective and a noun it is the determiner
101
Table 4.1 German case inection
nominative der liebe Mann the dear man
accusative den lieben Mann
genitive des lieben Mannes
dative dem lieben Mann(e)
that displays the maximum amount of differentiation. A paradigm of German noun
phrases is given in Table 4.1. It shows a masculine singular noun of the strong
declension with an adjective and determiners.
7
In Maba (Nilo-Saharan) a nominativeaccusative distinction is made by tone only
on the determiner: ` am ar` a-g` u lion-NOM versus ` am ar` a-g u lion-ACC (Tucker and
Bryan 1966: 199).
In some Balto-Finnic languages there is concord with some case markers and
not with others. Kilby relates this difference to the age of the case markers. The
older ones display concord, the newer ones do not. In the following examples from
Estonian, the elative shows concord (24), but not the comitative (25) (Kilby 1981:
115).
(24) Puu on murtud tugev-ast tuul-est
tree be broken strong-EL wind-EL
The tree was broken by a strong wind.
(25) . . . ise aralise elavuse-ga
unexpected liveliness-COM
with unexpected liveliness
In a few languages such as Georgian and Hurrian concord extends from the
head of a noun phrase to dependent noun phrases bearing adnominal case marking
(Plank 1990). The effect of this is to produce double case marking. (26) is from Old
Georgian (Melcuk 1986: 69).
(26) sarel-ita man-isa-jta
name-INST father-GEN-INST
with fathers name
Double case marking is not uncommon in Australia. It is found not only in languages
with concord but in languages that mark case phrase-nally. In (27), which is from
Alyawarra, -kinh marks the relationship of artwa ampu to ayliyla, and -ila marks
the external relationship of the phrase as a whole (after Yallop 1977: 117f ).
102
The sufx -(i)la represents a syncretism or merger of the ergative and the locative.
It expresses A, instrument and location, and is glossed OBL(ique).
(27) Artu-la ayliyla artwa ampu-kinh-ila atu-ka
man-OBL boomerang man old-GEN-OBL hit-PAST
The man hit [him] with the old mans boomerang.
Quechua is a typical subjectobjectverb (SOV) language with preposed modi-
ers in noun phrases and phrase-nal case marking. A dependent genitive precedes
the head noun and is cross-referenced on it (Weber 1989: 254).
(28) Hwan-pa wasi-n-ta rika-a
John-GEN house-3SG-ACC see-1SG
I see Johns house.
If the genitive appears on its own either because the head noun has been elliptically
deleted (29) or because the possessor and possessed are represented as separate
phrases (30), then the genitive case marker can be followed by an adverbal case
marker.
(29) Hwan-pa-ta rika-a
John-GEN-ACC see-1SG
I see Johns (house).
(30) Hipash-nin-ta kuya-a Hwan-pa-ta
daughter-3SG-ACC love-1SG John-GEN-ACC
I love Johns daughter.
Double case marking in noun phrases is logical in that it arises naturally from
the operation of two independent case-marking strategies, namely the marking
of dependents and the use of concord. The fact that it is relatively unusual sug-
gests that languages tend to employ some special principle to bar it. In Indo-
European languages generally there can be no second or outer layer of case mark-
ing; concordial case marking is suppressed where it would give rise to double case
marking. In Latin, for instance, if a noun in, say, the accusative has a dependent
genitive, then there is no extension of the accusative to the genitive-marked de-
pendent: Vd c onsulis ux orem (saw.1SG consul.GEN wife.ACC) I saw the consuls
wife.
Another way of suppressing double case marking in noun phrases is to allow
concordial case to supplant the expected genitive. This happens in Armenian where
103
the genitive is supplanted by both the ablative and instrumental (H ubschmann 1906:
479):
(31) bazmnteamb zaraukn Hayoc
mass.INST troops.INST Armenian
with the mass of Armenian troops
For an analogous example of concordial case marking replacing relational case
marking at the clause level see (55) and (56) below.
4.4 Within the word
4.4.1 Types of marking
As noted in the previous section case is most often realised via sufxes.
The almost total absence of prexes is striking.
8
One solid example of case prexing
is to be found in Nungali, a Northern language of Australia (Hoddinott and Kofod
1976: 397). In this language there are noun classes. Prexes mark case and class
cumulatively (Table 4.2).
In some languages the case marking is not linearly separable from the stem.
This is true of the classical Indo-European languages though the fusion of stem
and case/number sufx is not too great. One can almost separate stem and sufx
as a perusal of the Latin case forms in Table 1.2, which are fairly typical, reveals.
An example of greater fusion can be found in the umlauted datives and plurals of
Germanic. In Old English the dative of f ot foot is f et which derives from *f oti via
f ti and f t, the original -i sufx having induced an assimilatory fronting in the stem
vowel before being lost. In the Slavonic languages case is basically marked by suf-
xation, but certain distinctions in some languages are realised by suprasegmental
phonological properties. In Serbo-Croat the genitive/dative and locative of i-stems
are distinguished by tone: stv ari thing (GEN/DAT) versus stv ari (LOC) (Corbett 1987:
398f). In Russian the genitive singular and nominative plural are distinguished in
some instances by stress: ruk (GEN.SG) versus r uki (NOM.PL) (Comrie 1987: 333).
The ultimate union of stem and case is suppletion as evidenced, for example, in the
Table 4.2 Nungali class/case prexes
absolutive oblique
Class 1 di-gal nyi-gal water
Class 2 nya-ngarrung nganyi-ngarrung woman
Class 3 nu-ngulud nyu-ngulud camp
Class 4 ma-yadayn nyi-yadayn
9
skin
104
4.4 Within the word
pronouns of Indo-European languages, witness English I (nominative) versus me
(accusative).
Another category of inection found frequently on nouns is number. The marking
for number, like case marking, is usually via sufxation, and the number sufx
normally precedes the case sufx as in Tamil: aRu-kal
.
-ai (river-PLUR-ACC) rivers
(Steever 1987: 737).
10
A partial exception is to be found in Classical Armenian
where in some paradigms the number marking follows the case marking: am year,
am-k years; amaw (INSTR.SG), amaw-k (INSTR.PL) (Coleman 1991: 206).
Where pronominal possessors are marked on the noun these usually appear be-
fore the case marking, as in Turkish, where they appear between the number mark-
ing and the case marking: adam-lar-m-la (man-PL-1SG.POSS-LOC) with my men;
similarly in Hungarian: haj o-i-m-on (ship-PL-1SG.POSS-LOC) on my ships (Abon-
dolo 1987: 584). In the Balto-Finnic languages, however, the possessor marking
usually follows the case marking. In Finnish, for instance, we nd: kirkolla-mme
(church-ADESSIVE-1PL.POSS) at our church (Branch 1987: 610). The adessive
case expresses the sense of near or at.
11
Another category sometimes marked on nouns, and often on their dependents, is
gender or class. Where there is a separable sufx as in Dravidian or Semitic, this
usually appears before the case marking as in Arabic: mudarris-at-u-n (teacher-
FEM-NOM-INDEF) a female teacher (cf. Kaye 1987: 672). This example also illus-
trates a further category that can be marked on nominals, namely a specication of
deniteness or indeniteness. Dyirbal (Pama-Nyungan) is exceptional in marking
noun class by post-case sufxation on determiners: ba-gu-l that-DAT-MASC versus
ba-gu-n that-DAT-FEM (Dixon 1972: 44).
4.4.2 Stem formatives
In quite a number of languages some case marking is added to stems
that include a stem-forming element, often identiable as a case marker. This is
not uncommon in the Daghestan languages. The paradigm displayed in Table 4.3
is a partial paradigm from Archi (Kibrik 1991: 256). The genitive and dative are
representative of all the other cases in that they are based on a stem which coincides
Table 4.3 Archi case marking
singular plural
nominative qIin qIonn-or
ergative qIinn-i qIonn-or- caj
genitive qIinn-i-n qIonn-or-c e-n
dative qIinn-i-s qIonn-or-c e-s
105
Table 4.4 Yuwaalaray case marking
nominative dhayn ngali
ergative/instrumental dhayndu ngali
genitive dhayn.gu ngalingu
dative dhaynda ngalingunda
ablative dhayndi ngalingundi
with the ergative. One can say that they are built on the ergative stem, but it needs
to be noted that the forms -i and - caj/- ce do not function as ergatives when another
case sufx follows. The ergative would appear to be the unmarked oblique case (see
also Melcuk 1986: 63).
In Yuwaalaray (Pama-Nyungan) the dative and ablative of the pronouns are built
on a genitive stem. A typical noun paradigm and typical pronoun paradigm are dis-
played in Table 4.4 (Williams 1980: 38, 47). The formative -ngu(n)- in ngalingunda
and ngalingundi does not function as a genitive and -ngunda and -ngundi can be
taken to be unanalysable case markers. That is not to say that the presence of
-ngu (n)- is insignicant. It marks all the noncore cases. Such a generalisation can
be captured in a feature analysis of the cases (see Table 2.15), but such an analysis
is not dependent on the presence of any marker.
As mentioned in section 2.3.4(c), in Tamil singular nouns have a stem distinct
from the nominative to which all oblique case marking is added, but the stem-
forming element does not appear without a following case marker. Maram tree,
for example, has a nominative maram and an oblique stem marratt- to which the
case markers are sufxed: maratt-ai ACC, maratt-ukku DAT, etc. (Steever 1987: 737).
Sometimes the differentiation between nominative and oblique is the result of
phonological processes as with the consonant stems of several of the older Indo-
European languages. In Ancient Greek, for instance, the oblique stem is original
and the nominative results from cluster simplication, e.g. elpis NOM hope has an
oblique stem elpid- as in elpida ACC, elpidos GEN, elpidi DAT. In the nominative the
nal d of the stem has been lost before s.
4.4.3 Compound case marking
It is possible for a stem to bear more than one inectional case marker.
In some languages, especially the Daghestan languages, there are numerous local
cases the marking for most of which is segmentable into a component for location,
destination, path or source on the one hand, and a component for relative position
on the other. In Avar, for instance, -de indicates to the top of and may be broken
down into -d indicating on or over and -e representing away from. A fuller
exposition is included in section 5.6.
106
4.4 Within the word
As Melcuk points out, the stem-forming use of case markers and compound
case marking are logically independent. They co-occur in a number of Northeast
Caucasian languages including Lezgian where, for instance, the word vaxa-q
h
-di
moving to behind the sister consists of a base vax sister augmented by the ergative
-a used as a stem-forming case, plus q
h
behind and -di to (Melcuk 1986: 64).
4.4.4 Multiple case
Multiple case is not a common phenomenon. It involves two or more
cases with different scope and often with different functions. Most but not all
instances involve an inner layer of adnominal case plus an outer layer of adverbal
case, as in the examples from Old Georgian, Alyawarra and Quechua quoted in
section 4.3.2 (see (25) to (30)). However, some Australian languages evince double
adverbal case. In Warlpiri, for instance, a locally marked adjunct may take ergative
case marking in a transitive clause (Hale 1982: 266). Consider the contrast between
the following sentences. In (32a) the noun phrase in the role of destination is marked
for allative case as one would expect. However, if a verb for carry is substituted
for a verb meaning send, then it is possible to further mark the allative-marked
phrase for ergative:
(32)
a. Ngarrka-ngku ka maliki ngurra-kurra yilya-mi
man-ERG PRES dog.NOM camp-ALL send-NPST
The man is sending the dog to the camp.
b. Ngarrka-ngku ka kuyu ka-nyi ngurra-kurra (-rlu)
man-ERG PRES meat.NOM carry-NPST camp-ALL(-ERG)
The man is carrying the meat to the camp.
An inner local phrase normally has the patient as its scope (and this is normally
encoded in the absolutive relation, i.e. as S or P (Blake 1982: 76)). In Warlpiri the
use of the ergative on the locally marked phrase is to indicate that the agent (A) is
also within its scope. With carrying, the agent moves to the same destination as the
patient.
Where a local phrase occurs with a two-place intransitive verb that takes a nom-
inative subject and dative complement, the locally marked phrase can show dative
concord (Hale 1982: 269):
(33)
Ngaju ka-rna-rla kurdu-ku mari-jarri-mi ngurra-kurra-ku
I.NOM PRES-1SG-IO
12
child-DAT sorry-INCH-NPST camp-ALL.DAT
I feel sorry for the child (who is on its way) to camp.
107
Another Australian language, Kayardild, has what look like spectacular examples
of multiple case. In this language there is adnominal case, adverbal case, and two
outer layers of what are etymologically case marking. In the following example the
outer layers are glossed as M(odal)ABL(ative) and OBLIQUE (Dench and Evans
1988: 34).
(34) Maku-ntha yalawu-jarra-ntha yakuri-naa-ntha
woman-OBL catch-PAST-OBL sh-MABL-OBL
dangka-karra-nguni-naa-ntha mijil-nguni-naa-ntha
man-GEN-INST-MABL-OBL net-INST-MABL-OBL
The woman must have caught sh with the mans net.
The adnominal case is the genitive realised by -karra and the adverbal case is the
instrumental realised by -nguni which appears on mijil net and via concord on
dangka-karra.
Of the two outer layers the modal ablative appears on all the dependents within the
verb phrase and the oblique appears on every word in the clause. These markers do
not mark the relationship of dependent nouns to their heads, or even of dependent
verbs to their heads; they have a modal value. They are not case markers, but
they derive from case markers. Their historical derivation involves the following
steps:
(a) case is marked on a dependent verb
(b) case spreads from the head of the dependent clause (the verb) to its
dependents via concord
(c) the governing clause is omitted and the formerly dependent clause
becomes an independent clause.
Consider, for instance, the ablative, which means from. It can often be marked on
a verb, or at least a nominalised verb, where it will normally refer to the domain of
time. Compare English From working as a waiter he graduated to becoming one of
the citys leading restauranteurs. In many Australian languages case marking on a
dependent verb will spread to the dependents of that verb, often producing multiple
case marking. An example from Yukulta, which is closely related to Kayardild,
is given as (53) below. It is also common in Australian languages for dependent
clauses to come to be used without their governing clauses, i.e. they develop into
independent clauses (an example is given in (24) in chapter 6). When this happens
the case in the formerly dependent clause no longer marks the relationship of a
dependent to a head, since there is no longer any head. An ablative, which formerly
marked the subordinate clause as referring to an event prior to the activity or state
108
4.5 Within the subordinate clause
described in the governing clause, comes to mark an event as prior to the speech
act.
13
In (34) the oblique, which may once have shown concord with an oblique con-
stituent in a higher clause, marks the clause as if it were a clausal argument of a
perception verb like SEE. The oblique has the effect of presenting the proposition as
an inference (Evans 1995: 450 ff, Dench and Evans 1988: 345). It should probably
be treated as a feature of the verb which spreads to the dependents via concord.
Like modal case it is synchronically ungoverned (see section 1.2.4) and therefore
it lies outside the scope of our original, central denition of case and our broader
denitions (see section 4.2). Nevertheless, the derivation of this marginal use of
case marking is clear.
Multiple case differs from compound case marking in that it cannot be cap-
tured in terms of a selection of unordered features from a particular set. This is
particularly obvious in examples involving two tokens of the same case. Follow-
ing Stephen Anderson (1982: 598ff), we could suggest that each layer of multiple
case be assigned to a different layer of word structure, so that a word like ngurra-
kurraku in (33) would be described in terms of a bracketed structure of the form
[[ngurra-ALLATIVE]DATIVE] (Anderson 1982: 598ff).
14
Where a clause is subordinate the notion of case is relevant to the
marking of relations within the clause and the marking of the relation of the clause
as a whole to its governing predicate.
Broadly the following strategies are employed within a subordinate clause:
(a) no change from the schema used in an independent clause
(b) a nonnite predicate
(c) nominalisation
Nominalisation may be partial or complete. Where it is complete, the result is a
noun phrase. Such a nominalisation is not a subordinate clause but the functional
equivalent of a subordinate clause.
Subordinate clauses, including their reduced equivalents, serve a number of
functions:
(a) outer adjunct (not closely related to the meaning of the governing
predicate)
(b) inner adjunct (e.g. a purpose clause where the subordinate clause makes
sense only with certain governing predicates)
(c) complement (e.g. complement of verbs of saying, wishing, etc.)
(d) core functions such as subject and object
109
There is a correlation between function and form. The further one goes from
outer adjunct along the function scale, the greater the likelihood of formal reduction
towards nominalisation.
15
(35) form: nite > nonnite > partial nominalisation > full nominalisation
function: outer adjunct > inner adjunct > complement > core function
4.5.1 Internal relations
A nonnite predicate is signicantly different from a nite one in that
it usually lacks any specication of subject either within the verb or via a noun
phrase.
16
The missing subject of a complement clause is normally interpreted as
being coreferential with the S of an intransitive governing verb or the P of a tran-
sitive one, and the subject of an innitival adjunct clause is supplied by pragmatic
principles.
17
Another possibility is for the expected subject of the nonnite verb to
be expressed as the object of the governing verb. These distinctions can be illus-
trated from Imbabura Quechua (Cole and Hermon 1981 via Foley and Van Valin
1984: 2756). In (36) the complement of yacha know is expressed by a clause
with the normal valency, nominative for subject and accusative for object, though
the verb has a marker ( j) sensitive to tense, which replaces any representation of
the subject in the verb (compare yacha-n) and forms a stem to which case can be
added. It is glossed as a nominaliser (PRES.NM), but there is no syntactic reection
of any nominalisation.
(36) Juzi yacha-n nuca Maria-ta juya-j-ata
Jose-NOM know-3SG 1SG.NOM Maria-ACC love-PRES.NM-ACC
Jose knows that I love Maria.
In (37) the subordinate clause lacks an overt subject, the missing subject being
interpreted as being coreferential with Juzi.
(37) Juzi muna-n lichi-ta ufya-na-ta
Jose-NOM want-3SG milk-ACC drink-FUT.NM-ACC
Jose wants to drink milk.
In (38) the rst person singular is understood to be the experiencer of juya love.
However, it is expressed as the direct object of yacha know. In some modern
frameworks it would be described as having been raised from the lower clause to
the higher clause.
18
(38) Juzi nuca-ta yacha-wa-n Maria-ta juya-j-ta
Jose-NOM 1SG-ACC know-1SG-3SG Maria-ACC love-PRES.NM-ACC
Jose knows me to love Maria.
110
Full nominalisation allows the propositional content of a clause to be expressed
as a phrase. The process can be illustrated from English:
(39) a. John read the book to the students quickly
b. Johns quick reading of the book to the students
c. He read the book to the students quickly
d. His quick reading of the book to the students
The subject is re-expressed as a determiner and the object is re-expressed as a com-
plement marked by the preposition of.
19
The adverb is re-expressed as an adjective.
Adjuncts like the to-phrase in (39a) carry over to the nominalised phrase.
As we can see from (39b) English has genitive-like phrases to express both the
subject and the object of the nominalised verb. Where both genitive-like phrases
are used the s-phrase must be used to correspond to the subject and the of-phrase
to correspond to the object. Where only one of the complements of the verb is
to be re-expressed with the nominalised verb, there is some possibility of using
the s phrase for either the subject or the object; the interpretation of a phrase
like the robots delivery is dependent on context. There is also some possibility of
using either the s-phrase or the of-phrase to encode the subject of a nominalised
intransitive though s is normal with animate agents (The babys crawling) and of
with inanimate patients (the dripping of the tap). However, English is unusual in
having two genitive-like possibilities; most languages have only one.
As we saw in section 4.3.1 Latin uses the genitive case in both subjective and
objective senses, which can lead to ambiguity where both occur in the one sentence.
A feature of nominalisation is that it typically involves a reduction in the range of
possibilities found with verbs. In Latin the genitive used with a nominalisation can
correspond to any adverbal case including prepositionally marked complements.
Thus one can have imperium provinci arum command of the provinces where
the base verb imper are governs the dative, and re p ublicae diss ensi o (matter-GEN
public-GENdisagreement) disagreement onpolitical matters where the correspond-
ing verb and its complement is d e r e p ublic a dissentre with the complement of the
verb marked by the preposition d e which governs the ablative.
Although in general there is a reduction of the range of case possibilities, there
are extra prepositional possibilities mainly used to resolve ambiguities which would
result if two genitives were used. The phrase amor domin dominae could mean the
love of the master for the mistress or the love of the mistress for the master,
although the rst-mentioned genitive tends to be taken to be subjective. The am-
biguity can be resolved by providing a prepositional alternative for the objective
genitive: amor domin erga dominam means the love of the master towards the
mistress.
111
The use of a nominalised verb does not always mean a wholesale switch to noun
phrase structure. In English, for instance, alongside Johns reading of the book as
in (39b) there is partial nominalisation as in John(s) reading the book surprised me
or His reading the book surprised me where the nominalised verb still takes a direct
object though it can take a determiner.
In Latin the present participle behaves morphologically as a noun in that it takes
case and number marking and syntactically as a nonnite verb in that it takes the
same range of complements and adjuncts as the underlying verb less the subject
(Virgil, Aeneid 2, 49):
(40) . . . time o Dana os et d ona fer-ent- es
fear.1SG Greek.ACC.PL and gift.ACC.PL bear-PRPART-ACC.PL
. . . I fear Greeks even bearing gifts.
In (40) ferent es agrees in case and number with Dana os and takes an accusative
object, d ona. However, some participles used with reference to ongoing states or
permanent properties came to be regarded as adjectives and then took a comple-
ment in the genitive which reects full nominalisation. The following example is a
standard one (Cicero, pro lege Manilia 3):
(41) Semper appetent es gloriae atque avid
always seeking.NOM.PL glory.GEN and avid.NOM.PL
laudis fuistis
praise.GEN were.2PL
You have always been thirsty for glory and eager for praise.
Latin is unusual in allowing the genitive to encode both the subject and the
oblique complement in nominalisations, and, as we have seen, English is unusual
in having two genitive-like dependents available in a noun phrase to encode sub-
jects and objects of nominalised verbs. Most languages have only a single genitive
strategy. In some languages the genitive encodes the object of the verb; in others it
encodes the subject. According to Giv on the latter is more common (Giv on 1990:
504).
In Dhalandji (Pama-Nyungan) the dative, which is the main adnominal case in
this language, is used to mark the complement of transitive verbs in nonnite relative
clauses. Contrast the accusative on kanyara and the dative on murla in (42) (Austin
1981c: 222). It is signicant that the dative is the specialised adnominal case in this
language corresponding to the genitive in Latin.
112
(42) Ngatha nhaku-nha kanyara-nha murla-ku warni-lkitha
I see-PAST man-ACC meat-DAT cut-REL.DS
I saw the man (who was) cutting meat.
The verbal sufx -lkitha relative, different subject marks a qualifying clause the
covert subject of which must be interpreted as being distinct from the main clause
subject (see also section 6.4).
In Turkish the genitive is used to mark the subject of a nominalised verb. The
object of such a verb if present takes the normal case marking (Kornlt 1987:
640).
(43) Ahmed-i ben-i sev-di g-in-i bil-iyor-um
Ahmed-GEN 1SG-ACC love-NM-3SG.POSS.ACC know-PRES-1SG
I know that Ahmed loves me.
The form -in is a third-person-singular possessive form in cross-reference with
Ahmed-i. In Turkish noun possessors are cross-referenced on possessed nouns:
Biz-im heykel-imiz (we-GEN statue-lPL.POSS) our statue. The accusative on the
nominalised verb marks it as the complement of biliyorum, and the accusative on
ben marks it as the complement of sevmek to love.
In Finnish the subject of an innitive is expressed in the genitive. There are a
number of different innitives used for different purposes. The following exam-
ples involve an innitive marked for the inessive (inside) case to express action
contemporaneous with that of the governing verb. In (44) the subject of the inni-
tive is not expressed by a noun phrase since it is the same as the subject of the
governing verb sanoa to say, but it is represented by a possessive pronominal
sufx which follows the inessive case marking on the innitive (Whitney 1956:
142).
(44)
Tule, sano-i t ati kahvi-a kuppe-i-hin kaata-e-ssa-an
come say-IMPF.3SG aunt coffee-PAR cup-PL-ILL pour-INF-IN-2SG.POSS
Come, said aunt, as she poured coffee into the cups.
In (45) the understood subject of the inessive-marked innitive is different
from that of the governing verb tuijotta to stare and it is expressed in the
genitive.
113
(45) T adi-n kaata-e-ssa kahvi-a kuppe-i-hin,
aunt-GEN pour-INF-IN coffee-PAR cup-PL-ILL
tuijo-tti Eeva ikkuna-sta ulos
stare-IMPF.3SG Eva window-EL out
While aunt was pouring coffee into the cups, Eva stared out of
the window.
Besides these widespread strategies of using nonnite verbs and nominalisation,
there are various other more particular strategies of which two will be described.
The rst is the accusative and innitive found in Ancient Greek and Latin. The
second is the use of detransitivisation to indicate certain patterns of coreference.
This strategy is found in a number of Australian languages including Dyirbal and
Kalkatungu.
In Ancient Greek and Latin there is a subordinate clause type known as the ac-
cusative and innitive used to complement verbs of speaking, thinking, believing,
etc. The verb is in the innitive, but whereas nonnite verbs do not normally have
overt subjects, the innitive in this construction takes an accusative noun phrase
expressing S and A. If the verb is transitive, it still takes a direct object in the
accusative, so there can be two accusative noun phrases as in the following Latin
version of a prophecy from the priestess at Delphi to Pyrrhus, king of Epirus:
(46) Ai o t e R om an os vincere posse
say.1SG you.ACC Romans.ACC conquer.INF be.able.INF
Since word order is an uncertain guide in Latin, the string in (46) can be interpreted
as I say that you can conquer the Romans or I say that the Romans can conquer
you. Pyrrhus optimistically chose the rst alternative, which proved to be the wrong
choice.
20
In a handful of Australian languages a detransitivised construction is used in
purpose clauses andincertainother dependent clauses tosignal coreference between
the A of a lexically transitive verb and the absolutive of the governing clause. The
best-known exemplar of this phenomenon is Dyirbal (Dixon 1972). The following
examples are from Kalkatungu (cf. Blake 1979a). As illustrated in section 3.2,
the core case marking operates on an ergativenominative basis. (47) illustrates an
intransitive clause and (48) a transitive clause.
(47) Kaun muu-yan-ati
dress.NOM dirt-PROP-INCH
The dress is dirty.
114
(48) Kuntu wampa-ngku kaun muu-yan-puni-mi
not girl-ERG dress.NOM dirt-PROP-CAUS-FUT
The girl will not dirty the dress.
There is a detransitivised alternative to the transitive construction, namely the an-
tipassive. In this construction the A of the transitive verb is recoded as S (in the
nominative) and P is recoded in the dative (also illustrated in (1b) in chapter 3).
In independent clauses the antipassive signals reduced semantic transitivity, in this
instance habitual activity.
21
(49) Kuntu wampa kaun-ku muu-yan-puni-yi-mi
not girl.NOM dress-DAT dirt-PROP-INCH-AP-FUT
The girl will not dirty [her] dresses.
In dependent clauses the use of the antipassive is determined by syntactic rather
than semantic factors. The following illustrations involve lest clauses introduced
by the subordinating particle ana. In (50) the antipassive is used in the rst lest
clause signalling that the A of the transitive verb muuyanpuni is coreferential with
the S of the governing verb. This contrasts with the second lest clause where the
absence of the antipassive indicates that the covert A is referentially distinct from
the S of the governing verb.
(50) Wampa rumpi ana kaun-ku muuyanpuni-yi ana lhaa
girl.NOM fear LEST dress-DAT dirty-AP LEST hit
The girl is frightened in case [she] dirties her dress and [someone]
hits her.
If we substitute a transitive governing clause with wampa as P the antipassive
will operate in the lest clauses with the same effect.
(51) Wampa nga-thu yarrka-puni-mi ana kaun-ku
girl.NOM 1-ERG far-CAUSE-FUT LEST dress-DAT
muuyanpuni-yi ana lhaa
dirty-AP LEST hit
Ill send the girl away in case [she] dirties her dress and [someone]
hits her.
The antipassive is used to indicate that the A of a lexically transitive verb is
coreferent with the absolutive (S or P) of the governing verb.
115
4.5.2 External relations
The relationship of a subordinate clause to its governing predicate is
more often than not signalled by the normal case marking found in the language, i.e.
sufx, preposition or postposition. In English this is partly true in that prepositions
like before and after can function as subordinating conjunctions (Before I go, Ill
phone. After you get there, phone me), but there are subordinating conjunctions
such as when, although and while that are peculiar to that function.
22
In a large
number of languages the local case markers are used with nominalised verbs to
refer to time. In Quechua, for instance, the ablative is used to indicate since and
the limitative (as far as) to indicate until (Weber 1989: 291). In Pitta-Pitta the
ablative indicates since, the locative while and the allative until. An example
of the Finnish inessive (in) case used to indicate while was given in (43) above,
and an example of the translative appears in (52) immediately below.
Where a clause rather than a noun phrase is a dependent, the same possibilities
for the distribution of case marking arise. In most instances the case marker appears
only on the head of the clause, namely the verb, as in the following example from
Finnish, where the translative case is found on the innitive. The translative means
into, mainly metaphorically as in Youll turn into a pumpkin, and purpose as in
mi-ksi what for. With a nominalised verb it indicates purpose. The actor of the
nominalised verb is expressed by the possessive pronominal sufx (compare (45)
above) (Branch 1987: 615).
(52) Osti-n karttakirja-n suunnitella-kse-ni automatka-n
bought-1SG atlas-ACC plan-TRANS-1SG .POSS car.trip-ACC
I bought an atlas in order to plan a car journey.
Another possibility is for the case marking of a dependent verb to spread to its
dependents by concord. The following example is from Yukulta (Northern Aus-
tralian). Note that the dative, which is appropriate to the verb warratj-, spreads
to the allative-marked complement to yield a second layer of case marking (Keen
1972: 270).
(53) Taamitya = ngandi tangka natha-rul-ngkurlu warratj-urlu
ask = 1SG.3SG.FUT.AUX man.NOM camp-ALL-DATIVE go-DATIVE
Ill ask the man to go to the camp.
This spreading may also occur where a clause modies a noun phrase. In the
following example from the Pama-Nyungan language Panyjima all words in the
modifying clause show concord with the accusative-marked object of the main
clause (Dench and Evans 1988: 28).
116
(54) Ngatha wiya-rna ngunha-yu maripa-yu
I.NOM see-PAST that-ACC man-ACC
paka-lalha-ku nharniwalk-ku warrungkamu-la-ku
come-PERF-ACC hither-ACC morning-LOC-ACC
I saw that man who came this way this morning.
In Vedic Sanskrit the case marking of nominalised verbs traditionally described
as innitives spread to the complement of the innitive (Burrow 1955: 365):
(55) dr
.
s aye s
ury- aya
see.NM.DAT sun-DAT
to see the sun
(56) Tr

adhvam kart
ad avap adah
save.IMP pit.ABL fall.NM.ABL
Save [us] from falling into a pit.
These examples illustrate two principles. The rst is case concord in a subordinate
clause (cf. (53)) and the second is the use of special case marking for the complement
of a nominalised verb (cf. (42)).
117
5
Survey of case marking
5.1 Introduction
This chapter provides a global perspective of case systems and their
marking. It is divided broadly into two parts. In section 5.2 the organisation of
the core or nuclear relations is surveyed and in section 5.3 the organisation of
the peripheral relations is described. As Nichols 1983 points out, it is difcult to
maintain a strict distinction between core grammatical cases, which encode S, Aand
P, andsemantic cases like locative, allative andinstrumental. Asingle case maycover
A and instrumental function or P and allative (destination) function. Nevertheless
there are signicant generalisations that can be made about cases that encode S, A
and P, even if they sometimes cover peripheral grammatical relations as well.
5.2 Organisation of the core
As noted in section 3.5.3, noun phrases bearing core relations are of-
ten unmarked altogether, while noun phrases in peripheral relations are marked by
inectional cases, adpositions or both. In languages like this the burden of dis-
tinguishing subject (SA) from object (P) or the absolutive relation (SP) from the
ergative (A) is borne by cross-referencing bound pronouns or word order. On the
other hand there are languages with inectional case only for the core relations, or
for these plus possessor function (genitive), with adpositions for other grammatical
relations.
In the majority of languages the core grammar is organised on an SA/P basis
(accusative system). In a minority of languages the core is organised on an SP/A
(ergative) basis. A few languages have a system which could be described roughly
as distinguishing agents from patients irrespective of transitivity (active system)
and a few others are mixed.
5.2.1 Accusative system
The familiar languages of Europe, whether Indo-European or Uralic,
exhibit accusative systems. In Latin, for instance, the subject is encoded in the
nominative and the object in the accusative. Moreover, the subject is represented
118
pronominally in the verb. In Romance languages like French and Spanish the un-
marked word order also picks out the SA alignment, the order being subjectverb
object as in English. In the Bantu languages typically there is no case marking, but
the cross-referencing agreement systemand the word order operate in an accusative
system. Both subject and object are represented by prexes on the verb and the basic
word order is subjectverbobject.
Pure accusative systems of marking noun phrases where the marking of the
object is always distinct from the marking for subject are rare. Korean provides an
example, but, as pointed out in section 1.2.5 (see (12)), the status of the grammatical
case markers vis-` a-vis the postposition/sufx distinction is controversial. Japanese,
which employs postpositions as case markers, provides another example (see (9)
in chapter 1). With most languages, however, some qualication is required with
respect to animacy and/or deniteness. In Latin and the other Indo-European case
languages there is no nominative/accusative distinction with neuter nouns. This is
related to animacy in that virtually all neuter nouns are inanimate, though inanimate
nouns are also plentiful in the masculine and feminine genders. In the Slavonic
languages the accusative is not always distinct from the nominative, but it tends to
be the same as the genitive and distinct from the nominative for animate nouns.
In Russian, for instance, animate nouns of the masculine o-class and all animate
plurals take the accusativegenitive form.
In many languages accusative case marking is not used if the direct object is
indenite. It would probably be more accurate to say nonspecic, but even that
requires some qualication. Consider the following examples from Turkish.
1
(1) a. Hasan ok uz- u ald
Hasan.NOM ox-ACC buy.PAST.3SG
Hasan bought the ox.
b. Hasan bir ok uz ald
Hasan.NOM a ox buy.PAST.3SG
Hasan bought an ox. (nonspecic)
c. Hasan bir ok uz- u ald
Hasan.NOM a ox-ACC buy.PAST.3SG
Hasan bought an ox. (specic)
In (1a) the accusative case marker appears indicating that the direct object is denite.
A denite referent is one that is taken to be identiable by the hearer. In English
the denite article is used with common nouns when they are used with denite
reference. In (1b) and (1c) the direct object is indenite as indicated by the indenite
article bir. However, it is possible to use the denite accusative, as it is sometimes
called, with the indenite article. The accusative, as in (1c), is likely to be used where
119
the reference is to a particular (specic) ox identiable by the speaker. Where there
is no accusative marker, as in (1b), the reference is more likely to be to a nonspecic
ox (I was looking for an ox, but I couldnt nd one). It might be thought that if buying
is the activity involved, then a specic entity must pass into the possession of the
buyer, but where the identity of this item is unimportant, no accusative would be
used.
In other languages the direct object is marked only if it is both specic and
animate. In Hindi, for instance, a postposition ko marks specic, animate patients,
a usage with parallels in other languages of the subcontinent (see also (26) to (29)
below). In Spanish the preposition a is used to mark specic animate patients.
Compare (2a) and (2b).
(2) a. Deseo un empleado
want.1SG an employee
I want an employee. [Anyone will do.]
b. Deseo a un empleado
want.1SG an employee
I want an employee. [I cant think of his name for the moment.]
The preposition a (from Latin ad ) means to, though not in this context. In Ruma-
nian pe on (from Latin per through) performs an analogous function (Mallinson
1987: 31516).
Personal pronouns are denite and mostly animate and typically share any mark-
ing for animate and/or specic patients. In some languages only pronouns bear
accusative marking. English is, of course, an example.
In all the examples discussed up to this point the marking of P is sensitive to
properties of P alone, but in a few languages the accusative is sensitive to the
relative status of A and P on a person hierarchy of the form 1, 2 > 3 or 1 > 2 > 3
(> means is higher than). In the Australian language Rembarnga the proclitic
representing P on the verb is sufxed with -n only when the person/number of P is
higher than the person/number of A. In this language the hierarchy is of the form
1 > 2 > 3 plural > 3 singular. Compare (3a) where A is lower than P with (3b)
where A is higher than P (based on McKay 1976).
(3) a. Nga-n-pa-na
1SG-OBJ-3PL-see
They saw me.
b. Pa-nga-na
3PL-1SG-see
I saw them.
2
120
5.2.2 Ergative system
Ergative systems are often considered rare and remote, but in fact they
make up at least twenty per cent of the worlds languages. Ergative systems are
to be found in all families of the Caucasian phylum, among the Tibeto-Burman
languages, in Austronesian, in most Australian languages, in some languages of the
Papuan families, in Zoque and the Mayan languages of Central America and in a
number of language families in South America: J e, Arawak, Tup-Guaran, Panoan,
Tacanan, Chibchan and Carib.
3
Outside these phyla and families where ergative
systems of marking are common, ergativity is also to be found in some other lan-
guages including Basque, Hurrian and a number of other extinct languages of the
Near East, Burushaski (Kashmir, Tibet), Eskimo, Chukchi (Kamchatka peninsula),
and Tsimshian and Chinook (these last two being Penutian languages of British
Columbia).
In an ergative system of case marking there is normally zero marking for SP
(absolutive) and positive marking for A. The morphologically unmarked case is
referred to by some as the nominative and by others as the absolutive. In this text
the term nominative is used, the term absolutive being reserved for the grammatical
relation embracing S and P. The Daghestan language Avar provides an interesting
example because the SP/A system shows up also in the cross-referencing bound
pronouns (Ebeling 1966: 77).
(4) W-as w-ek er-ula
M-child.NOM M-run-PRES
The boy runs.
(5) Inssu-cca j-as j- ecc-ula
(M)father-ERG F-child.NOM F-praise-PRES
Father praises the girl.
The nominative is unmarked. The case marker glossed as ERG(ative) is the one that
also indicates instruments. It is common to nd that in ergative languages the so-
called ergative marks a peripheral function such as genitive, locative or instrumental
as well as A. In fact in Avar A is peripheral relative to P. P is represented generically
on the verb by a prex for masculine, feminine or neuter. In (5) j- is the pronominal
prex for feminine. It also appears on the noun as child. A, on the other hand, is
not represented on the verb and is omissible. Note that S in (4) is represented on
the verb by w-, the pronominal prex for masculine.
In most languages the perceiver of verbs of seeing and hearing aligns syntactically
with the agent, and the experiencer of verbs of loving, hating, etc. often does too.
Avar is interesting in lacking such a conation. While agents are put in the ergative,
experiencers are put in the dative (as in (6)), perceivers in the locative (7) and
possessors in an adverbal genitive (8).
4
121
Experiencer
(6) nssu-je j-as j- o -ula
father-DAT F-child F-love-PRES
Father loves the girl.
Perceiver
(7) nssu-da j-as j- x-ula
father-LOC F-child F-see-PRES
Father sees the girl.
Possessor
(8) inssu-l j-as j- go
father-GEN F-child F-be
Father has a daughter. [lit. of father is a girl]
In a few languages the core case marking is sensitive to the relative positions of
A and P on a hierarchy. An example was given above from Rembarnga (see (3))
where the bound pronoun in P function was marked only if A was third person and
P was second or rst, or A was second and P rst. In Fore (Papuan), the use of the
ergative case is determined by the following hierarchy: pronoun, personal name,
kin term > human > animate > inanimate. If A is higher than P, no case marking is
used, but if A is lower than P, then A appears in the ergative. So if a man kills a pig,
then no ergative is used, but if a pig kills a man, then the ergative marker -wama is
used (Scott 1978: 10017).
5
(9) Yagaa w a aeg uye
pig man.NOM 3SG.hit.3SG
The man kills the pig.
(10) Yagaa-wama w a aeg uye
pig-ERG man.NOM 3SG.hit.3SG
The pig kills the man.
Where A and P both occupy the same position on the hierarchy, the use of -wama
is optional. In the absence of marking, the rst noun phrase will normally be inter-
preted as A.
In languages with ergative case marking on nouns it is true more often than
not that the ergative marking is lacking from rst- and second-person pronouns
and sometimes from third. This latter situation is found in Yupik Eskimo, for
instance, where nouns exhibit nominativeergative marking but pronouns remain
unmarked for all core functions (Reed et al. 1977). In Kiranti and Gyarong (both
122
Table 5.1 Konjo clitic pronouns
erg abs poss
1 ku- -a -ku
2 familiar nu- -ko -nu
2 honoric ki- -ki -ta
3 na- -i -na
Tibeto-Burman) rst- and second-person pronouns lack ergative marking, but all
third persons exhibit it (Delancey 1987: 806).
6
Only a very small number of languages have an ergative system of cross-referen-
cing. Examples are to be found among the Caucasian languages, the Austrone-
sian languages and the Mayan languages. In Avar (see (4) and (5) above) only the
absolutive is represented on the verb, but in Abaza and Abkhaz (both Northwest
Caucasian) both A and SP are separately represented and there is no case marking
on noun phrases (Hewitt 1979).
7
A similar conguration is found in the Mayan
languages (Larsen and Norman 1979) and the South Sulawesi languages of Indone-
sia. The paradigm above (Table 5.1) and sentence examples are from the South
Sulawesi language Konjo (Friberg 1991). In this language the ergative and abso-
lutive relations are distinguished by both form class (Table 5.1) and order class,
the ergative bound pronoun preceding the verb and the absolutive following. Note
in passing that the ergative series of bound pronouns is the same as the possessor
series except in the second person honoric where the system distinguishes a direct
or core form from a possessor form.
8
(11) A-lampa-a
INT-go-1.ABS
I go.
(12) Na-itte-a
3.ERG-see-1.ABS
S/he sees me.
(13) Ku-itte-i balla-na
l.ERG-see-3.ABS house-3.POSS
I see his/her house.
(14) Na-itte-i balla-ku
3.ERG-see 3.ABS house-l.POSS
S/he sees my house.
Only a very small number of languages identify S and P in word order. One is
Kuik uro, a Carib language of Brazil, where the neutral word order is SV, PVA. SP
123
is also distinguished from A by stress and in the clitic system (Franchetto 1990:
4079). In Balinese the word order with basic, underived verbs is SV, PVA thereby
identifying P with S. There is a derivation yielding verbs with an AVP order (Artawa
and Blake 1997).
5.2.3 Active system
Some languages, perhaps no more than a few score, organise their core
grammar so that the argument of some one-place predicates is marked like the A
of a two-place verb while the argument of the other one-place predicates is marked
like the P of a two-place verb. Such languages have been called active languages
(Klimov1973), split-intransitive languages or split-Slanguages (Dixon1979: 805).
Examples can be found in the Kartvelian languages of the Caucasus. The following
sentences are from Laz (Harris 1985: 52f). Note that the sufx -k, glossed ergative
on the basis of its appearance on A in transitive clauses like (17), also appears on
the agent of the intransitive verb in (15). On the other hand the subject of the
intransitive verb in (16) is unmarked like the P of (17).
(15) Bere-k imgars
child-ERG 3SG.cry
The child cries.
(16) Bere oxori-s doskidu
child.NOM house-DAT 3SG.stay
The child stayed in the house.
(17) Baba-k me ccaps skiri-s cxeni
father-ERG 3SG.give.3SG.3SG child-DAT horse.NOM
The father gives a horse to his child.
This pattern also occurs in Georgian, but it applies only to certain classes of verbs
in the aorist tense group. In the present tense all subjects are in the nominative case
and the direct object in the dative (Harris 1981: 46).
9
The active system is also found in the Americas where it usually shows up in the
bound pronouns on the verb. It has been reported from Guaran (Andean); Lakhota
and other Siouan languages; the Pomoan languages; Caddo, Arikara and other Cad-
doan languages; and Mohawk, Seneca and other Iroquoian languages. It also occurs
in Acehnese (Austronesian) where agent clitics precede the verb and patient
clitics follow; the A of a transitive clause is marked by a preposition l e (Durie
1985: 18095). The position of the clitics with respect to the verb in this language
obviously invites comparison with the Konjo examples in (11) to (14) above.
The semantic motivation for splitting the one-place predicates into two groups
varies from language to language. In Guaran predicates meaning go, walk,
124
dance, swim, fall and die take agent pronominal prexes (e.g. a-x a I go),
whereas one-place predicates with meanings such as be sick, be dead, be wise
and be grey-haired take patient pronominal prexes (e.g. s e-ras
I am sick).
The distinction, according to Mithun (1991: 513), is primarily one of lexical aspect
or Aktionsart. Verbs in the agent group denote activities, accomplishments and
achievements. They involve change over time. Verbs in the patient group denote
states. InLakhota, onthe other hand, the markingis sensitive towhether the argument
is one that performs, effects and instigates an activity or not. The verbs for walk,
dance and swim take agent-pronominal prexes, but verbs for fall and die
take patient pronominal prexes (Mithun 1991: 51516).
5.2.4 Mixed systems
Some languages employ both ergative and accusative case, though it
is rare for a language to have both across-the-board ergative marking and across-
the-board accusative marking, i.e. a full tripartite system with S, A and P distin-
guished for all classes of nominal. The only example reported in the literature is
the Australian language Wangkumara (Breen 1976). The more common situation is
for ergative or accusative or both to be lacking from some classes of nominal. This
is the situation in Nez Perce (Penutian) where rst and second persons lack erga-
tive marking. The following examples illustrate the nominativeergativeaccusative
opposition (Rude 1985: 83, 228).
10
(18) Hi-p aay-na h aama
3SG-arrive-PERF man.NOM
The man arrived.
(19) H aamap-im aayato-na p ee-nehne-ne
man-ERG woman-ACC 3SG.3SG-take-PERF
The man took the woman away.
In the Hokan language Eastern Pomo (McLendon 1978) there is an interesting
variation on the active theme with complementarity between agent and patient in that
pronouns, kinship nouns and personal names may take patient marking and other
nominals may take agent marking (see Table 5.2) (I eschew the terms absolutive
and ergative in view of the distribution to be described below). Verbs fall into ve
classes with respect to whether they take one argument or two and whether they
take agent marking or patient marking:
1. One argument marked for agent and optionally a second argument
marked for patient: (kill, bite)
2. One argument markedfor patient: (fall, be burned, bleed, sneeze,
become sick)
125
Table 5.2 Case marking in Eastern Pomo
Patient Agent
pronouns -al -
kin -al -
personal proper names -yiy -
other - -la
3. Two arguments both marked for patient: (love, hate)
4. One argument marked for agent or patient: (slip/slide)
5. One argument marked for agent with pronouns, kin and proper nouns
and for patient with common nouns (i.e. no overt marking): (sit/dwell,
go/walk, stand up).
This distribution is interesting with respect to the distinction between case and
case marking. If we follow the traditional method of taking a case to be the set of
forms that together express one or perhaps more than one relation, then the columns
in Table 5.2 are cases. When it comes to describing the marking with class 5 verbs,
we can say that the patientive is used with common nouns and the agentive with
pronouns, kinship nouns and personal proper names.
An alternative would be to call the overt patient-marking forms patientive, the
overt agent-marking forms agentive and the unmarked forms nominative. This
would mean recasting all the statements made above about marking with the vari-
ous verb classes. Class 2 verbs, for instance, would have to be described as taking
patientive with pronouns, kinship nouns and personal names and nominative with
common nouns. The statements about marking would all be more complicated
except with class 5 where one could now say class 5 verbs take the nominative
case.
In a number of languages there is a split in the core case-marking system
along tense/aspect lines. Georgian has the same active pattern of marking as its
fellow Kartvelian language Laz (see (15), (16) and (17) above), but only in the
past tense (Harris 1981).
11
In the present it has a nominativeaccusative sys-
tem, though it should be noted that P is in the dative not the accusative. It is
common to refer to a system that operates on an SA/P basis as a nominative
accusative system, but it would probably be preferable to talk of an SA/P system,
since there is not necessarily any accusative case.
12
The Pama-Nyungan language
Pitta-Pitta has an unusual three-way system of marking in the past and present
with an unmarked nominative for S, an ergative for A and an accusative for P.
In the future, however, there is an SA/P system with a marked form for SA and
another marked form for P, namely the dative. Sentences (20) and (21) illustrate
126
the nominativeergativeaccusative system in the nonfuture (based on Roth 1897,
Blake 1979b).
(20) Muyutyu nhan-pa-ka nyunukana-ya
old.woman.NOM she-NOM-HERE tired-PRES
The old woman is tired.
(21) Muyutyu-lu nhan-tu-ka watyama-ka nhan-(nh)a-ka
old.woman-ERG she-ERG-HERE wash-PAST she-ACC-HERE
takuku-nha thupu-lu
baby-ACC soap-ERG
The old woman washed the baby girl with soap.
The following pair illustrate the future-tense versions of these same sentences.
Note that there is a common form -ngu for S function, A function and instrumental
function. It is glossed as F(uture) A(ctor) which is an ad hoc label for a case with an
unusual range of functions. P appears with dative marking (compare (24) below).
The future tense is morphologically unmarked.
(22) Muyutyu-ngu nhan-ngu-ka nyunukana
old.woman-FA she-FA-HERE tired.FUT
The old woman will be tired.
(23) Muyutyu-ngu nhan-ngu-ka watyama nhan-ku-ka
old.woman-FA she-FA-HERE wash.FUT she-DAT-HERE
takuku-ku thupu-ngu
baby-DAT soap-FA
The old woman will wash the baby with soap.
One would naturally wonder whether a sentence like (23) should be taken to be
intransitive since the patient is in the dative. However, there is another construc-
tion in which the patient appears in the dative, namely the antipassive, and there
is a syntactic opposition between a future-tense sentence with a dative-marked pa-
tient and an antipassive with a dative-marked patient in that the antipassive allows
the patient to be deleted but the future does not. For this reason I suggest that
the future construction is transitive and the antipassive intransitive. The dative in
Pitta-Pitta is the case used to express through or across, in other words it has
the perlative function. Its claim to be called dative rests on the fact that it marks
the patient complement of a handful of two-place verbs such as yatha- to like,
tiwa be jealous of and wapa- to look for. Example (24) illustrates one of these
verbs, wapa- to look for. The subject is in the unmarked nominative and the
127
other complement in the dative. Example (25) illustrates the antipassive version of
(21) above. The effect of the antipassive in this language is to express desiderative
modality.
(24) Muyutyu nhan-pa-ka wapa-ya
old.woman.NOM she-NOM-HERE seek-PRES
nhan-ku-ka takuku-ku
she-DAT-HERE baby-DAT
The old woman is looking for the baby girl.
(25) Muyutyu nhan-pa-ka watyama-li-ka nhan-ku-ka
old.woman she-NOM-HERE wash-AP-PAST she-DAT-HERE
takuku-ku thupu-lu
baby-DAT soap-ERG
The old woman was wanting to wash the baby girl with soap.
The motivation for using the dative for two-place verbs that are low in semantic
transitivity like wapa-, for the future tense and for the desiderative modality, is
clear. In each of these usages the transitivity is unfullled, i.e. there is no carrying
over of the activity to the patient. With the future and the antipassive this notion of
nonimpingement is morphologically marked; with verbs like wapa- the notion is
inherent or lexical.
13
A number of Indo-Aryan languages including HindiUrdu, Marathi and Punjabi
and some Iranian languages such as Pashto and Kurdish are described as hav-
ing an ergative construction only in the perfect. Typical Indo-Aryan languages
are described as having a direct/oblique case system where the direct case en-
codes S, A and P and the oblique is governed by postpositions. However, if P
is animate and specic, it is usually marked by a postposition. There is also
subjectverb agreement as in the following examples from Marathi (Rosen and
Wali 1989: 4f ).
(26) Ti keel
.
khaa-t-e
she banana eat-PRES-3SG.F
She eats a banana.
(27) Ti Ravi laa chal
.
-l
.
-a
she Ravi ACC torture-PRES-3SG.F
She tortures Ravi.
In the perfect, however, A is marked by a postposition. The verb agreement is with
P unless P is marked by the postposition for specic, animate nouns, with the verb
128
Table 5.3 Core case marking in Indo-Aryan
Nonperfect Perfect
S agreement agreement
A agreement ERG
P [+specic][+animate] ACC ACC
P [specic] or [animate] agreement
then remaining in its neutral form. In Marathi the ergative postposition is ni (Rosen
and Wali 1989: 5).
(28) Ti ni kel
.
i khaa-ll-it
she ERG banana.PL eat-PERF-3PL
She ate bananas.
(29) Ti ni Ravi laa chal
.
-l
.
-a
she ERG Ravi ACC torture-PERF-NEUT
She tortured Ravi.
The postposition laa, glossed as ACCusative, marks indirect as well as direct objects.
The typical Indo-Aryan patterns of core case marking and agreement are sum-
marised in Table 5.3.
5.2.5 Directinverse system
There is one other systemthat needs to be mentioned in this context and
that is the directinverse system of marking. In this system, which is characteristic
of the Algonquian languages, the marking on the verb indicates whether an activity
is in line with the person hierarchy or contrary to it. In the examples given earlier of
marking that is sensitive to the relative positions of Aand P on the person hierarchy,
the marking is identiable as accusative (as in Rembarnga, see (3)) or ergative (as
in Fore, see (9) and (10)). In the Algonquian languages, however, the marking can
simply be identied as direct (in line with the hierarchy) or inverse (contrary to it).
Compare the following example from Plains Cree.
14
(30) a. Ni-w apam- a-w
1SG-see-DIRECT-3SG
I see him.
b. Ni-w apam-ik
1SG-see-INVERSE.3SG
(ik <

ekw-w INVERSE.3SG)
He sees me.
129
In transitive clauses with two third-person participants, the direct and inverse mark-
ers distinguish whether a more topical participant (proximate) is A, which gives
a direct combination, or a less topical participant (obviative) is A, which gives an
inverse combination. A more topical participant will be chosen on the basis of
discourse principles and will tend to be the last-mentioned person or the discourse
topic. The less topical person is marked by the obviative sufx -wa.
(31) a. N ap ew atim-wa w apam- e-w
man dog-OBV see-DIRECT-3SG
The man saw the dog.
b. N ap ew-(w)a atim w apam-ik
man-OBV dog see-INVERSE.3SG
The man saw the dog.
(32) a. Atim n ap ew-(w)a w apam- e-w
dog man-OBV see-DIRECT-3SG
The dog saw the man.
b. Atim-wa n ap ew w apam-ik
dog-OBV man see-INVERSE.3SG
The dog saw the man.
As can be seen, there are two ways of expressing the same propositional content
according to which participant is chosen as topic.
5.2.6 Interpretations of core marking
In many languages where there are local cases such as locative, allative
and ablative, there is a straightforward relationship between case marking, case
and role or function. Typically there will be a marker peculiar to a case (though
the marker may have variant forms determined by the shape of the stem), and the
case will have a clear semantic function. With the grammatical cases, however, the
function is not always so clear, and the case marking, as we have seen, is often not in
a one-for-one correspondence with the cases. The rst point that we need to discuss
is the meaning or function of the core grammatical cases and then the signicance
of the distribution of the marking.
Obviously the core grammatical cases express the core grammatical relations, so
the rst question that needs to be answered is why these relations exist. Basically
the core relations of subject and direct object in an accusative (SA/P) system and
absolutive and ergative in an ergative (SP/A) system are syntactic rather than se-
mantic, though, with the exception of subject, they could be said to have a semantic
basis (see below). Languages do not mark the roles of their complements directly.
Apart from anything else it would be extremely uneconomical, since the roles are
130
predictable from the meaning of the predicate, though with two-place predicates
there needs to be some way of indicating which complement encodes which argu-
ment (compare the discussion of (27) in chapter 3). So irrespective of whether the
core grammar groups S with A or with P, we can expect purely syntactic relations
simply because of the redundancy that exists between predicates and the roles of
their arguments.
Subject
The most common system for organising the core is the SA/P or
nominativeaccusative system. This system has a discourse-pragmatic basis rather
than a semantic one. It is clear that the subject is not a semantic entity since it
embraces diametrically opposed semantic roles, agent in examples like It bit me
and patient in examples like I got bitten. On the other hand one can show that the
subject is characteristically associated with given as opposed to new material,
and that where there is a referential subject a clause can normally be described in
terms of being a predication about the subject even where the subject is new.
Let us take the association with given information rst. Cross-language studies
of the distribution of given and new information reveal the following givenness
hierarchy:
15
(33) A > S > P > peripheral
This is to be interpreted as indicating A is more often given than S which is more
often given than P which is more often given than the peripheral relations. The basis
for the association of A and to a lesser extent S with given information is the fact
that discourse, particularly narrative, typically involves the successive actions of
humans and animals. In fact, if one splits S into animate and inanimate or actor
(agent, perceived experiencer) and patient, then the hierarchy becomes:
(34) A > S
actor
> S
patient
> P > peripheral
which points even more clearly to the fact that discourse involves higher animates
being involved in a succession of clauses.
The subject treats the rst two positions on hierarchy (33) alike and some of
the grammatical characteristics of the subject can be related to its association with
givenness. The most common form of agreement is subject agreement and it is
generally accepted that the bound pronouns that gure in agreement derive from
unstressed pronouns that have become attached to the verb. If A and S are most fre-
quently represented by given information, one would expect them to be represented
very frequently by pronouns, in particular, unstressed pronouns.
131
The subject is typically used to encode what the clause is about. This is clearest
with symmetrical predicates such as resemble, collide with or be opposite in English.
If a cathedral and a kiosk stand on opposite sides of the city square, we can describe
this as in (35a) or (b).
(35) a. The kiosk is opposite the cathedral
b. The cathedral is opposite the kiosk
(35a) is a statement about the kiosk, about its position in relation to the cathedral.
Conversely (35b) is a statement about the cathedral, about its position in relation to
the kiosk. Of course (35b) is an odd sentence from the discourse point of view, but
that is because we do not normally take an insignicant kiosk as a reference point
for saying something about the position of a large building, which highlights the
fact that (35b) is a statement about the cathedral.
The fact that a sentence typically embodies a predication about its subject is
also evident in co-ordination. It is the subject that can be deleted from a noninitial
sentence and the identity of the deleted subject is established by reference to the
nearest preceding subject in the sequence of co-ordinated clauses. This is true even
where the subject of the preceding clause represents new information as in the rst
clause of (36b).
(36) a. What happened to the worms [ for shing]?
b. A kookaburra took them and ew off with them
In some languages there are constraints on what can appear as subject; in par-
ticular, there are bans on indenite subjects. In English there are circumstances
where an indenite subject is strongly disfavoured as in the following stock
example:
(37) Waiter! Theres a y in my soup
Grammatically one could say A y is in my soup, but this would be to say something
signicant about a y, and this is not a likely sentiment in the situation referred to.
English grammar demands a subject, so the nonreferential pronoun there is used as
a subject, which allows the situation to be described without making a statement
about a y or about my soup. For the latter perspective My soups got a y in it would
be appropriate.
In sum the subject is a purely syntactic relation characteristically associated with
presenting given and topical information.
132
Objects
The accusative is the case that encodes the direct object of a verb. It
will encode both the objects where there is a double-object construction and it may
mark the object of some or all adpositions. The direct object has both semantic and
discourse-pragmatic properties:
(a) Its core function is to express the role of patient in a two-place con-
struction.
(b) Where a nonpatient is expressedas direct object the activityis presented
from the point of view of its effect on the direct object.
(c) The direct object holds a position on the givenness hierarchy interme-
diate between the subject and the peripheral relations.
The direct object encodes the patient by denition. To identify a direct object
we need rst to identify the patient of a two-place verb, providing it is treated
differently from the sole argument of a one-place verb; in other words we want
to identify a P that is distinct from S. Not all two-place verbs treat the patient
alike. If we want to identify the relation that matches up with the direct ob-
ject of Ancient Greek and Latin, the reference point for our labelling, we need
to specify two-place predicates of maximum semantic transitivity, predicates like
SMASH, KILL or DESTROY, which take an affected patient, rather than semantically
weaker two-place predicates like HELP or TRUST, which take a patient nearer to
the neutral or unaffected end of the patient spectrum. This will ensure that we
identify a relation comparable to the direct object of the classical languages. If
we deal in verbs of lower semantic transitivity we frequently nd that some of
these take complements that are not comparable to the direct object of the clas-
sical languages. In Latin and Ancient Greek some verbs with meanings like HELP
or TRUST took an indirect object in the dative case, and this treatment of two-
place predicates of lower transitivity has parallels in other languages (see
section 5.3).
The signicance of the direct object can be appreciated if we contrast it with the
noncore or peripheral relations. Consider, for instance, the difference between (38a)
and (b):
(38) a. The old man walked in the streets of the village
b. The old man walked the streets of the village
(38a) tells us where the old man did some walking, but (38b) tells us that the man
traversed a lot of streets, perhaps most of the streets. Where a role other than patient
is expressed as P there is often an added sense of affectedness, or a holistic interpre-
tation becomes likely (compare the discussion in sections 3.4.2, 3.4.4 and 3.4.5).
133
The difference between expression as direct object and expression as a peripheral
relation is even more apparent in three-place constructions. Consider, for instance,
the following pair:
(39) a. The vandals stripped the branches off the tree
agent patient source
b. The vandals stripped the tree of its branches
agent source patient
The roles appropriate for (39a) are noncontroversial. The vandals are clearly the
agent, the branches patient and the tree source. Of the two sentences (39a) would
appear to be unmarked, since it exhibits a normal association of role and relation:
patient with direct object and source with a noncore or peripheral relation. The
ascription of roles in (39b) is more problematic. Under a common interpretation of
constructions like this the tree in (39b) would still be the source and the branches
the patient. But note the difference in meaning. (39a) presents the situation fromthe
point of viewof the effect of the activity on the branches, whereas (39b) emphasises
the fate of the tree. In fact, the phrase of its branches can be omitted from (39b).
This makes it problematic for those who claim that every clause has a patient. The
problem could be avoided by claiming that the tree in (39b) has been reinterpreted
as a patient, but that creates the further problem of nding a different role for the
branches. As with the previous example it seems that the role normally associated
with the direct object is patient and that the encoding of another role as direct object
may involve some added sense of affectedness.
In the double-object construction, where one could say that a recipient or bene-
ciary is advanced to direct object, the patient object remains as a secondary object;
it is not demoted as the branches is in (39b). It is clear in these constructions that
the recipient or beneciary object is not reinterpreted as a patient, but the effect of
using the double-object construction is to present an activity from the point of view
of its effect on the recipient or beneciary.
(40) a. I gave her a ver and she was pleased as punch
b. I made her a new frock for the party and she was ever so pleased
These sentences are appropriate ways of reporting the effect on the recipient and
beneciary respectively.
It is signicant that in most languages the passive allows as its subject only the
direct object of the corresponding active. The main function of the passive is to
allow a nonsubject of the active to be promoted to subject where it can serve as
topic (the subject almost always has the function of topic in the passive). Only noun
134
phrases expressing an entity that is seen to be of central concern, like the branches
in (39a), the tree in (39b), or her in (40a), are suitable choices for promotion.
Part of the motivation for the double-object construction as an alternative to the
single object plus prepositional phrase is to allowa redistribution of roles in terms of
given and new. The recipient or beneciary object tends to be given and the patient
object new, whereas in clauses with a direct object and a prepositional phrase there
is a tendency for the direct object to be given and the prepositional phrase new.
Outside the roles of recipient and beneciary the association of new information
and prepositional phrases remains strong. Entities in roles such as instrument and
location tend not to recur much in discourse. They come up an odd time as new
information expressed as a peripheral relation and languages do not provide for the
possibility of advancement to direct object nearly so often as they do with recipient
and beneciary. The direct object, on the other hand, is more frequently given and
this is reected in the fact that languages often have a series of bound pronouns for
direct object.
Absolutive and ergative
At rst blush it might seem difcult to nd a semantic basis for an
SP/A system, since the absolutive relation (SP) will include agents as in I swim
and patients as in It grabbed me. However, a number of linguists including John
Anderson (1977, 1979), Kibrik (1979) and Wierzbicka (1981) have in fact claimed
that the SP/A distinction has a semantic basis. A is seen as an entity acting on or
towards an entity external to itself, while the absolutive complements this. It encodes
a range of patients from strongly affected (The meat was eaten), through weakly
affected (The spear was taken) to unaffected or neutral (The spear lay there) and,
for a few verbs, agents who do not impinge on another entity (John swam). Since
the absolutive also encodes various nonpatients that are advanced to the absolutive
through various verbal derivations (see, for instance, (3) in chapter 3 where an
instrumental is advanced to P), it would be best to recognise that the absolutive is a
purely syntactic grammatical relation with the expression of patient being its core
function. This interpretation is similar to the interpretation of direct object offered
above, the difference being that the direct object encodes the patient when there is
an A in the clause, whereas the absolutive encodes the patient irrespective of the
presence of A.
It is interesting to note that in languages where there is number marking in
the verb, this marking almost always refers to the absolutive irrespective of the
organisation of the core grammar. The marking referred to here is pure number
marking, which is often marked suppletively, not the more familiar person/number
marking. In Dyirbal, for instance, the sufx -dyay added to the verb nyinya to sit
refers to the plurality of the sitters, but when added to a transitive verb like gundal
135
to put it refers to the plurality of the items put.
16
Dyirbal is a morphologically and
syntactically ergative language, but the point is that this pattern holds irrespective
of any other manifestation of ergativity. This would appear to reect the fact that the
absolutive is the relation most intimately connected with the verb and conversely
that the ergative is not so closely connected with the meaning of the verb. It should
also be noted that in many ergative languages the ergative relation is very much
an agent. There is often not much conation of other roles with the agent, often
only a conation of the perceiver of a few verbs like SEE/LOOK AT and HEAR/LISTEN
TO, and in no language is there is any possibilty of other roles being advanced to A.
To claim that an SP/A system has a semantic basis does not entail claiming
that the distinction is devoid of discourse-pragmatic signicance. In fact it would
appear that all the points made above about the difference between encoding a
participant as a direct object rather than in a peripheral relation apply to encoding
a participant in the absolutive as opposed to a peripheral relation. The signicant
difference seems to be P versus peripheral. Moreover, the distribution of marking
in an ergative system has discourse-pragmatic signicance. If we accept that A >
S > P > peripheral as in (33) is a universal hierarchy of givenness, then ergative
systems present a pattern of marking that is in conict with this hierarchy insofar
as they have positive marking for the function that is at the top of the hierarchy and
unmarked forms for the middle of the hierarchy.
The marking
Now we come to the question of the distribution of core case marking.
There are two aspects of this that are remarkable. The rst is that languages with
simple across-the-board ergative or accusative systems are a distinct minority. As
noted earlier in this chapter, a majority of languages have partial ergative marking,
partial accusative marking, or both partial ergative marking and partial accusative
marking. The second remarkable aspect of core case marking is that partial ergative
and partial accusative marking dene a hierarchy of nominal categories (pronoun,
kinship term, etc.).
Linguistically naive common sense would predict the occurrence of accusative
and ergative systems. If we have three entities, S, A and P, with two of them in
syntagmatic contrast, namely A and P, then the simplest marking system is one
that provides positive marking just for P (accusative system) or positive marking
just for A (ergative system). An S/AP system would obviously be dysfunctional
and an S/A/P system though functional is uneconomic. This discriminatory view
of the function of case marking fares fairly well. Accusative systems are common;
ergative systems are reasonably common; three-way S/A/P marking is rare, and no
across-the-board S/AP system has been reported. What this logical approach fails
to predict is the high frequency of partial marking.
17
136
Silverstein 1976 demonstrated that the distribution of partial ergative marking
and partial accusative marking dened a hierarchy of lexical content. The hierarchy
is as follows:
(41) 1st person (speaker)
2nd person (addressee)
3rd-person pronoun
personal name, kin term
human
animate
inanimate
18
Partial accusative marking will always run from the top of the hierarchy covering a
continuous segment. Partial ergative marking on the other hand will always run from
the bottom of the hierarchy. A language may have accusative marking only on rst-
and second-person pronouns (Dyirbal), only on personal pronouns (English), only
on pronouns, kin terms and personal names (Nhanda), and so on. All the positions on
the hierarchy can be dened on the basis of the distribution of accusative marking,
i.e. seven different cut-off points are attested. Ergative case marking tends to be
found on all nouns or all nouns and third-person pronouns.
19
In languages with
a mixture of ergative and accusative marking, the two may overlap in the middle
of the hierarchy giving a three-way nominativeergativeaccusative opposition for
some category such as third-person pronouns.
If this hierarchy is to have any explanatory value with respect to case mark-
ing, it needs to be established on an independent basis. It has been referred to as
an agency hierarchy, but this is only applicable to part of the hierarchy, namely
human > animate > inanimate where we can safely assume that humans act on
animates (in effect animals) and inanimates, more than the other way round, and
similarly animals act on inanimates more than the other way round. It has also been
characterised as an animacy hierarchy, but this too applies only to the last three
categories in the hierarchy. Mallinson and Blake suggest the hierarchy reects a
relative centre of interest, that events tend to be seen from the point of view of the
speech-act participants (1981: 86). With respect to the lower end of the hierarchy,
what is involved is the interest of categories for humans in general. Humans are
most interested in other humans, and more interested in animals than inanimates.
With respect to the top of the hierarchy, the degree of interest is related to the
speech act.
Silverstein 1981 claims that the hierarchy represents the extent to which enti-
ties are given or presupposed in language use. The act of speaking presupposes a
speaker, and communication presupposes an addressee. Anaphoric pronouns are
137
presupposed on the basis of reference earlier in the discourse and demonstrative
pronouns presuppose the presence of entities in the ambit of the speech event. Per-
sonal names (Maria, John) and kin terms (mothers brother, grandpa) are known to
the speech act participants. All the categories down to this point are denite. The
last three categories are hardly presupposed on the basis of inherent features, but
where accusative marking extends to these categories it is usually only where they
are denite or at least specic.
The hierarchy manifests itself in a number of areas besides case marking. These
include the following:
(a) rules of order
(b) number marking
(c) agreement
(d) advancement
(a) Order
In some languages the order of bound pronominal elements is sensitive to the
hierarchy. In Gunwin
y
gu (Northern Australian) a rst- or second-person bound
pronoun always precedes a third irrespective of whether it is subject or object
(Carroll 1976).
(42) a. Nga-be-n-bun
1SG-3PL-OBJ-hit.FUT
Ill hit them.
b. Nga-n-di-bun
1SG-OBJ-3PL-hit.FUT
Theyll hit me.
In French a direct object clitic generally precedes an indirect object clitic: Marie
les lui donne Mary gives them to him. But if the indirect object is rst or second
person, the indirect object clitic precedes the direct object clitic: Marie me les donne
Mary gives me them.
Kalkatungu (Pama-Nyungan) is an ergative language with pragmatically deter-
mined word order. In transitive clauses the orders APV and AVP predominate, but
where P is rst or second person and A inanimate it is unusual not to have P rst
as in (43).
(43) Ngai unthayi-nha kuu-ngku
me soak-PAST water-ERG
I got soaked by the rain.
138
(b) Number
In many languages there are obligatory number distinctions with pronouns, but not
with nouns. In Chinese, for instance, pronouns are obligatorily marked for plural
(t a s/he, t a-men they), but there is no bound form for plural that can be used
with nouns, though -men can be used with human nouns providing the stem is
polysyllabic: laosh -men teachers. In some languages number distinctions are
made only with pronouns and kin terms (e.g. Kalkatungu, see (4) in chapter 3). In
some noun-classifying languages of northern Australia a human plural class occurs,
but there is no corresponding nonhuman plural class. In Wappo (Penutian) plural
marking is obligatory for human nouns but only optional for nonhuman nouns;
adjectives show number concord only with human nouns (Li and Thompson n.d.).
In Turkish the third-person-plural pronominal agreement is dropped if the subject
is inanimate.
(c) Agreement
Agreement is normally on the basis of grammatical relations: subject only
(Indo-European, Uralic, Altaic, Dravidian), subject and object (Chukchi, numer-
ous Australian, Turkana), absolutive (Avar) or absolutive and ergative (Mayan).
However, in some languages agreement is with the argument higher in the hier-
archy irrespective of grammatical relations. In Dargwa (Northeast Caucasian) the
verb agrees with rst or second person not third irrespective of whether rst or sec-
ond is A or P. Where both A and P are speech-act participants, agreement is with P.
In Sora (Munda) a rst- or second-person object is marked on the verb but not third.
In Pintupi and Walmatjari (Pama-Nyungan) the subject and object are represented
by bound pronouns enclitic to the rst constituent of the clause. Noncore relations
may be represented too provided that they are animate.
20
In most languages with clitic pronouns for P only specic objects can be repre-
sented by the clitic. The Swahili example given as (14) in chapter 1 implies that Ali
loves a specic beautiful woman. Without the object clitic it would mean that Ali
loves any beautiful woman.
(d) Advancement
Verb derivations such as passive are normally based on grammatical relations, but
in some languages the pronoun-animacy hierarchy is involved. In the Wakashan
languages the passive is obligatory where the agent is third person and the patient
rst person. In Tiwa (Aztec-Tanoan) only a third person can appear as the agent
of a passive and the passive is obligatory where a third person is agent and rst or
second person patient.
The relevance of the hierarchy is also apparent with the double-object construc-
tion. In English, for instance, it is practically conned to constructions with an
139
animate recipient or beneciary: He sent the refugees food but not

He sent the
station food; He made his guest a weak cocktail but not

He made the lemon tree a
weak emulsion of white oil. In Relational Grammar the double-object construction
is considered to be derived and to reect the advancement of an indirect object or
a beneciary to direct object (see section 3.4.3). Not everyone would agree with
this analysis, but the construction certainly involves two object-like noun phrases
as opposed to a direct object and a prepositional phrase and it is certainly true that
the extra object is almost always animate. The double-object construction and its
counterpart in syntactically ergative languages (see section 3.2) is not uncommon
across languages and it is true that it is generally conned to animate advancees.
In languages where there are bound pronouns for P but not for indirect object, it is
almost always true that in the double-object construction the P series represents the
recipient or beneciary, a fact that supports an advancement analysis.
Case marking apart, these manifestations of the hierarchy are consistent with the
view that events are seen from the point of view of the speaker and hearer. In most
languages the subject, which typically encodes the topic, precedes the object either
as the result of a discourse-pragmatic strategy or a grammatical rule. Rules placing
rst and second person ahead of third can be interpreted as reecting a topic-
rst principle. Rules of advancement favouring the top end of the hierarchy can
be given a similar interpretation, as can preferential treatment in agreement since
representation by a bound pronoun reects givenness. The tendency to have more
number marking at the top end of the hierarchy is surely an example of speakers
making more distinctions in categories of greater interest to them.
If we have then a hierarchy based on the point of view of the speech-act partic-
ipants, how is it that partial accusative marking is regularly distributed from the
top of this hierarchy and partial ergative marking from the bottom? Silverstein and
Comrie see language as taking as most natural an event in which the agent is the
speaker or addressee and the patient a nonspecic inanimate.
Silverstein comments on the naturalness of the distribution as follows (1981:
243):
It would appear that . . . language structure is grammaticalising (with charac-
teristic asymmetry) the perspective from which a state of affairs is predicated
of referents, the most natural being that which grows out of the conguration
of the ongoing speech event, the informational givens of which, the maximally
presupposable entities, are of course the same orderings as above [see (41)].
Hence, speaker Agent-of and Subject-of in direct (as opposed to inverse)
predicate schemata.
In similar vein Comrie talks of the most natural kind of transitive construction
being one in which A is high on the pronoun-animacy hierarchy and P low and
140
nonspecic. He sees deviations from this pattern as leading to a more marked
construction (1989: 128). On this interpretation accusative marking at the top of
the hierarchy reects the fact that it is less than maximally natural for a category
such as a pronoun or a personal name to be the object than for lower categories
such as nonspecic animates or inanimates. Conversely the conning of ergative
marking to nouns reects the fact that it is less natural for a noun to be an agent of
a transitive verb than for a pronoun. This might seem strange outside the context
of discourse; after all, sentences of the pattern The man hit the dog are common
as linguistics examples. But in discourse sentences with nouns in A function are
not so common as one might think (see the givenness hierarchy in (33)). The A
function will be lled by a noun only where A is new, but in any coherent text most
of the activities will be carried out by an A (or S) that is given on the basis of prior
mention or because it is a speech-act participant. In both these circumstances Awill
be represented by an unstressed pronoun, by a bound pronoun if one is available.
Where ergative-marked noun phrases occur, they are often encoding nonspecic
agents or the forces of nature (as in (43) above).
In most languages with partial ergative or accusative case marking, A and P are
marked independently, i.e. simply according to the category of nominal instantiating
them, e.g. A is marked ergative if it is a noun and P is marked accusative if it is
a pronoun. In a few languages, as we have seen, the marking is determined by the
relative positions of A and P on the hierarchy and these are the most convincing
examples of the sensitivity of marking to hierarchical considerations. The examples
given in previous sections include the following types:
(a) accusative marking used only where A is lower than P. See examples
(3a) and (b) involving bound pronouns in Rembarnga.
(b) ergative marking used only where A is lower than P. See examples (9)
and (10) involving noun phrases in Fore.
(c) marking for direct combinations (A higher than P) or inverse
combinations (A lower than P) that cannot be linked either to A or
P. See examples (30) to (32) from Cree.
The notion that partial ergative and partial accusative marking appear where
there is a deviation from a kind of ideal transitive predication has the advantage of
supplying a single explanation for both ergative and accusative marking. There are,
however, some other attempts at explanation, one relevant primarily to accusative
marking and one relevant primarily to ergative marking.
Hopper and Thompson make cross-language generalisations about the way
greater semantic transitivity is reected in formal transitivity. Full semantic tran-
sitivity involves an agent affecting a patient. There is less than full transitivity if
the predicate is one of perception (SEE) or emotion (LOVE) or if the tense is future,
141
the polarity negative, the aspect imperfective and so on. Languages usually have
two ways of treating two-place predicates. There will normally be a large class of
transitive verbs and another smaller class of intransitive verbs and adjectives that
take their complement in a noncore relation. The formally transitive verbs will en-
code the predicates of greater semantic transitivity and the intransitive verbs and
adjectives will naturally correlate with less than full semantic transitivity. Hopper
and Thompson consider that there is greater semantic transitivity where a patient
is specic rather than nonspecic, denite rather than indenite, human or animate
rather than inanimate. They see positive marking for P, whether it be signalled by
cross-referencing, position or case marking, to be indicative of greater semantic
transitivity.
Aviewparticularly relevant to ergative case marking is to be found in Wierzbicka
(1980: 129f, 1981: 66ff). She sees ergative constructions as ones in which the action
is seen from the point of view of the patient rather than of the agent; the patient
is presented as the central, unmarked entity and the agent as a peripheral, marked
entity. First and second persons especially and pronouns in general are most likely
to drop their peripheral marking, since they are given and topic-worthy. One could
note that in English and various other languages passives with pronominal agent
phrases are generally disfavoured. The ergative construction, which is transitive, and
the passive, which is intransitive and in marked opposition with an active transitive
construction, cannot be fully identied. Nevertheless there is some analogy between
the marked A of the ergative and the marked agent of the passive.
The approaches to ergative marking of Comrie and Silverstein on the one hand
and Wierzbicka on the other are perhaps not too far apart. In the former pronouns
do not bear ergative marking because they are the most natural exponents of A and
languages tend to leave what is most natural least marked. In the Wierzbicka view
pronouns, especially rst- and second-person pronouns, are natural topics and it is
therefore appropriate that they be encoded as central, unmarked entities.
21
One last point. Where there is an SA/P system and an SP/A system in the one
clause, then it is always true that the SA/P system is found with bound pronouns,
which naturally will be carrying given information rather than new. It should also
be noted that interclausal rules of syntax such as the marking of switch-reference
usually operate on an SA basis even where the case-marking system is nominative-
ergative. These rules involve given information in at least one clause.
22
5.3 Dative
In Ancient Greek the case used to express the indirect object with
verbs like didonai to give was called the pt osis dotik e or giving case. The la-
bel dative is the Latin translation applied to the Latin case with the corresponding
function: Cassius Br ut o librum dat (Cassius.NOM Brutus.DAT book.ACC give.3SG)
142
5.3 Dative
Cassius gives a book to Brutus. However, in both Greek and Latin the dative also
marked the nonsubject complement of certain intransitive verbs. These comple-
ments were taken to be indirect objects like the dative complements of three-place
verbs. Here are some examples of intransitive verbs that take an indirect object in
the dative:
(44) Greek Latin
bo ethein auxiliar help
peithesthai par ere obey
pisteuein dere trust
orgizesthai r asc be angry with
Cases are not isomorphic across languages, i.e. we do not nd that languages
generally have the same number of cases with each case in one language matching
the extension of a corresponding case in other languages. The dative in Greek, for
instance, covers a greater range than the dative in Latin. Greek does not have an
ablative as Latin does and the Greek dative covers the instrumental and locative
functions of the Latin ablative.
23
This lack of isomorphy can lead to difculties in
applying case labels. What does one do in describing language Xif one case encodes
the recipient of giving and another case the complement of some intransitive verbs?
This problem confronted me in describing Kalkatungu (Blake 1979a). One case
marked by -ku encoded the complement of a handful of intransitive verbs as well as
the roles of purpose and beneciary. Another case marked by -kunha encoded the
recipient of the verb to give as well as the role of destination. I chose to call the -ku
case dative, since it marked the complement (indirect object) of verbs analogous
to verbs in Ancient Greek and Latin like those illustrated in (44), and the -kunha
case allative (see Table 2.14). This strategy for identifying a dative gives good
cross-language comparability. The dative under this interpretation emerges as the
main noncore case used to mark complements. Its range of functions will typically
include those listed as (a), (b) and (c) in (45) and quite frequently functions (d) to (h).
(45) a. indirect object of some two-place verbs low on the transitivity
scale (e.g. verbs such as HELP, SEEK or LIKE)
b. indirect object of a few three-place verbs such as GIVE
and SHOW
c. the roles of purpose (She went for sh) and beneciary
(She went for (on behalf of ) her mother). In a few languages
there is a separate purposive case (e.g. Irula). Basque has a purpo-
sive case and a benefactive case (Saltarelli et al. 1988: 15666)
d. possessor (frequently expressed by the genitive)
143
e. destination (sometimes expressed by a separate allative
case as in many Australian languages)
f. the indirect object of a detransitivised construction as in the
antipassive of various languages (see (25) above from Pitta-Pitta)
g. the direct object in certain aspects or tenses (see (23) above from
Pitta-Pitta and (53b) below from Georgian)
h. the indirect subject of certain verbs or of all verbs in certain
aspects. This usage is somewhat different from the rest and is
described at the end of this section. See (51), (52) and (53).
The accusative is a syntactic case which can encode a variety of semantic roles,
but one could take the central and dening function to be that of encoding the
affected patient of activity verbs. The dative is likewise a syntactic case that can
encode a variety of roles, but I would suggest that its central function is to encode
entities that are the target of an activity or emotion. Traditional denitions refer
to the entity indirectly affected as opposed to the entity directly affected, which
is encoded by the direct object (at least in the active).
24
The accusative and the
dative may be in syntagmatic contrast or in paradigmatic opposition. With verbs
like Latin d are to give, monstr are to show and mand are to entrust, the two
cases are in syntagmatic contrast with the accusative encoding the entity that is
directly affected in the sense that it is moved or transferred to new ownership
and the dative encoding the sentient destination, the one to whom the transfer is
directed.
(46) Br utus rem mihi mand avit
Brutus.NOM matter.ACC me.DAT entrust.PERF.3SG
Brutus entrusted the matter to me.
With a two-place verb like auxili ar to help, par ere to obey, dere to trust
or r asc to be angry with the dative complement is in paradigmatic opposition
with the accusative with verbs like nec are to kill, tract are to pull and mov ere
to move. The accusative encodes entities that are directly affected whereas the
dative encodes entities that are not directly affected. It should be added that one
cannot predict the case from the role, since while it is true that entities directly
affected by activities are always encoded in the accusative (at least in the active
voice), entities that are not affected are encoded as accusatives with some verbs and
datives with others. Am are to love, for instance, is a transitive verb which takes
an accusative direct object encoding the neutral (unaffected) patient, so is vid ere
to see. This generalisation about accusativedative alternation carries over to a
great number of other languages, although it needs to be reinterpreted somewhat for
144
5.3 Dative
languages that lack a dative case. In English, for instance, some predicates low on
the transitivity scale are instantiated as verbs that take prepositional complements
(look for, look like, etc.) and others are instantiated by adjectives that take prepo-
sitional complements with of as the unmarked or default choice (be jealous of, be
supportive of ).
In many languages some verb roots appear both as transitive verbs with a di-
rect object and as intransitive verbs with an indirect object in the dative. A few
verbs in Latin are like this; temper are to temper or to restrain is one; moderor
which has much the same meaning is another. When these verbs refer to restrain-
ing or controlling an entity external to the agent, they are used transitively (47a),
but where they refer to exercising self-restraint they take a dative complement
(47b).
(47) a. Ego moderor equum meum
I.NOM moderate.1SG horse.ACC my.ACC
I control my horse.
b. Ego moderor . or ati on meae
I.NOM moderate.1SG speech.DAT my.DAT
I moderate my speech.
With self-restraint the restraint is directed towards oneself and the entity specied in
the dative is only indirectly involved. In some Australian languages a particular verb
root may cover look for and nd. When such a verb takes a dative complement, it
means look for, but when it is used transitively it means nd. Here the difference
seems to be between activity directed towards a target and successful achieving of
the target.
In many languages alternations between a transitive construction and a detran-
sitivised antipassive-type construction are not conned to particular roots, but
express some general difference of aspect or modality. Many of these detransi-
tivised constructions involve the dative case. Examples from two Australian lan-
guages, Kalkatungu and Pitta-Pitta, have been given earlier. In Pitta-Pitta the dative
is used as follows:
(a) the indirect object of a handful of verbs such as wapa- to look for,
yatha to like and tiwa to be jealous of (24)
(b) the indirect object in the antipassive construction which expresses
desiderative modality (25)
(c) the direct object in the future tense. A tentative example from Pitta-
Pitta was given as (23) above. A stronger example from Georgian is
given as (53b) below.
145
As noted in the discussion of the Pitta-Pitta examples, the use of the dative across
this range makes sense. It marks entities that are unaffected either because of the
lexical meaning of the predicate (as with tiwa be jealous of ) or because of a
modality (desiderative) or a tense (future) that logically overrides any sense of
direct impingement inherent in a particular predicate. Of course it should be added
that this logic is not normal language practice. More often than not case marking
is simply insensitive to tense, aspect, modality and the like.
In the examples discussed up to this point the clause containing the dative has
also contained a subject expressing an agent or an experiencer In these examples
the alignment of roles and relations has been normal. However, it is also possible
to nd examples where the dative encodes the roles normally assigned to subject.
In many languages the dative is used to express the experiencer with a handful of
verbs often including the translational equivalent of English please. This can be
illustrated from Latin where plac ere takes an experiencer in the dative. It may be
used with a referential subject (48) or impersonally (49).
(48) N on placet Ant oni o meus c onsul atus
not please.3SG Antonius.DAT my.NOM consulship.NOM
My consulship does not please Antonius.
(49) Vener placet mittere . . .
Venus.DAT please.3SG send.INF
It pleases Venus to send . . . .
Latin also affords examples of a dative that is not limited lexically, but which
can apply to any verb. This is the so-called dative of agent. It is used with the
gerundive, a passive-like participle with a modal force of obligation. An example
was given in chapter 3 in connection with identifying the role of agent across
different cases (see (32)). It is repeated here as (50b) with its active counterpart
(50a).
(50) a. Ego hanc provinciam d efend o
I.NOM this.ACC province.ACC defend.1SG
I defend this province.
b. Haec mihi provincia est d efendenda
this.NOM me.DAT province.NOM is.3SG defend.GER.NOM
This province is to be defended by me.
In (50b) the direct object of the active verb has been re-expressed as the subject of
the gerundive. If we use the gerundive construction with a dative-taking verb like
146
5.3 Dative
moderor in (47), the dative complement cannot be re-expressed as subject, so an
impersonal construction must be used. In (51) the verb has a third-person-singular
neuter nonreferential subject expressed in the verb. Note that the original dative of
the base verb is retained so that putting the subject of the base verb into the dative
results in two datives with a syntagmatic contrast between an indirect object and
what is sometimes referred to as an indirect subject.
(51) Moderandum est mihi or ati on meae
moderate.GER.3SG.NEUT is.3SG me.DAT speech.DAT my.DAT
I must moderate my speech.
The constructions illustrated in (48), (49) and (50b) are all examples of inver-
sion constructions. The label inversion implies that the alignment of role and
relation is inverted, in particular it implies that the roles normally encoded as
subject are encoded in the dative (see also the discussion of inversion in
section 3.4.3). These indirect subjects, as they are sometimes called, are not uncom-
mon; they are an areal feature of the Indian subcontinent. The following example
is from Malayalam (Dravidian) (McAlpin 1976: 191):
(52) Avalkku avane itikkaam
her.DAT him.ACC can/may.hit
She can hit him.
In the Malayalam example the dative in conjunction with the potential marker
-aam on the verb signals physical ability or permission. In some Kartvelian lan-
guages such as Georgian the use of the dative to mark the agent indicates the
evidential mode: Evidently she did so-and-so (Harris 1984). The following group
of examples illustrate the ergative construction used in the past tense (53a), the
accusative construction with dative-marked direct and indirect object used in the
present tense (53b), and the inversion construction used to indicate lack of direct
evidence (53c) (Harris 1984: 263).
(53) a. Rezo-m ga cuka samajuri ( sen)
Rezo-ERG you.gave.3S.it bracelet.NOM (you-DAT)
Rezo gave you a bracelet.
b. Rezo ga cukebs samajur-s ( sen)
25
Rezo.NOM you.give.3S.it bracelet.DAT (you.DAT)
Rezo gives you a bracelet.
147
c. Turme Rezo-s u cukebia samajuri sen-tvis
apparently Rezo-DAT gave.3S.it.EVID bracelet.NOM you-BEN
Apparently Rezo gave a bracelet to you.
Indirect subjects are rather like ergative-marked dependents in lacking the
supercial properties of subject such as case, obviously, and control of agreement.
However, they often exhibit syntactic properties of subjects. In Italian the dative
complement of piacere to be pleasing to can control the understood subject of a
nonnite complement just as an unequivocal subject of a noninversion verb can. In
(54) which contains a normal noninversion verb, the subject Giorgio is understood
as the subject of the innitive far(e). In (55) it is still Giorgio that is understood
as the subject of the innitive lasciar(e), even though Giorgio is not the subject of
the main verb, una compagna dufcio is (examples adapted from Perlmutter 1982:
31618).
26
(54)
Giorgio mi ha rimproverato tante volte da farmi paura
George me has reproved so.many times to make.me fear
George rebuked me so many times that he scared me.
(55)
A Giorgio ` e talmente piaciuta una compagna dufcio da lasciarci
To George is so pleased a companion-of-ofce to leave.us
[lit.] To George was so pleasing an ofce co-worker that he left us.
George was so taken with a woman at the ofce that he left us.
In Old and Middle English there were inversion verbs where the dative comple-
ment could be treated as a subject from the point of view of co-ordination. In the
following example from Chaucer, thoughte (< O.E. thyncan to seem), which is
an inversion verb, is conjoined with graunted, which is a normal verb. The pronoun
us is an oblique form, as it is in modern English. It represents a syncretism of ac-
cusative and dative. The subject of graunted, if it had been expressed, would have
been we (Canterbury Tales, the Prologue A 7857).
(56) Us thoughte it was noght worth to make it wys
And graunted him withouten more avys,
And bad him seye his verdit, as him leste.
The verb leste (< O.E. lystan to please) is also an inversion verb used imper-
sonally, literally as to-him it lusted, i.e. as it pleased him.
It is quite common for the dative to be used to express the possessor. In languages
where there is a dative but no separate genitive the dative is used adnominally.
148
5.4 Genitive
A dative of possession occurred in Latin as well as a genitive of possession and a
possessive adjective (examples from Woodcock 1959: 46).
(57) Illi duae fu ere liae
that.DAT two.NOM were.3PL daughters.NOM
He had two daughters.
(58) Illius duae fu ere liae
that.GEN two.NOM were.3PL daughters.NOM
The two daughters were his.
With the dative the existence of the daughters is asserted; with the genitive the
existence of the daughters is taken as given and the identity of the possessor asserted.
5.4 Genitive
A genitive case is widespread. On the basis of Latin one would nor-
mally ascribe the label genitive to the most common or unmarked adnominal case,
although one would not expect such a case to be exclusively adnominal. In Latin the
genitive marks the complement of a handful of verbs such as oblv sc to forget and
miser er to pity. In other languages, Old English for instance, the genitive marks
the complement of some scores of verbs and vies with the dative as the case used
to encode the complement of intransitive verbs. In such a situation the adnominal
function would be crucial for allotting the label genitive. The unmarked adnominal
case normally covers the sense of possessor, and the label possessive case is a com-
mon alternative. A genitive is found not only in Indo-European languages, but also
in Uralic, Caucasian, Altaic, Dravidian and Semitic languages. In some Australian
languages a single case covers the range of the Latin dative and genitive cases and
the label dative is preferred.
In a few languages there is a case common to A function and possessor func-
tion. These include Zoque (Mexican), a number of Caucasian languages and the
Eskimo languages, where the case concerned is called the relative case.
27
The fol-
lowing example is from Yupik Eskimo (Reed et al. 1977: 83). An explanation for
the complex gloss on the possessed noun is given below in connection with (63)
and (64).
(59) Angute-m nera-a neqa
man-REL eat-3SG.3SG sh
The man is eating the sh.
(60) Angute-m qimugta-i
man-REL dog-3PL.ABS.3SG.ERG
the mans dogs
149
In a number of languages the pronominal afxes that mark the person and number
of the possessor are the same as those used for subject function or the same as those
used for Afunction. An example of an ergativegenitive series of bound pronouns in
Konjo was given above (Table 5.1). An ergativegenitive series is also found in the
Mayan languages and in Abaza. To appreciate what is going on here it is interesting
to look at languages that use bound forms for both the possessor and possessed.
The rst example is from Abaza (Northwest Caucasian), a language which has
extensive cross-referencing of pronominal elements on both the verb and the noun.
The example is taken from a paper by W. Sidney Allen, who was perhaps the rst to
draw attention to parallels in the expression of A and possessor (Allen 1964: 340).
In Abaza there is an ergativeabsolutive distinction in third-person bound pronouns.
(61) D-l-pa-b
3SG.MASC.ABS-3SG.FEM.ERG-son-is
He is her son.
(62) D-l- s
e
-d
3SG.MASC.ABS-3SG.FEM.ERG-kill-PAST
She killed him.
In (61) the prex d- acts as a class marker to the noun, and in relation to -l-, which
marks the possessor, it represents the possessed.
In Yupik Eskimo, where the relative case marks both A function and possessor
function as illustrated in (59) and (60) above, our (dual) boats (dual) is expressed
thus:
(63) angya-g-puk
boat-3DU.ABS-1DU.ERG
our boats
The bound forms used are the same as for a transitive verb with a rst-person dual
A and third-person dual P (Reed et al. 1977: 140).
(64) Ce nirciiqa-g-puk
visit.FUT-3DU.ABS-1DU.ERG
We shall visit them.
It appears from these examples that the possessive schema is transitive with the
possessor encoded as Aand the possessed as P. This is obvious where there is a bound
formcorresponding to the possessed noun, but it may be that possession is conceived
of as transitive also in other languages where a subject or an ergative bound form
150
5.5 Partitive
is used to encode the possessor. As we observed in describing nominalisation, the
genitive case is often used to encode a complement of a nominalised verb, most
often the subject. See sections 4.3.1 and 4.5.1.
A large number of languages have possessive afxes for rst, second and third
person. In many of these the possessive afx cross-references the possessor. An
example from Turkish was given in section 1.1, namely adam-n ev-i man-GEN
house-3SG.POSS (the mans house). Here of course there is a genitive, but in many
languages the cross-referencing is the sole means of marking possession.
5.5 Partitive
In eastern Europe it is common to nd that partly affected patients are
put in the partitive case as in Hungarian (Moravcsik 1978: 261).
(65) Olvasta a k onyvet
read.3SG the book.ACC
He read the book.
(66) Olvasott a k onyvb ol
read.3SG the book.PARTITIVE
He read some of the book.
In Estonian and Finnish the partitive is used for the patient if it represents part
of a whole or an indenite quantity, or if the action is incomplete, or if the polarity
of the clause is negative. In Polish and Russian similar conditions determine the
choice of genitive as opposed to accusative (Moravcsik 1978: 265, 269; see also
Plank (ed.) 1984).
5.6 Local cases
The term local in this context refers to place. Local cases express
notions of location (at), destination (to), source (from) and path (through).
In languages with inectional case systems usually at least two different cases
are employed to make local distinctions, though not necessarily two exclusively
local cases. Indo-European languages originally used two local cases, locative and
ablative, with the accusative expressing destination and path. In Turkish there is
a locative and an ablative, with destination expressed via the dative. An allative
case expressing destination is found in a number of languages including Uralic and
Australian languages. A separate case for path is not so common but such a case is
found in a few Australian languages (Blake 1987: 40), and is part of the local case
system in some Northeast Caucasian languages including Avar (Table 5.4).
Where languages have large case systems it is always through the elaboration
of the local cases. Large case systems are characteristic of the Finno-Ugric branch
151
Table 5.4 Avar local cases
locative allative ablative perlative
(location) (destination) (source) (path)
on (top of) -da -d-e -da-ssa -da-ssa-n
at -q -q-e -q-a -q-a-n
under - - -e - -a - -a-n
in, among - - -e - -a - -a-n
in a hollow - - -e - -ssa - -ssa-n
object
of the Uralic family and of the Northeast Caucasian languages. They arise from a
combination of markers for relative orientation (above, beside, etc.) with markers
for location, destination, source and path. One dialect of Tabassaran (Northeast
Caucasian) is reported to have fty-three cases (Comrie 1981: 209, but see Comrie
and Polinsky 1998 and comments below). In Avar (also Northeast Caucasian) there
are twenty-seven cases, including twenty local cases deriving froma combination of
four cases and ve orientation markers. The combinations of orientation markers
and case markers are morphologically agglutinative and semantically transparent.
(See Table 5.4, which is based on Ebeling 1966.) In a language like this one could
simply recognise a layer of orientation markers plus a layer of case markers. If
one analyses a further outermost layer of forms as case markers rather than as clitic
particles, the system is even larger (Hjelmslev 1937: 225).
In Avar, as in other Northeast Caucasian languages, the locative (usually called
essive in sources on Caucasian languages) is literally unmarked and the perlative is
doubly marked, consisting of the ablative -a plus -n. This distribution of marking in
Northeast Caucasian languages ts Hjelmslevs notions of markedness and perhaps
played a part in shaping them (see Table 2.11).
The method of labelling local cases builds on the model of the Latin abl atvus
ablative which is made up of the preposition ab from plus a stem l ativus the
root of which is l at. This root supplies the perfect participle of fer o I bear. Other
labels are formed by varying the preposition or by combining prepositions with the
stem essvus from esse to be (cf. Mel cuk 1986:7275). Some local case labels
are presented in Table 5.5.
More often than not the transparency of large local systems is obscured by some
fusion of the orientation marker and what could be called the case marker proper,
by innovating forms or by semantic shifts. In Finnish, for instance, there are fteen
cases of which nine are local. As a rst approximation one can say that there are
local cases for location, destination and source, which can appear on their own or
appear with interior (inside) and exterior markers (outside). These are displayed
152
5.6 Local cases
Table 5.5 Labels for local cases
Latin root Meaning Case label Meaning
ad to allative to(wards) (the exterior of)
in into illative into
ab from ablative from (the exterior of)
e(x) out of elative from (the inside of)
super above superlative to the top of
trans through translative through
per through perlative through, along
essive at (cf. locative)
in in inessive in(side)
ad to adessive at
super above superessive above
sub under subessive below
in Table 5.6. The proto-forms for the cases are

-na locative,

-ta ablative and

-ne
allative. As can be seen, the formative for interior is -s- and for exterior -l-. The n
of the locative and allative assimilated to the preceding consonant to yield the array
in (67), which corresponds to Table 5.6.
(67) -na -ta -ne
-ssa (< -
sna) -sta -sse (< -
sne)
-lla (< -
lna) -lta -lle (<
lne)
A comparison of (67) with Table 5.6 reveals that -ksi and -(h)Vn /-sVVn have been
substituted for the forms expected on comparative-historical grounds. Semantically
Table 5.6 Finnish local cases
Location Source Destination
-na -tta -ksi
essive partitive translative
Interior -ssa -sta -(h)Vn, -sVVn
inessive elative illative
in from (inside) into
Exterior -lla -lta -lle
adessive ablative allative
at from (outside) to(wards)
153
the system is not as transparent as one would gauge from combining the labels
for the rows and columns in Table 5.6. The -tta form has a partitive function, the
translative -ksi is used to mark the category into which something has been changed
(Youll turn into a pumpkin, Translate it into English), and the allative -lle has dative
functions.
If one recognises that the orientation markers in any of the truly agglutinative
systems as inAvar forma separate systemfromthe case systemproper, thennaturally
the allegedly very large systems reduce to no more than a score or so. Comrie and
Polinsky (1998: 99) analyse northern dialects of Tabassaran as having seven cases
and eight orientation markers. Finnish is not entirely agglutinative nor semantically
transparent so it is appropriate to take the fteen case markers as constituting a
single system.
5.7 Other cases
A variety of languages including Basque, Zoque (Mexican), Ossete
(Indo-Iranian), Archi (Northeast Caucasian) and Finnish have a comitative case
expressing accompaniment. It is also found in Tamil, Telugu and a majority of
Dravidian languages, where it is known as sociative.
An instrumental case encodes the instrument with which an action is carried
out as in She wiped the screen with a cloth. It is found in the Indo-European family
(e.g. Sanskrit), Uralic (e.g. Komi-Permyak), Altaic (e.g. Nanai), Dravidian (e.g.
Tamil), Basque, Tonkawa (Uto-Aztecan), Tarascan, Australian and elsewhere. It is
sometimes used to encode the agent of the passive in an accusative language (e.g.
Russian). In a number of ergative languages the ergative case covers instrumen-
tal as well as A function (e.g. Avar, Tibetan and the majority of Pama-Nyungan
languages).
In the Uralic language a case called abessive (Latin ab-esse to be absent) or
privative (Latin prv are to deprive) is found. It means lacking, not having.
In Finnish, for instance, rahta-tta is (money-AB) moneyless. This case is also
found in Australia where it is matched by a having case called the concomitant
or proprietive. These two categories tend to show up in many Australian lan-
guages in lexicalised formations. In Kalkatungu, for instance, putu-yan (stomach-
PROP) means pregnant. In light of this and other evidence the having and
lacking sufxes have been taken to be derivational. On the other hand they
exhibit concord in languages with case concord which suggests they are inec-
tional and should be considered cases (Blake 1987: 87f; Dench and Evans 1988:
713).
The labels aversive (lit. turning from), evitative (avoiding) and causal have been
used for a case category common in Australian languages. It indicates what is to
be feared or avoided, as in Kalkatungu: Yanyi-ngkungu rumpi (ghost-AVERSIVE fear)
154
5.8 Inectional case hierarchy
S/he is afraid of ghosts or the cause of a state or activity, again from Kalkatungu:
palpir-tungu uli-nyin (poison-AVERSIVE die-PART) dying from poison.
The only other case not mentioned up to this point that occurs with any frequency
is the comparative (than), which occurs, for instance, in some Dravidian and some
Northeast Caucasian languages. There are certainly more than a dozen or so other
case labels to be found in the literature, quite apart from systematic local labels of
the type illustrated in Table 5.5. Not all of these are described in detail and some
may turn out on closer inspection to be variants of some of the labels presented
here. We shall not pursue the matter.
Morphological case systems range from two members to a dozen or
so. However, if one takes all the combinations of orientation markers and case
markers proper in some Finno-Ugric and Northeast Caucasian languages as cases,
then the gure runs to forty or so (see section 5.6). The question addressed here
is whether morphological (inectional) case systems grow and decay in a certain
order. In attempting to answer this question I shall consider rst case marking
on independent nominals and ignore bound pronouns and word order. However,
bound pronouns and word order are relevant to morphological case in that they
represent alternative ways of encoding grammatical relations that lie at the top end
of the relational hierarchy and therefore preclude the possibility of what otherwise
might be implicational relationships. One might consider that a peripheral case like
ablative is not likely to be found in an accusative language unless a core case like
accusative is also found, but this will not hold in a language where the object is
represented pronominally in the verb or only by position after the verb. Similarly
one might propose that a language will not have a comitative case (which is only
moderately common across languages) unless it has a genitive (which would appear
to be more common), but a possessor in many languages is represented by a bound
pronoun standing, where necessary, in cross-reference with a noun as in Warndarang
(Northern Australian) (Heath 1980).
Comparing cases across languages is problematic from the methodological point
of view. If we compare nonisomorphic case systems, then strictly no case in one
system will correspond with any case in the other system. As remarked previously,
the dative in Ancient Greek, for instance, does not correspond closely with the dative
of Latin. Greek lacked the ablative of Latin and the source function (from) of the
Latin ablative was expressed by the genitive, while the locative and instrumental
functions of the Latin ablative were expressed by the dative. The Greek dative is
a more comprehensive case than the Latin dative. So in comparing cases across
languages we need to consider the functions covered by a particular case and we
must not accept traditional labels at face value.
155
If we look at a sample of case systems, ignoring for the moment those languages
where some relations are marked exclusively by bound pronouns or word order, we
nd that indeed they do tend to be built up in a particular order, i.e. a hierarchy
emerges.
(68) nom acc/erg gen dat loc abl/inst others
This hierarchy is to be interpreted as follows. If a language has a case listed on the
hierarchy, it will usually have at least one case from each position to the left. Thus
if a language has a dative case it will have a genitive, an accusative or ergative or
both, and a nominative. In a small system of two, three, four or ve cases the lowest
ranked case will usually have a large range of functions, i.e. it will be a kind of
elsewhere case. The hierarchy as it is presented in (68) gives the impression that
a language could have a two-case system in which the non-nominative case was
accusative or ergative. While it is true that the second case is likely to cover P or
A function, it is also likely to cover such a wide range of functions that the label
oblique would be more appropriate than accusative or ergative.
A number of languages have a two-case system. Chemehuevi, a language of the
Numic branch of Uto-Aztecan, has a nominativeoblique system in which nomi-
native marks subject and oblique marks object and possessor. There are also post-
positions which attach directly to the stem and perhaps could be said to belong
to the case system, but unlike the oblique case marker they cannot gure in con-
cord (Press 1979). A nominativeoblique system plus postpositions is typical of the
Uto-Aztecan family (Langacker 1977).
Kabardian (Northwest Caucasian) has a two-case system in which the marked
case covers A and possessor function. Two-case systems are also found among the
Iranian languages. Yaghnobi has both an accusative construction and an ergative
construction, the latter used in tenses based on the past stem. Within the core
the marked case is used for a specic direct object in the accusative construc-
tion and for A in the ergative construction. Outside the core the marked case
is used for possessor, indirect object and object of a preposition (Comrie 1981:
16970).
28
In a three-case systemthere is usually a nominative, an accusative and a genitive
oblique. The Semitic languages have systems like this. Classical Arabic provides
an example, though there is accusativegenitive syncretism in the plural. The
Nubian languages (Nilo-Saharan) also have a nominativeaccusativegenitive sys-
tem. Modern Greek has a nominativeaccusativegenitive system and so has
Comanche (Uto-Aztecan) (Robinson and Armagost 1990). Greek also has a voca-
tive, but since the vocative is functionally so different it is not considered a case here
(see section 1.2.3).
156
As the hierarchy in (68) stands, it would appear that a three-case system could
contain nominative, accusative and ergative. Accusative and ergative cannot be
separated on the hierarchy, but as far as I know there are no languages that
have both a case covering A and a case covering P unless a fourth case is
present.
A number of languages have the following four-case system:
(69) nom acc gen dat/obl
The last case often has a variety of functions and is variously named. Languages
with this system include Ancient Greek (which also had a vocative), a number of
Germanic languages including German, Icelandic and Old English (where there was
also a vestigial instrumental), Yaqui and several Nilo-Saharan languages including
Fur, Nuer and languages of the Didinga-Murle group where the accusative rather
than the nominative is unmarked.
The next case to be distinguished from the elsewhere case is the dative. In Latin,
for instance, we have the following system of ve cases (plus a vocative) (see also
Table 1.2):
(70) nom acc gen dat abl/obl
The lowest case in Latin is called the ablative, but the Latin ablative represents a
syncretism of an earlier ablative, instrumental and locative (still vestigially distin-
guishable in Classical Latin). The label is somewhat arbitrary. Old and Middle High
German had a similar system with an instrumental where Latin had an ablative.
The OHG and MHG instrumental expressed locative and ablative functions as well
as instrumental.
The only other case that can be placed on the hierarchy with any condence
is the locative. Systems of six or more cases almost always have a locative. In a
number of Slavonic languages including Polish, Czech, Slovak and Serbo-Croatian
the following system is found:
(71) nom acc gen dat loc inst
The next cases for consideration are the ablative and the instrumental. It does not
seem that they can be distinguished hierarchically. Both are quite common, and
there are languages with an instrumental but no ablative (like the Slavonic languages
where source is normally expressed by a preposition governing the genitive) and
there are languages like the Altaic languages where there is an ablative but no
157
instrumental. The case system in Turkish is as follows. The instrumental function
is expressed by a postposition.
(72) nom acc gen dat loc abl
Some Altaic languages have both an instrumental and an ablative as does Classical
Armenian which reects the system of proto-Indo-European minus the vocative:
(73) nom acc gen dat loc abl inst
It is doubtful whether the hierarchy can be developed much further. Outside the
cases discussed up to this point the most common would appear to be the comitative
(accompanying), the purposive, the allative (to), the perlative (through) and
the comparative (than). Tamil, for instance, has the seven cases listed in (73) plus
the comitative (called sociative in Dravidian linguistics):
(74) nom acc gen dat loc abl inst com
Two other Dravidian languages, Toda (Sakthivel 1976: 43548) and Irula (Perialwar
1976: 495519), have a purposive as well (plus a vocative):
(75) nom acc gen dat loc abl inst com purp
Dravidian languages mostly have largish case systems. Kodaga (Balakrishnan 1976:
42134) and Kasaba (Chidambaranathan 1976: 46780) also have a comparative
(than) case.
The Pama-Nyunganlanguages of Australia mostlyhave case systems witheight to
ten members. There is normally a common case for Aand for instrumental function,
most often referred to as the ergative, and there is frequently an allative and often an
evitative (indicating what is to be avoided). Some Pama-Nyungan languages lack a
genitive, the possessor function being expressed via the dative (Blake 1977, 1987).
In illustrating the hierarchy I have so far avoided languages which encode gram-
matical relations solely by word order or by bound pronouns, since the use of these
alternative mechanisms can mean there are gaps in the morphological case hierar-
chy. However, the use of word order as the sole means of distinguishing subject and
object is practically conned to subjectverbobject (SVO) prepositional languages
like French, Cambodian and Thai. The only real competitor for morphological case
is the use of bound pronouns. Bound pronouns may be used for subject, object,
indirect object and possessor, and in some ergative languages for the absolutive
158
and ergative relations. Naturally there will be gaps where bound pronouns are used
instead of morphological case. In Nanai (Tungusic), for instance, there are seven
cases but no genitive, the possessor relation being expressed by a cross-referencing
bound pronoun (Nichols 1983):
(76) nom acc dat loc abl inst all
Northern Australian languages have cross-referencing for subject and object and
some have no core case marking. In Warndarang, for instance, the case inventory
is as follows (Heath 1980: 268):
(77) nom loc abl inst all purp
The missing cases are genitive, dative and a second core case, ergative or accusative.
All of these gaps can be accounted for by reference to the cross-referencing bound
pronouns which encode subject, object and possessor. With the verb for GIVE the
object series of bound pronouns encodes the recipient. There is also a verbal prex
ma- to indicate that the object series is encoding a beneciary. In this way the
functions that would be associated with a dative are covered.
If morphological case systems are built up on hierarchical lines and if most of
the gaps in the hierarchy can be attributed to the use of bound pronouns, then
morphological cases and systems of bound pronouns between them dene a hi-
erarchy of functions or relations. It seems languages place more importance on
having some morphosyntactic means of expressing relations such as subject and
object than on having some morphosyntactic means of expressing relations such
as locative or instrumental. Lower ranked relations like these are often expressed
via prepositions or postpositions. Adpositions, although forming closed classes of
some dozens or perhaps some scores of members, are mostly lexical rather than
grammatical.
There are instances where a number of functions are expressed by a single case.
As mentioned above, it is not uncommon in Australian languages to nd a genitive
dative case and in Kannada there is syncretism of the ablative/instrumental distinc-
tion even though there is a comitative which one would expect to be ranked lower
on the grounds of its sparser distribution across languages. In Tarascan (Chibchan)
there is also a comitative even though there is no dative or ablative. In this language
there is no indirect object; the recipient and beneciary are expressed in the ac-
cusative. The locative covers not only location, but the other local notions such as
from and to (Suarez 1983: 87):
(78) nom acc gen loc inst com
159
The fact that syncretisms occur and gaps, mostly attributable to an alternative
grammatical device, precludes the possibility of establishing implicational rela-
tionships between cases, but the syncretisms and gaps do not render the hierarchy
vacuous. Case systems tend to be built up in a certain sequence with overwhelm-
ingly greater than chance frequency, to borrow a phrase from Greenberg (1963).
29
On the point of one case expressing more than one function, it is interesting to note
an asymmetry between the treatment of ablative and allative. An ablative is much
more widely distributed than an allative. The to-function is frequently expressed
by another case such as the accusative (Latin and some other Indo-European), the
dative (Turkish and other Altaic) or the locative (Yidiny).
30
Of course, instances like
these where a case expresses more than one function or role raise the question of how
these cases are labelled. It seems that it is widely accepted practice to label on the
basis of the higher function on the hierarchy. Where a particular case expresses, say,
A function and instrumental function, it is usually labelled ergative. This practice
will of course create apparent gaps in the hierarchy at the point where the lower
function occurs.
160
6
Life cycle of case systems
6.1 Origin
It is generally recognised that in the historical development of language
grammatical forms develop from lexical forms and bound forms from free forms.
There are two common lexical sources for case markers, one verbal and the other
nominal, of which the verbal is probably the more fruitful. Adverbial particles also
provide a source. The sequence of development is from noun, verb or adverb to
preposition or postposition, and from postposition to sufx.
1
In many areas case marking is of long ancestry as witnessed by formal complexity
(suppletive variants and fusion) and semantic complexity. This is true of case mark-
ing in Indo-European languages, for instance, and in such a situation the prospects
for determining the origin of particular case markers are not too promising. On the
other hand there are some languages, including the Indo-Iranian branch of Indo-
European, where the development of new case markers is attested, and there are
many languages in which we can plausibly reconstruct the origin of postpositions
and phrase-nal postposition-like case markers.
6.1.1 Verb to case marker
The key to understanding the verbal origin of adpositions lies in the
notion of an argument implied by a predicate. A predicate like COME is a two-place
predicate implying a comer and a destination. A predicate like LEAVE is a two-place
predicate implying a leaver and a source. A predicate like FLY is a one-place predi-
cate implying a yer, but there will be a need for languages to express X ies to Y
or X ies from Y. One way to do this is to use pairs of predicates. To express y
to the predicates FLY and COME can be used; to express y from the predicates FLY
and LEAVE can be used. Here are examples from Thai:
(1) a. Th an c` a bin maa krungth eep
he will y come Bangkok
He will y to Bangkok.
161
b. Th an c` a bin c` aak krungth eep
he will y leave Bangkok
He will y from Bangkok.
These constructions are usually described as serial verb constructions, but note that
the second verb cannot be modied independently of the rst for tense, mood, aspect
or polarity (in Thai these modications are via free forms like c` a in (1a) and (1b)).
The second verb is nonnite. It cannot take an expressed subject; its missing subject
must be interpreted as being that of the rst verb. As is obvious fromthe translations,
the second verb is functionally equivalent to a preposition. A preposition is rather
like a nonnite verb. Consider the translation of (1a), for instance. The preposition
to is a two-place predicate. One argument has the role of destination (Bangkok);
the other argument is not expressed and must be interpreted as being coreferential
with S.
Now consider the word h ay. It can occur as the sole lexical verb in the clause,
where it has the meaning give (2a), or it can be used following another verb, where
it has either the meaning give or on behalf of (2b) (Kullavanijaya 1974: 51, 85).
(2) a. Ph h ay ng
e
n P` uk
father give money Pook
Father gives money [to] Pook.
b. M y ep s ya h ay P` uk
mother sew garment give Pook
Mother sews the garment [and] gives [it to] Pook.
Mother sews the garment on behalf of Pook.
In (2b) h ay cannot be modied for tense, mood, aspect or polarity. It could be
analysed as a nonnite verb or as a derived preposition. Kullavanijaya suggests that
it is a nonnite verb when it means give and a derived preposition when it means
on behalf of .
The reanalysis of a verb as an adposition is clearer in a language where verbs
host bound forms, but adpositions do not. In Ewe, a language of the Kwa branch
of the Niger-Congo family, the form n a can appear as a verb meaning give (3a)
or as a preposition meaning for (benefactive or purposive) (3b). In (3a) n a bears
person/number marking for the subject; no such marking is possible in (3b) (Heine
et al. 1991: 1).
(3) a. Me-n a ga kof
1SG-give money Ko
I gave Ko money.
162
6.1 Origin
b. Me-w d v ev e n a dod okp l a
1SG-do work hard give exam DEF
I worked hard for the exam.
The morphological difference between n a in (3a) and n a in (3b) matches the
difference in meaning and conrms that while the rst n a is a verb the second is a
preposition.
The development of serial verbs into adpositions is a widespread phenomenon.
It is particularly common in the languages of West Africa, New Guinea, Southeast
Asia and Oceania. There are also similar developments in a wide variety of lan-
guages involving verb forms morphologically marked as nonnite. In English, for
instance, there are examples of prepositions deriving from the present participle
such as regarding, following and concerning. Similarly in German where betre-
ffend concerning derives from a present participle. Betreffend is a postposition
since the originating participle is clause-nal, German subordinate clauses being
verb-nal.
2
Similar developments have occurred in the Dravidian languages where
the postposition p arttu towards, which governs the accusative, derives from the
participle p arttu looking at (Steever 1987: 736), and also in Sanskrit where the
postposition sahita with derives from a past participle of a verb meaning join
(Andersen 1979: 26).
Not surprisingly a survey of the development of adpositions reveals clear ten-
dencies. The semantic roles that are in frequent demand are the local notions of
location, source and destination plus a few others such as instrument, purpose, ben-
eciary and accompaniment. A verb meaning come provides an obvious source
for an adposition meaning to (compare (1a) above); a verb meaning leave is a
likely source for an adposition meaning from (compare (1b) above), and a verb
meaning give is a regular source for an adposition meaning for (see (2) and (3)
above). There are also examples of adpositions with a purely grammatical function
deriving from verbs. The most widely quoted example involves b a in Mandarin
Chinese. Mandarin is basically an SVO language with no adposition or afx to
mark the object. However, there is a construction in which the notional object is
preceded by b a and the b a-phrase precedes the verb (Li and Thompson 1981: 464).
(4) T a b a f` antng sh oushi-g anjing le
3SG dining.room tidy-clean PERF
S/he tidied up the dining room.
b a derives from a verb meaning take hold of or grasp and the construction in
(4) derives from a construction of the form X take Y and (X ) verb (Y ). The original
verb b a has lost its literal sense of take hold of , but though it is generally described
163
as an object-marking preposition, it is signicant that it is conned to denite or
generic patients and cannot generally be used with verbs such as ` ai to love and
xi ang to miss (someone). A similar restriction appears with the object-marking
preposition k` e in the West African language G a. This preposition also derives from
a verb meaning take (Lord 1982: 2868).
Verbs meaning take often come to mark instruments. A construction that is
literally Xtaking axe chopped wood becomes reinterpreted as Xwith axe chopped
wood.
Here are some examples of recurrent types of development of verbs toadpositions.
They are grouped according to the function of the resulting adposition:
3
(5) object
Yatye aw a take, Yoruba gb` a get, Mandarin b a take hold of
dative/benefactive
Yoruba f un give, Mandarin g ei give, Ewe n a give, Thai h ai
give
allative
Nupe l o go, Mandarin d` ao arrive, Ewe d
.
e reach
comitative
Tamil kon
.
t
.
u (< *kol take), Mandarin g en follow
instrumental
Mandarin y` ong use, n a take, Ewe ts
take
locative
Mandarin z` ai be at, Ewe le be at
ablative
Mandarin c ong follow, Ewe t s o come from
perlative
Ewe t o pass by
6.1.2 Noun to case marker
Adpositions and case afxes of nominal origin are common in the
expression of the local case relations. As mentioned in section 1.3.3, languages
often specify relative orientation by using a relator noun such as top in It stands on
top of the cupboard. Top is the head of a noun phrase that has a prepositional phrase
(of the cupboard ) as a post-head modier. The essential predication is between it,
stand and cupboard. Although top is syntactically the head of the phrase top of
the cupboard it is semantically subordinate in that its function is to specify more
precisely the relationship of cupboard to stand. (On is not too important either here,
164
6.1 Origin
since stand expects a locative complement.) Not surprisingly relator nouns like top
become adpositions and sometimes afxes (as noted in section 1.3.3, one could
interpret on-top-of as a compound preposition).
This kind of development is common in the Dravidian languages. In Kannada
there are forms like horage as in (6) (Madtha 1976: 143):
(6) Kl asina horage h ogu
class.GEN outside go
Get out of the class!
Forms like horage, sometimes referred to as secondary case markers, are postposi-
tions. The form hora derives from a noun meaning exterior. It betrays its nominal
origin in governing a complement in the genitive, the main adnominal case (compare
HindiUrdu example given as (11d) in chapter 1).
This kind of development is also common in Uralic languages. In Finnish kohta
place in the adessive kohdalla at the place is used with a genitive dependent as
in talo-n kohdalla (house-GEN place-AD) at the house. Here kohdalla functions as
a postposition.
In the Finnic language Karelian, the comitative is marked by -ke as in velle- -ke
with the brother where -ke derives from *ker a-lla at a turn, at time of used
with genitive attributes. Compare Finnish koira-n keralla (dog-GEN with) with the
dog. The genitive -n is reected as - in velle ke (Anttila 1972: 149). In develop-
ments like this the original head (*ker alla in this instance) eventually becomes an
afx and the original dependent a stem. In another Finnic language, Komi-Permiak,
there are ve cases where the rst formative is -vv-: -vva superlative to the top of ,
-vvi s sublative down from, -vvyn superessive on, -vv
e
superterminative
up to and -vv
e
t perlative via. The formative can clearly be related to the noun
vyv top (Austerlitz 1980: 237).
4
The Hungarian case sufx -bel-ba into derives from b el innards, guts in its
allative case form: vilag-bel-e (world-insides-to) yields modern Hungarian vil agba
(in)to the world where -be/-ba is the illative case marker.
In Persian a noun root r adi meaning reason or goal gave rise to a postposition
r adi by reason of, concerning. It reduced to r a and came to indicate cause, purpose
or reference (as for . . . ) and then later to mark indirect and direct objects. In
Modern Persian it is used with indirect objects and specic direct objects (Windfuhr
1987: 534, 541f ).
Where nouns develop into prepositions, postpositions and case sufxes, the
development is almost always to a local form, which may later develop more abstract
or grammatical functions. Here are some examples grouped under local labels. The
most common source would appear to be body parts (cf. Heine et al.).
165
(7) inside, into
Swahili ndani (< da guts), Mixtec ini heart, chi stomach (also
under), Cakchiquel -pan stomach, Kpelle -l a mouth
to
Kpelle -t ue front, Mixtec nuu face (also develops other
local functions)
in front of
Swahili mbele (from a root meaning breasts), Hausa g` aban < g` ab a
front, Jacaltec sat face
on top of, above, on
Hausa k an < kai head, Tamil m el sky, Kpelle -m a surface,
Jacaltec wi < wie head
near
Kpelle -k
e
l vicinity
behind, at the back of
Kpelle -polu back, Hindi pche rear, Ewe megb e back,
Gujerati put
.
he back
below
Tamil n ed
.
e ground, Swahili chini (< *ci ground)
from
Kpelle -y ee hand
6.1.3 Adverbial particle to case marker
Some adpositions have their immediate origin in adverbs or particles.
In the Indo-European languages there appear to have been free forms which de-
veloped into prepositions on the one hand and components of compound verbs on
the other. In Latin, there are forms like per through, in in, trans across and
e/ex out of which show up as prepositions and as the rst member of compound
verbs. Compare Ambul a per aedes (walk.IMP through house.ACC) Walk through the
house! where per is a preposition governing the accusative and Perambul a aedes
where perambul are is a transitive verb governing the accusative.
5
In Sahaptin (Penutian) there are two ergative sufxes. Rigsby calls one (-in)
obviative (it is used when Ois third person) and the other (-nim) inverse (it is used
when Ais third person and Orst or second person (Rigsby 1974)). Rude argues that
-nimis derived froma particle meaning hither (1988). Both the putative source and
the derived function are illustrated in the following example where the hypothesised
particle is reected as a verb sufx and a noun sufx (Jacobs 1931: 126).
166
6.2 Developments within case systems
(8)

Aw=na s xwi -saat-ni -m i-tw ana-m-a s
now=1SG old.man-ERG 3SG-follow-hither-IMPF
Now the old man is following me.
In the closely related Nez Perce there is an ergative marker -nim cognate with the
Sahaptinform. Facedwithanergative anda hither formof similar shape, one would
not be inclined to see them as related. In fact one would probably expect an ergative
to have an origin in a from formrather than a to or hither form, but the Sahaptin
evidence is highly suggestive. It makes obvious sense to use a hither form to mark
action directed towards the speech act participants (cf. the discussion in section
5.2.6). Rude suggests the inverse ergative has widened its scope to become a general
ergative in Nez Perce. He also suggests that the accusative sufx in Sahaptin (-na or
-nan) and Nez Perce (-ne) derives froma particle meaning to, toward or thither.
When a verb or noun becomes an adposition, it does not necessarily
lose its lexical character, though it becomes a member of a relatively small closed
class of a score or two of members. When a postposition becomes a case sufx, two
signicant changes occur. First of all it is liable to develop variant forms according
to the phonological properties of its host, and secondly it becomes a member of a
much smaller set of forms, usually no more than ten or so. Since all noun phrases,
no matter what semantic role they bear in relation to their governor, must be marked
by one member of a small set, it is inevitable that at least some case sufxes will
cover a broad semantic range. We will discuss the phonological factors rst, then
the semantic ones.
6.2.1 Phonological factors
It is a general principle of language that grammatical forms are less
prominent than the lexical forms they relate. As nouns and verbs develop into
adpositions, they become unstressed and pronounced as if they were an unstressed
syllable of a neighbouring word. The neighbouring word will be either the nominal
that is the head of the phrase governed by the adposition or simply the nearest word
in the phrase. In a consistent modierhead language like Turkish a postposition
will always be adjacent to the noun that is head of the noun phrase it governs. But
adpositions and the heads of noun phrases are not always adjacent. In English, for
instance, the preposition is not necessarily adjacent to the head of the noun phrase.
Postpositions are likely to become sufxes and develop variant forms according
to the phonological shape of the stem to which they attach. In this way paradigms
develop. The Pama-Nyungan languages provide a convenient example. The ergative
167
Table 6.1 Djaru case markers
disyllabic longer consonant
vowel stems vowel stems stems
water food man
nom ngapa mangarri mawun
erg/inst ngapa-ngku mangarri-lu mawun-tu
loc ngapa-ngka mangarri-la mawun-ta
dat ngapa-wu mangarri-wu mawun-ku
tends to be realised as -lu following vowel stems of more than two syllables and -Tu
following a consonant-nal stem. T stands for a stop which assimilates in point
of articulation to the nal consonant. An original l has hardened to a stop and
then assimilated to the point of articulation of the nal consonant of the stem.
Vowel stems of two syllables take the variant -ngku. According to an hypothesis
originally advanced by Hale, disyllabic vowel stems at one stage of their history
had a nal velar nasal. The l of the ergative hardened to a stop and assimilated to the
velar nasal in point of articulation. Subsequently the velar nasal was lost in word-
nal position so that a word like *marang hand with an ergative *marang-ku
became mara with an ergative mara-ngku. The proto-Pama-Nyungan locative is
reconstructed as *-la. It developed a parallel series of variants (Hale 1976). Some
examples of these are shown in Table 6.1 from Djaru (Tsunoda 1981). The dative
is also included. Here an original *-ku has developed a weakened variant -wu after
vowels.
6
Once paradigms develop there is scope for exchange of forms between paradigms.
Distinctions made in one paradigm may be levelled by analogy with another
(see section 6.3 below) or an important functional distinction made in one paradigm
may be copied in another (see (9) below).
The model of adpositions developing into afxes leaves one manifestation of
inectional case unexplained, namely case concord. One way concord could come
about is through the discontinuous expression of the content normally expressed
through a noun phrase in a language like English. As illustrated earlier in the text
with examples from Latin and Warlpiri (see (8), (9) and (10) in chapter 4), it is
common in some languages to represent an English phrase like that tall woman by
two or even three separate phrases. Naturally each of these phrases will require its
own marking. We could consider that any such matching in marking constitutes a
system of concord, but it may be that all such phrases are sisters, each governed by
the same head, usually the verb. If the possibility of discontinuous representation
was lost and the marking of the erstwhile phrases retained, then the result would be
multiword phrases with concord.
168
Afxes are always liable to fall prey to the reductive phonological processes of
assimilation, loss, vowel centring to schwa and fusion. The effect of such changes
is to reduce the differentiation between cases. In some Pama-Nyungan languages
vowel reduction to schwa or vowel loss has obliterated the distinction between the
ergative and the locative. As we shall see below, reductive phonological changes
can all but destroy a morphological case system. Languages do not normally take
prophylactic measures to restrain the eroding effects of phonological change on
a case system, they simply employ an alternative means such as using word or-
der to distinguish subject and object or using prepositions or postpositions. How-
ever, occasionally a language takes therapeutic measures to repair the damage done
by phonological change. In Latin there were syncretisms resulting from reductive
phonological processes at different periods of its history. The reconstructed ac-
cusative plural is -*ns. However, in early Latin n dropped out before s in nal
syllables and the preceding vowel lengthened.
(9) lst declension *domin a-ns > domin as mistresses
2nd declension *domino-ns > domin os masters
3rd declension *consul-

ns > consul es consuls
4th declension *manu-ns > man us hands
5th declension *di e-ns > di es days
This produced nominativeaccusative syncretism in the plural. In the second de-
clension a form-oi (later ) was introduced fromthe demonstrative paradigm, which
restored the nominativeaccusative distinction. The form was then adopted in the
rst declension where the sequence a reduced to a diphthong (ae in the orthography;
see Table 1.2).
6.2.2 Nonphonological factors
An inectional case system is small. It typically contains half a dozen
or so members and rarely more than a dozen. The whole point of having a case
system, or any morphological system for that matter, is to provide brief signals
for broad categories that will be sufcient for communication most of the time.
Where a case system does not provide a sufciently explicit indication of the se-
mantic role of a dependent, then the secondary system of adpositions or relator
nouns can be employed, or, in rarer instances, a more elaborate paraphrase can be
used.
The small case system with its broad categories is sufcient most of the time
because of the redundancy inherent in lexical choices. A predicate implies argu-
ments, so all a case system is required to do with complements, at least from
the functional point of view, is to distinguish arguments (recall the discussion in
169
section 5.2.6). With adjuncts the role will often be determinable from the choice
of noun. Suppose we have a sentence in which the lexical items are MAN, DOG, HIT
and STICK. There is hardly any need for case marking to express the propositional
content. In the absence of an indication to the contrary one naturally assumes that
MAN will be agent, DOG patient and STICK instrument.
The notionthat case categories are broadandthat theyare able tofunctionas broad
categories because of a certain amount of redundancy arising from lexical choice
is fundamental to understanding how a case system evolves. A schematic outline
of some of the major developments is given in Figure 6.1. When nouns, verbs or
adverbial particles develop into adpositions and when postpositions develop further
into sufxes they are usually semantically homogeneous with a concrete meaning,
typically a local meaning. There is evidence that adpositions and case sufxes
extend to cover more abstract meanings and to mark syntactic categories (shown by
the horizontal arrows to the left in Figure 6.1). When a case system withers away
(see section 6.3 below), the marking remains on the bound pronouns, which mark
the purely syntactic functions.
What is not shown in Figure 6.1 is the tendency for cases to merge or syncretise.
The reverse of this, namely a case splitting into two, is rare but an example from
Pitta-Pitta is given at the end of this section.
7
It is generally agreed that semantic cases, especially local cases, can expand
their territory and come to cover syntactic relations such as direct object, thereby
becoming grammatical cases, but hard evidence is difcult to come by (Lehmann
1982: 111). The development is to be expected on the grounds that the expression of
abstract categories in general derives from the extension of morphemes or combi-
nations of morphemes with a concrete meaning. To take one lexical example from
Figure 6.1 Development of appositions, sufxes and bound pronouns
170
thousands, the word education derives from the Latin e-duc- ati o (out-lead-NM),
literally a physical leading out.
With adpositions it is easier to nd documented examples of the development of
a local marker to the marker of a syntactic relation. This kind of development is
captured by the lower row of left-pointing arrows in Figure 6.1.
The Latin preposition ad to is reected in Spanish as a. It still retains its local
meaning (10), but it has come to mark the indirect object (11) and the direct object,
the latter where the object is animate and specic (12).
(10) Juan vuelve a so hotel
Juan return.3SG to his hotel
Juan returns to his hotel.
(11) Le expliqu e el caso a mi hermano
3SG.IO explain.PAST.3SG the case to my brother
He explained the matter to my brother.
(12) Vi a mi hermano
saw.1SG my brother
I saw my brother.
In Rumanian the Latin preposition per through is reected as pe and it too has
come to be used as a direct-object marker.
The Latin preposition d e from is reected as a function form in Romance
including di in Italian and de in Spanish. These forms mark adnominal dependents
and the complements of certain verbs low on the transitivity scale. The following
examples are Spanish:
(13) a. la casa de la se nora Garcia
Mrs Garcias house
b. corrida de toros
bullght
c. asustarse de la luz
to be afraid of the light
d. cambiar de vestidos
to change clothes
English of and off reect Old English of from. The Old English word with its
unstressed pronunciation [
e
v] has developed into a grammatical marker while the
same word with its full pronunciation has been retained in its local function. Modern
English of is rather like Spanish de and Italian di in that it has become the default
171
marker for the complement of nouns and adjectives (student of linguistics, full of
emotion).
When we turn to morphological case systems, we are dealing with systems of
marking that are of some antiquity and it is naturally more difcult to nd clear
examples of semantic cases developing into grammatical cases. Certainly there is
no shortage of languages where a grammatical case expresses a purely syntactic
relation plus a semantic role that lies outside this relation. In various Indo-European
case languages the accusative expresses destination as well as direct (and sometimes
secondary) object. In Latin, for instance, one can say Ms leg at os R omam (sent.I
legates.ACC Rome.ACC) where leg at os is the direct object and R omama complement
expressing destination. In many ergative languages the ergative case covers a non-
core function. In Pama-Nyungan languages the ergative usually covers instrumental
function as well as A (see section 3.2.1); in Sumerian the ergative also expressed
location. Adative will often express destination as well as indirect object and a gen-
itive sometimes expresses a local function as in Ancient Greek where it expresses
source.
Examples where there is some comparative-historical evidence that a local case
assumed a syntactic function include the development of -e/-a in Turkish from
an allative to a dativeallative, the analogous development of -man in Quechua
(Lehmann 1982: 109), and the assuming of genitive function by the Pama-Nyungan
locative in Pitta-Pitta (see below).
Where a passive develops in a language with a morphological case system one
of the noncore cases will be chosen to express the demoted subject. Passives pro-
vide a class of clear examples of cases assuming a new syntactic relation. The
demoted subject is expressed in the genitive in Ancient Greek, the instrumental in
Russian, Czech and several of the Dravidian languages, and the ablative in Latin
and Armenian.
In all of these examples of local or other concrete forms extending their range
to cover syntactic relations, the redundancy that arises from lexical choices plays a
part. Destinations can often be distinguished from roles such as patient subsumed
under direct object by the fact that they will normally only occur with verbs of
motion; instruments can be distinguished from roles such as agent subsumed under
the ergative relation on the grounds that instruments are almost always inanimate
and agents animate, and so on. Lexical redundancy also plays a part in other mergers.
In some languages there is a locativeallative case and one suspects there has been
a merger. In Hittite there is positive evidence of such a merger. In Late Hittite the
to case (called allative in this text but directive in some sources) merged with the
dativelocative, which resulted from an earlier merger of the dative and the locative
(Luraghi 1987). An allativelocative merger is possible because destination will
normally occur only with verbs of motion.
172
With some mergers it can be said that the cases in question had some overlap to
start with in the sense that there are extra-linguistic situations that can be interpreted
as belonging to one case or the other. Take ablative and instrumental, for instance.
Besides situations in which there is a clear source (She walked all the way from the
station) and situations where there is an obvious instrument (She swatted the y with
a rolled up newspaper), there are situations which could be interpreted as ablative
or instrumental, mainly with reference to a resultant state. Suppose a natural spring
provides irrigation for a garden. One could speak of the garden being watered from
a spring (source) or by a spring (means, instrument or perhaps agent). Here are
some other possibilities, but it should be noted that they are not meant to illustrate
the interchangeability of the English prepositions from and by.
(14) battered by/from the gale
wounded by/from shrapnel
smoothed by/from rubbing
warped by/from the heat
Ablativeinstrumental syncretism occurred in a number of Indo-European lan-
guages including Hittite, where the two cases came to be used interchangeably
before the formal distinction was lost (Luraghi 1996: 46f ).
As noted in section 6.2.1, Pama-Nyungan languages usually reect an erga-
tive *-lu and a locative *-la. In most of these languages the ergative covers A
and instrumental relations, but in some Pama-Nyungan languages the locative
case markers have come to express instrumental as well as locative, leaving the
-lu to express just A. In English the preposition by, formerly a purely local form
(He stood by the window), came to acquire a sense of means or instrument. The
Oxford English Dictionary suggests that it acquired its instrumental sense via ex-
pressions such as She read by candlelight where the by-phrase, originally a locative
(Where did she read?), was reinterpreted as instrumental (How did she read it?).
It is not hard to nd situations that allow a locative or instrumental interpretation
and which could facilitate a locative or instrumental form adopting both functions.
Here are some examples:
(15) wash the cloth in/with water
cook meat on/in/with re
carry wood on/with shoulder
come on/by horse
go along/by means of the back road
see someone in/by means of the relight
mix our in/with water
173
A merger of instrumental and locative also occurred in Middle Indo-Aryan, but
there the cases merged under the instrumental forms. In Latin there was a syn-
cretism of the Indo-European instrumental and ablative and then the syncretism
of these two with the locative (the latter merger not quite complete as we saw in
section 2.2.2). The forms of the Latin ablative reect markers of the three syncre-
tised cases. The - o of the second declension singular is from an Indo-European
ablative, the -s of the second declension plural (borrowed into the rst declen-
sion also) is of instrumental provenience and the -e of the third declension sin-
gular of the consonant stems reects an Indo-European locative (Palmer 1954:
241ff ).
8
Of a slightly different kind is the merger of the genitive and dative under the
dative form in a number of modern German dialects. As mentioned in section 5.3, a
dative can be used to signal possession. In some languages all possessive datives are
predicative; in others an attributive use develops which overlaps with the genitive. In
Pennsylvania German the dative has become the only way of expressing possession
so that there is effectively a genitivedative case. Note that the construction in the
following example involves a possessive adjective cross-referencing the possessor
(Burridge 1990: 41).
(16) Des is em Gaul sei(n) Schwans
this is the.DAT horse.DAT his tail
This is the horses tail.
As mentioned earlier it is possible for a case to split, though this would ap-
pear not to be too common. In the Pama-Nyungan languages there is usually a
locative case which is represented by -la with vowel-nal stems of more than two
syllables, by -ngka with disyllabic vowel stems, and by -Ta with consonant-nal
stems (compare section 6.2.1 above). In Pitta-Pitta there are no consonant stems
so not surprisingly there is no -Ta. The case markers -ngka and -la appear, the
former as -nga (a sound change of -ngk to -ng is independently attested). How-
ever, each marks a different case. The marker -nga signals the genitive/purposive
and -la the evitative (what is to be avoided). The locative is marked by -ina, an
apparent innovation. What appears to have happened is that two allomorphs of a
locative have come to mark two separate cases. The locative in some Australian
languages covers purposive and evitative functions as well as location. It is likely
that at some earlier stage in the history of Pitta-Pitta the locative covered this
range of functions and an association came to be made between particular allo-
morphs and particular functions. The evitative is semantically similar to the ab-
lative. The ablative simply refers to motion away from, but the evitative has the
sense of keeping away from an entity or moving away from an entity because it is
174
6.3 Loss of case marking
a source of danger and the like. If someone simply walked away from a re, this
would be expressed in the ablative (maka-inya). If, however, they moved away to
avoid the heat or smoke, this would be expressed in the evitative (maka-la). There
are some instances of the evitative being added to the ablative (maka-inya-la).
Perhaps this usage is responsible for the association of -la and evitative. If at
some earlier stage the locative expressing avoidance was added to the ablative
expressing motion away from, the -la allomorph would have been used because
the ablative-marked form would have been a vowel-nal stem of more than two
syllables.
As we have seen in the previous section, cases may merge and in this
way a case system will be reduced unless new members are recruited. And as we
saw in section 6.2.1, phonological change, which tends to be reductive especially
in unstressed syllables, may obliterate case distinctions. The case system of Clas-
sical Latin was severely eroded by phonological change (the reader should consult
Table 1.2 in reading this section). Word-nal -m, which was the characteristic marker
of the accusative singular, must have been reduced, probably to nasalisation. It is
occasionally omitted in inscriptions, and word-nal syllables in -mare elided before
vowels in verse: d onumexiti ale doomed gift is pronounced as if it were d onexiti ale.
In Vulgar Latin in the rst three centuries of the Christian era the distinction be-
tween short a and long a was lost in unstressed syllables (hence the ablative was
no longer distinguished from the nominative and accusative in the rst declension
singular) and i merged with e (as e) and u merged with o (as o). The merging of
u and o, together with the loss of word-nal -m, meant that in the second declension
singular there was a common form for accusative, dative and ablative.
As these reductive phonological changes began to accumulate and severely reduce
the number of distinctions, a number of other changes took place. The two small
declensions merged with larger ones, the fourth with the second and the fth with
the rst. The neuter gender was eliminated with neuters that declined like bellum
war taking on the nominative in -s, thus caelus sky for earlier caelum, f atus fate
for earlier f atum. Some neuter plurals in -a were reinterpreted as feminine singulars
of the rst declension, e.g. folium a leaf had a plural folia which was taken to
be a feminine singular and is reected in Italian la foglia, Spanish la hoja, French
la feuille, etc.
As the case distinctions begantobe reduced, the use of prepositions increaseduntil
the situation found in the modern Romance languages emerged where grammatical
relations outside the core are expressed only via prepositions. The use of reexes
of d e from came to mark adnominal noun phrases and reexes of ad to came
to mark indirect objects as well as destination, these innovations replacing the
175
genitive and dative respectively. Compare the following sentence from the Vulgate
(Luke XV. 15) and its Italian (17b) and Spanish (17c) translation (in the Latin
of the Vulgate all the morphological distinctions of Classical Latin are retained).
The literal translation of the Italian is He put himself to service of one of the
inhabitants of that region and of the Spanish He put himself to serve to a citizen
of that land.
(17)
a. Adhaesit un cvium regi onis illius
adhere.PERF.3SG one.DAT citizen.PL.GEN region.GEN that.GEN
He attached himself to one of the citizens of that region.
b. Si mise a servizio di uno degli habitanti di quella regione
c. Se puso a servir a un ciudadano de aquella tierra
By the time the Romance languages began to emerge in the early Middle Ages no
noun displayed more than two distinctions, nominative and oblique. Table 6.2 gives
representative forms of the rst, second and third declensions. In the upper part of
the paradigmforms fromWestern Romance are shown where -s was retained. In the
lower part of the table Italian forms are shown representing Eastern Romance where
-s was not retained. Rumanian was more conservative than Italian in retaining to
the present day the neuter gender and developing a directoblique case system, i.e.
one where the direct or nominative covers core functions andthe other case covers the
functions formerly expressed by the genitive and dative. (See note 28 in chapter 5.)
The history of English provides another example of a case system withering
away. In Old English there were four cases (nominative, accusative, genitive and
Table 6.2 Early Romance noun declensions
1 2 3
moon year dog
Western Romance
sg nom luna annos canes
obl luna anno cane
pl nom lunas anno canes
obl lunas annos canes
Italy
sg nom luna anno cane
obl luna anno cane
pl nom lune anni cani
obl lune anno cani
176
Table 6.3 Old English case inection
stone ship
singular
nom se st an tht scip
acc thone st an tht scip
gen ths st anes ths scipes
dat th m st ane th m scipe
plural
nom th a st anas th a scipu
acc th a st anas th a scipu
gen th ara st ana th ara scipa
dat th m st anum th m scipum
tale name
singular
nom s eo talu se nama
acc th a tale thone naman
gen th re tale ths naman
dat th re tale th m naman
plural
nom th a tala th a naman
acc th a tala th a naman
gen th ara tala th ara namena
dat th m talum th m namum
dative) plus a vestigial instrumental. Representative nouns fromthe more numerous
paradigms are shown in Table 6.3. St an stone represents a masculine noun, scip
ship a neuter and talu tale a feminine. Nama represents the -an declension which
contained masculine and feminine nouns and the odd neuter. Nama is masculine.
The cases in Old English are only weakly differentiated, with more differentiation
in the demonstrative than in the noun. For this reason the demonstrative that has
been included in the table (compare the German case forms in Table 4.1).
The Old English case paradigms are strikingly similar to those of German,
whether old or Modern, but while German has lost only the instrumental from
Old High German to Modern German, English lost almost the entire system be-
tween the late tenth and thirteenth century. With nouns only the genitive remained
distinct from the nominative. With pronouns the genitive was reinterpreted as a
possessive adjective and a nominativeoblique distinction remained, the oblique
reecting a merger of the accusative and dative.
Two phonological changes destroyed the system. One was the reduction of un-
stressed vowels to schwa which involved the loss of a number of distinctions. The
177
other was the loss of word-nal n in inections which affected not only the -an
paradigm but all datives in -m, since n had been substituted for m. Virtually the
only forms to survive these changes were those ending in -s, the genitive singular of
nouns in the general masculine and neuter classes and the nominativeaccusative
plural of nouns in the general masculine class. Both these forms were extended to
all paradigms so that by the end of the Middle English period the system was as in
Modern English:
(18) nom sg the stone, ship, tale, name
gen sg the stones, shippes, tales, names
plural the stones, shippes, tales, names
The nominative forms of the demonstrative with initial s (se, seo) adopted the th
of the rest of the paradigm. The loss of inection entailed the loss of grammatical
gender and tht, formerly the nominativeaccusative singular demonstrative with
neuters, remained as a singular demonstrative for all genders while the, which
reects the rest of the demonstrative paradigm, weakened semantically to become
a denite article.
The near elimination of case in English and Romance makes an interesting com-
parison. In both areas the use of prepositions increased as the case system was
eroded so that all adverbal relations outside the core eventually had to be expressed
via prepositions. Here is an Old English version of the verse from Luke given as
(17) above and Wyclif s fourteenth-century translation.
(19)
a. He folgode an-um burg-sitt-end-um menn thaes rices
he followed one-DAT town-dwell-ing-DAT man.DAT that.GEN land.GEN
b. He clevede [= attached himself ] to oon of the citizeins of that contr e
In both areas SVO word order emerged and served as the normal means of
distinguishing A from P.
9
In both areas case distinctions were retained more on pronouns than nouns. To
start with, Latin and Old English made stronger case distinctions in pronouns than
nouns in the sense that the marking consisted in some instances of a change to the
stem (rather than a sufx) including suppletion. Consider, for instance, the rst-
and second-person-singular pronouns given in Table 6.4. In English a three-way
distinction was retained, ic/me/min etc., but the genitive forms even in Old English
were reinterpreted as stems and came to be used as possessive adjectives (my, thy)
and possessive pronouns (mine, thine).
10
In Romance there is always a distinction
between direct and indirect object forms at least in the third-person clitics. The
178
Table 6.4 Latin and Old English pronouns
Latin Old English
1 2 1 2
nom ego t u ic th u
acc m e t e m e th e
gen me tu mn thn
dat mihi tibi m e th e
abl m e t e
singular pronouns of French and Italian are given in Table 6.5. Free pronouns are in
bold. French has clitic pronouns for subject, object and indirect object plus a series
of free, stressable pronouns which can be used in isolation (moi), with prepositions
(avec moi with me), predicatively (L etat cest moi The state, it is me) or when
stress is required (Moi, je le vois I see it). In Italian, which is more representative
of Romance, there are no subject clitics, the subject being represented inectionally
in the verb.
It is generally true that case distinctions are retained more on pronouns, especially
clitic pronouns, than on nouns, hence the arrows in Figure 6.1 indicating that the
nal resting place of case marking is on bound pronouns where bound includes
clitics and inection.
It is common for nouns in oblique cases to be reinterpreted as adverbs, particularly
adverbs of place, time and manner. In Latin, for instance, there are adverbs of degree
such as multum much and quantum how much which are accusative case forms
in origin that have been reanalysed as adverbs; similarly there are adverbs that
derive from ablative case forms such as prm o in the rst place and qu ar e why
which is a fossilised ablative phrase qu a r e by what matter. When a case system
disappears these fossilised case markers may remain. English why is from OE hw y,
the instrumental of the interrogative root (cf. hw a who, hwt what). Even in Old
Table 6.5 French and Italian singular pronouns
French Italian
1 2 3 1 2 3
m f m f
subj. je tu il elle io tu lui lei
obj. me te le la mi ti lo la
ind.obj. me te lui lui mi ti gli le
oblique moi toi lui elle me te lui lei
179
English the instrumental was vestigial, but this lexicalised formation has survived.
Examples of the Old English genitive survive in adverbs like besides and sideways
as well as in once, twice, thrice and since where the voiceless sibilant [s] is spelled
differently from the voiced [z].
11
6.4 Derived functions of case marking
Forms used to mark case may come to mark other functions. It some-
times happens, for instance, that case is marked on verbs, that the case function is
reinterpreted, and that the marking remains and signals the new function. It is also
possible for inectional case to be lost and for the marking to remain only for one
or more derived functions.
Local expressions normally cover the domain of time as well as space, so not
surprisingly local case marking often comes to mark tense or aspect with verbs.
Archaic English expressions like The times, they are a-changing illustrate a local
preposition (OE/ME an/on on) used with the present participle to indicate ongo-
ing activity. In French the expression venir de to come from refers to time as in
Je viens de manger (I come from eat.INF) Ive just eaten, similarly in Pitta-Pitta
Tatyi-ka-inya nganytya (eat-NM-ABL I) Ive just eaten, where the ablative corre-
sponds to the French preposition de.
Perhaps the neatest example of the transference of case marking to verbs is to
be found in Kala Lagau Ya, an Australian language of the Western Torres Strait.
Practically the whole system of cases and their markers have been transferred to the
domain of time and they show up as part of the system of tense/aspect markers on
the verb. There is one problem of identication. In Kala Lagau Ya -n appears as an
ergative marker with nouns and as an accusative marker with pronouns. Table 6.6
displays the identications. Given this data one can easily explain the correspon-
dences albeit in a post hoc way. In fact one could justify a transference of either
ergative or accusative to completive, since both are associated with high semantic
transitivity, one of the correlates of which is perfective aspect (see section 5.2.6).
Table 6.6 Kala Lagau Ya case and tense/aspect
matchings
marker case tense/aspect
-n ergative completive
accusative
-pa dative/allative incompletive
-pu comitative habitual
-ngu ablative yesterday past
-nu locative immediate past
180
6.4 Derived functions of case marking
Examples (20) and (21) illustrate -pa, the dative with nouns and incompletive with
verbs respectively (Kennedy 1984):
(20) Nuy ay-pa amal-pa
he food-DAT mother-DAT
He [went] for food for mother.
(21 ) Ngoeba uzar-am-pa
1DU.INC go-DU-INCOMPLETIVE
We two will go (are endeavouring to go).
A dative indicates activity directed towards a goal, so it is appropriate for an aspect
that marks incomplete actions in the present, including present continuous actions
[and] ... actions in the past where the speaker wishes to emphasise purposiveness
or goal orientation (Kennedy 1984: 160).
The innitives of some languages derive fromthe reinterpretation of case-marked
verbs. The innitives in Indo-European languages arose in this way. In Vedic
Sanskrit, for instance, a verb could take the accusative d atu-m giving, dative
bhuj- e enjoying, genitiveablative nidh ato-s putting down or locative nes
.
an
.
-i
leading. In a sentence such as Vas
.
ti
.
ar abham (desire.3SG begin.ACC) He wants
to begin, he wants a beginning the accusative is identiable as such. In Classi-
cal Sanskrit, however, the accusative-marked form was the only one to survive. It
extended its scope to cover the functions previously marked by the others, so that
in a sentence like Avasth atu-m, sth an antara-m cintaya (stay-INF place.other-ACC
think.IMP) Think of another place to stay the -m on the verb is hardly to be identi-
ed with the accusative, but rather as a marker of a category of nonnite verb, the
category that came to be called the innitive (Burrow 1959: 3646).
In Indo-European the innitive normally remains a nonnite, subordinate form,
but in Australian languages such forms come to be used as independent nite verbs.
Most examples involve dative-marked forms which come to be used as purposive
forms. The following examples are from Kalkatungu. In (22) there is a dative-
marked noun. In (23) there is a verb plus nominalising sufx -nytya- plus dative
(-aya). This verb is nonnite, but since Kalkatungu is a syntactically ergative lan-
guage, it is the absolutive relation (SP) that remains unexpressed. In (24) a verb
marked by -nytyaaya is used as an independent nite verb.
(22) Yurru ingka-nha natha-aya
man.NOM go-PAST nurse-DAT
The man went for the nurse.
181
(23) Yurru ingka-nha natha-ngku nanyi-nytya-aya
man.NOM go-PAST nurse-ERG see-NM-DAT
The man went so the nurse could have a look at him.
(24) Natha-ngku yurru nanyi-nytyaaya
nurse-ERG man.NOM see-PURP
The nurse is going to have a look at the man.
The use of an originally subordinate verb form as a main verb probably arises from
the ellipsis of semantically weak governing verbs with meanings like go (cf. the
aspectual-like force of go in English Im going to wash).
As mentioned at the beginning of the section, case marking that has acquired
a new function may remain even though inectional case is lost. The innitive in
Latin, for instance, reects a formation consisting of a verb root plus derivational
sufx -s- plus locative i, e.g. *deike-s-i > *dikese > dikere i.e. dicere to speak. In
Latin -re is analysable as the marker of the innitive and it is this -re that survives
in Italian innitive forms like dire.
In some languages there is formal identity of case markers and switch reference
markers. This identity can be seen in the Muskogean, Yuman and Uto-Aztecan
languages (Jacobsen 1983: 151) as well as in Australian languages (Austin 1981b).
The following examples are from the Yuman language Diegue no (Langdon 1970:
1504). In (25) -(v)c marks the subject; in (26) it marks the fact that the unexpressed
subject is the same as the subject of its sister clause and is glossed SS (same subject).
(25) Sin
y
-c
e
cw
e
yu-w-m
e
yip-s
woman-NOM 3SG.sing-DS 1SG & 3SG.hear-ASSERT
I heard the woman sing.
(26) amp n
y
a-ta am-c
e
wu-w-s
1SG.walk when-1SG.be.around-SS 1SG & 3SG.saw-ASSERT
As I was walking, I saw him.
The sufx -(v)m marks direction away from the point of reference and toward
the object as in
e
wa-v
e
m away to the house as well as instrumental and accom-
paniment. It is also used to indicate that a verb has a different subject from that of
a sister clause. This is illustrated in (25) where it is glossed D(ifferent) S(ubject).
There is general agreement that same-subject and different-subject markers derive
from case-marking forms.
Austin has shown that in Australia locative case markers have developed into
markers for different subject in one large area and into markers for same subject
in another large area. Allative markers have also developed into different-subject
182
6.5 Finale
markers. Austin has also demonstrated that the development of same-subject and
different-subject markers covers a continuous area involving languages that are not
all closely related. This suggests that the development is subject to diffusion (Austin
1981b: 32932).
6.5 Finale
Seen in the context of the way language works, case makes sense. In
general, lexical forms have a certain substance and carry some stress. They are
linked in constructions by less substantial function forms or by order. Case is the
system of function forms that provides most of the linking. It marks the orienta-
tion of predicates, i.e. it indicates which lexical item represents which argument,
and it indicates in broad terms the semantic relationship of other dependents to
heads. It frequently consists of marking on dependents in the form of prepositions,
postpositions or sufxes. But arguments are frequently given information and are
themselves represented by less substantial pronominal forms. It is not surprising
that these pronominal forms are likely to absorb any case marking they carry, and to
be absorbed into a predicate to produce what is known as head marking (following
Nichols 1986).
All forms are subject to phonetic erosion. Head marking and dependent marking
may be reduced to a point where they can no longer function. But new case markers
become available to take their place as verbs and nouns are grammaticised through
phonetic reduction and semantic bleaching. In this way case is cyclically renewed.
The workings of case also illustrate the economy with which language works. A
case system is typically small, having no more than half a dozen or so members.
Each case has a broad function or set of functions sufciently explicit for most
purposes but able to be supported by more elaborate alternatives when necessary. It
is economical in that it represents the orientation of a predicate rather than the roles
of its arguments, economical in that it usually marks only one argument (accusative
or ergative) or it tends to reduce accusative and ergative marking in accordance
with expectations associated with animacy and topic-worthiness. As stated at the
beginning, case is a systemfor marking dependent nouns for the type of relationship
they bear to their heads. This is true, but, as we have seen, especially in chapter 5,
it marks both relationships of propositional content and of discourse-pragmatic
viewpoint.
183
NOTES
1 Overview
1 The symbol indicates a high, back, unrounded vowel.
2 Lewis 1967 refers to this case as the absolute case.
3 It also has postpositions that govern the genitive with singular pronouns but the
nominative with other nominals. Such a postposition is gibi like: kim-in gibi
(who-GEN like) like whom, but bu adam gibi (this man.NOM like) like this man.
4 The singular base can be inected for plural and/or for the person and number of the
possessor: adam-lar-m-la (man-PL-1SG.POSS-LOC) with my men. If one takes both
of these types of inection into account, there are 84 word forms. However, since the
case system remains separate from the number system and the possessor system, i.e.
there is no fusion of the case marking and any other marking, one can say there is a
case paradigm of six members rather than a paradigm of 84 members.
5 In Turkish there is a syllable-nal devoicing of stops. Compare kitaba to the book
and kitapta on the book.
6 I use concord to refer to case/number/gender matching between nouns and their de-
pendents and agreement to refer to person/number matching between the verb and
its subject or other complements. This represents majority practice, but some writers
do not distinguish between the two terms. I adopt here the view that the traditional
noun phrase has a noun as its head and that a determiner is a dependent. There is an
alternative view that is gaining currency which takes the determiner to be the head.
For a discussion of this view see Lyons 1999: ch. 8. See also section 4.3.2, especially
the discussion of (21).
7 A nominal in the vocative can take dependents as in Catullus poem addressed
to Cicero (xlix), which begins Disertissime Romuli nepotum (eloquent.most.VOC
Romulus.GEN grandson.GEN.PL) Oh, most eloquent of the descendants of Romulus
where disertissime is a noun in the vocative with a genitive dependent which in turn
has a genitive dependent.
8 Mohawk has no case marking, but there is, or was, an address form (Marianne
Mithun, personal communication).
In Maori bisyllabic names, but not longer names, used as terms of address are
preceded by a preposition e which otherwise marks the demoted subject in the
passive.
9 These secondary postpositions are nominal in origin, hence the use of the genitive
ke. Compare English in the middle of the night, at the bottom of the stairs. See
also section 6.1.
184
Notes to pages 1126
10 With the less-used co-ordinator kuliko all co-ordinands take case marking (Jae Jung
Song, personal communication).
11 I take a particle to be a free form belonging to a closed set. Most particles have
adverbial functions. A particle does not govern a noun; a postposition does.
12 See, for instance, Bresnan and Mchombo 1987. See also section 3.2.2.
13 See note 2 in chapter 5.
14 Other valency-changing derivations are illustrated in section 3.2.1.
15 An alternative is to take expressions such as on top of and in front of to be complex
prepositions.
2 Problems in describing case systems
1 I am using complement for all the phrases that realise arguments including the
subject. In most modern usage the subject is excluded in principle. In some traditional
usage subject and objects tended to be excluded simply because they bore a label for
a particular kind of complement.
2 It is difcult to demonstrate this directly, since place names are not normally modied
directly, rather a common noun is used in apposition as in (1). However, there are
odd examples of place names being modied by adjectives referring to places, e.g.
Suessae Auruncae at Auruncan Suessa (Woodcock 1959: 36).
3 The use of pre-theoretical syntactic entities derives originally from Dixons descrip-
tion of Dyirbal (Dixon 1972: xxii) and they have proved particularly useful in com-
paring case marking within languages and across languages. Dixon uses S, A and
O; others, including Comrie, prefer S, A and P (see, for instance, Comrie 1978). P is
more logical in that it bears a relationship to patient that is analogous to that borne
by A to agent. It needs to be remembered that P covers the patient of a transitive verb
and extends to other roles that are treated the same way grammatically. It follows
from this that it does not cover the patient in I got hit by a car, since the patient in
this construction is not treated the same way as the patient argument of a transitive
verb; nor does it cover the patient in constructions where another role is advanced
and treated as P while the patient is demoted to an adpositional phrase or peripheral
case. See, for instance, (3) in chapter 3 where the role of instrument is treated as P
and the patient appears in the dative. For a denition of transitive verb, see the Guide
to terminology.
4 My preference is to use nominative for the case that is used to encode S irrespective
of whether this case covers S + A or S + P or indeed whether it is exclusive to S or
covers S + A + P. This case will normally be unmarked and will be the case used
in isolation from constructions. I will reserve the term absolutive for a grammatical
relation that subsumes S and P.
5 Some Pama-Nyungan languages mark case only on the nal word of a phrase (phrase-
marking languages), others mark case on all nominals (word-marking languages).
In some word-marking languages the nominals translating an English noun phrase
tend not to cohere. It may be that there are no noun phrases, only nominals in parallel.
Under this interpretation there would be no concord in the sense of dependent-to-head
concord.
6 Where nominals appear in parallel as in (3), it is likely that only one bears the
grammatical relation and that the other is an adjunct. In (3) the pronoun would be
185
Notes to pages 2651
a likely candidate for direct object since it can be replaced by an object clitic. The
construction is described in semantic terms (patient of a transitive verb) to leave
the syntactic interpretation open.
7 For a discussion of the differences between Australianist practice and traditional
practice, see Goddard 1982, Blake 1985.
8 Wierzbicka (1983: 249ff) denies that there is a partitive case and takes the -u to be
the genitive case marker used with certain nouns to indicate uncountable stuff , i.e.
it is a case marker that cumulatively represents another semantic notion.
9 The nominative is the morphologically unmarked case and is referred to as the
absolute case in some grammars of Turkish. See also note 2, chapter 1.
10 Compare (11) in chapter 1.
11 The Roman writer Varro took the nominative to be the canonical word form from
which the oblique cases could be developed (Robins 1967:52), though technically
this is not always true of Latin (nor of the other Indo-European case languages). For
instance, with a stem in -g, there will be assimilation of g to the voiceless -s in the
nominative so that the word for ock has nominative grex (/greks/), genitive gregis.
However, since there are stems in /-k/, there are words like d ux (/du:ks/) nominative,
d ucis (/du:kis/) genitive. This means that one cannot predict the genitive or any
oblique form from the nominative. For this reason dictionaries give the genitive as
well as the nominative since the genitive shows the oblique stem.
12 In Latin, many of the verbs that take a peripheral- or indirect-marked complement are
deponent verbs; in fact, all the verbs that take the ablative rather than the accusative
are deponent: fung to busy oneself with, to perform(a duty), fru to enjoy (oneself
in), ut to use, to avail oneself of , potr to have possession of and vesc to use
as food, feed upon. Deponent verbs are described as passive in form but active in
meaning. Historically they derive from reexives and in some of them the sense of a
patient coreferent with the subject is still apparent. It is because of this patient that
the overt complement has come to be represented peripherally.
Potr can also govern the genitive and occasionally the accusative where it tends
to mean get control over as opposed to be in control of (de Carvalho 1982:263ff).
13 See the discussion in section 5.2.6.
14 Examples include:
acc ad bene vvendum for living well
gen ars scrbend the art of writing
dat p ar disserend o equal to arguing
abl d e bene vvend o about living well
15 The dative is not taken to be local in this system even though the dative could be said
to encode an abstract destination (compare the systems of Simon (Table 2.8), Martin
(Table 2.9) and Jakobson (Table 2.13)). In some languages the dative rather than the
accusative expresses destination and there is no question that the dative would have
to be taken to be [+ local].
3 Modern approaches to case
1 The antipassive inKalkatungualsohas a syntactic function. See(50)(51) inchapter 4.
2 Bresnan and Mchombo (1987) allowfor pronominal marking on the verb to alternate
between agreement (where a cross-referenced NPis present) or representation (where
no cross-referenced NP is present).
186
Notes to pages 5463
3 Some linguists use the terms subjective case andobjective case rather thannominative
and accusative. This is not well-motivated terminology when one considers that cases
are multifunctional.
4 For the original version of the model see Chomsky 1981. Case theory is described
in various sources including Chomsky 1985, van Riemsdijk and Williams 1986,
Baker 1988 and Haegeman 1994. See Sch utze 1993 for an analysis illustrating the
difference between structural and morphological case. See Roberts 1997: ch. 2, for
an account of recent work on case in the Chomskian framework.
5 An example of structural case assignment in LexicalFunctional Grammar was pre-
sented in section 2.3.4.
6 The term predicate is used in two different but overlapping senses. In traditional
grammar the predicate is opposed to the subject and covers the verb, its complements
and its adjuncts, less the subject of course, and less any constituents moved out of
the predicate phrase, typically to the front of the sentence. In logic predicate is op-
posed to argument where predicate is a term expressing a property of an entity or a
relation between entities. The entities are the arguments of the predicate. Apredicate
can be simple or complex. Where a predicate takes in various complements, as in
the traditional predicate versus subject, it is a complex one-place predicate with the
subject as its argument.
7 Yip, Maling and Jackendoff (1987) present a theory to account for case assignment
in accusative and ergative languages. In their theory (Case in tiers) syntactic case
is assigned directionally, rst the unmarked case, then the marked case (accusative,
ergative). In an accusative language the direction of assignment is left-to-right, in an
ergative language right-to-left. Bythis methodof assignment the unmarkedcase is ap-
plied to S or Ain an accusative language and to S and P in an ergative language. Con-
sider the following examples from West Greenlandic and their English translations.
(i) Kaali pisuppoq
Karl walks
NOM
(ii) Kaali-p (ERG) Hansi takuaa
ERG NOM R-to-L
Karl sees Hans
NOM ACC L-to-R
However, this system makes no distinction between syntactic ergativity and mor-
phological ergativity; it makes certain assumptions about word order that do not apply
to all ergative languages; it cannot handle pragmatic as opposed to grammatical word
order; it cannot handle languages that use ergative and accusative case marking on
the basis of lexical content or aspect (see section 5.2.4), and it cannot handle active
systems (see section 5.2.3).
8 Stowell (1981) proposes a Case Resistance Principle to the effect that those word
classes that assign case cannot receive it.
9 In a more widely accepted analysis him would be taken to have been raised into the
higher clause and to be the direct object of the higher clause. See also note 19.
10 Sources for LexicalFunctional Grammar include Bresnan (ed.) 1982, Bresnan and
Kanerva 1989 and Neidle 1988 which contains a neat outline of the theory.
187
Notes to pages 6785
11 Fillmore treated the effected patient (She built a sand castle) as factitive (1968) and
later as goal (1971).
12 But see section 3.5.4 on role hierarchies. Some would claim that roles are relevant
to the grammaticality of reexives and passives.
13 See (8) and (10) above where a whole and a part share a role.
14 See discussion in Jackendoff 1972: 346; also Somers 1987.
15 The gerundive is a passive-like adjective derived froma verb. It has the sense of Xis
requiring to be Verb-ed. There are fossilised Latin gerundives in English: Amanda
female requiring to be loved, memoranda things requiring to be remembered,
agenda things requiring to be done.
16 For a similar interpretation of analogous pairs of sentences see Dik 1978: 99f.
17 Sources include Perlmutter (ed.) 1983, Perlmutter and Rosen (eds.) 1984 and Blake
1990.
18 In a sentence like John took himself off to the pictures the [erg] and [abs] are repre-
sented separately.
19 In the sentence The child seems to hear the noun phrase the child belongs seman-
tically with the verb in the nonnite, dependent clause. We can demonstrate this by
substituting an inanimate noun and reecting on the source of the semantic incon-
gruity: The stone seems to hear. We can also compare It seems that the child/the
stone hears. One way to analyse a sentence like The child seems to hear and to relate
it to It seems that the child hears is to posit an underlying structure seem [child
hear]. The sentence It seems that the child hears can be derived directly simply by
meeting certain grammatical requirements (adding it, that and verb agreement), and
the sentence The child seems to hear can be derived by raising the child to become
the subject of seem. In the sentence The doctor expected the child to recover the
logical object of expect is the proposition [child recover] (compare The doctor ex-
pected that the child would recover), but the noun phrase the child is grammatically
the object of expect, at least in the most widespread interpretation, though not in the
Government and Binding interpretation (see (19) in this chapter). Under this inter-
pretation the child is described as having been raised into the higher clause (Postal
1974).
It is common to talk about subject-to-subject raising, but in fact there is never
raising to A, only to S. In Relational Grammar raising verbs like seem and appear
are taken as having an initial 2, so the appropriate generalisation is that the target of
raising is always a 2. This covers the seemtype as in (53) and the expect type as in (54).
20 The missing subject of the innitival adjunct is not grammatically controlled. If there
is an agent in the governing clause, the missing subject will be interpreted as being
coreferential with the agent, but it is possible to construct examples that show that
the missing subject is not necessarily coreferent with a particular role or relation in
the governing clause:
Sand was spread to prevent slipping
It was sanded to prevent slipping
It rained to complete the miserable scene
The following example was actually uttered with reference to the manuscript of this
book:
Chapter one was moved across there to look up something
188
Notes to pages 8894
A few verbs like promise are exceptions to the rule that the PATIENT controls the
missing subject of an innitival complement, at least in varieties of English in which
one can say He promised me to mow the lawn. Here the AGENT appears to control
the missing subject, but Starosta, following Jackendoff 1972, takes promise to be
intransitive. Independent motivation would appear to be weak.
21 In some languages such as Dyirbal and Kalkatungu only the absolutive can be rela-
tivised. See (11), (12) and (13) in this chapter.
22 One could take the s to be derivational marking the derivation of a determiner;
compare his dog and the man over theres dog. However, this would mean taking the
whole noun phrase as a word.
23 For further examples of various patterns of cross-referencing see sections 5.2.2
and 5.4.
24 Double-object constructions are common across languages. It is interesting to note
that verbs for GIVE are practically never marked for what would be, in Relational
Grammar terms, an advancement of indirect object to direct object. It would appear
that though the double object is a marked construction, it is unmarked or natural GIVE.
25 For a discussion of the felicity of passives see Wierzbicka 1980: 4969.
4 Distribution of case marking
1 In Classical Arabic to express X is Y one uses a nominal predicate that shows
concord with the subject. However, to express X was Y one uses a transitive verb
kaana, which takes an accusative complement.
(i) Muh
.
ammad-u kabiir-un
Muh
.
ammad-NOM old-NOM
Mohammed is old.
(ii) kaana muh
.
ammad-u kabiir-an
was Muh
.
ammad-NOM old-ACC
Mohammed was old.
2 In traditional descriptions of European languages the verb is normally said to agree
with the subject, but in languages with cross-referencing pronominal representation
in the verb it makes more sense to characterise the verb as requiring certain categories
of noun phrase to match the representation on the verb. In Latin, for instance, a
verb form like v enit S/he has come is marked as including a third-person-singular
subject. It can be characterised as expecting a third-person-singular noun phrase in
the nominative with which it can be construed. This expectation need not be met, but
the so-called agreement can be described in terms of the head (the verb) controlling
the dependent (the subject noun phrase).
The concord between a predicative noun or adjective and a subject would nor-
mally be described as concord of the predicative word with the subject, since it
typically involves inherent features of the subject being marked on the predicate. In
the Latin sentence Maria acuta Maria is sharp the inherent properties of Maria
(feminine singular) show up on the predicative adjective acuta.
3 One analysis that would take [Cicer onem c onsulem] to be a constituent of the
higher clause is the small clause analysis of Chomskys Government and Binding
189
Notes to pages 98110
Theory (Haegeman 1994: 1236). For a description of a raising analysis see note 19
in chapter 3. For control of the missing subject of nonnite clauses see section
3.4.5.
4 With pronouns there are genitive forms that are used in objective function but not
subjective function. These forms (me, tu, nostr and vestr) resolve some ambigu-
ities.
5 For Gooniyandi see McGregor 1990.
6 Pitta-Pitta, which is fairly closely related to Diyari, also has determiners morpholog-
ically identical to third person pronouns. See (20) to (25) in chapter 5.
7 Compare the Old English case inections in Table 6.3.
8 See Cutler, Hawkins and Gilligan 1985. Frachtenberg (1922: 324f, 462f, 570) re-
ported a discriminative case prex in Coos and Siuslaw, possibly an ergative case
marker, and a modal or instrumental prex in Coos. The case status and the formal
status (prex versus proclitic) is not entirely clear. Both these languages have cases
marked by sufx. See also Sherzer 1976: 69f, 261f.
9 Prexed class/case marking may arise from the use of free-form generic nouns pre-
posed to more specic nouns, a practice not uncommon in Australian languages.
The generic noun takes case marking in parallel with the more specic noun. It pre-
sumably takes less stress than the specic noun and undergoes phonological attrition
and absorption into the specic noun. In some languages there is transparent case
marking on the class prex. The following paradigm is from Mangarayi (Merlan
1982: 57). Gardugu is woman and nga- is the feminine class marker. The form
nga- is cognate with the nga- of nganyi-, the class/case oblique prex for class 2
(feminine) nouns in Nungali.
nom ngarla-gardugu
acc ngarlan-gardugu
dat ngaya-gardugu
loc ngaya-gardugu-yan
all ngaya-gardugu-rlama
abl ngaya-gardugu-wana
10 An example of number marking preceding case marking in Turkish is given in (1)
in chapter 1. A paradigm showing case and number marking in Archi is given in
Table 4.3.
11 See Comrie 1980 for a discussion of how this order arose.
12 The sufx -rla, glossed as indirect object, indicates that the object enclitic is to be
interpreted as encoding an indirect object. There is no overt form, however, for third
person singular object.
13 Normally modal case covers the whole of the verb phrase including the verb itself. It
is not proposed to discuss the absence of the modal case from the verb in (34), since
this would take us too far aeld.
14 The expression of multiple case need not be agglutinative. Dench and Evans 1988
give an inventory of sequence conditions and illustrate portmanteau case markers,
case haplology (one marker where two identical markers are expected), the use of
empty morphs to separate case markers, etc.
15 See Silverstein 1976: 163 for a more elaborate hierarchy of clause types.
190
16 In a syntactically ergative language it is SP, the absolutive, that remains unexpressed
with nonnite verbs. See (23) in chapter 6.
For a nonnite verb with a subject see (20) in chapter 3 and (46) in this chapter.
17 See section 3.4.5.
18 See note 19 in chapter 3.
19 Possessive adjectives like his clearly belong to the class of determiners, but the analy-
sis of s-phrases is controversial. I amtaking s to mark the derivation of a determiner
from a noun phrase.
20 Another example of the Latin accusative-and-innitive construction is given in (20)
in chapter 3. In connection with an analogous example in English (19) it was pointed
out that under a popular analysis the noun phrase that expresses the subject of the
innitive is interpreted as having been raised to become the direct object of the
governing verb. See also note 19 in chapter 3.
21 See the discussion of Hopper and Thompson in section 5.2.6.
22 For a discussion of an analysis that takes subordinating conjunctions to be preposi-
tions in English see Huddleston 1984: 33841, 390.
5 Survey of case marking
1 The examples are from Comrie (1989: 1326) which contains a useful discussion of
the labels denite and specic.
2 As mentioned in section 1.3.1 it is sometimes possible to isolate an apparent case
marker in a bound pronoun system. A number of Northern Australian languages are
like Rembarnga in having rst-person-singular subject nga and rst-person-singular
object ngan. However, it is not really appropriate to call -n an accusative case marker,
since it occurs in a system that marks particular grammatical relations. It is better
to describe the bound pronoun system in terms of subject forms and object forms.
Where a marker can be isolated, as with -n, it can be called an object marker. All this
of course assumes that the bound pronouns are taken to be clitics. If they are analysed
as person/number/relation inection, it would be quite inappropriate to refer to case
marking.
3 For languages of the Amazon basin see Derbyshire and Pullum 1986, 1989, 1990,
1991, and Dixon and Aikhenvald 1999. For information on the distribution of ergative
languages see Dixon 1994.
4 The systemreported here is not peculiar to Avar. Asimilar lack of conation between
agent, experiencer and perceiver is found in other Northeast Caucasian languages.
5 Fore has both ergative and accusative marking and therefore falls within the type
described in section 5.2.4. It is mentioned here since it provides a clear example of
ergative marking determined by the relative position of A and P on the hierarchy.
The accusative marking in this language appears as the rst formative in the verb
induced by a preceding P.
6 See also Table 2.4 Core case marking in Pama-Nyungan.
7 See (61) and (62) for Abaza examples.
8 See section 5.4 on the genitive.
9 See (53) for an example.
10 See also the case-marking schema in Pama-Nyungan illustrated in Table 2.4.
191
11 See (47) below.
12 A language like Thai with SV, AVP word order can be described as accusative, but
there is no case.
13 De Carvalho, describing the dative independently of the data from Pitta-Pitta and
other languages that use it in the future, characterises it thus, Le datif est, en quelque
sorte, lhomologue du futur (de Carvalho 1983: 64).
14 The original source is Wolfart 1973. These examples have been taken from Foley
and Van Valin 1984: 333ff.
15 This hierarchy is adapted from the ndings of DuBois 1987. DuBois links SP/A
alignment in grammar with the universal tendency for new information to appear
preferentially in S or P function, at least within the core. See also Giv on 1976 for
various hierarchies of topicality and their signicance for agreement patterns.
16 For Dyirbal see Dixon 1972: 249. For general discussion see Durie 1986.
17 Dixon(1994: 202f) gives anexample of a language withanS/APsysteminthe present
tense. It is Rushan, a Pamir (Iranian) language. The Pamir languages typically have
an oblique case marker just for P in the present but just for A in the past. In Rushan
the oblique has come to be used for P in the past to produce an S/AP opposition.
18 Following Silverstein we could give a more conservative version of the hierarchy in
which 1 and 2 share the rst position. In a few languages 2 outranks 1 but usually
in the matter of priority of representation. In many languages one nds that in clitic
combinations involving 1 and 2, only one person can be directly represented. Against
the few languages in which there is evidence that 2 outranks 1 there are many more
affording evidence that 1 outranks 2.
19 In the Australian language Mangarayi the ergative is distinguished only with inan-
imates. With higher categories there is a nominativeaccusative system. For a
paradigm of feminine nouns see note 9, chapter 4.
20 For Pintupi see Hansen and Hansen 1978. For Dargwa see Wierzbicka 1981. The
other sources for sections (c) and (d) are in Mallinson and Blake 1981: 889.
21 Another interpretation of split ergativity not dissimilar to the views presented here
can be found in Delancey 1981.
22 For examples see (42) in chapter 4 and (25) and (26) in chapter 6.
23 See also section 3.2.
24 See also Croft 1991: 171.
25 The reader might wonder why (53b) is considered to be an accusative construc-
tion with a direct object in the dative rather than an intransitive construction. The
grammatical relations can be established from the bound pronouns on the verb. Note
too that if samajurs in (53b) were an indirect object, there would be a problem in
accounting for the grammatical relation of sen. In Harriss Relational Grammar anal-
ysis of (53c) the dative-marked constituent is an initial subject that has been demoted
to indirect object pushing the initial indirect object ( sen) into ch omage where it is
marked by the benefactive ( sent vi s). The fact that two datives are allowed in (53b)
is consistent with samajurs being direct object.
26 Una compagna dufcio is subject by the criteria of (a) control of agreement and (b)
case (the distinctively nominative pronoun lei can be substituted for una compagna
dufcio). A Giorgio has one subject property, namely control of the missing sub-
ject of various types of innitival complement. In a multistratal analysis of the type
used in Relational Grammar, different participants can be subject in different strata.
192
Giorgio would be taken to be the initial stratum subject and the nal stratum indirect
object. Una compagna dufcio would be the initial direct object. However, it is not
taken to be the nal stratumsubject despite the fact that it controls the verb agreement
and appears in the nominative case. The construction is taken to be impersonal (see
Blake 1990: 84 for references). The statement in the text that una compagna dufcio
is the subject is based on the conventional analysis. See the discussion of inversion
in section 3.4.3.
27 There is a common case for A and possessor in Kabardian and Ubykh (Northwest
Caucasian), Lak (Northeast Caucasian) and Laz (South Caucasian).
28 Rumanian is said to have a two-case system (plus a vocative). One case, which
Mallinson calls the nominativeaccusative, covers subject and object, and the other,
which Mallinson calls genitivedative, covers the noncore functions. There are sep-
arate pronoun forms, both clitic and free, for subject, object and indirect object, but
free object pronouns are governed by a preposition pe, which also governs nouns in
object function that are high on the animacy scale (Mallinson 1988).
noun pronoun clitic pronoun
star rst person singular
subject stea eu eu
object stea pe mine m a
indirect object stele mie mi
Old French is usually described as having a nominativeoblique case system,
but masculine demonstratives show a three-way nominativeaccusativedative dis-
tinction (cil/cel/celui that, cisti/cest/cestui this).
Clitic pronouns in Old French, as in the Romance languages generally, have
three series of forms, one for subject, one for object and one for indirect object; the
direct and indirect object forms are distinct only in the third person.
29 Croft (1991: 237) gives a useful list of roles that are subsumed under a single form
using a forty-language sample.
30 The source for Yidiny is Dixon 1977, which contains a useful discussion of the
classication of cases (124ff).
6 Life cycle of case systems
1 As noted in section 4.3.2 prepositions seem not to develop into prexes. They can,
however, fuse with determiners as in French au (=` a le to the), German zur (=zu
der to the.DAT).
2 See also section 4.3.2.
3 Sources include Giv` on 1975: 93, Norman 1988: 163, Lord 1982, Heine et al. 1991.
4 Compare the local forms discussed in section 5.6, especially Table 5.4.
5 The use of the transitive verb can express a higher degree of transitivity than the
intransitive verbplus preposition. Perambul are aedes canmeantoinspect the house.
6 Sands 1996 posits a proto-Australian form*-THUfromwhich gu derives via Hales
hypothesis, and she shows that there are plausible reexes of *-THU scattered over
the continent. She takes -lu to reect a proto-Pama-Nyungan *-lu, a formof unknown
provenience.
7 See also the Margany example given in Table 2.7.
193
8 In Early Latin the dative singular of the o-stems was oi which developed into - o. The
ablative was - od, but word-nal d was lost to produce the dativeablative syncretism
shown in Table 1.2. In the singular of the a-stems there was an inherited genitive
ablative syncretism with the form -s. However, the genitive -s was replaced by the -i
of the second declension and a new ablative in - ad was formed on the model of the
second declension singular (Palmer 1954: 233ff).
rst declension second declension
gen as a > ae o >
dat ai > ae oi > o
abl as ad > a od > o
9 Although subjectverbobject is normal in transitive clauses, verbsubject is normal
in intransitive clauses in some Romance languages (not French) with verbs of motion,
location, state and change of state.
10 See section 1.3.4.
11 An interesting example of reinterpretation of case marking as part of the stemis to be
found among some central Australian languages. The example is interesting because
it involves case marking that is of purely phonological provenance. Among the Pama-
Nyungan languages nouns in SP function are generally unmarked (see Table 2.5), but
in some languages there is a phonotactic requirement that all words end in a vowel.
Some of these languages have augmented consonant-nal stems with a syllable
-pa (alternatively -ba, there not being any voiced/voiceless opposition). In the table
below, the left-hand column represents a language that allows consonant-nal stems
and the middle column illustrates the augmentation. Nowin some languages the case
marker -pa is reinterpreted as part of the stem to produce the paradigm shown in the
right-hand column of the following table. The root illustrated means tongue (Hale
1973a, Dixon 1980: 209f).
Luritja Pitjantjatjara Warlpiri
SP t
y
arlin
y
t
y
arlin
y
pa t
y
alan
y
pa
A t
y
arlin
y
t
y
u t
y
arlin
y
t
y
u t
y
alan
y
parlu
dative t
y
arlint
y
ku t
y
arlin
y
ku t
y
alan
y
paku
194
GUI DE TO TERMI NOLOGY
The following is a guide to the terminology used in this text. It covers the terms
relevant tocase, grammatical relations andsome relatedconcepts suchas transitivity.
In general it does not contain the terms that appear in one section only, nor does
it contain labels for semantic roles, which are listed in section 3.4. Terms in bold
within an entry appear in the guide as headwords.
A. The agent argument of a transitive verb plus any other role that is treated in the same way
grammatically. See also S and P. See section 2.2.3.
abessive. This label has been applied to cases such as the Finnish abessive which means
lacking: rahta-tta money-less. It is based on Latin ab-esse to be away from, i.e. to be
absent.
ablative case. The case that expresses the role of source, which is expressed by from in
English. See also elative case.
absolutive. In this book absolutive is used for the grammatical relation embracing S and
P. In some sources the term absolutive is used for the case that expresses S and P.
accusative case. The case in which the direct object is expressed. In some works this case
is referred to as the objective case.
accusative language. Alanguage in which morphosyntactic rules identify Awith S opposing
SA (the subject) to P (the object).
accusative system. A system of inection, adpositions, clitics, etc. that identies S with A
opposing SA to P.
active language. A language in which the core grammar is organised on a basis that could
be described roughly as agent versus patient irrespective of transitivity. The agent-like
argument of a one-place verb is treated like the A of a transitive verb and the patient-like
argument of a one-place verb is treated like the P of a transitive verb. Other terms include
split-S and split intransitive. See section 5.2.3.
active voice. The active is opposed to the passive in the voice system of the verb. In the active
voice A is the subject of a transitive verb. In the passive, which is a derived intransitive,
P is advanced to become S, and A is demoted to adjunct status.
adessive. A case expressing at or near.
195
Guide to terminology
adjunct. Adependent that is not a complement, i.e. a dependent not representing an argument
and so not implied by the governing predicate. An adjunct is always optional as with the
prepositional phrase in He did it with ease.
adnominal. An adnominal dependent is a dependent of a nominal.
adposition. Cover term for preposition and postposition.
advance, advancement (promotion). The termadvancement was popularisedbyRelational
Grammar and is now in general use. It refers to the promotion of a participant up the
hierarchy of grammatical relations by such means as using a derived verb. The passive
can be described in terms of the advancement to subject of the participant holding the
direct object relation.
adverbal. An adverbal dependent is a dependent of a verb.
adverbial. An adverbial dependent is one that has a function frequently expressed by an
adverb, e.g. manner, time, location. It can be applied to words, phrases or clauses. Applied
to cases it can cover any noncore case or even an adverbial function of a core case such
as the adverbial accusative in Latin: n udae lacert os (nude.NOM.PL arm.ACC.PL) bare as to
the arms.
afx. Cover term for prex, sufx and inx.
agreement. The marking of the person, number and sometimes gender or class of arguments
on the verb is usually referred to as agreement. The term implies agreement in person,
number and gender or class between the marking on the verb for a particular argument
and any noun phrase representing the same argument. This marking is always organised on
the basis of grammatical relations and it is the subject whose properties are most often
represented. In some languages the person and number of a possessor in a noun phrase
is marked on the head noun, which encodes the possessed (section 5.4). Agreement is
almost always of a type that has been called cross-referencing. In this type the pronominal
marking can represent an argument and a noun phrase representing the subject or possessor
or whatever relation is cross-referenced can be omitted. In a few languages such as English
there is non-cross-referencing agreement where the noun phrase controlling the agreement
cannot be omitted (compare He runs with They run where the sibilant inection in runs
agrees with a third-person-singular subject that cannot be omitted).
allative. A case expressing to.
analytic case marker. Adpositions are analytic in the sense that they are free forms and they
are case markers inthat theyserve tomarkthe relationthat their dependent nounbears tothe
governor of the adposition. In the sentence He put it in the drawer the preposition in marks
the relation between put and drawer. A relator noun can also be an analytic case marker.
analytic language. The perfect example of an analytic language would be one in which each
morpheme was a lexeme, i.e. one where there were no bound morphemes. No such language
exists, but languages like Chinese and Vietnamese have relatively few bound morphemes.
antipassive. A detransitivised derived verb in an ergative language. In the antipassive con-
struction the A of the transitive verb becomes S and P is expressed as a noncore relation.
See also sections 3.2, 4.5 and 5.3.
In Relational Grammar the term antipassive is not conned to ergative languages, but
a distinction is made according to the fate of the demoted P. If P is demoted to indirect
196
object, this is called direct-to-indirect object retreat. If P is not demoted to indirect ob-
ject, then it is considered to have been demoted to ch omeur status and the construction is
referred to as antipassive.
applicative. An applicative construction is one in which a participant is advanced to assume
the grammatical relation held by the patient. Typically this is the direct object relation, but
it can be the absolutive in an ergative system or the subject in a language like Balinese
where the patient is encoded as the subject of an unmarked verb.
argument. See predicate.
aversive. This label is sometimes used for a case indicating what is to be avoided. Another
term that has been used is evitative.
bound pronouns. In this text the term bound pronouns refers to any kind of pronominal
representation on the verb whether by forms analysable as clitic pronouns or by inection.
In a few languages bound pronouns are enclitic to the rst constituent of the clause.
case. 1. A system of marking dependent nouns for the type of relationship they bear to their
heads. Originally applied to inectional systems, but sometimes used of other systems such
as systems of postpositions.
2. A member of such a system, e.g. the dative case in Turkish.
case form. Where a lexeme can bear case the word forms expressing the lexeme in the various
cases will be case forms. In Latin manum is the accusative case form of the lexeme for
hand (Table 1.2).
case grammar. Fillmores 1968 conception of a generative grammar was one in which
semantic roles or deepcases were basic. This approachtogrammar andsimilar approaches
are referred to collectively as case grammar.
case language. A language with an inectional case system.
case marker. An afx that marks a case. In Turkish -da in Istanbul-da is the case marker
for the ablative case.
case relation. This term is sometimes used as an alternative to grammatical relation. In
this text it is used for the syntacticsemantic relations found in Andersons Localist Case
Grammar (section 3.4.4) and Starostas Lexicase (section 3.4.5).
case system. The set of cases in a language.
ch omeur. In Relational Grammar a participant holding a core relation may be demoted to
ch omeur status. In the passive the subject of the active is said to be demoted to hold the
ch omeur relation where it is realised as a peripheral adjunct. See section 3.4.3.
clitic. Aformthat is treated as a separate element in the syntax, i.e. as a word, but as part of an
adjacent word in the phonology, i.e. as an afx. A form that attaches to the following word
is a proclitic as in French Jarrive I arrive, and a form that attaches to a preceding word is
an enclitic as in Italian comprandolo buying it where lo is the enclitic. See section 1.2.5.
comitative. A case expressing with whom an entity is located. Usually used of animates. In
the sentence The dog is with his master the phrase his master would be encoded in the
comitative in a language with such a case.
comment. See topic.
complement. In this book a complement is taken to be a dependent representing an argument.
This means that the nite verb in a language like English is described as having a subject
197
as a complement. The direct object, indirect object and some prepositional phrases are
also complements. In the following examples the complements are in bold:
She sent a cheque to the missionary.
She sent the missionary a cheque.
She put the money in the bank.
He was ready for a try at the title.
A student of the classics.
In some traditional treatments the subject, the direct object and the indirect object were
excluded simply because they merited more particular labels. This left complement as a
term for little more than the nonsubject complement of verbs like be and become (She is
a swimmer. She became a champion). In the Chomskian tradition subject is excluded on
principle and this usage is dominant in contemporary linguistics.
See also adjunct.
compound case marking. Where a case is marked by more than one segment it can be
described as having compound case marking. See Tables 2.14 and 5.4, and section 4.4.3.
concord. Where a determiner or an attributive adjective displays marking for categories
of the head noun such as case, number or gender, it is said to exhibit concord with the
head noun (section 1.2.1). There may also be concord between a predicative adjective
and a controlling nominal (section 4.2). This phenomenon is sometimes referred to as
agreement.
control, controller. The terms control and controller are useful in two contexts. Where
there is concord or agreement the word that provides the source may be referred to as the
controller. In the following examples the controller is in small capitals and the controllee
in bold:
This BOOK is on the table.
These BOOKS are on the table.
The terms control and controller are also used with reference to pronouns and to unex-
pressed subjects. In the following example the noun in small capitals controls the pronoun
in bold and the covert subject represented by [ ].
TOM saw himself in the mirror and [ ] began to [ ] worry.
In contemporary formal grammar Tom would be said to bind the (reexive) anaphor
himself.
core. The core grammatical relations cover various organisations of S, A and P including
subject, direct object or absolutive and ergative. Where a three-place verb takes two
P-like relations (as in English She gave me the apples), both these relations are included.
core case. One of the cases used to encode the core relations: nominative, ergative, ac-
cusative. See also absolutive.
cross-referencing, cross-referencing agreement. See agreement.
cumulation, cumulative exponence. Where two or more grammatical categories have fused
representation throughout the language, the categories can be said to have cumulative
198
exponence (Matthews 1974/1991). In Latin number marking and case marking can never
be separated (Table 1.2).
dative case. The case that encodes the indirect object.
declension. Declension is inection for case, or case and number. Sometimes the set of word
forms expressing case or case and number is described as a declension; more often the
term is used to refer to a class of words that inect in the same or a similar way. Latin
nouns that inect like manus in Table 1.2 are said to belong to the fourth declension.
default. The default value or setting is the one that will be in force if no value or set-
ting is specially assigned. The term has been popularised recently in the wake of the
widespread use of computers. It is very similar in meaning to unmarked as opposed to
marked.
denite. An entity that is specic and which the speaker assumes can be identied by the
hearer is referred to as denite. In English a common noun may be marked as denite by
the use of the denite article (the), a demonstrative (this, that), or a possessive determiner
(my book).
dependent. See head.
dependent marking. See head marking.
derivation. Where one lexeme is made up on the basis of another, it is said to be derived
from that lexeme. For example, foolish is derived from fool, and foolishness is derived
from foolish. In this book derivation is used mostly with reference to the forming of a verb
with a new valency.
detransitivise. To derive an intransitive verb from a transitive one.
direct case. In a number of Indo-Aryan languages the unmarked case covers S, A and P and
is often referred to as the direct case.
directinverse. The terms direct and inverse refer to the hierarchy of person. In some
languages the speech-act participants (rst and second person) are given some kind of
priority over third persons or rst person may be given priority over second as well. In some
languages there are also distinctions within third person between a more topical person
(proximate) and a less topical person (obviative). If A is higher than P on the hierarchy, the
combination is direct. If A is lower than P on the hierarchy, the combination is inverse.
See also person.
direct object. The grammatical relation that covers P and does not cover S or any part of S.
In this book the recipient or beneciary object is generally taken to be the direct object
in double-object constructions like She gave him bread. She bought him a scarf. This is
controversial in English but clearly true of some other languages. See sections 3.2.2 and
3.4.3.
double-object construction. A construction like She gave the beggar some money where
two noun phrases have pretensions to being considered direct object or where one noun
phrase can be identied as direct object and where there is a second noun phrase that is
marked in the same way. See also direct object, secondary object.
elative. A case expressing out of .
enclitic. See clitic.
ergative. 1. A case encoding A but neither S nor P.
199
2. A grammatical relation corresponding to A. In this text the term A is applicable to
any language, but the term ergative relation is considered appropriate only where there
is morphological or syntactic evidence to demarcate A.
3. A system of core grammar that identies S and P as opposed to A. The term ergative
can be applied to case marking, cross-referencing agreement or any other system. The term
ergative syntax is applied to syntactic rules that identify S and P, e.g. a requirement that
only the absolute relation (SP) can be relativised. The term ergative may also be used
with reference to a language that has an ergative system.
essive. A case indicating location. The more common term for such a case is locative.
evitative. See aversive.
extension. See intension.
nite, nonnite. In the classical languages a verb that was marked for the person and number
of its subject and which could take a noun phrase as subject was described as nite. Forms
such as the innitive and the participle took no person/number marking for the subject and
normally took no noun phrase as subject, at least not a nominative noun phrase (see (46) in
chapter 4). These forms of the verb were described as nonnite. In section 6.1.1 the term
nonnite is used of noninitial verbs in serial verb constructions. These verbs cannot take a
noun phrase as subject.
function. In a grammatical system each member can be said to have a function. The dative
case in some languages may be said to have the function of expressing purpose and the
function of expressing the demoted object in a derived antipassive construction. In this text
function is used informally where it is not clear exactly what grammatical relations are
involved. It may be that on a consideration of more evidence it would be more appropriate to
subsume the purpose function and demoted object function under the grammatical relation
of indirect object.
genitive case. The case that encodes the adnominal relation that subsumes the role of pos-
sessor.
Gesamtbedeutung. Cases have been described as having a generalised meaning (German
Gesamtbedeutung) as well as various particular meanings (Sonderbedeutungen). Among
these particular meanings it is common to be able to nd a principal meaning or Hauptbe-
deutung.
given (information). At any point in discourse the speaker will make certain assumptions
about what the hearer knows and treat this as given information as opposed to new in-
formation. Given information is typically represented in terms of pro-forms, i.e. pronouns,
pro-verbs and pro-adverbs: I asked Bill to leave and he did so immediately. The speech-
act participants are given by the very act of communication, though they may represent
new information in certain predications.
govern. Where the head of a construction determines the presence of a dependent or its
presence and form, it is said to govern the dependent. In some languages a transitive verb
may be said to govern a direct object in the accusative case.
grammatical cases. Cases that encode a purely syntactic (as opposed to semantic) relation.
Also called syntactic cases (see section 2.3.2). The class includes nominative, accusative,
ergative, genitive and dative.
200
grammatical relations. The morphosyntactically determined relations borne by dependents
to their heads. They include purely syntactic relations like subject and semantic relations
such as locative.
head. Most constructions can be described in terms of an obligatory member (the head) and
an optional member (the dependent). In the phrase young woman the word woman is the
head and young the dependent. In this book the verb is taken to be head of the clause and
an adposition to be head of an adpositional phrase.
head marking. Marking grammatical relations on the head of a construction rather than on
the dependent (dependent marking). This almost always involves using cross-referencing
bound pronouns.
hierarchy. A classication involving ranking into successively subordinate levels, e.g. a
person hierarchy: 1st > 2nd > 3rd.
holistic interpretation. This term is used in linguistics to mean wholly affected in the
context of comparing the expression of a role as P or as a noncore relation. Compare He
read the book (holistic) and He read from the book (partial).
hyponym. Where the meaning of a morpheme is included in that of another, the rst is a
hyponym of the second. Lion is a hyponym of animal and so is tiger. Lion and tiger are
co-hyponyms.
illative. A case meaning into.
indirect object. The indirect object is the major grammatical relation for encoding comple-
ments outside the core. It will typically encode the nonsubject complement of two-place
intransitive verbs or the noncore complement of three-place verbs. It is traditionally de-
scribed as encoding entities that are indirectly affected by the activity or state denoted
by the verb. In a language with a case system of four or more members the associated
case is the dative. In this book the term indirect object is not applied to the recipient or
beneciary object in the double-object construction, but this is a traditional usage that
has support.
indirect subject. Where a noun phrase with morphological properties of an indirect object
(usually the dative case) exhibits syntactic properties characteristic of the subject, it is
sometimes referred to as the indirect subject . See section 5.3, especially (52), (53) and
(55).
inessive. A case meaning inside.
innitive. See nite.
inection. A system for marking grammatical categories on a word, categories such as case,
number, aspect and tense.
instrumental. A case expressing the means by which an activity is carried out.
intension, extension. The intension or the intensional denition is the set of dening proper-
ties, whereas the extension is the set of entities to which a term may refer. The intensional
denition of locative case may specify that it refers to location, but the case may cover
situations that lie outside this denition.
inverse. See directinverse.
inversion. Typically a predicate that takes an experiencer argument encodes the experiencer as
subject and the neutral patient as a direct object, indirect object or peripheral relation.
201
Where the patient of such a predicate is encoded as subject, the assignment of roles to
grammatical relations is said to be inverted and the predicate is described as an inversion
predicate or as exhibiting inversion.
normal: The trainer liked the horses performance.
inversion: The horses performance pleased the trainer.
See sections 3.4.3 and 5.3.
isomorphic. Two systems are isomorphic if the distinctions made in one correspond point-
for-point with the distinctions in the other. The hierarchy of ranks in an army could be
described as being isomorphic with the ranks in a navy. The case system of one language
could be isomorphic with that of another. The case paradigm of one class of nouns could
be isomorphic with that of another class, the two paradigms differing only in details of
realisation.
local cases. Cases encoding roles relating to position (location) or change of position (source,
path, destination).
localist. An approach which interprets the marking of grammatical case as being a metaphor-
ical extension of local case marking.
locative case. The case that expresses the role of location.
marked, unmarked. The unmarked member of an opposition is the normal or default
member. In a two-way opposition of number, singular vs. plural, the unmarked member is
the singular. The terms marked and unmarked refer to the fact that the normal member
will usually lack any marking. Compare English book vs. books.
meaning. In this text cases are described as having meanings or functions. A meaning is
semantic, but a function can be syntactic or semantic. The dative in a particular language
may be described as having the function of expressing the indirect object and as having a
purposive meaning or function, a beneciary meaning or function, etc.
modier. See dependent.
multiple case. The appearance of more than one case on a noun. See section 4.4.4.
new. See given.
nominal. Nominal is a cover term for noun and pronoun. It may also cover other words
that have some properties of nouns but not all. In traditional grammar noun and pronoun
were considered separate parts of speech. In contemporary theories noun and pronoun are
subclasses of the same category. This category is often referred to as noun and abbreviated
N, but readers need to be aware of noun in the traditional sense and noun in the wider
sense.
nominalisation. The derivation of a noun, most often from a verb. In English destruction is
a nominalisation of destroy.
nominative case. In Ancient Greek and Latin the nominative or naming case was used
for nouns used in isolation and for nouns expressing the subject relation. In this work it
is used for the case that encodes S. It may encode only S, S plus A (subject), S plus
P (absolutive) or all core relations. Some writers use absolutive for a case that en-
codes SP (the absolutive relation) and some use direct for a case that covers all core
relations.
202
nonautonomous case. Melcuks term for a case that does not have any distinctive forms of
its own, but which can be recognised by seeing how a function is treated across a number
of different nonisomorphic paradigms. See section 2.2.2.
nonnite. See nite.
object. This termusually refers to the direct object. It may be used to refer to both the objects
in the double-object construction.
objective case. See accusative case.
objective genitive. Where there is a noun that has the same lexical meaning as a verb and
a genitive-marked dependent encodes what would have been the object or any other non-
subject complement of the corresponding verb, the genitive is described as an object(ive)
genitive. See (14) in chapter 4.
oblique. Originally the term oblique referred to all the non-nominative cases (section 2.1).
It may be used for a single multifunctional case opposed to the nominative (section 5.8)
and it is sometimes used for a multifunctional case in a system of three or four cases. In
recent linguistics it has sometimes been used to refer to all noncore cases. In Relational
Grammar a distinction is made between the term relations (subject, direct object and
indirect object) and the oblique relations (semantic relations such as locative and
instrumental).
obviative. See person.
P. The patient argument of a transitive verb plus any other role that is treated in the same
way grammatically.
paradigm. A paradigm is a set of grammatically conditioned forms based on a particular
root such as the set of case forms of a particular noun.
paradigmatic opposition. Aparadigmatic opposition can be said to exist between elements
in a system, e.g. a tense system or a case system. In Turkish one could say that there is
a paradigmatic opposition between case forms like adam, adama, etc. (see Table 1.1),
or between case markers like zero, -a, -da, etc. or between the number markers zero
and -lar-.
participant. Participant can be used with reference to the content of a message or with
reference tothe speechact. Speechact participant is describedunder person. Participant is
a cover termfor the entities represented by the complements and adjuncts in a construction.
Some participants are arguments and others are not. In the sentence She bought bread at the
convenience store both she and bread represent arguments and she, bread and convenience
store represent participants, each with its own role.
partitive. A case that indicates an entity partly affected.
passive voice. See active voice.
peripheral. In this text peripheral is used to cover the noncore grammatical relations and
the cases that encode them.
perlative. A case expressing through, across or along.
person. The participants in a speech event are classied as rst person (speaker) and second
person (addressee) as opposed to all nonparticipants, who are collectively known as third
person. In some languages a distinction is made between an inclusive rst person (including
the addressee) and an exclusive rst person (excluding the addressee). In some languages
203
a distinction is made within third person between a proximate (more topical) and obviative
(less topical) person. See also directinverse and section 5.2.5.
phrase-marking language. In this book the term phrase-marking language is used to refer
to a language in which case is marked only once in a phrase as opposed to a word-marking
language where case is marked on the head of a noun phrase and on the dependents via
concord.
postposition. A postposition differs from a preposition only in that it follows rather than
precedes the nominal it governs. See preposition.
predicate. A predicate is a relational term signifying a property of an entity or a relation
between entities. The entities are the arguments of the predicate. FLY is a one-place pred-
icate, i.e. a predicate that implies one argument, the entity that ies. BASH is a two-place
predicate and GIVE a three-place predicate. The use of capitals is to distinguish abstract
predicates from the verbs that express them. In traditional grammar a sentence was divided
into subject and predicate, where the predicate included the verb and its arguments less
the subject. In a sentence like Vikings attack Wessex the phrase attack Wessex realises a
predicate with Vikings as its argument. However, it is a complex predicate, which can itself
be analysed into a predicate (represented by attack) with Wessex as its argument.
predicative. A phrase instantiating a predicate can be described as predicative. In They
elected me chairperson the noun chairperson instantiates a predicate, similarly the adjective
in They painted it red. Both the noun chairperson and the adjective red can be described
as predicative.
preposition. Apreposition is a word that is itself a dependent, but which precedes and governs
a nominal. In the sentence She put the sash on him the preposition on is governed by the
verb put and in turn governs the following pronoun him. The preposition serves to mark
the relationship of him to put.
proclitic. See clitic.
promotion. See advance, advancement.
proximate. See person.
relator nouns. Relator nouns are a specialised subclass of nouns that behave like adpositions
in relating a predicate to a noun phrase. In He is standing on top of the cupboard the word
top could be analysed as a relator noun showing the relationship of cupboard to stand. See
section 1.3.3.
role. The semantic relation borne by a dependent to its head. In She washed the tablecloth
the subject she could be described as having the role of agent and the tablecloth as having
the role of patient. Other terms used are semantic role, case role, thematic role and theta
role. See section 3.4.
S. 1. The single argument of a one-place predicate.
2. In the context of word order it refers to subject as in the abbreviation SOV
referring to subjectobjectverb order.
3. In phrase-structure rules S stands for sentence.
secondary object. Where a direct object can be identied in a double-object construction
the other object can be referred to as the secondary object.
semantic cases. Cases that encode purely semantic relations. See section 2.3.2.
204
semantic relation. In this book semantic relation is synonymous with (semantic) role.
specic. In linguistics the term specic carries its normal meaning of particular. It is
frequentlyusedwithreference tothe entityencodedas P, where specic means a particular
token or particular tokens identiable by the speaker.
split-S, split intransitive. See active language.
stem-forming case marking. In some languages forms identiable as case markers in one
context are used in other contexts to mark the stem to which other case markers are added.
See section 4.4.2.
structural case. In some modern theories such as Lexical Functional Grammar (see (12)
in chapter 2) and Government and Binding Theory (see section 3.3), some cases, notably
nominative for subject and accusative for direct object, are assigned on the basis of the
position which the relevant nominal holds in clause structure.
subessive. A case meaning under or below.
subject. The grammatical relation that covers S plus A. In various languages it manifests
itself in terms of being represented by a series of clitic pronouns or agreement on the verb,
by being in the nominative case, or by occupying the initial position in the clause proper.
subjective genitive. Where there is a noun that has the same lexical meaning as a verb and a
genitive-marked dependent encodes what would have been the subject of the corresponding
verb, it is described as a subjective genitive. See (13) in chapter 4.
superessive. A case meaning on top of , over or above.
superlative. A case meaning to a position on top of or over.
syncretism. 1. The merging of formerly distinct grammatical categories.
2. The neutralisation of grammatical categories in a particular environment. Sense 1 is
older and is most often used with reference to the merging of formerly distinct inectional
forms.
syntagmatic. Syntagmatic refers to relations between the constituents of a construction. A
word can be said to bear syntagmatic relations to the other words in the same construction,
but paradigmatic relations with other words that can be substituted for it. In I saw him
there are syntagmatic relations between the three words and paradigmatic relations between
each word and all the other words that can be substituted for them. The verb saw is in a
paradigmatic relation with heard, watched, etc., etc.
system. As used in this book system is a set of items that can occupy a position in a word,
phrase, clause or sentence structure. Normally only one member of the system can occupy
the position in question at one time. Case is a system. In an inectional-case language, for
instance, each noun in a sentence must bear one case from the system.
theme. 1. In the Functional Sentence Perspective analysis of the Prague School the term
theme meant approximately topic.
2. Gruber introduced the term for the semantic role that others called patient. Theme,
patient and patient/theme are now competing terms. See under patient in the list of
roles in section 3.4.
topic, comment. The topic is what is spoken about as opposed to the comment, which is what
is said about the topic. The topic is normally given information and typically expressed as
the subject. There can also be a topic that stands outside the clause proper, usually before
205
the clause, set off by an intonation break. There are also discourse topics, topics that are
established at some point and hold over some subsequent stretch of discourse. Consider
the following sentence from an article about the French singer Edith Piaf:
The English too, they admired her.
The English represents the clause-external topic; they, the subject, represents given infor-
mation (given by the clause-external topic) and provides the clause-internal topic, and her,
which is part of the comment, represents the discourse topic of the passage.
transitive verb. A two-place verb with an agent and an affected patient or any other verb that
has the same valency. In some languages there are alternative valencies (e.g. antipassive
and passive). Only the unmarked valency qualies as transitive. Averb like put that requires
the expression of three arguments is often referred to as a ditransitive verb.
translative case. A case label meaning through, but note that the translative case in Finnish
refers to the end point of a change of state. See Tables 5.5 and 5.6.
unmarked. See marked.
valency (American: valence). A specication of the arguments of a predicate and their
expression. The specication would include the number of complements, their role, their
grammatical relation and details of their expression (case, clitic, etc.).
vocative. A case used to mark a nominal as an addressee. See section 1.2.3.
word-marking language. In this book the term word-marking language is used to refer to
a language in which case is marked on the head of a noun phrase and on the dependents
via concord.
206
GUI DE TO FURTHER READI NG
Despite the fact that case plays a prominent part in European languages there are surprisingly
few works devoted exclusively to case, particularly works in English. Among the most useful
papers on inectional case are Melcuks 1986 paper Toward a denition of case and two
papers by Comrie: On delimiting cases (1986) and Form and function in identifying cases
(1991). The latter is in a volume edited by Frans Plank entitled Paradigms: the economy of
inection. This volume is largely about case paradigms and provides a concentrated source
for a largely neglected area.
For semantic roles or deepcases there is one particularlyuseful reference, namelySomers
1987 book Valency and case in computational linguistics. Although one could not guess it
from the title, most of the book is a clear, concise summary of the case grammar tradition
including Fillmore, Dik, J. Anderson, Chafe, etc.
The description of case inevitably involves grammatical relations and the various gram-
matical mechanisms languages employ to package the propositional content according to the
demands of discourse. There are numerous works to choose from in this eld including Foley
and Van Valin 1984, Giv on 1984, 1990 and Croft 1991. Ergative systems gure prominently in
the recent literature. Comries 1978 paper Ergativity and Dixons 1979 paper with the same
title provide a good coverage. Dixons 1994 book Ergativity provides an updated coverage of
manifestations of ergative phenomena in numerous languages.
For those who read Spanish, Ana Aguds Historia y teora de los casos (1980) is a compre-
hensive account of how case has been dealt with over the last two and a half thousand years.
Volume 1 (Casos y preposiciones) of Rubio (1966) Introducci on a la sintaxis estructural
del latin provides a solid structural analysis of case in Latin. For readers of Italian, Calbolis
Linguistica moderna e il latino: i casi is a useful summary of analyses of the Latin case
system.
207
REFERENCES
Abondolo, D. 1987. Hungarian. In B. Comrie (ed.), 57792
Agesthialingom, S. & K. Kushalappa Gowda (eds.). 1976. Dravidian case system. Anna-
malainagar: Annamali University
Agud, A. 1980. Historia y teora de los casos. Madrid: Gredos
Allen, W. S. 1964. Transitivity and possession. Language 40: 33743
Andersen, P. K. 1979. Word order typology and prepositions in Old Indic. In B. Brogyanyi
(ed.), Studies in diachronic, synchronic and typological linguistics, 2 vols., vol. 2, part
1, 2334. Amsterdam: John Benjamins
Anderson, J. M. 1971. The grammar of case: towards a localistic theory. Cambridge
University Press
1977. On case grammar. London: Croom Helm
1979. On being without a subject. Bloomington: Indiana University Linguistics Club
1997. A notional theory of syntactic categories. Cambridge University Press
Anderson, S. 1982. Wheres morphology? Linguistic Inquiry 13: 571612
Anttila, R. 1972. An introduction to historical and comparative linguistics. New York:
Macmillan
Artawa, K. &B. J. Blake. 1997. Patient primacy in Balinese. Studies in Language 21: 483508
Austerlitz, R. 1980. Typology and universals on a Eurasian east-west continuum. In G.
Brettschneider & C. Lehmann (eds.), Wege zur Universalienforschung. Sprachwis-
senschaftliche Beitr age zum 60 Geburtstag von Hansjakob Seiler, 23544. T ubingen:
Gunter Narr
Austin, P. 1981a. A grammar of Diyari, South Australia. Cambridge University Press
1981b. Switch-reference in Australia. Language 57: 30934
1981c. Case marking in southern Pilbara languages. Australian Journal of Linguistics 1:
21126
Baker, M. C. 1988. Incorporation: a theory of grammatical function changing. University of
Chicago Press
Balakrishnan, R. 1976. Kodagu case system. In S. Agesthialingom & K. Kushalappa Gowda
(eds.), 42134
Blake, B. J. 1977. Case marking in Australian languages. Canberra: Australian Institute of
Aboriginal Studies
1979a. A Kalkatungu grammar. Canberra: Pacic Linguistics
208
References
1979b. Pitta-Pitta. In R. M. W. Dixon & B. J. Blake (eds.), 182242
1982. The absolutive: its scope inEnglishandKalkatungu. InP. J. Hopper &S. A. Thompson
(eds.), 7183
1983. Structure and word order in Kalkatungu: the anatomy of a at language. Australian
Journal of Linguistics 3: 14376
1985. Case markers, case and grammatical relations: an addendum to Goddard. Australian
1987. Australian Aboriginal grammar. London: Croom Helm
1988. Redening Pama-Nyungan: towards the prehistory of Australian languages. In N.
Evans &S. Johnson (eds.), Aboriginal Linguistics 1: 190. Armidale: University of New
England, Department of Linguistics
1990. Relational grammar. London: Routledge
Bloomeld, L. 1933. Language. New York: Holt, Rinehart & Winston
Boas, F. (ed.) 1922. Handbook of American Indian languages, part 2 (Bureau of American
Ethnology, Bulletin 40). Washington, DC: Smithsonian Institution
Bowe, H. 1990. Categories, constituents and constituent order in Pitjantjatjara: an Aboriginal
language of Australia. London: Routledge
Branch, M. 1987. Finnish. In B. Comrie (ed.), 593617
Brecht, R. D. & J. S. Levine (eds.) 1986. Case in Slavic. Columbus, OH: Slavica
Publishers
Breen, J. G. 1976. Ergative, locative and instrumental inections in Wangkumara. In R. M.
W. Dixon (ed.), 3369
1981. Margany and Gunya. In R. M. W. Dixon & B. J. Blake (eds.), vol. 2: 274
393
Bresnan, J. (ed.) 1982. The mental representation of grammatical relations. Cambridge,
MA: MIT Press
Bresnan, J. & J. Kanerva. 1989. Locative inversion in Chichewa: a case study of factorization
in grammar. Linguistic Inquiry 20: 150
Bresnan, J. & S. A. Mchombo. 1987. Topic, pronoun, and agreement in Chichewa. Language
63: 74182
Burridge, K. 1990. Sentence datives and grammaticization and the dative possessive: evidence
from Germanic. La Trobe Working Papers in Linguistics 3: 2948
Burrow, T. 1955. The Sanskrit language. London: Faber & Faber
Calboli, G. 1972. Linguistica moderna e il latino: i casi. Bologna: Riccardo Patron
Cardona, G. 1976a. P an
.
ini: a survey of research. In W. Winter (ed.), Trends in linguistics,
State-of-the-Art reports (6). The Hague: Mouton
1976b. Subject in Sanskrit. In M. K. Verma (ed.), 138
Carroll, P. J. 1976. Kunwinjku (Gunwinggu): a language of Western ArnhemLand. Canberra:
Australian National University masters thesis
Carvalho, P. de. 1980. Reexions sur les cas: vers une th eorie des cas latins. LInformation
Grammaticale 7: 311
1982. Cas et personne. Propositions pour une th eorie morpho-s emantique des cas latins.
Revue des Etudes Anciennes 82: 24374
209
References
1983. Le syst` eme des cas latins. In H. Pinkster (ed.), Latin linguistics and linguistic theory,
5971. Amsterdam: John Benjamins
1985. Nominatif et sujet. In C. Touratier (ed.), 5578
Chafe, W. 1970. Meaning and the structure of language. University of Chicago Press
Chidambaranathan, V. 1976. Case system of Kasaba. In S. Agesthialingom & K. Kushalappa
Gowda (eds.), 46780
Chomsky, N. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press
1981. Lectures on government and binding. Dordrecht: Foris
1985. Knowledge of language: its nature, origin and use. New York: Praeger
1993. A minimalist program for linguistic theory. In K. Hale and S. J. Keyser (eds.), The
view from building 20. Cambridge, MA: MIT Press, 152
Cole, P. &G. Hermon. 1981. Subjecthood and islandhood: evidence fromQuechua. Linguistic
Inquiry 12: 130
Coleman, R. 1985. The Indo-European origins and Latin development of the accusative with
innitive construction. In C. Touratier (ed.), 30744
1987. Latin and the Italic languages. In B. Comrie (ed.), 180202
1991 (ed.). New studies in Latin linguistics: selected papers from the 4th international
colloquium on Latin linguistics, Cambridge, April 1987. Amsterdam: John Benjamins
1991a. The assessment of paradigmstability: some Indo-European case studies. In F. Plank
(ed.), 197213
1991b. Latin prepositional syntax in Indo-European perspective. In R. Coleman (ed.),
32338
Comrie, B. 1976. The syntax of causative constructions: cross-language similarities and
divergences. In M. Shibatani (ed.), The grammar of causative constructions (Syntax and
Semantics 6), 261312. New York: Academic Press
1978. Ergativity. In W. P. Lehmann (ed.), Syntactic typology, 32994. Sussex: Harvester
Press
1980. The order of case and possessive sufxes in Uralic languages: an approach to the
comparative-historical problem. Lingua Posnaniensis 23: 816
1981. The languages of the Soviet Union. Cambridge University Press
1986. On delimiting cases. In R. D. Brecht & J. S. Levine (eds.), 86105
1987. Russian. In B. Comrie (ed.), 32947
1987 (ed.). The worlds major languages. London: Croom Helm
1989. Language universals and linguistic typology. Oxford: Blackwell
1991. Form and function in identifying cases. In F. Plank (ed.), 4156
Comrie, B. &Polinsky, M. 1998. The great Daghestan case hoax. In A. Siewierska &J. J. Song
(eds.), Case typology and grammar in honor of Barry J. Blake. Amsterdam: Benjamins,
95114
Cook, W. 1979. Case grammar: development of the matrix model. Washington: Georgetown
University Press
Corbett, G. 1987. Serbo-Croat. In B. Comrie (ed.), 391409
Croft, W. 1991. Syntactic categories and grammatical relations. University of Chicago Press
210
References
Crowley, T. 1983. Uradhi. In R. M. W. Dixon & B. J. Blake (eds.), 306428
Cutler, A., J. A. Hawkins & G. Gilligan. 1985. The sufxing preference: a processing expla-
nation. Linguistics 23: 72358
Delancey, S. 1981. An interpretation of split ergativity and related pattems. Language 57:
62657
1987. Sino-Tibetan languages. In B. Comrie (ed.), 797810
Dench, A. & N. Evans. 1988. Multiple case-marking in Australian languages. Australian
Derbyshire, D. C. & G. K. Pullum (eds.). 1986, 1989, 1990, 1991. Handbook of Amazonian
languages, vols. 14. Berlin: Mouton de Gruyter
Dik, S. 1978. Functional grammar. Amsterdam: North Holland
Dixon, R. M. W. 1972. The Dyirbal language of North Queensland. Cambridge University
Press
1976 (ed.). Grammatical categories in Australian languages. Canberra: Australian Institute
of Aboriginal Studies; New Jersey: Humanities Press
1977. A grammar of Yidiny. Cambridge University Press
1979. Ergativity. Language 55: 59138
1980. The languages of Australia. Cambridge University Press
1987 (ed.). Studies in ergativity. Amsterdam: North Holland
1994. Ergativity. Cambridge University Press
Dixon, R. M. W. & A. Aikhenvald (eds.) 1999. The Amazonian languages. Cambridge
University Press.
Dixon, R. M. W. & B. J. Blake (eds.). 1979, 1981, 1983, 1991, 2000. Handbook of Australian
languages, vols. 15. Canberra: ANU and Amsterdam: John Benjamins (vols. 13);
Melbourne: Oxford University Press (vols. 45)
Du Bois, J. 1987. The discourse basis of ergativity. Language 63: 80555
Durie, M. 1985. A grammar of Acehnese on the basis of a dialect of North Acehnese. Dor-
drecht: Foris
1986. The grammaticalization of number as a verbal category. Proceedings of the Twelfth
Annual Meeting of the Berkeley Linguistics Society, 35570
Ebeling, C. L. 1966. Review of Chikobava & Cercvadzes The grammar of literary Avar.
Studia Caucasica 2: 58100
Evans, N. D. 1995. A Grammar of Kayardild. Berlin: Mouton de Gruyter
Filliozat, P-S. 1988. Grammaire Sanskrite P anin eenne. Paris: Picard
Fillmore, C. J. 1968. The case for case. In E. Bach & R. T. Harms (eds.), Universals in
linguistic theory, 188. London: Holt, Rinehart & Winston
1971. Some problems for case grammar. Working Papers in Linguistics, Ohio State
University 10: 24565
Foley, W. A. &R. D. Van Valin Jr. 1984. Functional syntax and universal grammar. Cambridge
University Press
Frachtenberg, L. J. 1922a. Coos. In F. Boas (ed.), 297429
1922b. Siuslawan. In F. Boas (ed.), 431629
211
References
Franchetto, B. 1990. Ergativity and nominativity in Kuikuro and other Carib languages. In
D. L. Payne (ed.), Amazonian linguistics: studies in lowland South American languages,
40727. Austin: University of Texas Press
Friberg, B. 1991. Ergativity, focus and verb morphology in several south Sulawesi languages.
In R. Harlow (ed.), 10330
Giridhar, P. P. 1987. A case grammar of Kannada. University of Mysore: doctoral dissertation
Giv on, T. 1975. Serial verbs and syntactic change: Niger-Congo. In C. Li (ed.), Word order
and word order change, 47111. Austin: University of Texas Press
1976. Topic, pronoun and grammatical agreement. In C. Li (ed.), Subject and topic, 14988.
New York: Academic Press
1984a. Syntax: a functional-typological introduction, vol. 1. Amsterdam: John Benjamins
1984b. Direct object and dative shifting: semantic and pragmatic case. In F. Plank (ed.),
15182
1990. Syntax: a functional-typological introduction, vol. 2. Amsterdam: John Benjamins
Goddard, C. 1982. Case systems and case marking in Australian languages. Australian Journal
of Linguistics 2: 16796
Greenberg, J. H. 1963. Some universals of grammar with particular reference to the order
of meaningful elements. In J. H. Greenberg (ed.), Universals of language, 73113.
Cambridge, MA: MIT Press
Groot, W. de. 1956. Classication of cases and uses of cases. For Roman Jakobson, 18794.
The Hague: Mouton
Gruber, J. 1965. Studies in lexical relations. Cambridge, MA: MIT doctoral dissertation
[reproduced by Indiana University Linguistics Club, Bloomington, Indiana]
1976. Lexical structures in syntax and semantics. Amsterdam: North Holland
Haegeman, L. 1994. Introduction to government and binding theory, Second edition. Oxford:
Blackwell
Hale, K. L. 1973a. Deep-surface canonical disparities in relation to analysis and change:
an Australian example. In T. A. Sebeok (ed.), Current trends in linguistics, vol. 11:
Diachronic, areal and typological linguistics, 40158. The Hague: Mouton
1973b. Person marking in Walbiri. In S. R. Anderson & P. Kiparsky (eds.), A festschrift
for Morris Halle, 30844. New York: Holt, Rinehart & Winston
1976. On ergative and locative sufxial alternations in Australian languages. In R. M. W.
Dixon (ed.), 41420
1982. Some essential features of Warlpiri verbal clauses. Working Papers of the Summer
Institute of Linguistics Australian Aborigines Branch. A6: 217315
Halliday, M. A. K. 1985. An introduction to functional grammar. London: Edward Arnold
19678. Notes on transitivity and theme in English, Parts 1, 2 and 3. Journal of Linguistics
3: 199244, 4: 189215
Hansen, K. C. & L. E. Hansen. 1978. The core of Pintupi grammar. Alice Springs: Institute
for Aboriginal Development
Harlow, R. (ed.). 1991. VICAL 2. Western Austronesian and contact languages. Papers from
the fth international conference on Austronesian linguistics, Auckland, New Zealand
1991. Auckland: Linguistic Society of New Zealand
212
References
Harris, A. 1981. Georgian syntax: a study in relational grammar. Cambridge University Press
1984. Inversion as a rule of universal grammar: Georgian evidence. In D. M. Perlmutter &
C. Rosen (eds.), 25191
1985. Diachronic syntax (Syntax and Semantics 18). Orlando: Academic Press
Harris, M. & N. Vincent (eds.). 1988. The Romance languages. Oxford University
Press
Heath, J. 1980. Basic texts in Warndarang: grammar, texts and dictionary. Canberra: Pacic
Linguistics (ANU)
Heine, B., U. Claudi & F. H unnemeyer. 1991. Grammaticalization. University of Chicago
Press
Hewitt, B. G. 1989. Abkhaz. London: Routledge
Hinnebusch, T. J. 1979. Swahili. In T. Shopen (ed.), Languages and their status, 20993.
Cambridge, MA: Winthrop
Hjelmslev, L. 1935. La cat egorie des cas: Etude de grammaire g en erale I (Acta Jutlandica:
Aarsskrift for Aarhus Universitet 7.1). Copenhagen: Munksgaard
1937. La cat egorie des cas: Etude de grammaire g en erale II (Acta Jutlandica: Aarsskrift
for Aarhus Universitet 9.3). Copenhagen: Munksgaard
Hoddinott, W. G. & F. M. Kofod. 1976. Djamindjungan. In R. M. W. Dixon (ed.), 397401
Hopper, P. J. & S. A. Thompson. 1980. Transitivity in grammar and discourse. Language 56:
25199
1982 (eds.). Syntax and semantics, vol. 15: Studies in transitivity. New York: Academic
Press
Hubschmann, H. 1906. Armenica 7: Kasus-Attraktion im Armenischen. Indogermanische
Forschungen 19: 47880
Huddleston, R. 1984. Introduction to the grammar of English. Cambridge University
Press
Hudson, R. 1992. So-called double objects and grammatical relations. Language 68:
25176
Jackendoff, R. 1972. Semantic interpretation in generative grammar. Cambridge, MA: MIT
Press
Jacobs, M. 1931. Asketch of Northern Sahaptin grammar. University of Washington [Seattle]
Publications in Anthropology 4: 85291
Jacobsen, W., Jr. 1983. Typological and genetic notes on switch-reference systems in North
American Indian languages. In J. Haiman & P. Munro (eds.), Switch-reference and
universal grammar, 15183. Amsterdam: John Benjamins
Jakobson, R. 1936/1971. Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der
russischen Kasus. In R. Jakobson, Selected writings II: words and language, 2371.
The Hague: Mouton
1958/1971. Morphological inquiry into Slavic declension: structure of Russian case forms
[text in Russian, summary in English in 1971 version]. In R. Jakobson, Selected writings
II: words and language, 15483. The Hague: Mouton
Jensen, J. T. 1991. Case in Yapese. In R. Harlow (ed.), 21530
Juret, A-C. 1926. Syst` eme de la syntaxe latine. Paris: Les Belles Lettres
213
References
Kaye, A. S. 1987. Arabic. In B. Comrie (ed.), 66485
Keen, S. L. 1972. A description of the Yukulta language: an Australian Aboriginal language
of Northwest Queensland. Monash University (Melbourne): MA Thesis
1983. Yukulta. In R. M. W. Dixon & B. J. Blake (eds.), 190304
Keenan, E. L. &B. Comrie. 1977. Nounphrase accessibilityanduniversal grammar. Linguistic
Inquiry 8: 6399
Kennedy, R. 1984. Semantic roles: the language speakers categories (in Kala Lagaw Ya). In
Papers in Australian Linguistics 16, 15370. Canberra: Pacic Linguistics
Kibrik, A. E. 1979. Canonical ergativity and Daghestan languages. In F. Plank (ed.), 6177
1991. Organising principles for nominal paradigms in Daghestanian languages: compar-
ative and typological observations. In F. Plank (ed.), 25574
Kilby, D. 1981. On case markers. Lingua 54: 10133
Klaiman, M. H. 1987. Bengali. In B. Comrie (ed.), 490513
Klimov, G. A. 1973. O cerk ob s cej teorii ergativnosti (Outline of ageneral theory of ergativity).
Moscow: Nauka.
Kornlt, J. 1987. Turkish and the Turkic Languages. In B. Comrie (ed.), 61944
Kullavanijaya, P. 1974. Transitive verbs in Thai. Honolulu: University of Hawaii doctoral
dissertation [available from University Microlms, Ann Arbor, Michigan]
Kurylowicz, J. 1949. Le probl` eme du classement des cas. Biuletyn PTJ 9: 2043 [Reprinted
in Esquisses linguistiques, 13150. WroclawKrakow: Ossolineum. 1960]
1964. The inectional categories of Indo-European. Heidelberg: Carl Winter
Langacker, R. W. 1977. Studies in Uto-Aztecan grammar, vol. 1: An overview of Uto-
Aztecan grammar. Arlington: Summer Institute of Linguistics and University of Texas at
Arlington
Langdon, M. 1970. A grammar of Diegue no: the Mesa Grande dialect. Berkeley: University
of California Press
Larsen, T. W. & W. M. Norman. 1979. Correlates of ergativity in Mayan grammar. In F. Plank
(ed.), 34770
Lehmann, C. 1982. Thoughts on grammaticalization: a programmatic sketch, vol. 1 (Arbeiten
des k olner universalien Projekts 48). Cologne: Instit ut f ur Sprachwissenschaft
1985. Latin case relations in typological perspective. In C. Touratier (ed.), 81104
Levin, B. 1987. The middle construction and ergativity. In R. M. W. Dixon (ed.), 1732
Lewis, G. L. 1967. Turkish grammar. Oxford University Press
Li, C. N. & S. A. Thompson. 1981. Mandarin Chinese: a functional reference grammar.
Berkeley: University of California Press
n.d. Wappo. Typescript
Lindenfeld, J. 1973. Yaqui syntax. Berkeley: University of California Press
Longacre, R. E. 1976. An anatomy of speech notions. Lisse: Peter de Ridder
Lord, C. 1982. The development of object markers in serial verb languages. In P. J. Hopper
& S. A. Thompson (eds.), 27799
Luraghi, S. 1987. Patterns of case syncretism in Indo-European languages. In A. Ramat et al.
(eds.), Papers from the 7th international conference on historical linguistics, 35571.
Amsterdam: John Benjamins
214
References
1989. The relationship between prepositions and cases within Latin prepositional phrases.
In G. Calboli (ed.), Subordination and other topics in Latin, 25371. Amsterdam,
Philadelphia: John Benjamins
1991. Paradigmsize, possible syncretisms, andthe use of adpositions withcases ininective
languages. In F. Plank (ed.), 5774
1996. Studi su casi e preposizioni nel greco antico. Milan: Franco Angeli.
Lyons, C. 1999. Deniteness. Cambridge University Press.
Madtha, W. 1976. Case system of Kannada. In S. Agesthialingom & K. Kushalappa Gowda
(eds.), 17354
Mallinson, G. 1987. Rumanian. In B. Comrie (ed.), 30321
1988. Rumanian. In M. Harris & N. Vincent (eds.), 391419
Mallinson, G. & B. J. Blake. 1981. Language typology: cross-linguistic studies in syntax.
Amsterdam: North Holland
Marantz, A. 1984. On the nature of grammatical relations. Cambridge, MA: MIT Press
Matthews, P. H. 1974. Morphology: an introduction to the theory of word-structure.
Cambridge University Press (second edition 1991)
McAlpin, D. W. 1976. Dative subject in Malayalam. In M. K. Verma (ed.), 18394
McGregor, W. 1990. A functional grammar of Gooniyandi. Amsterdam: John Benjamins
McKay, G. R. 1976. Rembarnga. In R. M. W. Dixon (ed.), 494505
McLendon, S. 1978. Ergativity, case, and transitivity in Eastern Pomo. IJAL 44: 19
Melcuk, I. A. 1986. Toward a denition of case. In R. D. Brecht & J. S. Levine (eds.), 3585
Merlan, F. 1982. Mangarayi (Lingua Descriptive Series No. 4). Amsterdam: North Holland
Mithun, M. 1991. Active/agentive case marking and its motivation. Language 67: 51046
Moravcsik, E. A. 1978. On the case marking of objects. In J. H. Greenberg (ed.), Universals
of human language, vol. 4: Syntax, 24989. Stanford University Press
1984. The place of direct objects among the noun phrase constituents of Hungarian. In
F. Plank (ed.), 5585
Neidle, C. 1982. Case agreement in Russian. In J. Bresnan (ed.), 391426
1988. The role of case in Russian syntax. Dordrecht: Kluwer
Nichols, J. 1983. On direct and oblique cases. Berkeley Linguistics Society 9: 17092
1986. Head-marking and dependent-marking grammar. Language 62: 56119
Norman, J. 1988. Chinese. Cambridge University Press
OGrady, W. 1991. Categories and case: the sentence structure of Korean. Amsterdam: John
Benjamins
Palmer, L. R. 1954. The Latin language. London: Faber & Faber
Perialwar, R. 1976. Irula case system. In S. Agesthialingom & K. Kushalappa Gowda (eds.),
495519
Perlmutter, D. M. 1982. Syntactic representation, syntactic levels, and the notion of subject.
In P. Jacobson & G. Pullum (eds.), The nature of syntactic representation, 283340.
Dordrecht: Reidel
1983 (ed.). Studies in relational grammar 1. University of Chicago Press
Perlmutter, D. M. & P. M. Postal. 1983. Some proposed laws of basic clause structure. In D.
M. Perlmutter (ed.), 81128
215
References
1984. The 1-advancement exclusiveness law. In D. M. Perlmutter & C. Rosen (eds.),
81125
Perlmutter, D. M. & C. Rosen (eds.). 1984. Studies in relational grammar 2. University of
Chicago Press
Pinkster, H. 1985. Latin cases and valency grammar: some problems. In C. Touratier (ed.),
16390
Plank, F. (ed.) 1979. Ergativity. London: Academic Press
1984 (ed.). Objects: towards a theory of grammatical relations. London: Academic Press
1990. Sufx copying as a mirror-image phenomenon. European Science Foundation,
Programme in Language Typology: Working Paper 1: 111
1991 (ed.). Paradigms: the economy of inection. Berlin: Mouton de Gruyter
Postal, P. M. 1974. On raising: one rule of English grammar and its theoretical implications.
Cambridge, MA: MIT Press
1977. Antipassive in French. Proceedings of the Seventh Annual Meeting of the North
Eastern Linguistics Society, 273313
Press, M. L. 1979. Chemehuevi: a grammar and lexicon (University of California Publications
in Linguistics 92). Berkeley and Los Angeles: University of California Press
Ramarao, C. 1976. Markedness in case. In S. Agesthialingom&K. Kushalappa Gowda (eds.),
22140
Reed, I., O. Miyaoka, S. Jacobsen, P. Afcan & M. Krauss. 1977. Yupik Eskimo gram-
mar. University of Alaska: Alaska Native Language Center and the Yupik Language
Workshop
Riemsdijk, H. van. 1983. The case of German adjectives. In F. Heny & B. Richards (eds.),
Linguistic categories: auxiliaries and related puzzles, vol. 1: 22352. Dordrecht: Reidel
Riemsdijk, H. van & E. Williams. 1986. Introduction to the theory of grammar. Cambridge,
MA: MIT Press
Rigsby, B. 1974. Sahaptin grammar. Typescript
Roberts, I. 1997. Comparative syntax. London: Arnold
Robins, R. H. 1967. A short history of linguistics. London: Longman
Robinson, L. W. & J. Armagost. 1990. Comanche dictionary and grammar. Arlington: Sum-
mer Institute of Linguistics and University of Texas at Arlington
Rosen, C. 1984. The interface between semantic roles and initial grammatical relations. In
D. M. Perlmutter & C. Rosen (eds.), 3880
Rosen, C. & K. Wali. 1989. Twin passives, inversion and multistratalism in Marathi. Natural
Language and Linguistic Theory 7: 150
Roth, W. E. 1897. Ethnological studies among the north-west-central Queensland Aborigines.
Brisbane: Government Printer
Rubio, L. 1966. Introduc oion a la sintaxis estructural del Latin, vol. 1: Casos y preposiciones.
Barcelona: Ariel
Rude, N. 1985. Studies in Nez Perce grammar and discourse. University of Oregon: doctoral
dissertation
1988. On the origin of the Nez Perce ergative NP sufx. (Conference on Grammaticaliza-
tion, University of Oregon, 1215 May 1988)
216
References
Sakthivel, S. 1976. Toda case system. In S. Agesthialingom & K. Kushalappa Gowda (eds.),
43548
Saltarelli, M. et al. 1988. Basque (Croom Helm Descriptive Grammars). London: Croom
Helm
Sands, K. 1996. The ergative in proto-Australian. Munich: Lincom Europa
Schooneveld, C. H. van. 1986. Jakobsons case system and syntax. In R. D. Brecht & J. S.
Levine (eds.), 37385
Sch utze, C. 1993. Towards a minimalist account of quirky case and licensing in Icelandic. In
C. Phillips (ed.), Papers on Case and agreement II (MIT Working Papers in Linguistics.
19), 32175. Cambridge, MA: MIT Press
Scott, G. 1978. The Fore language of Papua New Guinea. Canberra: Pacic Linguistics
Seidel, H. 1988. Kasus: zur Explikationeines sprachwissenschaftlichenTerminus (amBeispiel
des Russischen). T ubingen: Gunter Narr
Serbat, G. 1981a. Cas et fonctions. Paris: Presses Universitaires de France
1981b. Le syst` eme des cas est-il syst ematique? Revue des Etudes Latines 59: 298317
Sherzer, J. 1976. An areal-typological study of American Indian languages north of Mexico.
Amsterdam: North Holland
Silverstein, M. 1976. Hierarchy of features and ergativity. In R. M. W. Dixon (ed.), 11271
1981. Case marking and the nature of language. Australian Journal of Linguistics
1: 22744
Sittig, E. 1931. Das Alter der Anordnung unserer Kasus und der Ursprung ihrer Bezeichnung
als F alle (T ubinger Beitr age zur Altertumswissenschaft 13). Stuttgart: Kohlhammer
Smith, W. 1888. A rst Greek course. London: John Murray
Somers, H. L. 1987. Valency and case in computational linguistics. Edinburgh University
Press
Starosta, S. 1971. Some lexical redundancy rules for English nouns. Glossa 5: 167201
1985. Relator nouns as a source of case inections. In R. L. Leed & V. Z. Acson (eds.), For
GordonFairbanks (Oceanic Linguistics Special Publications No. 20), 11133. Honolulu:
University of Hawaii Press
1988. The case for lexicase. London: Pinter
Steever, S. B. 1987. Tamil and the Dravidian languages. In B. Comrie (ed.), 72546
Stokes, B. 1982. A description of Nyigina, a language of the West Kimberley, Western
Australia. Canberra: ANU doctoral dissertation
Stowell, T. 1981. The origins of phrase structure. Cambridge, MA: MIT doctoral dissertation
Suarez, J. A. 1983. The Mesoamerican Indian languages. Cambridge University Press
Torrego, M. E. 1991. The genitive with verbal nouns in Latin: a functional analysis. In
R. Coleman (ed.), 28193
Touratier, C. 1978. Quelques principes pour l etude des cas (avec application ` a lablatif
latin). Languages 50: 98116
1979. Accusatif et analyse en morph` emes. Bulletin de la Soci et e de Linguistique de Paris
74(1): 4392
1985 (ed.). Syntaxe et Latin: actes du II` eme Congr` es International de Linguistique Latine,
Aix-en-Provence, 2831 Mars 1983. Aix-en-Provence: Universit e de Provence
217
References
Tsunoda, T. 1981. The Djaru language of Kimberley, Western Australia. Canberra: Pacic
Linguistics (ANU)
Tucker, A. N. & M. A. Bryan. 1966. Linguistic analyses: the non-Bantu languages of North-
Eastern Africa. Oxford University Press
Van Valin, Robert Jr. 1993. A synopsis of Role and Reference Grammar. In R. Van Valin
Jr (ed.), Advances in Role and Reference Grammar, 1164. Amsterdam: Benjamins
Vennemann, T. 1973. Explanation in syntax. In J. P. Kimball (ed.), Syntax and semantics, vol.
2: 150. New York: Seminar Press
Verma, M. K. (ed.). 1976. The notionof subject inSouthAsianlanguages. Madison: University
of Wisconsin, South Asian Studies
Weber, D. J. 1989. A grammar of Huallaga (Hu anuco) Quechua. Berkeley: University of
California Press
Whitney, A. H. 1956. Finnish. London: Hodder and Stoughton
Wierzbicka, A. 1980. The case for surface case. Ann Arbor: Karoma
1981. Case marking and human nature. Australian Journal of Linguistics 1: 4380
1983. The semantics of case marking. Studies in Language 7: 24775
1988. The semantics of grammar. Amsterdam: John Benjamins
Williams, C. J. 1980. A grammar of Yuwaalaraay. Canberra: Pacic Linguistics (ANU)
Windfuhr, G. L. 1987. Persian. In B. Comrie (ed.), 52346
Wolfart, H. C. 1973. Plains Cree: a grammatical study (Transactions of the American Philo-
sophical Society 63, part 5). Philadelphia: American Philosophical Society
Woodcock, E. C. 1959. A new Latin syntax. London: Methuen
Yallop, C. 1977. Alyawarra: an Aboriginal language of Central Australia. Canberra:
Australian Institute of Aboriginal Studies
Yip, M., J. Maling & R. Jackendoff. 1987. Case in tiers. Language 63: 21750
218
AUTHOR I NDEX
Abondolo, 101, 105
Agud, 32, 72, 207
Aikhenvald, 191
Allen, 150
Andersen, 163
Anderson, J., 62, 71, 803, 135, 197, 207
Anderson, S., 109
Anttila, 165
Aristotle, 18, 30
Armagost, 156
Artawa, 89, 124
Austerlitz, 165
Austin, 12, 101, 112, 1823
Baker, 187
Balakrishnan, 158
Blake, 15, 67, 78, 89, 107, 11415, 124, 1278,
137, 143, 151, 154, 158, 186, 188, 1923
Bloomeld, 13, 63
Bopp, 37
Bowe, 26
Branch, 105, 116
Breen, 28, 125
Bresnan, 47, 92, 185, 186, 187
Bryan, 102
Burridge, 174
Burrow, 117, 181
Caesar, 19, 73, 98
Calboli, 18, 19, 31, 48, 207
Cardona, 64, 65
Carroll, 138
Carvalho, de, 31, 40, 186, 192
Catullus, 184
Chafe, 66, 80, 207
Chaucer, 148
Chidambaranathan, 158
Chomsky, xiii, 47, 57, 58, 61, 62, 186, 187, 189
Cicero, 23, 112
Cole, 110
Coleman, 105
Comrie, 878, 104, 1401, 142, 152, 154, 185,
190, 191, 207
Cook, 66, 80
Corbett, 104
Croft, 192, 193, 207
Crowley, 100
Cutler, 99, 190
Delancey, 123, 192
Dench, 1089, 116, 154, 190
Derbyshire, 100, 191
Dik, 62, 66, 68, 912, 188, 207
Dionysius Thrax, 19
Dixon, 56, 105, 114, 124, 185, 191, 192,
193, 194, 207
DuBois, 192
Durie, 124, 192
Ebeling, 121, 152
Evans, 1089, 116, 154, 190
Filliozat, 64
Fillmore, 47, 623, 668, 74, 80, 82, 91,
188, 197, 207
Foley, 72, 92, 110, 192, 207
Frachtenberg, 190
Franchetto, 124
Friberg, 123
Gilligan, 99, 190
Giridhar, 100
Giv on, 112, 192, 193, 207
Goddard, 186
Greenberg, 15, 160
Groot, de, 30
Gruber, 679, 205
Haegeman, 59, 187, 190
Hale, 96, 107, 168, 194
Halliday, 66
Hansen, 193
Harris, 124, 126, 147, 192
Hawkins, 99, 190
Heath, 155, 159
Heine, 162, 165, 193
219
Author index
Hermon, 110
Hewitt, 123
Hinnebusch, 13
Hjelmslev, 8, 19, 30, 369, 956, 101, 152
Hoddinott, 104
Hopper, 1412, 191
Horace, xv
H ubschmann, 104
Huddleston, 191
Hudson, 76
Jackendoff, 92, 187, 188, 189
Jacobs, 166
Jacobsen, 181
Jacobson, 31, 37, 3941, 186
Jensen, 8
Juret, 30
Kanerva, 92, 187
Kaye, 14, 105
Keats, 16
Keen, 116
Keenan, 88
Kennedy, 181
Kibrik, 105, 135
Kilby, 102
Klaiman, 101
Klimov, 83, 124
Kofod, 104
Kornlt, 113
Kullavanijaya, 162
Kurylowicz, 31, 32, 33, 34
Langacker, 156
Langdon, 182
Larsen, 123
Lehmann, 172
Levin, 57
Lewis, 27, 184
Li, 16, 91, 139, 163
Lindenfeld, 101
Livy, 43
Longacre, 66
Lord, 164, 193
Luke, 176, 178
Luraghi, 1723
Lyons, 184
Madtha, 165
Maling, 187
Mallinson, 15, 120, 137, 192, 193
Marantz, 57
Martin, 367, 186
Matthews, 3, 18, 199
McAlpin, 147
McGregor, 190
Mchombo, 185, 186
McKay, 120
McLendon, 125
Melcuk, 9, 22, 27, 41, 102, 1067, 152,
203, 207
Merlan, 191
Mithun, 125, 184
Moravcsik, 151
Neidle, 40, 456, 187
Nichols, 13, 33, 34, 118, 159, 183
Norman, 123, 193
OGrady, 11
Palmer, 174, 194
P an
.
ini, 636, 74
Perialwar, 158
Perlmutter, 47, 7580, 148, 188
Pinkster, 89
Plank, 102, 151, 207
Planudes, 35, 36
Plato, 7
Pliny, 94
Polinsky, 152, 154
Postal, 47, 7580, 188
Press, 156
Pullum, 100, 191
Quintilian, 48
Ramarao, 43, 90
Rask, 37
Reed, 122, 14950
Riemsdijk, van, 60, 187
Rigsby, 166
Roberts, 187
Robins, 19, 31, 186
Robinson, 156
Rosen, 79, 1289, 188
Roth, 127
Rubio, 31, 35, 40, 73, 207
Rude, 125, 1667
Sakthivel, 158
Saltarelli, 100, 143
Sands, 193
Schooneveld, van, 39
Sch utze, 187
Scott, 122
Seidel, 3
Serbat, 35, 36
Sherzer, 190
Silverstein, 56, 63, 137, 140, 142,
190, 192
Simon, 356, 41, 186
Sittig, 18, 19
Smith, 29
220
Author index
Somers, 188, 207
Song, 185
Starosta, 16, 62, 836, 189, 197
Steever, 105, 106, 163
Stokes, 100
Stowell, 187
Suarez, 159
Tacitus, xv
Thompson, 16, 91, 139, 1412, 163, 191
Torrego, 98
Touratier, 334
Tsunoda, 168
Tucker, 102
Van Valin, 72, 92, 110, 192, 207
Varro, 19, 186
Vennemann, 98
Virgil, xv, 93, 96, 112
Wali, 1289
Weber, 103, 116
Whitney, 113
Wierzbicka, 40, 135, 142, 186, 189, 192
Williams, C., 106
Williams, E., 187
Windfuhr, 14, 165
Wolfart, 192
Woodcock, 31, 149, 185
W ullner, 37
Wyclif, 178
Yallop, 102
Yip, 187
221
LANGUAGE I NDEX
Abaza, 69, 123, 150, 191
Abkhaz, 123
Acehnese, 124
African, Southern, 16
African, West, 1634
Algonquian, 129
Altaic, 139, 149, 154, 1578, 160
Alyawarra, 102, 107
Amazonian, 100, 191
Arabic, 14, 105
Arabic, Classical, 156, 189
Arawak, 121
Archi, 105, 154, 190
Arikara, 124
Armenian, 103, 172
Armenian, Classical, 105, 158
Asian, South-east, 16, 163
Australian, xv, 11, 15, 26, 28, 56, 88, 96, 100,
1079, 11415, 120, 125, 139, 143, 145, 149,
151, 154, 159, 174, 1802, 186, 190, 1924
Australian, Northern, 14, 104, 116, 155, 159, 191
Austronesian, 9, 889, 121, 1234
Avar, 106, 1213, 139, 1512, 154, 191
Aztec-Tanoan, 139
Balinese, 89, 124, 197
Balto-Finnic, 99, 102, 105
Bantu, 13, 167, 31, 88, 119
Basque, 101, 121, 143, 154
Bengali, 101
Burushaski, 121
Caddo(an), 124
Cakchiquel, 166
Cambodian, 91, 158
Carib, 121, 123
Caucasian, 121, 1234, 149, 152
Caucasian, Northeast, 68, 71, 107, 139, 1512,
1545, 191, 193
Caucasian, Northwest, 123, 150, 156, 193
Caucasian, South, 193
Celtic, 60
Chemehuevi, 156
Chibchan, 121, 159
Chinese, 157, 139, 196
Chinook, 121
Chukchi, 121, 139
Comanche, 156
Coos, 191
Cree, 12930, 141
Czech, 157, 172
Daghestan, 27, 30, 95, 1057, 121
Dargwa, 139, 192
Dhalandji, 11213
Didinga-Murle, 157
Diegueno, 182
Diyari, 11, 101, 190
Djaru, 168
Dravidian, 30, 90, 100, 105, 139, 147, 149,
1545, 158, 163, 165, 172
Dyirbal, 567, 88, 105, 11416, 1357, 185,
189, 192
English, 2, 1011, 1416, 324, 53, 601, 66, 71,
746, 789, 818, 912, 968, 108, 11112,
116, 11920, 1324, 137, 13940, 145, 148,
163, 1678, 1713, 1768, 180, 182, 184,
1889, 191, 1959, 206
English, Middle, 148, 178
English, Old, 17, 104, 1489, 157, 171,
17680, 190
Eskimo, 38, 1212, 14950, 187
Estonian, 102, 151
European, 15, 118, 189
Ewe, 162, 1646
Fennic, 71
Finnic, 165
Finnish, 105, 11314, 116, 1514,
165, 206
Finno-Ugric, 151, 155
Fore, 122, 129, 141
French, 123, 70, 79, 87, 119, 138, 158, 175,
17980, 1934, 197
Fur, 157
222
Language index
G a, 164
Georgian, 90, 102, 124, 126, 143, 145, 1478
Georgian, Old, 102, 107
German, 11, 13, 35, 58, 79, 95, 98, 102, 157,
163, 174, 177, 193, 200
German, Middle High, 157
German, Old High, 157, 177
German, Pennsylvania, 174
Germanic, 13, 101, 104, 157
Gooniyandi, 100, 190
Greek, Ancient, xv, xvi, 3, 78, 1819, 26, 2830,
367, 39, 44, 46, 48, 51, 63, 90, 95, 99, 106,
114, 133, 1423, 155, 157, 172, 202
Greek, Modern, 91, 156
Guaran, 1245
Gujerati, 166
Gunwinygu, 138
Gyarong, 122
Hausa, 166
Hindi, 10, 28, 120, 166
HindiUrdu, 10, 128, 165
Hittite, 172
Hokan, 125
Hungarian, 14, 101, 105, 151, 165
Hurrian, 102, 121
Icelandic, 157
Indo-Aryan, 10, 1001, 1289, 199
Indo-Aryan, Middle, 174
Indo-European, Proto, 158
Indo-European, xvi, 7, 1819, 30, 40, 51, 88, 90,
99, 1036, 11819, 139, 149, 151, 154, 1601,
166, 1726, 181
Indo-lranian, 154, 161
Indonesian, 1415, 66, 87
Iranian, 128, 156, 192
Iroquoian, 124
Irula, 143, 158
Italian, 79, 87, 148, 171, 176, 179, 1923, 197
Jacaltec, 166
Japanese, 9, 11, 119
J e, 121
Kabardian, 156, 193
Kala Lagau, Ya, 1801
Kalkatungu (Kalkadoon), xv, xvi, 412, 4756,
63, 66, 767, 79, 86, 889, 11415, 1389,
143, 145, 1545, 1812, 186, 189
Kannada, 100, 159, 165
Karelian, 165
Kartvelian, 124, 126, 147
Kasaba, 158
Kayardild, 108
Kiranti, 122
Kodaga, 158
Komi-Permyak, 154, 165
Konjo, 123, 150
Korean, 11, 119, 185
Kpelle, 166
Kuikuro, 123
Kurdish, 128
Kwa, 162
Lak, 27, 194
Lakhota, 1245
Latin, xv, xvi, 310, 14, 1724, 26, 28, 317, 39,
425, 46, 48, 514, 612, 66, 724, 80, 89, 90,
938, 1034, 11114, 11820, 133, 1427,
149, 1525, 157, 160, 166, 1689, 1712,
1746, 1789, 182, 1846, 18991, 194,
1967, 199, 202
Laz, 124, 126, 193
Lezgian, 107
Luritja, 194
Maba, 102
Malayalam, 147
Mandarin Chinese, 1634
Mangarayi, 190, 1923
Maon, 184
Marathi, 1289
Margany, 28, 193
Mayan, 121, 123, 139, 150
Mexican, 149, 154
Mixtec, 166
Mohawk, 124, 184
Mongolian, 99
Munda, 100, 139
Muskogean, 182
Nanai, 154, 158
Near Eastern, 121
Nez Perce, 125, 167
Nhanda, 137
Niger-Congo, 162
Nilo-Saharan, 102, 1567
Nubian, 156
Nuer, 157
Numic, 156
Nungali, 104, 190
Nupe, 164
Nyigina, 100
Oceania, 163
Ossete, 154
Pama-Nyungan, 13, 246, 28, 96, 99, 101, 1056,
112, 116, 126, 1389, 154, 158, 1679, 1726,
185, 191, 1934
Pama-Nyungan, proto, 168
Pamir, 192
Panoan, 121
223
Language index
Panyjima, 116
Papuan (New Guinea), 99, 1212, 163
Pashto, 128
Penutian, 121, 125, 139, 1667
Persian, 14, 165
Pintupi, 139, 192
Pitjantjatjara, 26, 52, 194
Pitta-Pitta, 15, 42, 116, 1268, 143, 1456, 170,
172, 174, 180, 190, 192
Polish, 151, 157
Pomoan, 124
Pomo, Eastern, 1256
Punjabi, 128
Quechua, 99, 103, 107, 110, 116, 172
Rembarnga, 120, 122, 129, 141, 191
Romance, 119, 171, 1759, 194
Rumanian, 120, 171, 176, 193
Rushan, 192
Russian, 9, 22, 27, 29, 33, 3941, 456, 104,
119, 151, 154, 172
Sahaptin, 1667
Sanskrit, 51, 636, 74, 117, 154, 163, 181
Semitic, 14, 99, 105, 149, 156
Seneca, 124
Serbo-Croatian, 104, 157
Siouan, 124
Siuslaw, 191
Slavonic, xvi, 104, 119, 157
Slovak, 157
Sora, 139
Spanish, 79, 11920, 1716
Sulawesi, South, 123
Sumerian, 172
Swahili, 135, 17, 87, 90, 139, 166
Tabassaran, 956, 101, 152
Tacanan, 121
Tagalog, 89
Tamil, 42, 105, 154, 158, 164, 166
Tarascan, 154, 159
Telugu, 43, 90, 154
Thai, 14, 87, 91, 158, 1612,
164, 192
Tibetan, 154
Tibeto-Burman, 121, 123
Tiwa, 139
Toda, 158
Tonkawa, 154
Tsimshian, 121
Tungusic, 99, 158
Tup-Guaran, 121
Turkana, 139
Turkic, 99
Turkish, 13, 9, 12, 278, 87, 90, 105,
113, 119, 139, 151, 158, 160, 167,
172, 184, 190, 197, 203
Ubykh, 193
Uradhi, 100
Uralic, 118, 139, 149, 1512, 154, 165
Uto-Aztecan, 13, 101, 154, 156, 182
Vietnamese, 145, 196
Wakashan, 139
Walmatjari, 139
Wangkumara, 125
Wappo, 90, 139
Warlpiri, 96, 107, 109, 168, 194
Warndarang, 155, 159
Western Malayo-Polynesian, 89
Yaghnobi, 156
Yalarnnga, 57
Yapese, 8
Yaqui, 101, 157
Yatye, 164
Yidiny, 160, 193
Yoruba, 164
Yukulta, 108, 116
Yuman, 182
Yupik Eskimo, 69, 122, 14950
Yuwaalaray, 106
Zoque, 121, 149, 154
224
SUBJ ECT I NDEX
A, 259, 4857, 1301, 1356, 185, 195
abessive, 154, 195
ablative, 29, 19, 358, 425, 803, 1523,
15760, 164, 186, 195
absolute, 186
absolutive, 259, 547, 7983, 13542,
185, 195, 199, 202
accusative, 16, 9, 20, 25, 2930, 41,
15760, 195
accusative and innitive, 114
accusative language, 25, 88, 11820,
13142, 195
active language, 1245, 195
active voice, 195
adessive, 153, 195
adjunct, 334, 42, 188, 196
adnominal, 149, 196
adposition, 912, 71, 98, 1614, 170, 1758, 196
advancement, 76, 82, 88, 139, 196
adverb, 156, 24, 1667
adverbal, 42, 196
adverbial, 196
afx, 196
agent, 68
AGENT, 846
agglutinative, 3, 154
agreement, 1314, 21, 13940, 184, 186, 189,
196, see also concord
allative, 38, 42, 133, 164, 196
analytic case marker, 912, 196
analytic language, 196
animate object, 120
antipassive, 49, 88, 115, 196
applicative, 197
apposition, 23, 512, 71
argument, 1, 187, 196, 197
ascriptive, 40, 45
aversive, 28, 42, 154, 197
benefactive, 143
beneciary, 69, 143
bound pronoun, 124, 513, 191, 197,
see also clitic
case, 1, 18, 48, 197
abstract case, 5762
adnominal case, 42, 968, 14951
compound case, 41, 1067, 198
core case, 334, 43, 505, 11842, 198
grammatical case, 313, 200
inherent case, 5861
local case, 33, 45, 1514, 202
multiple case, 1023, 1079, 202
nonautonomous case, 224, 27, 203
peripheral case, 324, 43, 50, 203
prexed case, 104, 190
semantic case, 313, 170, 204
structural case, 456, 5862, 205
ungoverned case, 9
universal case, 62
case features, 3446
case form, 3, 1829, 197
case grammar, 47, 62, 6686, 197
case in tiers, 187
case inection, 19
case language, 1, 197
case marker, 2, 912, 1617,
1901, 197
analytic case marker, 912
synthetic case marker, 9
case marking, 93117
exceptional case marking, 61
loss of case marking, 17580
case relation, 3, 54, 63, 806, 197
case resistance, 187
case system, 2, 197
case/number form, 18
casus, 18
causal, 154
ch omage, ch omeur, 7580, 197
clitic, 12, 26, 523, 96, 123, 197, 199
coherence, 38
comitative, 154, 15860, 164, 197
comment, 197, 205
comparative case, 155
complement, 334, 42, 185, 197
concomitant, 154
225
Subject index
concord, 5, 7, 93117, 168, 184, 189, 198,
see also agreement
control, 198
core, 11842, 198, see also case, core
CORRESPONDENT, 846
cross-referencing, 134, 512, 1213,
14950, 198
cumulative exponence, cumulation, 3,
18, 198
dative, 16, 357, 3944, 1428, 15760,
164, 174, 197, 199
dative of agent, 73
declension, 4, 18, 199
declinatio, 18
default, 199
denite, 199, see also specic
dependent, 1, 7, 199
deponent verb, 186
derivation, 1617, 199
destination, 69, 144
detransitivise, 114, 199
direct case, 10, 128, 199
direct object, see object
direct-inverse, 12930, 199
direction, 40
discriminative case, 190
double object, 25, 29, 53, 756, 13940,
189, 199
elative, 153, 199
eloignement, 378
enclitic, see clitic
equative, 38
ergative, 249, 38, 42, 4857, 7983,
88, 1214, 13542, 156, 191, 199
essive, 153, 200
evitative, 154, 200
experiencer, 68, 146
extension(al), 39, 200
extent, 69
nite, 162, 200
function, 3, 200
fusion, 34, 18, 161
gender, 4, 1314, 17, 105
genitive, 2, 4, 40, 434, 14951,
15660, 174, 184, 200
gerund, 44, 186
gerundive, 73, 96, 1467, 188
Gesamtbedeutung, 35, 3940, 200
given/new, 131, 140, 200
govern, 200
Government and Binding, 5862
grammatical relation, 3, 8, 48,
92, 201
Hauptbedeutung, 39
head, 1, 1314, 198, 201
head marking, 1314, 201
hierarchy, 201
case, 8990, 15560
case assigners, 60
case marking, 901
givenness, 1312
grammatical relations, 75, 869
person/animacy, 122, 13741, 192
person/number, 120
roles, 912
holistic, 201
hyponym, 201
illative, 153, 201
incoherence, 38
indirect object, 1, 6, 756, 1449, 201
inessive, 153, 201
innitive, 182
inection, 201
instrument(al), 25, 28, 3842, 4757,
68, 15460, 164, 201
intension(al), 39, 201
inverse, 12930, 201
inversion, 1479, 201
isomorphic, 19, 202
k araka, 636
Lexical Functional Grammar, 457, 187
Lexicase, 3, 836
localist, 36, 81, 202
Localist Case grammar, 3, 803
location, 68
locative, 2, 224, 3843, 803, 1523,
1569, 164, 202
LOCUS, 836
manner, 69
marginal, 3940, 456
markedness, 3446, 202
meaning, 202
MEANS, 836
merger, 170, 174
Minimalism, 62
mixed (erg/acc) system, 1259
modal case, 1089, 190
modier, see dependent
morpheme, 18
neutralisation, 20, see also syncretism
new, see given/new
nominal 202
nominalisation, 978, 11017, 202
nominative, 14, 301, 15660,
1856, 202
226
Subject index
nonnite, see nite
number marking, 34, 105, 139
O, 185
object, 1, 5, 889, 1335, 164, 199, 203
double object, 189, 199
secondary object, 204
object marking, 191
objective case, 187
objective genitive, 17, 190, 203
oblique case, 10, 19, 301, 42, 44, 156,
186, 203
oblique relation, 75, 203
obviative, 130, 203
onoma, 31
P, 259, 4857, 185, 203
paradigm, 2, 168, 203
participant, 203
partitive, 22, 151, 153, 186, 203
passive 6, 32, 59, 65, 72, 756, 139
path, 68
patient, 678
PATIENT, 836
peripheral, 43, 203
perlative, 38, 153, 164, 203
person, 203
phonological factors, 1679, 1758, 194
phrase-marking language, 100, 185, 204
possessive adjective, 17, 178, 191
possessor, 2, 69, 105, 143, 14951
possessor ascension, 77
postposition, 912, 204, see also adposition
secondary postposition, 10
predicate, 58, 187, 189, 204
predicative, 5, 25, 93, 101, 189, 204
prepostion, 5, 1758, 204, see adposition
privative, 154
proclitic, see clitic
proprietive, 154
proximate, 130
pt osis, 18
purpose, 69, 143
purposive, 143, 15860
quantifying, 40, 45
raising, 83, 188
rapprochement, 378
recipient, 1, 69
redundancy, 70, 130, 16970
reexive 49
Relational Grammar, 3, 47, 7580
relative case, 149
relator noun, 157, 1645, 204
rhema 31
role, 2, 6292, 204
role assignment, 704
S, 259, 4857, 186, 204
semantic roles, 6292, 205
serial verbs, 1634
sociative, 154
Sonderbedeutungen, 39
source, 68
specic (object), 12, 119, 205
specier, 58, 62
split (of case), 1745
split-intransitive, 124
split-S, 124
stem formation, 105, 205
subessive, 153, 205
subject, 1, 5, 557, 1312, 205
subjective case, 187
subjective genitive, 17, 191, 205
subordinate clause 109, 117
superessive, 153, 205
superlative 153, 205
supine 45
suppletion, 161
suppositum, 36
switch-reference 1823
syncretism, 1929, 41, 434, 169, 205
syntactic relation, 2
syntagmatic 81, 205
system 2, 25, 205
term, 75
theme, 67, 205
topic 31, 207
transitive 206
translative 206
Universal Alignment Hypothesis, 779
unmarked, see markedness
valency 206
vocative, 45, 8, 21, 43, 185, 206
wholepart construction, 26, 52
word-marking language, 99, 185, 206
word order, 1415, 138
227

J. Blake, Barry. Case

Uploaded by

Copyright:

Available Formats

J. Blake, Barry. Case

Uploaded by

Document Information

Original Title

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

J. Blake, Barry. Case

Uploaded by

Copyright:

Available Formats

What is the book about?

What are some of the concepts discussed in the book?

This page intentionally left blank

See (8), (9) and (10)

sna) -sta -sse (< -

lna) -lta -lle (<

You might also like