Serial Verbs in White Hmong (PDFDrive)

Serial Verbs in White Hmong
Brill’s Studies in Language,

Cognition and Culture
Series Editors
Alexandra Y. Aikhenvald (Cairns Institute, James Cook University)

R.M.W. Dixon (Cairns Institute, James Cook University)
N.J. Enfield (Department of Linguistics, University of Sydney)
VOLUME 12
The titles published in this series are listed at brill.com/bslc

Serial Verbs in White Hmong
By
Nerida Jarkey
LEIDEN | BOSTON
Cover illustration: Hmong working life in Laos, embroidery by an anonymous artist. Photograph by
Thomas A. Roberts.
Library of Congress Cataloging-in-Publication Data
Jarkey, Nerida, author.

Serial verbs in White Hmong / by Nerida Jarkey.
pages cm. — (Brill’s studies in language, cognition and culture, ISSN 1879-5412; volume 12)
Includes bibliographical references and index.
Summary: “Investigates verb serialization, a highly productive grammatical strategy in this dynamic
Southeast Asian language in which multiple verbs are simply concatenated within a single clause to depict
a single event. The investigation identifies four major types of serial verb construction (SVC) in White
Hmong and finds that the key function of all these types is to depict a single event in an elaborate and
vivid way, a much-favoured method of description in this language. These findings concerning the nature
and function of SVCs in White Hmong contribute to broader discussions on the nature of events as both
cognitive and cultural constructs”—Provided by publisher.
ISBN 978-90-04-29240-6 (hardback : alk. paper) — ISBN 978-90-04-29239-0 (e-book) 1. White Hmong
dialect—Verb phrase. I. Title.
PL4072.95.W45J37 2015
495.9’72—dc23
2015014547
This publication has been typeset in the multilingual “Brill” typeface. With over 5,100 characters covering
Latin, ipa, Greek, and Cyrillic, this typeface is especially suitable for use in the humanities.
For more information, please see www.brill.com/brill-typeface.
issn 1879-5412
isbn 978-90-04-29240-6 (hardback)
isbn 978-90-04-29239-0 (e-book)
Copyright 2015 by Koninklijke Brill nv, Leiden, The Netherlands.

Koninklijke Brill NV incorporates the imprints Brill, Brill Hes & De Graaf, Brill Nijhoff, Brill Rodopi and
Hotei Publishing.
All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system,
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise,
without prior written permission from the publisher.
Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill nv provided
that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive,
Suite 910, Danvers, ma 01923, usa. Fees are subject to change.
This book is printed on acid-free paper.

In memory of my mum, Bette
∵
Contents
Acknowledgements xi
List of Figures xii
List of Tables xiii
List of Abbreviations xiv
Introduction 1
1 Introduction to White Hmong Language 7

1.1 About the Language Family and the Locations of its Speakers 7
1.2 Phonology and Orthography 11
1.2.1 The Consonant Phonemes and their Allophones 11
1.2.2 The Vowel Phonemes and their Allophones 13
1.2.3 Tones 14
1.2.4 Syllable Structure 22
1.3 Syllables, Morphemes and Words 23
1.3.1 Sub-Syllabic Morphotonemes 23
1.3.2 Polysyllabic Morphemes 26
1.3.3 Multi-Morphemic Words 26
1.4 The Noun Phrase 31
1.4.1 The Basic Structure of the Noun Phrase 31
1.4.2 Classifiers and Measures 33
1.5 The Locative Phrase and Spatial Deictics 40
1.6 Simple Clause Types 43
1.6.1 Existential Clauses 43
1.6.2 Nominal Subject Complement Clauses 45
1.6.3 Intransitive Clauses 46
1.6.4 Reciprocal Clauses 47
1.6.5 Transitive Clauses 48
1.6.6 Ditransitive Clauses 53
1.7 Operations on the Simple Clause: Mood, Modality, and Aspect 56
1.7.1 Imperative 56
1.7.2 Interrogative and Uncertainty with puas 57
1.7.3 Negative with tsis and txhob 58
1.7.4 Certainty with twb 59
1.7.5 Future, Likelihood, and Intention with yuav 59
1.7.6 Likelihood and Obligation with yuav-tsum and yuav-tau 62
1.7.7 Deontic Modality with a Complementation Strategy 63
viii contents
1.7.8 Deontic Modality with Postverbal tau 63

1.7.9 Achievement with Preverbal tau 65
1.7.10 Imperfective Aspect 67
1.7.11 The Perfect with lawm 68
1.8 Conclusion to Chapter One 69
2 Introduction to Serial Verb Constructions 70

2.1 The Term ‘Serial Verb Construction’ 72
2.1.1 The Term ‘SVC’ Used for Distinct Actions 74
2.1.2 The Term ‘SVC’ Used for a Single Action/Event 75
2.2 The Notion of a Single Event 76
2.2.1 Construal as a Single Event 78
2.2.2 Packaging as Single Event 80
2.3 The Notion of Joint Predication 82
2.4 Syntactic Framework for the Analysis of SVC Types in
White Hmong 84
2.4.1 Key Notions for the Analysis 85
2.4.2 The Syntactic Analysis of White Hmong SVCs Proposed 92
2.5 Conclusion to Chapter Two 101
3 Productive Serial Verb Constructions in White Hmong 102

3.1 Cotemporal Serial Verb Constructions 106
3.1.1 Cotemporal Motion Serial Verb Constructions 106
3.1.2 Cotemporal SVCs Involving Actions 115
3.1.3 Nexus and Level of Juncture in Cotemporal SVCs 118
3.2 Cause-Effect Serial Verb Constructions 124
3.2.1 Constraints on the Verbs Appearing in Cause-Effect SVCs 127
3.2.2 The Type of Causation Described by Cause-Effect SVCs in White
Hmong 134
3.2.3 Nexus and Level of Juncture in Cause-Effect SVCs 136
3.2.4 Cause-Effect SVCs: One Subject or Two? 139
3.3 Attainment Serial Verb Constructions 147
3.3.1 Introduction 147
3.3.2 Types of Predicate in Attainment SVCs 148
3.3.3 Components Common to All Attainment SVCs 159
3.3.4 Comparison of Attainment SVCs with ‘Resultative Verb
Compounds’ in Mandarin 160
3.3.5 Rules for Deriving Attainment SVCs: Syntactic or Lexical? 161
3.3.6 Nexus and Juncture in Attainment SVCs 164
3.3.7 Grammaticalization of V2 tau to Express Deontic Modality
‘can’ 166
Contents ix
3.4 Disposal Serial Verb Constructions 170

3.4.2 The Semantic Relationship Between the Verbs 171
3.4.3 The Position of the Shared Object 173
3.4.4 Nexus and Juncture in Disposal SVCs 174
3.4.5 Disposal SVCs Compared with Other SVC Types 175
3.4.6 Disposal SVCs with muab ‘take’ 177
3.4.7 V2 Introducing an Unshared Argument 179
3.5 Conclusion to Chapter Three 181
3.5.1 SVCs Express Various Facets of a Single Event 181
3.5.2 Linked Serial Verb Constructions 183
3.5.3 Relationship Between Lexical Complexity and Serialization 186
4 Valency-Increasing SVCs and Grammaticalization 188

4.1 The Phenomenon of Valency-Increasing SVCs 188
4.1.1 Adposition or Verbs? 188
4.1.2 Criteria for Part-of-Speech Classification 190
4.1.3 Part-of-Speech Constraints vs Environmental Constraints 192
4.1.4 Part-of-Speech Change: Discrete or Gradual? 193
4.2 Valency-Increasing Serialization in White Hmong 196
4.2.2 Criteria Relevant to Part-of-Speech Classification in White
Hmong 198
4.2.3 Examination of Valency-Increasing Serialization in Hmong 202
4.3 Conclusion to Chapter Four 230
5 Superficially Similar Constructions 233

5.1 Elaborate Expressions 233
5.2 Juxtaposed Clauses 237
5.3 Complementation Resembling Cause-Effect SVCs 241
5.3.1 Complements of Non-Implicative Manipulative Verbs that Appear
without a Complementizer 242
5.3.2 Complements of the Causative Verb ua and the Permissive
Verbs 246
5.3.3 Complements of Perception Verbs 248
5.4 A Complementation Strategy with Modal Verbs 249
5.5 Purpose Constructions 253
5.5.1 Semantic and Lexical Differences Between SVCs and Purpose
Constructions 254
5.5.2 Syntactic Differences Between SVCs and Purpose
Constructions 261
x contents
5.5.3 Syntactic Similarities Between SVCs and Purpose

Constructions 264
5.6 State Verbs Used Attributively and Adverbially 266
5.6.1 State Verbs Used Attributively 266
5.6.2 State Verbs Used Adverbially/Depictively 267
5.7 Transitive Verbs Used Descriptively and Adverbially 270
5.7.1 Transitive Verbs Used Descriptively 270
5.7.2 Transitive Verbs Used Adverbially 272
5.8 Conclusion to Chapter Five 273
6 Conclusion 275
6.1 The Notions of a Single Event and of Joint Predication 275
6.2 The Syntactic Analysis of SVCs in White Hmong 276
6.3 The Common Properties of SVCs in White Hmong 277
6.4 The Functions and Semantics of SVCs 278
6.5 The Relationship Between Function and Part of Speech 281
6.6 Superficially Similar Constructions 282
6.7 Directions for Further Research 284
Sources for White Hmong Examples 287

Bibliography 288
Index of Authors 298
Index of Languages and Language Families 300
Index of Subjects 302
Acknowledgements
This study began as my PhD project in 1985. I’m so pleased to have this oppor-
tunity to say thank you again to my dear Hmong friends who acted as my lan-
guage teachers and consultants at that time. My most sincere gratitude goes
to Cua Lee for her kind patience and her joyful laugh as we worked together
with our babies and toddlers tumbling around us. Special thanks also to Cua’s
brother-in-law, Yeeb, for many fascinating discussions on Hmong language.
For additional enthusiastic help, I am obliged to Cua’s husband Sao, his broth-
ers Gary (Nkaj Yia) and Yue (Ntxawg), their much-loved, late mother Maiv Yaj,
to Ge and Niaj Pov Lis, Charlie Sayaxang, Sourivan Thongpao, and Maiv Dub
Yaj. More recently I have received further generous support from Jenny Ly and
William Lee, from Zoo Lis and Thaiv Thoj, and from Mai Vue.
I was extremely privileged to be guided during the course of my PhD candi-
dature by three most insightful and generous supervisors: Bill Foley, Marybeth
Clark and Michael Walsh. My gratitude also goes to three other outstanding
scholars in the field, David Bradley, Martha Ratliff, and David Strecker. Sincere
thanks, too, to all my linguistics teachers, particularly Bob Dixon, Tim Shopen,
and Anna Wierzbicka, who opened my eyes to the thrill of language.
Without the incredibly kind encouragement, guidance and mentorship of
Bob Dixon and Sasha Aikhenvald, I may never have returned to my study of
Hmong, after a hiaitus of many years. My time with them as a visiting scholar
at the Research Centre for Linguistic Typology at La Trobe University in 2006,
and at the Language and Culture Research Centre at James Cook University
in 2013, allowed me to revise and extend the work I had done for my Phd the-
sis and enabled me to produce this book. Bob and Sasha’s input and support
in this project have been invaluable. I am also indebted to Nick Enfield for
his extremely useful suggestions on the penultimate draft of the manuscript.
Thanks to all the editorial team at Brill, for their patience and expertise in
bringing this book to publication.
To my family and friends I cannot begin to express my gratitude, nor fully
list the numerous ways in which they have helped me. Most of all I am grateful
for their encouragement and love. This I received from them all, but especially
from my dear husband Aubrey, my late mum Bette and my great dad Max, my
wonderful brother Don, and my sweet sons Alex and Dom.
My initial research on White Hmong during my PhD candidature was sup-
ported by an Australian Postgraduate Research Award.
List of Figures
Figure caption
1 Monophthongs 14
2 Opening diphthongs 14
3 Closing diphthongs 14
4 Diagrammatic representation of a simple clause 86
5 Diagrammatic representation of nuclear juncture 88
6 Diagrammatic representation of core juncture in which the point
of intersection is a shared core argument 89
7 Diagrammatic representation of core juncture in which one junct
is a subset of the core of the other 89
8 Diagrammatic representation of clausal juncture in which
the intersection of the two peripheries may be null, partial, or
total 89
9 Diagrammatic representation of clausal juncture in which one
junct is a subset of the periphery of the other 90
10 Diagrammatic representation of SVCs involving nuclear
coordination 94
11 Diagrammatic representation of SVCs involving nuclear
cosubordination 94
12 Diagrammatic representation of SVCs involving core
cosubordination 97
13 Diagrammatic representation of core-level coordinate juncture in
which two cores share a single periphery 99
List of Tables
Table caption
1 The consonant phonemes and their allophones I 12

2 The consonant phonemes and their allophones II 13
3 Tone symbols, pitch, voice quality and length 15
4 Tone sandhi changes 17
5 Productive SVC types—verbs and argument sharing 103
6 Verb types and order in Cotemporal SVCs 111
7 Verb types and order in Disposal SVCs 172
List of Abbreviations
1 first person
2 second person
3 third person
A transitive subject
ACC accusative
ACHV achievement
AGT agent
ASP aspect
ATTR attribute
AUX auxiliary
CAUS causative
CC copula complement
CLF classifier
CLF.PL plural classifier
COMP complementizer
COMPL completive
CONJ conjunction
COP copula
CS copula subject
DECL declarative
DEF definite
DEM demonstrative
DIR directional
DNH Dab Neeg Hmoob (Johnson, 1985)
DU dual
ELEV elevational
EMPH emphatic
FACT factitive
HAB habitual
HFT Hmong folk tales (Johnson, 1981)
IMM immediate
IMP imperative
INCL inclusive
INDF indefinite pronoun
INTNS intensifier
INTR intransitive
IP illocutionary particle
List Of Abbreviations xv
IPFV imperfective
IRR irrealis
IWH ‘Introduction to White Hmong’ (Southeast Asian Summer Studies
Institute, 1985)
KNH Kawm Ntawv Hmoob (Overseas Missionary Fellowship, 1976)
Lit. literal translation
LOC locative
LP locative phrase
M masculine
NEG negative
NP noun phrase
NS new situation
NUM numeral
PAT patient
O transitive object
ONMT onomatopoeia
PFV perfective
PL plural
PN proper noun
POSS possessive/possessor
PRF perfect
PST past
Q question marker
QUANT quantifier
REAL realis
REC recent
RECP reciprocal
RDUP reduplicated form
REL relative marker
REM remote
RRG Role and Reference Grammar
S intransitive subject
SG singular
SVC serial verb construction
TAM tense, aspect, mood or modality
t.c. tone change
THW The Hmong World (Tsawb & Strecker, 1986)
TOP topic
TR transitive
U undergoer
xvi list of abbreviations
UNCERT uncertain
V Verb
VIS visual
VLT Vim leejtwg (Lis, N.P. (n.d.))
VP verb phrase
WHD White Hmong Dialogues (Strecker & Vang, 1986)
WHG ‘Notes on White Hmong grammar’ (Lis, Y. (n.d.))
Introduction
About This Study
This study is a substantially revised version of work that originally began

as my PhD thesis, submitted in 1991. It examines the phenomenon of serial
verb constructions in White Hmong, a language spoken in the mountainous
regions of Southern China and Mainland Southeast Asia, and by a number of
diasporic communities around the world. The study adopts the view that, in
cross-linguistic terms, verb serialization should be thought of as a strategy or
technique rather than as a ‘construction’ (Aikhenvald, 2006b, p. 2), and seeks
to understand how this strategy is utilised and for what purposes, in which
construction types in this particular language.
The term ‘serial verb construction’ (SVC) has been used in very different
ways to describe a diverse array of construction types in many languages, all
of which involve the ‘simple’ juxtaposition of verbs. This study follows major
typological research (Aikhenvald and Dixon (2006); Bisang (2009); Durie
(1997); Foley and Olson (1985); Givón (1991)) in using the term to refer only to
constructions in which verbs are juxtaposed within a single clause to describe
a single event. In White Hmong, and in many other languages in which the
simple juxtaposition of verbs is extremely common, this narrows down the
candidates for inclusion in the class of serial verb constructions considerably.
However, even scholars who adopt the ‘single clause, single event’ criteria
still tend to include quite a varied assortment of constructions, all of which
involve the simple juxtaposition of verbs, within the class of ‘serial verb con-
structions’. For example, in the languages of Mainland Southeast Asia, in par-
ticular, there are many cases of juxtaposed verbs in which one verb functions
as the complement of the other, but which do assert the occurrence of only
one event, and in which the complement exhibits few features of an indepen-
dent clause.
A careful investigation of the full range of constructions containing jux-
taposed verbs in White Hmong reveals the importance of an additional
criterion—often mentioned in the literature—for delimiting the class of serial
verb constructions in this language: that the two or more verbs work together
as joint predicators to express a single proposition. This criterion excludes con-
structions in which one verb is part of a complement, and therefore sub-prop-
ositional. In this way, the study identifies four major serial verb construction
types in White Hmong—Cotemporal SVCs, Cause-Effect SVCs, Attainment
SVCs, and Disposal SVCs—and distinguishes these from superficially similar
© koninklijke brill nv, leiden, ��5 | doi ��.��63/9789004292390_002

2 Introduction
constructions in the language, including juxtaposed clauses, Purpose Con-

structions, and various types of complementation involving manipulative and
modal verbs.
By considering these four major constructions together as types of serial
verb constructions, and by observing similarities between them in their uses
in narrative texts, the study succeeds in identifying a common function for
which the strategy of serialization is used in this language. A key finding of this
study is that these four SVC types in White Hmong all work in different ways to
provide detail and elaboration concerning the single event they described. In
some cases, this elaboration relates entirely to an action or process; in others
it relates not only to an action or process, but also to its outcome. In all cases
the strategy of verb serialization supports the use of two or more verbs work-
ing as joint predicators of a single proposition, to express a single event within
a single clause.
After briefly describing the data sources for this research below and mak-
ing a note on the glossing and translation of examples, the study begins in
Chapter 1 with a short introduction to the White Hmong language, including a
very brief discussion of its genetic relationships and the location of speakers,
of its phonology and orthography, and of issues relating to the distinction in
this language between the syllable, the morpheme and the word. Chapter 1
goes on to describe the key components of the Noun Phrase, and to introduce
the Locative Phrase and Spatial Deictics. The description of grammar goes only
as far as the level of the clause, with an overview of simple clause types and
of operations on the simple clause. More detailed discussion of verb types, of
the valency-increasing function of some verbs, and of certain types of comple-
mentation is reserved for the later chapters in which these issues are most per-
tinent to the focus of the study.
Chapter 2 provides an introduction to use of the term ‘serial verb construc-
tion’, and to the notions of a single event and of joint predication. With a view
to better understanding the way in which the verbs work together in a single
clause, Chapter 2 also describes the syntactic framework that has been found
useful for the analysis of serial verb constructions in White Hmong. This frame-
work is based on the Role and Reference Grammar model of the clause and of
the linkage of clausal components (Foley & Van Valin, 1984; Van Valin, 1987,
2005; Van Valin and LaPolla, 1997), but with an important modification that is
necessary to account for the tightly packaged syntactic structure of serial verb
constructions.
In Chapter 3, the four major types of serial verb constructions in White
Hmong, along with a number of interesting sub-types, are described and
compared in detail. This description focuses particularly on the nature of the
Introduction 3
shared arguments, on the verb types that appear in each position in each of
these constructions, and on the ways in which these variables work within the
syntactic structure to portray the various facets of different event types in each
case. This chapter also briefly investigates the phenomenon of linked SVCs, a
strategy that is used together with a range of others in narrative texts to portray
an unfolding event in the intricate detail much favoured in the narrative genre
in this language.
The use of serial verb constructions as valency-increasing devices is the
topic of Chapter 4. The analysis focuses on understanding which type of SVC
each of these uses derives from, and on examining the features of the usage in
the context of that construction type in order to determine the extent to which
a process of grammaticalization may be underway in each case. It finds that
there is clear evidence of grammaticalization in some cases—that is, those in
which the erstwhile verb has lost its predication relationship with the subject
of the sentence. In many other cases, however, the fact that a verb is function-
ing in a valency-increasing role is seen to be simply another way in which the
strategy of serialization is used in White Hmong to provide detail and elabo-
ration within a single clause, and there is no reason to assume that part-of-
speech change has taken place.
The final chapter before the Conclusion to this study—Chapter 5—exam-
ines a wide range of constructions that are superficially similar to serial verb
constructions in this language. These include Elaborate Expressions, juxta-
posed clauses, various kinds of complementation without a complementizer,
Purpose Constructions, and State Verbs and Transitive Verbs used both attribu-
tively and adverbially.
As noted above, an issue of considerable interest arising from the investiga-
tion in Chapter 5 is the way in which certain types of complementation bear a
strong resemblance to serial verb constructions in White Hmong. This resem-
blance is found to be due to a high degree of ‘binding’ between the main and
complement verbs (Givón, 1980, 1985, 2001b). Binding relates to the extent to
which the proposition described by the main verb influences the likelihood
of the realisation of that described by the complement verb. In various types
of complement constructions, factors such as implicature, the degree of influ-
ence of a manipulator over the other participant, and the degree of emotional
attachment of the subject to the outcome all affect the degree of binding.
A high degree of binding can result in some complement verbs not having a
truth-value independent of the main verb, and/or not being able to express any
kinds of temporal operations independently of the main verb. Another con-
sequence of binding is that the complement is more syntactically integrated
into the main clause. The distinction between highly bound complementation
4 Introduction
types and certain serial verb construction types in this language is thus blurred
to a considerable extent.
This issue is just one example of the way in which the findings of this inves-
tigation, while responding to some important questions about the nature and
function of verb serialization in White Hmong, raise many more interest-
ing questions, both in relation to this language and of wider, cross-linguistic
significance.
About the Data Used for the Study
The primary source of data for this study was a volume of twenty-seven White
Hmong narrative texts collected and edited by Charles Johnson, entitled Dab
Neeg Hmoob: Myths. Legends and Folk Tales from the Hmong of Laos (Johnson,
1985). This volume is not only of immense cultural and literary worth but, for
the purpose of the linguistic analysis of oral Hmong texts in the narrative style,
it is extremely valuable.
There are three main reasons for its great value in this respect. Firstly, the
majority of the texts were furnished by Hmong elders with little or no formal
education; only five of the texts were collected from young or middle-aged
Hmong (p. xiii). It is reasonable to assume that the Hmong of these older
speakers would be less likely to exhibit any influences—either lexical or
grammatical—from languages of recent historical contact, with the possible
exception of Lao and Thai. The occurrence of a few slightly archaic lexical
items and formulaic expressions in the texts may make them less than ideal
for the purposes of some types of linguistic investigation, but are of marginal
relevance as far as the present study is concerned.
Secondly, the texts are of particular value because of the circumstances in
which they were collected. They were recorded in the (American) homes of
the narrators, with Hmong children and adults gathered around to listen; no
foreigners were present. There is no doubt that such a relaxed setting for story
telling would have elicited the most natural possible style of language.
Finally, the transcription of the stories was accomplished entirely by native
Hmong speakers, and is completely faithful to the spoken word. In his intro-
ductory remarks (p. xiv) Johnson records:
Some of our Hmong consultants told us that our written Hmong texts
contained ‘bad Hmong’, ‘mistakes’ in grammar and syntax, and were not
literary texts worthy of being published for teaching proper Hmong lan-
guage. We had to persuade our most educated Hmong collaborators that
what we wanted was an accurate verbatim transcription of the spoken
Introduction 5
language, the vernacular spoken by our (generally) unschooled narra-

tors, with their vocalized pauses, repetitions, colloquial expressions and
rustic regional vocabulary faithfully represented, without any ‘cleaning
up’ or ‘improvements’, above all without rewriting or rephrasing in order
to achieve a literary style. We insisted that the narrator’s spoken text is
authoritative and that the oral style must be preserved in the written text.
The problem of Hmong language consultants, particularly (though not only)

those with a higher level of formal education, wanting to ‘correct’ examples
of natural, spoken Hmong was one that was encountered frequently in the
course of this investigation. The reasons for this are complex, and cannot be
discussed fully here. One factor seems to be simply the low esteem in which
speakers often hold unwritten forms of language, or any aspects of the lan-
guage, written or unwritten, that seem in any way vague or imprecise. Clark
(1992, p. 147) notes:
. . . the phenomena of NP zero anaphora and serialization of unmarked

constituents are characteristic of informal language of the sort that is not
readily available in print and is often denied or rejected by native speak-
ers in response to query.
She continues in a footnote:
Tony Diller (personal communication) tells of recording natural Thai

conversations, but when the recordings were transcribed by Thai speak-
ers they ‘corrected’ the recorded version, adding conjunctions, preposi-
tions, referents, etc., to make the language ‘correct Thai’.
Another important reason why natural language data is often either ‘corrected’
or rejected by Hmong language consultants is related to the fact that utter-
ances in Hmong, as in many of the other languages in Southeast Asia, are
highly context dependent. If a single sentence is presented in isolation from
the context in which it was uttered, it is often difficult, or even impossible, to
interpret. As pointed out by Li (n.d., p. 4):
The problem of interpretation is particularly serious for the analytic (iso-

lating) type of language which lacks inflection, agreement, morphologi-
cal markers, and uses zero anaphora freely. In general, one can say that the
meaning of a sentence in such analytic languages as Hmong and Chinese
tends to be much more sensitive and dependent on the discourse context
than that of a sentence in a polysynthetic language.
6 Introduction
It is for this reason that an attempt has been made to base the present study pri-
marily on natural language data obtained from reliable and recorded sources.
Probably the least reliable of the sources used were primers and other peda-
gogical materials. However, these were occasionally found useful (particularly
in Chapter 1, in which the structure of the simple sentence in White Hmong is
briefly described) because the simplicity of examples from such sources facili-
tates the clear illustration of the point at issue. Whenever a written or tran-
scribed source has been utilised, an abbreviated reference is given with the
example, so that a reader wishing to refer to the context can do so. (See the List
of Abbreviations for the abbreviated reference codes, and the list of ‘Sources
for White Hmong Examples’ for bibliographic details of each reference cited).
Where no reference is given with an example, it can be assumed that the
sentence was either recorded from natural conversation or directly elicited.
Wherever possible, direct elicitation from language consultants was accom-
panied by discussion of an appropriate context, and the examples were dis-
cussed with more than one native speaker. However, even when the original
context discussed was reintroduced, it was found that consultants often dis-
agreed regarding the acceptability of certain examples, and even that the same
consultant would give varying grammaticality judgements, and feel a greater
or lesser need to ‘correct’ and ‘clarify’, on different days.
About the Glossing and Translation of Examples
Examples from languages other than Hmong are generally glossed and trans-
lated as in the reference cited. Modifications to glosses have sometimes been
necessary, simply to bring them into line with the conventions and abbrevia-
tions used elsewhere in this book. Very occasionally, the original translations
given have also been modified slightly to emphasise a certain point, but only
when the accuracy of such modification is beyond question.
Hmong language examples taken from secondary sources have sometimes
been used. In this case, just as with examples taken from primary sources, the
glosses and translations given are my own. It is the exception rather than the
rule that these glosses and translations agree exactly with those given in the
secondary source cited. When Hmong language examples are included as part
of a quote from the work of another author, the glosses are generally amended
according to my own conventions and abbreviations (such changes being indi-
cated in square brackets). However, the original translation is not changed in a
quote, as this translation is usually relevant to the issue being discussed.
CHAPTER 1
Introduction to White Hmong Language
1.1 About the Language Family and the Locations of its Speakers
The language investigated in this study is White Hmong [mww].1 The autonym
used by the speakers consulted—[m̥ ɔ̘̃ ŋ˦ dɐ͜ə̆˦]—is transcribed as Hmoob Dawb
in the orthography used here, known as the Roman Popular Alphabet (RPA).
Speakers of White Hmong and other closely related dialects live in the moun-
tainous regions of south-west China (Sichuan, Guizhou, and Yunnan provinces
and Guangxi Autonomous Region) and also in the northern parts of Vietnam,
Laos, Thailand and Myanmar. As the result of displacement from Laos since
the mid-1970s in the aftermath of the so-called ‘Laotian Civil War’ (also often
referred to as the ‘Secret War’), many speakers also now live in diasporic com-
munities outside China and Southeast Asia. By far the greatest number are
in the United States, with smaller communities in Canada, France, Germany,
Australia, French Guiana, and Argentina. The data used for this study are from
speakers or writers who live in Laos or whose origins are in Laos. The principal
language consultants live in Luang Prabang province, or were born in Xieng
Khouang province and now live in Australia.
In China and Southeast Asia, some groups of speakers of the White Hmong
language/dialect use a different autonym (such as the ‘Striped (Armband)
Hmong’ in Laos and Thailand).2 Quite a number of other groups speak a lan-
guage/dialect sufficiently similar to White Hmong to be mutually intelligible to
at least some extent (including the Hmong Sua (suav ‘Chinese, other’) [hmz]
in Southwest China (Tapp, 2002) and the Hmong Dô [hmv] in Vietnam (Lewis
et al., 2013)).3
1 A three-letter code given in square brackets after a language name is an ISO-639 code. These
codes are created by the Ethnologue (Lewis, Simons, & Fennig, 2013) in cooperation with the
International Organization for Standardization (ISO) to uniquely identify languages.
2 Autonyms are [m̥ ɔ̘̃ ŋ˦ qɷ̆͜ ɐ˨ mbɐ˦], written Hmoob Quas Npab in RPA orthography, or [m̥ ɔ̘̃ ŋ˦
t̪sɐ̪ ͜ɩ˥˩̆ (mbɐ˦)], Hmoob Txaij (Npab), both of which mean ‘striped (arm) Hmong’.
3 The situation regarding which groups speak which language/dialect is quite difficult to clar-
ify, particularly in China, Vietnam, and Myanmar. Often, groups speaking the same dialects
use different autonyms, related to non-linguistic differences such as costume. This is the case
with the Striped (Armband) Hmong, who speak White Hmong, but whose self-designation
refers to the distinct blue and black stripes on the arms of the women’s clothing. Alternatively,
those speaking different dialects may use the same autonym. There is a group in Myanmar,

8 Chapter 1
Outside Asia, the best known of these mutually intelligible dialects is Mong
Njua [hnj], translated as either ‘Green Mong’ or ‘Blue Mong’, and also often
referred to as Mong Leng.4 While there are some lexical and grammatical dif-
ferences, the major differences between White Hmong and Mong Njua are
phonological, relating to certain vowel and tonal distinctions as well as to
some consonants, including the loss of the distinction between voiced and
voiceless initial nasals in Mong Njua. The name of the language/dialect group
is thus variously written in English as ‘Hmong/Mong’, ‘(H)mong’, or ‘HMong’,
in an attempt to acknowledge the differing pronunciations of the initial bila-
bial nasal by these two dialect groups. In this work, when the term ‘Hmong’
is used without a modifier, the reference is to the whole language/dialect
group; ‘White Hmong’ is used to refer to the specific dialect investigated here,
and ‘Mong Njua’ is used in preference to either ‘Green Mong’, ‘Blue Mong’, or
‘Mong Leng’.
Groups that speak dialects mutually intelligible with White Hmong dif-
fer further with regard to marriage and funeral customs, household architec-
ture, clothing and headdress. As with many other dialects and languages of
the Southeast Asian region, the autonyms of Hmong sub-groups often clearly
make reference to some prominent aspect of traditional costume. The name
White Hmong may refer to the white or undyed skirt worn by women (in most
regions now only on ceremonial occasions). However, Tapp (2002) suggests
that there may be other possible origins for this autonym.
Hmong belongs to the language family called ‘Hmong-Mien’ by most
Western linguists, but widely referred to as the ‘Miao-Yao’ language family by
for example, who refer to themselves as ‘White Hmong’ (K.-L. Yang (1998), cited by Tapp
(2002, p. 97)), but whom Lewis et al. (2013) categorizes linguistically as “Green Hmong.” Tapp
(2002: 81) also notes, “I have met Hmong from Yunnan, Guizhou and Sichuan who all claimed
to have been ‘White Miao’ and yet used what I recognised as the distinctively ‘Green Hmong’
consonant ‘gl’ or ‘dl’ to describe this.” In addition, exonyms rarely correlate precisely with
emic distinctions, and etic distinctions made by ethnographers and linguists may bear little
relation to either. It can be unclear whether names given in the literature are exonyms or aut-
onyms, Romanized transcriptions vary greatly, and tones are often ignored. See Culas (2009)
for an interesting discussion of some of these issues in a historical context in Vietnam.
4 Autonyms are [mɔ̘̃ ŋ˦ ndʒɷ̆͜ ɐ˦], written Moob Ntsuab in the RPA orthography, or [mɔ̘̃ ŋ˦ l̪ɛŋ̃ ˥˩̤],
Moob Leeg, or [mɔ̘̃ ŋ˦ l̪ɛŋ
̃ ˨], Moob Lees. English speaking scholars initially translated the word
ntsuab as ‘blue’, but began to use ‘green’ in the 1960s in response to negative associations with
the colour ‘blue’ expressed by speakers in Laos. Some community members in America have
come to feel that the term ‘green’ may have derogatory connotations, and now favour ‘blue’,
or prefer to simply use Mong Leng (Lee & Tapp, 2010, pp. xvii–xviii).
Introduction to White Hmong Language 9
many Western anthropologists and by all Chinese scholars.5 The terms ‘Miao’
and ‘Yao’ are also used to designate the official Chinese ‘nationalities’ (民族
mínzú) to which most speakers of these languages belong. It is important, how-
ever, to distinguish clearly between the terms Miao and Yao used to refer to
the language family and its sub-families, on the one hand, and to the political
entities on the other. By no means all of those who are officially identified as
belonging to the Miao or Yao nationalities in China speak a Miao-Yao (Hmong-
Mien) language. Furthermore, some grouped with the Miao nationality speak
a Yao (Mienic) language and some classed as Yao nationality speak a Miao
(Hmongic) or even a Tai-Kadai language (Ratliff, 2010, pp. 3–4).
Chinese linguists, and some outside of China (Greenberg, 1953; Shafer,
1964), have claimed that the Miao-Yao (Hmong-Mien) family is of Sino-Tibetan
stock. However, similarities may be attributable to long historical contact and
extensive borrowing and bilingualism (particularly with the various, domi-
nant Sinitic languages/dialects), rather than to common ancestry (Ratliff,
2010, p. 2). Even before the movement of Han peoples into southern China,
speakers of Miao-Yao/Hmong-Mien languages are likely to have had contact
not only with speakers of Tibeto-Burman languages, but also with speakers of
languages of the Austroasiatic, Tai-Kadai, and Austronesian families. Thus it
is not surprising that Haudricourt (1951, 1966, 1971) finds similarities between
the Miao-Yao (Hmong-Mien) family and both the Austroasiatic (Mon-Khmer)
and the Tibeto-Burman families, referring to it as a link between the two, and
that Benedict (1972, 1975) observes sufficient similarities between the Miao-
Yao (Hmong-Mien), Tai, and Austronesian languages to group them together
in one huge family: Austro-Thai. While expressing some doubts about the
Austronesian connection, Kosaka (2002) supports the idea of a link between
Miao-Yao (Hmong-Mien) and the Tai-Kadai languages, proposing a Miao-Dai
family. All of these suggestions are highly speculative.
The relationships between languages within the family are somewhat clearer.
On the basis of extensive comparative research, Ratliff (2010) groups Hmong
with the A-Hmao6 and Bunu7 languages, into the West Hmongic branch8
of the Hmongic sub-family of Hmong-Mien (Miao-Yao). This sub-family also
5 See Lemoine (2005, pp. 7–8), Ratliff (2010, pp. 3–4), and Tapp (2002, pp. 97–99) for interesting
points of view regarding this divergent terminology.
6 Also called ‘Large Flowery Miao’ and Diandongbei Miao among other names; ISO-639
code [hmd].
7 May include ISO-639 codes [bwn], [buh], [bwx].
8 West Hmongic is ‘Chuanqiandian’ in the Chinese terminology.
10 Chapter 1
includes the slightly more distantly related languages Qo Xiong9 and Hmu,10
along with Pa-Hng11, Ho Ne12 and Jiongnai,13 among others (Ratliff, 2010, pp. 3,
33–34). Some of these, such as Bunu, Qo Xiong, and Hmu, are really language
groups, with a number of ‘dialects’ that range from mutually intelligible to
unintelligible. Others, such as Pa-Hng, Jiongnai, and particularly Ho Ne, are
clearly endangered (Bradley, 2007; Lewis et al., 2013).
Largely in response to Han expansion over many centuries, the Miao-Yao/
Hmong-Mien speaking peoples have moved progressively further south and
west from what were probably their earliest locations in Central-southern
China, perhaps around the middle reaches of the Yangtze River (Wen et al.
(2005) and Fei (1999), cited therein). The majority now dwell in the southern
and south-western provinces of China: Hunan, Guizhou, Sichuan, and Yunnan,
and Guangxi Autonomous Region. The Hmong are among those who have
expanded the furthest southward, with some groups reaching northern Laos
possibly as early as 1810 or 1820 (Lee & Tapp, 2010, p. xxii), northern Vietnam by
1868 and northern Thailand around 1880 (Geddes, 1976, p. 29). As mentioned
above, many of the Hmong from Laos have relocated since 1975 to North
America, South America, Australia, and Europe.
Determining the number of Hmong speakers is extremely difficult. In China,
this is partly because census data refers only to the far more numerous and
linguistically diverse Miao minzu minority ‘nationality’, and it is often unclear
whether other estimates refer exclusively to speakers of the Hmong language
or not. In the United States, some problems with census data—related to pre-
ferred identification, language and cultural barriers and the like—have also
been reported (Pfeifer & Lee, 2004, p. 3). These kinds of problems are likely to
be relevant in all of the countries in which Hmong speakers live.
On the basis of information from a range of Chinese, English and French
language sources, as well as from census data and careful consideration
of likely population increases, Lemoine (2005) estimates that, there could
have been approximately 2.8 million Hmong speakers in China in 2000, with
roughly a further 1.26 million in Mainland Southeast Asia—the majority in
9 North Hmongic; classified as ‘Xiangxi Miao’ in Chinese terminology; includes ‘Western

Xiangxi Miao’ (autonym [ɕõ]; ISO-639 code [mmr]) and ‘Eastern Xiangxi Miao’ (ISO-639
code [muq]) (Sposato, 2012, p. 50).
10 East Hmongic; classified as ‘Qiandong’ in Chinese terminology; includes ISO-639 code
[hea] (the official standard), [hms], and [hmq] (Lewis et al., 2013).
11 Autonym [pa hŋ] (Bradley, 2007); ISO-639 code [pha].
12 Autonym [huo nte] (Bradley, 2007); ISO-639 code [shx]. Referred to as ‘She’ in Chinese
(Bradley, 2007; Ratliff, 1998, p. 97).
13 Autonym [kjɔŋ nai] (Bradley, 2007); ISO-639 code [pnu].
Vietnam—and perhaps around 320,000 outside Asia—the majority in the

United States. This suggests a worldwide population of Hmong speakers
approaching 4.5 million.14
1.2 Phonology and Orthography
The phonemic analysis outlined in the following pages is based on the one
devised by Smalley, Barney, and Bertrais (reported in Smalley (1976)), and the
phoneme symbols used correspond to those of the ‘Roman Popular Alphabet’
developed for Hmong by these authors. However, the analysis of the allophones
of the segmental phonemes (§1.2.1 and §1.2.2 below) and of the auditory quali-
ties of the tones (§1.2.3.2 below) differs somewhat from Smalley’s descriptions.
1.2.1 The Consonant Phonemes and their Allophones

White Hmong has a large inventory of consonants: fifty-eight in all, including
thirty-nine stops and affricates. In addition to great variation in place of articu-
lation, the inventory is further expanded by the possibility of pre-nasalisation,
lateral, affricated, and aspirated release, and voicing distinctions for fricatives,
nasals and laterals. Tables 1 and 2 show the consonant phonemes and their
allophones.
1.2.1.1 Notes on the Realisation of Consonants

– /f/ is reported to have allophone [ɸ] in all environments in some dialects of
White Hmong (Martha Ratliff, pers. comm.).
– The strong voicing during the hold phase of /d/ and the whispery voice dur-
ing the hold phase of /dh/ generally dies out briefly before the release of the
oral closure. In both cases, this produces the impression of a short ‘catch’
in the voice (Jarkey, 1987, p. 48). This cessation of voicing before release is
often referred to as ‘pre-glottalisation’ (e.g. Smalley, 1976).
– For /r/, /rh/, /nr/ and /nrh/, some speakers seem to have the following four
allophones instead: [ʈʂ], [ʈʂh], [ɳɖʐ ], and [ɳʈʂh]. See Lyman (1974, p. 34)
regarding the collapse of this series with two others in some idiolects in
Mong Njua.
14 Lewis et al. (2013) give figures specifically for ‘Hmong Daw’ (White Hmong; mww) in Asia,
but the sources are not always clear: China (2004) 233,000; Thailand (2000) 32,400; Viet
Nam (2009 census) 1,068,000; Laos (1995 census) 170,000. (Note that the 2005 census in
Laos did not distinguish between White Hmong and Mong Njua.) For Myanmar, Enwall
(1994) cites the figure of 10,000 for the ‘Hmong subgroup’, from Xiong (1992). However,
Lewis et al. (2013) indicate that these are Mong Njua (Myanmar, 2000).
12 Chapter 1
table 1 The consonant phonemes and their allophones I
Labial Apico-Dental Apico- Lamino-

Alveolar Alveolar
Lateral Affricated
release release
Stops & Unaspirated /p/ /pl/ /t/ /tx/ /d/ /ts/

Affricates [p] [pl] [t̪] [t̪s]̪ [d] [tʃ]
Aspirated /ph/ /plh/ /th/ /txh/ /dh/ /tsh/
[ph] [pɬ] [t̪h] [t̪sh̪ ] [d̤ h] [tʃh]
Pre-Nslzd /np/ /npl/ /nt/ /ntx/ /nts/
[mb] [mbl] [n̪ d̪] [n̪ d̪z̪] [ndʒ]
Pre-Nslzd /nph/ /nplh/ /nth/ /ntxh/ /ntsh/
aspirated [mph] [mpɬ] [n̪ th̪ ] [n̪ ts̪ h̪ ] [ntʃh]
Fricatives Voiced /v/ /z/
[v] [ʒ]
Voiceless /f/ /x/ /s/
[f] [s̪] [ʃ]
Nasals Voiced /m/ /ml/ /n/
[m] [ml] [n̪ ]
Voiceless /hm/ /hml/ /hn/
[m̥ ] [m̥ ɬ] [n̪̥ ]
Laterals Voiced /l/
[l̪]
Voiceless /hl/
[ɬ̪]
– The consonants /c/, /ch/, /nc/, and /nch/ all have slight frication at release.
This frication can be attributed to their dorsal articulation, and they are still
considered to be stops in this analysis. They are articulated with the tongue
tip behind the lower front teeth.
– The phoneme /g/ appears in parentheses in Table 2 because its status is
somewhat marginal. It occurs only in onomatopoeic words in the dialect of
my consultants from Laos. However, it is reported to occur in the word gus
‘goose’ in Thailand (Heimbach, 1979, p. 47).
table 2 The consonant phonemes and their allophones II
Apico Post- Dorso Velar Uvular Glottal

Alveolar (Rhotic) (Pre)-Palatal
Stops & Unaspirated /r/ /c/ /k/ /q/ [q] /ʔ/

Affricates [ʈɽ] [c] [k] [ʔ]
Aspirated /rh/ /ch/ /kh/ /qh/
[ʈɽh]
̥ [ch] [kh] [qh]~ [qχh]
Pre-Nslzd /nr/ /nc/ /nk/ /nq/
[ɳɖɽ] [ɲɟ] [ŋg] [NG]~ [Nq]
Pre-Nslzd /nrh/ /nch/ /nkh/ /nqh/
aspirated [ɳʈɽh]̥ [ɲch] [ŋkh] [Nqh]
Fricatives Voiced /y/
[ʝ]
Voiceless /xy/ /h/
[ɕ] [h]
Nasals Voiced /ny/ (/g/)
[ɲ] [ŋ]
Voiceless /hny/
[ɲ̥]
– The consonant /ʔ/ is not symbolised in the orthography. Its presence is gen-
erally implied when a word is written with no initial consonant. However,
vowel initial illocutionary particles are articulated without initial /ʔ/, except
in citation form. For example, the illocutionary particle os [o˨] can be com-
pared with the noun os [ʔo˨] ‘duck’.
1.2.2 The Vowel Phonemes and their Allophones

The vowel system of White Hmong is quite straightforward, with eight monoph-
thongs (two of which are nasalised), two closing diphthongs and three open-
ing diphthongs. Figures 1–3 show the vowel phonemes and their allophones.
1.2.2.1 Notes on the Realisation of Vowels

– /w/ is articulated with slack rather than spread lips.
– Different speakers seem to favour different pronunciations of /ee/, includ-
ing [ɛ̃ŋ], [ãŋ], and [ɐ̃ŋ].
14 Chapter 1
/i/ /w/ /u/

[i] [ɨ] [u̘ ]
/e/ /o/
[e] [ɔ̘ ]
/ee/ /oo/
[ɛ̃ŋ] [ɔ̘̃ ŋ]
/a/
[ɐ]
figure 1 Monophthongs
i u
/ia/ /ua/
[ɩɐ] [ɷɐ]
a
figure 2 Opening diphthongs
i w u
/ai/ /aw/ /au/

[ɐɩ] [ɐə] [ɐɷ]
a
figure 3 Closing diphthongs
1.2.3 Tones
There are seven phonemic tones in White Hmong. Five of these are articulated
in modal voice and characterised by variation in pitch and/or contour. Two
involve additional variation in voice quality.
1.2.3.1 Orthographic Representation of Tones

In the Romanised Popular Alphabet, tones are represented orthographically by
a single consonant symbol appearing at the end of the syllable. One tone—the
mid level tone—is not symbolised; it is simply indicated by default.
The use of consonant symbols syllable finally to mark tones is possible
because almost all syllables are open. The only syllable-final consonant phone
in White Hmong is [ŋ], and this is simply dealt with in the orthography as
part of the realisation of a nasalised vowel, symbolised by the doubling of the
vowel: ‘ee’ and ‘oo’.15
The seven phonemic tones in White Hmong are symbolised ‘b’, ‘j’, ‘g’, ‘v’, ‘s’,
‘m’, and ∅ (the orthographically unmarked mid-level tone). The orthographic
tone symbol ‘d’ represents an eighth phonetic tone, generally a syntactically
conditioned variant of tone ‘m’ (see §1.2.3.5.b below).
1.2.3.2 The Auditory Description of Tones in Citation Form

Table 3 gives the consonant symbols used for the eight phonetic tones along
with a brief description of their key characteristics in citation form: pitch and
contour, voice quality and (if relevant) length. Pitch and contour are also rep-
resented on a numerical scale, in which 5 symbolises the upper end of a speak-
er’s voice range, and 1 the lower end.
table 3 Tone symbols, pitch, voice quality and length
Symbol Pitch/Contour Brief Description
b 55 high level, modal, short

j 53 high falling, modal
g 42 high/mid falling, breathy
v 24 mid rising, modal
∅ (no symbol) 33 mid level, modal
s 22 low level, modal
m 21 low falling, creaky end, short
d 213 low falling rising, modal, long
1.2.3.3 Notes on the Realisation of Tones

– Auditorily, tone ‘b’ seems fairly level, and is generally characterised as such
(e.g. Huffman (1987); Ratliff (1992); Smalley (1976)). However, acoustic anal-
ysis indicates that it actually rises slightly in pitch, and also that it is com-
paratively short in duration (Esposito, 2012, p. 471). The auditory impression
of a level tone may be because contour is more likely to be perceived over a
15 This treatment of the final nasal as a feature of the vowel also makes sense in the light of
the fact that nasalisation never occurs with the high vowels; it occurs with only the mid
vowels in White Hmong, and with these and the low vowel [ɐ] in Mong Njua.
16 Chapter 1
long vowel than a short one, and because the rise in pitch does not seem to
be a distinctive feature of this tone.
– Tone ‘g’ does not occur in syllables beginning with voiceless aspirated plo-
sives or voiceless fricatives, nasals or liquids, presumably because the breathy
phonation of the vowel interferes with the perception of the distinctive
features of the consonant. Fulop and Golston (2008) demonstrate the pho-
netic distinction between the breathy phonation of this tone and the whis-
pery phonation of the consonant phoneme /dh/.
– The exact point within the voice range at which tone ‘g’ begins seems to
vary considerably from speaker to speaker. Ratliff (1992, p. 12) observes that
women tend to have a higher starting point than men, and this is confirmed
by acoustic measurements for the young female participants in a study by
Esposito (2012).
– Some scholars characterise tone ‘m’ as creaky (Andruski & Ratliff, 2000;
Esposito, 2012; Lyman, 1974, p. 38) and others simply as modal voice ending
in a glottal stop (Huffman, 1987, pp. 496, 501; Smalley, 1976, p. 100). There
may be some difference in the speed of the glottal closure depending on
vowel height (Andruski & Ratliff, 2000).
– There is often a slight fall or “natural tapering” (Ratliff, 1992, p. 11) at the end
of tone ‘s’, which may help to distinguish it from the mid level tone ‘∅’.
1.2.3.4 The Unmarked Tone

The low falling tone ‘s’ may be the ‘unmarked’ tone in White Hmong. It is the
tone that appears in nonsense syllables used to illustrate the pronunciation of
segmental phonemes, and is used quite often in syllables in borrowed words
from English.16 Note, however, that it is the mid level tone ‘∅’, rather than the
low falling tone ‘s’, that is unmarked in the orthography.
1.2.3.5 Tone Change

There are two distinct types of tone change in White Hmong: tone sandhi and
morphotonemic alternation.
a Tone Sandhi
When tone sandhi occurs in White Hmong, it is progressive and is triggered by
high level tone ‘b’ and the high falling tone ‘j’:
16 Riddle (1989) notes a number of recent loanwords from Chinese with tone ‘m’, and notes
the value of further study on ‘loan-tones’ in White Hmong.
table 4 Tone sandhi changes
Tone of Citation Form Tone Following ‘b’ or ‘j’
high falling ‘j’

low level ‘s’ → falling breathy ‘g’
low falling creaky ‘m’
mid level ‘∅’ → low level ‘s’
mid rising ‘v’ → mid level ‘∅’
As shown in Table 4, these changes result in the neutralisation of three tones

(‘j’, ‘s’, ‘m’) in this environment. Furthermore, neither the high level tone ‘b’ nor
the breathy voiced tone ‘g’ are influenced by the tone that precedes them.
There is no clear phonetic motivation for the kind of tone sandhi outlined
above. Ratliff (1986a, p. 30) notes that:
. . . the same historical tone categories are involved as triggers, targets,

and sandhi forms in . . . [many related dialects], regardless of the pho-
netic values of the reflexes of those categories. The fact that the trigger
forms in [White Hmong] are phonetically . . . high tones . . . is just acci-
dental . . . making a ‘pitch harmony’ theory implausible.
Furthermore, tone sandhi in White Hmong is not a thoroughgoing phonetic

feature; it is favoured in the environment of certain lexical items and word
classes, and restricted to certain grammatical constructions involving close
juncture. While common patterns can be observed, there are numerous excep-
tions to these tone sandhi ‘rules’ (see examples in i–iv below, and Ratliff (1987;
1992, pp. 26–62)). Even in contexts in which sandhi generally occurs, it is some-
times optional.
The main reason for the irregular behaviour of tone sandhi in White Hmong
is that it seems to be largely a historical relic in the modern language (Downer,
1967; Ratliff, 1987, 1992). The more extensive and regular sandhi system that
previously existed presumably ceased to be productive before the entry of
many loanwords from Chinese, and before the merger of the two historical
reflexes of tone ‘s’, partially accounting for many exceptions to sandhi ‘rules’.
Other exceptions are due to the non-occurrence of tone ‘g’ in morphemes
beginning with voiceless aspirated plosives, and voiceless fricatives, nasals and
liquids (see §1.2.3.3 above).
18 Chapter 1
Some of the conditions under which tone sandhi often occurs are described
briefly below. See Ratliff (1987; 1992, pp. 26–62) for a thorough analysis.
i. After the Numbers 1–5 and 9

(1) a. tus neeg
CLF person
‘the person’
b. ib tug neeg
one CLF(t.c.) person
‘one person’
c. rau tus neeg

six CLF person
‘six people’
(2) a. caum
tens
b. tsib caug
five tens(t.c.)
‘fifty’
c. xya caum
seven tens
d. Exception: cuaj caum

nine tens
‘ninety’
ii. After the Reciprocal Morpheme sib

(3) a. tom
bite
‘bite’
b. sib tog
RECP bite(t.c.)
‘bite/fight each other’
(4) a. qawm
show.affection
‘show affection’
b. sib qawg
RECP show.affection(t.c.)
‘show mutual affection’
(5) a. tham
chat
‘chat’
b. Exception: sib tham

RECP chat
‘chat together’
iii. In Head-Attribute Constructions

(6) kauj + ntsej → kauj ntseg
ring ear ring ear(t.c.)
‘ear lobe’
(7) a. nqaij + nyuj → nqaij nyug

meat cow meat cow(t.c.)
‘beef’
b. nqaij + npua → nqaij npuas

meat pig meat pig(t.c.)
‘pork’
c. Exception: nqaij + ntses → nqaij ntses

meat fish meat fish
‘fish’
(8) a. dej sov

water be(come).warm
‘The water is warm.’
b. dej so
water be(come).warm(t.c.)
‘warm water’
20 Chapter 1
(9) a. dej tsias

water be(come).cool
‘The water is cool.’
b. dej tsiag
water be(come).cool(t.c.)
‘cool water’
iv. In Cases of Noun Incorporation

(10) a. poob dej
fall water
‘fall into the water’
b. poob deg
fall water(t.c.)
‘drown’
(11) a. muaj txiaj

have money
‘have some money’
b. muaj txiag
have money(t.c.)
‘be rich’
Examples (8)–(11) above illustrate the fact that the syntactic relationship
between constituents can be reflected in the occurrence (or non-occurrence)
of tone sandhi.
b Morphotonemic Alternation
In addition to tone sandhi, the other type of tone change that occurs in White
Hmong is morphotonemic alternation; this takes place only between tone ‘m’
and tone ‘d’.
White Hmong has a closed class of Spatial Deictics that appear in Locative
Phrases directly before a Noun Phrase (see §1.5 below). The majority of these
Spatial Deictics exhibit tone ‘m’, for example pem (for a place up a slope), nram
(for a place down a slope), ntawm (for a place nearby), tom (for a place some
distance away), tim (for a place on the other side of some intervening feature)
etc. However, when one of these ‘m-tone’ morphemes constitutes a Locative
Phrase alone, or is used attributively after a Noun Phrase, it invariably exhibits
tone ‘d’:
(12) a. [tus tub] mus txog [nram hav-dej]

CLF boy go arrive down valley-water
‘The boy got down to the river valley.’ (KNH 2)
b. [tus tub] mus txog nrad

CLF boy go arrive down.there(t.c.)
‘The boy got down there.’
(13) a. thaum [tus tsov] dhia mus txog [sab [nraum

time CLF tiger jump go arrive side outside
[lub ces]]] lawm . . .
CLF log PRF
‘When the tiger had jumped around to the other side of the log . . .’
b. thaum [tus tsov] dhia mus txog [sab nraud] lawm . . .

time CLF tiger jump go arrive side outside(t.c.) PRF
‘When the tiger had jumped around to the other side . . .’ (LPV 2)
Ratliff (1992, pp. 112–120) refers to these tone ‘d’ words as ‘Demonstrative
Nouns’. She agrees with Downer (1967, pp. 597–598) that the change in tone
in this environment can probably be attributed to the loss of the demonstra-
tive no ‘this’ phrase finally, with only a trace of the mid-tone (33) remaining.
This yields the falling-rising contour (213) ‘d’ tone in place of the falling (21)
‘m’ tone.
Ratliff (1992, pp. 114–123) notes that “a few other –m tone words have been
attracted into the pattern . . . and have developed –d tone demonstratives as
well.” These include qaum/qaud ‘back, top part’, sim/sid ‘time’, npaum/npaud
‘measure’, and the third person dual pronoun nkawd ‘those two’, derived from
nkawm ‘a pair, a couple’.
This alternation occurs not only in the case of demonstratives but also in
vocatives, as in the use of niad instead of niam ‘mother’ below:17
17 Morphotonemic alternation does not seem to be entirely restricted to words with tone
‘m’. Alternation also seems to occur between tone ‘s’ and tone ‘d’ in limited contexts. The
conditions under which it does so are not fully clear, but the form tud can occur in place
of the animate classifier tus with the meaning ‘each one does V’:
(i) ces tud hle hlo [tud [nkawm khau]]
CONJ CLF(t.c.) take.off quickly CLF(t.c.) pair shoe
‘And each one took off his/her own shoes.’ (DNH 463: 17)
22 Chapter 1
(14) Niad kuv yuav noj mis

mother(t.c.) 1SG IRR eat breast
‘Mother, I want to nurse at your breast.’ (DNH 365:66)
1.2.4 Syllable Structure

Smalley (1976, p. 103) writes:
The typical Hmong syllable consists of two parts, an onset and a peak.
Onsets may be null (zero), simple, or complex. Peaks consist of a syllabic
and a tone. The syllabic may be simple or complex . . . Tones may be sim-
ple or complex . . .
Complex onsets are created by aspiration, prenasalisation, or both. Or,
they are created by lateral or affricated release. Complexity caused by aspira-
tion, nasalisation or lateral release is indicated in the symbolisation. Affri-
cation is not always so indicated. /r/, for example, has an affricated release.
Complex syllabics result from vowel glide [i.e. diphthongisation] or
from nasalisation . . . Complex tones are a combination of pitch with
breathiness or with glottalisation [i.e. creak].
The only aspect of the present analysis that differs from Smalley’s is in regard
to the possible types of ‘onset’. Here the underlying forms of all syllables are
considered to begin with a consonant, disallowing the possibility of Smalley’s
‘null’ onset. The small number of instances of syllables which lose their initial
/ʔ/ (a set of illocutionary particles, approximately six in all) should be dealt
with by a morphophonemic rule involving a close juncture condition. This
solution seems considerably more economical than proposing a separate syl-
lable type to deal with such a small number of lexical items, particularly as
they are pronounced with an initial [ʔ] in citation form.
Smalley claims that he avoids the problem of whether phonetically com-
plex consonants and vowels should be regarded as single phonemes or as clus-
ters. However, it is clear that he does regard them as single phonemes for the
purposes of his phonemic analysis; although the symbols used for many pho-
nemes are featural, they nevertheless do represent single phonemes. This has
the effect of producing a very large inventory of phonemes, particularly for
the consonants, and of necessitating a fairly uneconomical statement of the
phonetic realisation of these phonemes. For example, each instance of /Ch/
(where C = stop or affricate) in this system must be independently described
as having the allophone [Ch]. However, if such initials were regarded as com-
binations of /C/ + /h/, a simple rule could be given to predict the phonetic
realisation in virtually all cases. Only those phonemes with slightly aberrant
realisations (e.g. /qh/ → [qh] ~ [qχh]) need be treated separately.
However, although such an analysis would simplify and reduce the pho-
neme inventory, at the same time it would necessitate a more complex pho-
notactic statement, in order to constrain the possibilities for the combination
and ordering of the phonetically simple phonemes involved. Smalley’s solu-
tion of a large phoneme inventory is preferred.18
1.3 Syllables, Morphemes and Words
The units of syllable, morpheme, and word correspond directly in most cases
in White Hmong. However, there are a number of exceptions to this clear ten-
dency. These include: instances of ‘meaningful’ tone, borrowing, compound-
ing, class-nouns and prefixes, and reduplication.
1.3.1 Sub-Syllabic Morphotonemes

Tone works at the morphological level in a number of highly limited but impor-
tant ways in White Hmong. This topic is explored in detail by Ratliff (1986a,
1986b, 1987, 1992). In her discussion of the morphological functions of tone,
Ratliff notes the role of tone in cases of
– sandhi forms in compound word formation (1987; 1992, pp. 58–62) (see
also §1.2.3.5.a.iii and §1.2.3.5.a.iv above),
– tonally-defined word classes and (possibly) gender-marked kinship terms
(1986a; 1992, pp. 93–135), and
– tonal iconicity in expressive vocabulary (1986b; 1992, pp. 136–163).
Of considerable interest is the possibility of the use of tone to signal gen-

der in kinship vocabulary. While by no means applying to all kin terms, the
word pairs in examples (15)–(19) below seem to illustrate a somewhat regular
though minor pattern of tonal differentiation, in which tone ‘m’ is associated
with female kin, and tone ‘g’ primarily with male kin and secondarily with a
few other referents that are unmarked for gender.
18 Another reason for the preference for a large inventory of consonant phonemes is that
this aligns with the system adopted in the only orthography known to have been devel-
oped by a native speaker of White Hmong, Soob Lwj Yaj (Shong Lue Yang), often called
Niam Ntawv ‘Mother of Writing’. This fascinating orthography does not, however, employ
a featural system for writing consonants. See Smalley, Vang and Yang (1990) for a full
description of the system, and Vang, Yang, and Smalley (1991) for a first-hand account of
its development and early use in Laos. Today it is referred to as the Pahawh Script (Ntawv
Phajhauj or Ntawv Caub Fab) and used only by a now tiny group of Hmong in Laos, along
with some in refugee and diasporic communities.
24 Chapter 1
(15) a. ntxawm
‘youngest daughter’
b. ntxawg
‘youngest son’
(16) a. (poj) ntsuam

‘widow’
b. ntsuag
‘widow(er); orphan’
(17) a. yawm (txiv)

‘maternal grandfather’; ‘older men on mother’s side’
b. yawg
‘paternal grandfather; sir’; ‘older men on father’s side’
(18) a. niam (tais)

‘wife of yawm’ (see (17)a.)
b. pog
‘wife of yawg’ (see (17)b.)
(19) a. (muam) npaws

‘female first cousin, different clan’
b. npawg
‘male first cousin, different clan, peer, friend’
Also of interest are the clearly iconic tonal patterns that occur in many two-
word expressives. Expressives form a distinct and important word class in
White Hmong, occurring quite commonly in a wide range of genres, including
everyday conversation. Two-word expressives function primarily as mimetic or
sound symbolic manner adverbs, but also as phenomimes and psychomimes,
to portray elements of the sensory impression of a situation and its associa-
tions in the mind of the speaker. Ratliff (1992, p. 137) explains:
. . . two low level (-s) tones can be used to suggest humming, droning, or
an unending sight; a high falling (-j) tone followed by a mid rising (-v)
tone can be used to suggest a back and forth movement or mental vacil-
lation. The choice of initial consonant (which must be identical in the
two words) from the rich Hmong inventory of initials, and to a lesser
extent the choice of the second vowel (the first being determined by the
tone pattern . . .), together with the choice of tone pattern allows a great
deal of precision in fit between sound and meaning in these expressive
phrases.
Examples Ratliff provides of the two-tone patterns mentioned above include:
(20) Two low level (-s) tones: humming, droning, or continuous sensation
a. nrov lis loos
loudly
‘droning loudly / buzzing (of bees)’
b. tham chis chaus

chat
‘to chat continuously’
c. tsoo taub-hau nkhis nkhoos

crash head
‘to bump one’s head with a hollow sound’
(21) High falling (-j) tone then mid rising (-v) tone: back and forth
a. ua npaws txhuj txhoov
do fever
‘to have chills and fever’
b. quaj nyuj nyev

cry
‘to whimper/whine’
c. ua ntos nthuj nthav

do loom
‘to weave with the sound of the shuttle’
These examples represent just two of the six tone patterns that Ratliff finds to
have clear iconicity, and just a tiny sample of the numerous examples she gives
(1992, pp. 218–245).
26 Chapter 1
1.3.2 Polysyllabic Morphemes

While the examples described in §1.3.1 above involve more than one meaning-
ful unit within a single syllable, there are also some cases of more than one
syllable associated with a single meaningful unit. Some of these are modern
borrowings, others of less recent origin; for still others the origin is unclear.
Examples include doslas ‘dollar’ (<English), tajlaj ‘market’ (<Lao), phoojywg
‘friend’ (<Chinese), xibfwb ‘teacher’ (<Chinese), sawvdaws ‘everyone’, laujkaub
‘cooking pot’. Though polysyllabic, these words should be considered mono-
morphemic in White Hmong, and so are written here without a word break
or hyphen.
The examples above have shown minor violations of the correspondence
between the syllable and the morpheme: firstly in the case of the sub-syllabic
morphotonemes associated with gender and with sound symbolism, and sec-
ondly in that of polysyllabic morphemes that arise through borrowing. It is,
nevertheless, reasonable to conclude that the basic structure of White Hmong
morphemes is monosyllabic.
1.3.3 Multi-Morphemic Words

While there is a close association between the syllable and the morpheme in
White Hmong, and while the language generally lacks inflectional or deriva-
tional morphology, there are quite a number of instances of words that are
composed of more than one morpheme. These multi-morphemic words can
be attributed to three factors: compounding, class nouns and prefixes, and
reduplication.
1.3.3.1 Compounding
White Hmong is a language that utilises compounding to a considerable extent
in the process of word formation, for example: kawm-ntawv (study-paper/
writing) ‘study, attend school’; dav-hlau (hawk-iron) ‘aeroplane’, niam-txiv
(mother-father) ‘parents, a couple’; me-nyuam (little-little) ‘child’. The mor-
phemic components of such compounds are generally transparent, but their
idiomatic interpretations dictate that they should be regarded as single, multi-
syllabic and multi-morphemic words.
1.3.3.2 Class Nouns and Prefixes

A similar process to nominal compounding, which also produces multi-syllabic,
multi-morphemic words, is the use of ‘class nouns’—morphemes with a gen-
eral, superordinate meaning—appearing in combination with nouns of more
specific reference. Bisang (1993, pp. 42–45) gives a wide range of examples. Here
are listed just a few, to give a sense of just some of the kinds of morphemes used.
Class nouns frequently appear in the names of flora and fauna, such as trees,
fruit, and birds:
– ntoo ‘tree/wood’: ntoo-peeb-lab (tree-penis-red/monkey?) ‘a type of Areca

nut palm’,19 ntoo-ciab ‘a type of larch’20
– txiv ‘fruit’: txiv-quav-miv (fruit-faeces-cat) ‘tamarind’, txiv-taub-ntoo(s)
(fruit-gourd-tree) ‘papaya’, txiv-cuab-thoj ‘guava’
– noog ‘bird’: noog-daj (bird-yellow) ‘a type of oriole,21 noog-ncaws-nplej (bird-
hoe/peck-rice) ‘hoopoe’.22
Other types of nouns commonly used as class nouns include those for various
types of people and places:
– kws ‘specialist’ for artisans and experts: kws-ntoo (expert-tree/wood) ‘a car-

penter’, kws-kho-tsheb (expert-fix-car) ‘a mechanic’, kws-tov-tshuaj (expert-
mix.with.water-herbs/medicine) ‘a pharmacist’, kws-ntawv (expert-paper/
writing) ‘a scholar/writer’
– tsev ‘house’ for buildings associated with particular activities: tsev-kawm-
ntawv (house-study-paper/writing) ‘school’, tsev-kho-mob (house-fix-sick)
‘hospital’, tsev-teb (house-field) ‘a field shed’.
This kind of compounding with class nouns is a word formation strategy used
widely in the Mainland Southeast Asian region.
In White Hmong, as in many languages of the Hmong-Mien family (Ratliff,
2010, pp. 200–207), some class nouns have come to be used with increasingly
general meanings and have developed into nominal prefixes. Ratliff (2010,
p. 201) explains:
Nominal prefixes derive from class nouns as the first member of noun-
noun compounds. This process is independent and ongoing in different
languages of the family. For example, . . . tub /tu1/ ‘son/boy’ in Hmong
is now used as a class noun in compounds that refer to male humans,
whether young or not: for example, tub-rog /tu1 rɔ6/ ‘soldier’ (literally,
‘boy-war’).
19 Areca catechu.
20 Possibly Larix potaninii.
21 Possibly Oriolus chinensis.
22 Possibly Upupa epops longirostris.
28 Chapter 1
In fact, there is evidence to show that tub ‘boy’ has extended its meaning
even further pre-nominally, so that its usage is no longer confined even to
male referents. While the meaning components of ‘young’ and ‘male’ seem to
have been largely bleached, in many of the examples below we can, however,
see the perseverance of a sense of one who lacks agentivity or, in some cases,
lacks status or expertise.
– tub ‘son/boy’: tub-mob (boy-be.sick) ‘a patient’, tub-nkeeg (boy-be.lazy)

‘a lazybones’, tub-txib (boy-dispatch) ‘a messenger’, tub-txib-saum-ntuj (boy-
dispatch-place.above-sky) ‘angels’, tub-mab-tub-qhe (boy-stranger-boy-
slave/employee) ‘a servant/labourer’, tub-luam (boy-trade) ‘a merchant/
trader’, tub-sab (boy-?) ‘a thief’, tub-kawm-ntawv (boy-study-paper/writing)
‘a student’.
Regarding the development of nominal prefixes, Ratliff (2010, p. 201) continues:
An older, but still recoverable example of the same historical process

can be seen in the common White Hmong class noun or prefix pob
/pɔ1/, which serves as the first element in compounds that denote lumpy
things, such as stones, knots, tree stumps, and earlobes. Pob is a loanword
from Chinese 包 (Man. bāo) ‘to wrap; a bundle, lump’, and was probably
originally borrowed as a noun. Although this word can be used as a class
ifier as well, which is not typical of prefixes (ib pob mov /i1 pɔ1 mɔ3/ ‘one
CLF-clumps rice’), it is more typically used as a prefix attached to a noun
which requires yet another word as its classifier (ib lub pob-zeb /i1 lu1 pɔ1
ʐe1/ ‘one CLF-bulky things clump-stone’).
The relationship between nominal prefixes and classifiers will be discussed

further in §1.4.2 below.
1.3.3.3 Reduplication
Another common type of multi-morphemic word in White Hmong arises
through the morphological process of reduplication. Reduplication can be
applied to verbs, adjectives and adverbs, and its semantic effect can be broadly
described as augmentation or intensification. This broad meaning is inter-
preted in various ways, depending on the semantics of the reduplicated ele-
ment. With adjectives and with adverbs of quantity and manner, for example,
reduplication indicates a high/extenuated degree of the property, quality,
quantity or manner concerned (24); with adverbs of time (22) it emphasises
the length of time involved. Likewise, with verbs that have the property of
duration (states / activities (22) / accomplishments), reduplication expresses

extended duration or persistence, while with semelfactive verbs, it indicates
iteration.23 With many verbs expressing processes (24) and actions, redupli-
cation also adds a sense of energy or effort, and with those conveying emo-
tion (23), as well as with modal auxiliary verbs (23)/(24), a sense of strength
of feeling.
(22) aub pheej ntsim~ntsim [niag tsov]O ces [niag tsov]S

oh keep.on RDUP~sting great tiger CONJ great tiger
khooj ib-hnub~ib-hnub
bend.double RDUP~one-day
‘Oh, [the peppery liquid] kept on really stinging Tiger, and Tiger writhed
(in pain) the whole day long.’ (DNH 21:400).
(23) xav~xav tau nyabO ces txawm yuav nyabO

RDUP~want get son’s.wife CONJ then obtain son’s.wife
rau Yaj Yim Leej. Ces Yaj Yim LeejS kub~kub-siab ntawv
for PN CONJ PN RDUP~be.hot-liver writing
los, Yaj Yim LeejA ho mob~mob-siab [Yaj Yim Leej
CONJ PN actually RDUP~be.sick-liver PN
tus poj-niam]O thiab lau
CLF woman also INTNS
‘[Ya Yee Leng’s parents] really wanted to get a daughter-in-law, so [they]
got a wife for Ya Yee Leng. Ya Yee Leng was totally dedicated to study, but
his heart really ached for his wife as well.’ (DNH 461:2)
(24) kojA rau~rau-siab mus <txiav [taws ntau~ntau] O>PURP ;

2SG RDUP~put.in-liver go chop firewood RDUP~much
es thaum kojA/A txiav tau ntau~ntau lawm tiag
CONJ when 2SG chop get RDUP~much PRF really
ces kojA rhaub [ib yiag dej]O kub~kub, npau~npau
CONJ 2SG heat one pot water RDUP~hot RDUP~boil
‘Do your very best to go and chop lots and lots of firewood; and when you
have really chopped a lot, heat a pot of water absolutely boiling hot.’ (DNH
364:59)
23 Reduplication seems not to occur with achievement verbs in White Hmong. The sequence
tau tau commonly occurs, but this is probably a sequence of the preverbal marker of
achievement tau ‘get to, manage to’ and the achievement verb tau ‘get, come to have’,
rather than a reduplication of the achievement verb.
30 Chapter 1
Notice in the examples above that reduplication is often combined with other
expressions of duration and intensification (as in the auxiliary verb pheej ‘keep
on doing’ in (22), and the intensifying particle lau in (23)).
There are two reasons why the final element rather than the initial element
of a reduplicated form is considered the stem. First, as shown by the forms
kub~kub-siab (RDUP~be.hot-liver) ‘be totally dedicated to’ and mob~mob-siab
(RDUP~be.sick-liver) ‘have an aching heart for’ in example (23), and rau~rau-
siab (RDUP~put.in-liver) ‘do one’s very best to’ in (24), only the first morpheme
in compound adjectives and verbs (including modal verbs) is reduplicated.
Notice that this does not apply to reduplicated compound adverbs, as in
ib-hnub~ib-hnub (RDUP~one-day) ‘the whole day long’ in (22).
The second reason why it is the final element of a reduplicated form that
is taken as the stem relates to the shortening of the initial element(s) of redu-
plicated forms. This is especially noticeable with those containing opening
diphthongs.
(25) ces npuaS khiav~khiav lawm

CONJ pig RDUP~run PRF
‘And the pig ran away.’
In example (25) above, although the reduplicated form is almost invariably

written in full, it is very likely to be pronounced khi~khiav, with the initial syl-
lable reduced and pronounced without a tone.
One striking feature of spoken Hmong, especially when a speaker is telling a
narrative in a colourful or excited manner, is the number of reduplications that
can occur. Up to four have been observed:
(26) nwsS/A rov hlo los <nrhiav>PURP, nwsA

3SG return quickly come search 3SG
nrhiav~nrhiav~nrhiav~nrhiav tsis tau
RDUP~search NEG get
‘He came quickly back to search (for it); he searched and searched and
searched and searched (but) didn’t find (it).’ (DNH 462:12)
In spite of these minor violations of the correspondence between the word

and the morpheme, firstly in the case of compounding, secondly in the case
of class nouns, and finally in that of reduplication, it is reasonable to conclude
that the basic structure of White Hmong words is morphemic (see also Ratliff
1991, p. 701).
1.4 The Noun Phrase
1.4.1 The Basic Structure of the Noun Phrase

Ratliff (2010, p. 229) proposes the following core structure of the Noun Phrase
for all languages of the Hmong-Mien family:
[[NUM+CLF]QP+[(prefix)-NOUN+ADJ]NP]NP
Ratliff explains:
The hierarchical structure of the noun phrase in Hmongic languages, in

which the quantifier [= NUMeral above (NJ)] and the classifier form one
constituent (QP = quantifier phrase) and the noun and its modifiers form
another (NP = noun phrase), is reflected in those languages that have
tone sandhi relationships between certain elements of the noun phrase
but not others: the numeral may change the tone of the classifier; the
prefix may change the tone of the noun. But the classifier never changes
the tone of the prefix; these elements seem not to “see” each other.
Building on this core structure, the basic structure of the NP in White Hmong
is tentatively proposed as:
[POSS+[NUM+CLF/QUANT]QP+[(prefix-)NOUN+ATTR+DEM]NP]NP
Note that the Possessor is indicated here as separate from the Quantifier
Phrase, as there is no influence between a Possessor and a Numeral.
Examples (27)–(29) illustrate this structure:
(27) ob tug me-nyuam yaus no

two (animate) child young this
NUM CLF(t.c.) NOUN ATTR DEM
‘these two young children’
(28) kuv plaub tug nyuj ntawd

1SG four (animate) cow there.nearby
POSS NUM CLF(t.c.) NOUN DEM
‘my four cows there’
32 Chapter 1
(29) ntau yam tsiaj qus

many kind animal wild
QUANT CLF NOUN ATTR
‘many kinds of wild animals’
1.4.1.1 Comments on NP Constituent Order

a Non-numeric Quantifiers
The non-numeric quantifiers, such as coob ‘many (animate)’, ntau ‘much/
many (inanimate)’, and tsawg ‘little/few’, have been included in the same slot
as numerals, following Ratliff (1991, p. 695). However, as Ratliff points out (p. 2),
the non-numeric quantifiers, unlike numerals, do not seem to co-occur with
possessives. Furthermore, these quantifiers also differ from numerals in that
they can be ‘floated’, i.e. they can optionally appear following the noun. In this
case the noun appears without a classifier:
(30) muaj [nqaij tsawg~tsawg]S li

have meat RDUP~little INTNS
‘There’s very little meat.’ (Mottin, 1978, p. 52)
Compare:
(31) muaj [tsawg [lub tsev] <yug npuaO>REL]S

have few CLF house raise pig
‘There are few families who raise pigs.’ (Mottin, 1978, p. 52)
b Attributes that Precede the Noun

The normal position for an attribute in the NP in White Hmong is following the
noun, as shown in examples (27) and (29) above. However, for historical rea-
sons, there are a few common words that have an attributive function but that
directly precede the noun they modify, including: niag ‘great’, nyuag ‘little’, qub
‘old/former’. These are the only constituents that can appear between the clas-
sifier and the noun, sometimes even replacing the classifier after a numeral, as
in the (b) examples below.
(32) a. lub niag nroog

CLF great city
‘capital city’ (Heimbach, 1979, p. 140)
b. ib niag tsov
one great tiger
‘one big ol’ tiger’ (DNH 397: 3)
(33) a. peb tug nyuag hluas nkauj

three CLF(t.c.) little young unmarried.female
‘three little girls’ (DNH 130:8)
b. ib nyuag suab
one little sound
‘a little sound’ (DNH 34:42)
(34) lub qub zos

CLF old/former village
‘the old village’ (Heimbach, 1979, p. 265)
1.4.2 Classifiers and Measures

Classifiers and Measures in White Hmong appear before the noun, almost
always directly before it (see §1.4.1.1.b above regarding exceptions). However, a
classifier may appear without a following noun, in which case the interpreta-
tion depends on the context. If there is an antecedent, or if the referent is oth-
erwise clear from the context, then the classifier is interpreted anaphorically
as having definite reference:
(35) [koj lub tsheb]CS txawv [kuv lub]CC

2SG CLF car differ 1SG CLF
‘Your car is different from mine.’ (WHD 20:19)
If the reference is not clear from the context, however, then a classifier without
a following noun functions like an indefinite pronoun:
(36) tusA pom nwsO, los <qhia kuvO>PURP

CLF see 3SG come tell 1SG
‘(If) anyone sees him, come (and) tell me.’ (Mottin, 1978, p. 67)
1.4.2.1 Types of Classifiers: Sortal and Mensural

Two basic types of classifiers occur in White Hmong: sortal classifiers and men-
sural classifiers. Sortal classifiers delineate inherent properties of count nouns
(such as animacy, shape or function) and thus categorize according to kind.
Mensural classifiers delineate units of measure that are intrinsic to the refer-
ent concerned, and thus individuate/categorize both count and mass nouns
in terms of temporary features such as arrangement or quantity (Aikhenvald,
2000, pp. 114–115; Lyons, 1977, p. 463). Each mensural classifier is used only
with referents of a particular semantic type—those that naturally occur in
that arrangement or quantity. In this respect mensural classifiers differ from
34 Chapter 1
Measures, which delineate more general units of measure unrelated to the ref-
erents to be measured.
a Sortal Classifiers
Many of the most commonly occurring sortal classifiers in White Hmong are
linked with nouns according to some common feature of the shape of their
referents. For example:
– lub for round, bulky, or hollow objects and for machines (e.g. ntuj ‘the sun’,
tsev ‘house’, thoob ‘bucket’, vaj ‘tray’, tsheb ‘car’);
– tus for long, slender, cylindrical objects (e.g. tw ‘tail’, cwj mem ‘fountain pen’)
(tus is used not only for shape but also for type—see below);
– txoj for objects that come in lengths (longer than those for tus) (e.g. kev
‘road, path, way’, hlua ‘rope’);24
– daim for flat objects (e.g. teb ‘field’, ntawv ‘paper/writing’, nplooj ‘leaf’).
Three classifiers from this category are used for classifying most abstract nouns:
lub, tus and txoj. For example:
– lub for sij hawm ‘time’, npe ‘name’, neej ‘life’.

– tus for mob ‘sickness’, dab ntub ‘sleep’, nqe ‘price’;
– txoj for nominalized compounds with kev ‘way’ (e.g. kev txom nyem
‘suffering, poverty’, kev mob kev tuag ‘sickness and death’).
Some sortal classifiers delineate nouns according to their type rather than
their shape, that is according to some basic superordinate semantic category.
For example:
– leej for human beings (especially for family members) and spirits;
– tus for animate beings (e.g. neej ‘person’, dab ‘spirit, demon’, nees ‘horse’,
ntses ‘fish’) (tus is also used for shape—see above);
– rab for tools and weapons (e.g. riam ‘knife’, phom ‘gun’);
– tsab for messages (e.g. ntawv (paper/writing) ‘letter’, xov ‘message’);
– zaj for stories, speeches, songs, ceremonies (dab neeg ‘legend, folk tale’, hu
plig ‘soul-calling ceremony’).
24 Jaisser (1987) identified both length and flexibility as key features distinguishing txoj from
tus. In experimental conditions, Sakuragi and Fuller (2013) found a preference for the use
of txoj over tus for longer items of the same kind (e.g. zippers).
b Mensural Classifiers
Mensural classifiers that co-occur with count nouns serve as collectives, and
often indicate the way in which the referents are arranged with respect to one
another. For example:
– pab for groups of people or flocks of birds/sheep;

– cuam for lengths of thatch/roofing leaves;
– tsob for a grove or clump of plants (e.g. ntoo ‘tree’, nroj ‘herb’);25
– rev for stems or bunches of fruit, leaves, and flowers;
– zes for a clutch of eggs (from zes ‘nest’);
– yim for families.
Mensural classifiers that co-occur with mass nouns serve as partitives, describ-
ing a naturally or commonly occurring delimitation of a particular type of
referent:
– kob for showers of rain;

– tawb for pats of manure, (also for faeces, urine, and spit);
– rauv for squares of embroidery;
– maij for bolts of cloth.
1.4.2.2 Measure Words

Measure words can serve to divide count or mass noun referents of many dif-
ferent semantic types into fairly vague divisions or parts:
– hom/yam for kinds, classes;

– fab for sections, divisions, points of the compass;
– kwv for portions, parts.
Measure words that describe specific units of measure serve as partitives when
used with both count and mass nouns. Many of these quantifiers are mor-
phemes that are used elsewhere as nouns that refer to vessels, enclosures and
25 Bisang (1993, pp. 6, 13) includes tsob in his class of ‘intrinsic quantifiers’. However, its posi-
tion before rather than after Pre-Nominal Attributive niag ‘great’ suggests that it can func-
tion as a Unit Classifier:
(i) . . . qaws zoj zees [tsob niag tauj]O
roll.up hurriedly CLF great reeds
‘. . . hurriedly rolled up clumps of big ol’ reeds . . .’ (DNH 131:19)
36 Chapter 1
so on, that typically hold or bound certain quantities of the referent in ques-
tion. For example:
– khob for cupfuls (khob ‘small cup’);

– tais for bowlfuls (tais ‘bowl’);
– nrog for trayfuls (nrog ‘flat area inside winnowing tray’).
1.4.2.3 The Specification of ‘Underspecified’ Nouns

The system of classifiers in White Hmong clearly has a strong semantic basis;
classifiers are not simply paired with nouns in an idiosyncratic way that would
have to be learnt by rote. This point is reinforced by looking at a few examples
of certain semantically ‘underspecified’ nouns (Ratliff, 1991), such as ntawv
‘paper/writing’, duab ‘image’, or lus ‘speech/word’, which can co-occur with
several different classifiers, resulting in further specification of the referent.
For example:
– daim ntawv ‘sheet of paper’ (daim for flat things)

– nplooj ntawv ‘page’ (nplooj for leaves)
– tsab ntawv ‘letter’ (tsab for messages)
– phau ntawv ‘book’ (phau for piles)
– tsiaj ntawv ‘letters of the alphabet’ (tsiaj ‘animal’)
– tus duab ‘shadow’ (tus for long things and animate beings)
– daim duab ‘photograph’ (daim for flat things)
In spite of this slight addition to their functional load, the morphemes daim,
tsab, phau, etc. in the examples above are still considered classifiers. This analy-
sis is supported by their appearance between numeral and noun in expressions
such as: ib phau ntawv ‘one book’ etc. However, Ratliff (1991) gives evidence to
show that, in the environment of what is here termed the ‘universal’ collective
classifier, cov, these morphemes actually function as part of a nominal com-
pound with the ‘underspecified’ noun that follows them. If they were analysed
as classifiers in this environment, they would violate the constraint of one clas-
sifier per NP.
(37) cov phau ntawv

CLF.PL book paper/writing
‘the books’ (phau elsewhere classifier for piles)
(38) cov tsab ntawv

CLF.PL letter paper/writing
‘the letters’ (tsab elsewhere classifier for messages)
1.4.2.4 The ‘Universal’ Collective Classifier cov

The ‘universal’ collective classifier cov, meaning ‘group’, functions in much
the same way as other collective classifiers with count nouns with plural ref-
erents, except that it is used regardless of the arrangement or type of refer-
ent involved. Its basic use is thus to indicate plurals of an unspecified number.
Compare:
– tus noog (CLF-animate bird) ‘the bird’

– cov noog (CLF.PL bird) ‘the birds’
– ib pab noog (one CLF-flock bird) ‘a flock of birds’
– ib co noog (one CLF.PL(t.c.) bird) ‘a group of birds/some birds’
– lub txiv tsawb (CLF-bulky banana) ‘the banana’
– cov txiv tsawb (CLF.PL bananas) ‘the bananas’
– ib kuam txiv tsawb (one CLF-hand banana) ‘a hand of bananas’
– ib co txiv tsawb (one CLF.PL(t.c.) banana) ‘a group of/some bananas’
Unlike other collective classifiers, however, cov also appears with mass nouns:
– cov mov (CLF.PL rice) ‘the rice’

– kuv cov mov (1SG CLF.PL rice) ‘my rice’
– cov mov no (CLF.PL rice this) ‘this rice’
According to Ratliff (1991, p. 700),
cov meets syntactic criteria for membership in the classifier class, but
does not consistently meet semantic or functional criteria . . . It is best
thought of as a word that has undergone a degree of grammaticalization
toward an abstract plural.
1.4.2.5 The Grammatical Functions of Classifiers

Classification in White Hmong is related to instantiation; that is a classified NP
refers to one, or more than one, instance of the referent of that NP. An instanti-
ated NP is not necessarily definite, nor even specific. For example, quantified
NPs refer to a certain quantity or number of instantiations of the referent, but
may be indefinite, and even non-specific. Observe the following examples:
38 Chapter 1
(39) Indefinite/specific
puag thaum ub, muaj [[ib tug qav] thiab
INTNS time yonder have one CLF(t.c.) frog and
[ib tug tsov]]S
one CLF(t.c.) tiger
‘Long ago, there was a tiger and a frog . . .’ (LPV 1)
(40) Indefinite/specific
mas muaj [[cuaj dag nta] hos [yim
well have nine arm.span(t.c.) cross-piece CONJ eight
daj taub]]O
arm.span head
‘Well, [my crossbow] has a nine-arm-span cross-piece and an eight-
arm-span handle.’ (DNH 35: 50)
(41) Indefinite/non-specific
kuvA xav yuav [ob tug qaib]O
1SG want obtain two CLF(t.c.) chicken
‘I want to buy two chickens.’
(42) Indefinite/non-specific
Niam, kojA muab [ob lub nyuag vab]O los . . .
mother 2SG take two CLF little tray come
‘Mum, you bring two little trays . . .’ (DNH 382: 35)
Of course, any definite (and thereby specific) NP naturally constitutes an

instantiation, and so will include a classifier, as in examples (43)–(46) below:
(43) nws ob nthab nplej / nws nthab nplej

3SG two platform rice 3SG platform rice
‘his two platforms-worth of rice’ / ‘his platform-worth of rice’
(44) plaub tug neeg no / tus neeg no

four CLF(t.c.) person this CLF person this
‘these four people’ / ‘this person’
(45) [tus tsov]S tshaib~tshaib-plab li

CLF tiger RDUP~hungry-stomach INTNS
‘The tiger was very hungry.’ (LPV p. 1)
(46) cov piam-thaj no

CLF.PL sugar this
‘this sugar’
NPs that do not refer to instantiations of their referents appear without a clas-
sifier in White Hmong. Examples of these are generic NPs, as in examples (47)
and (48) below, and NPs with indefinite/non-specific reference that are not
quantified, as in example (49):
(47) plasCS y og noogCC

owl COP bird
‘Owls are birds.’ (IWH 12: 7)
(48) nwsA nyiam <noj zaubO>COMP

3SG like eat vegetable
‘She likes eating vegetables.’
(49) haus dejO

drink water
‘Drink [some] water!’
(50) yog tsis muaj tsuS mas pebA xuas laujkaubO <ncu>PURP
COP NEG have steamer TOP 1PL take pot simmer
‘If there isn’t a steamer, we take a pot to simmer [it].’ (IWH 17:7)
An exception to the rule that instantiated nouns appear with a classifier are
certain count nouns that are clearly instantiated, such as proper nouns, pro-
nouns, and certain kinship terms that can have only one possible referent (e.g.
kuv niam ‘my mother’, kuv txiv ‘my father’). These generally appear without a
classifier.26
An exception to the rule that NPs that are not instantiated do not contain a
classifier is the occasional appearance of a generic NP with a classifier:
26 The word txiv (man), used without a classifier means ‘father’, as distinct from tus txiv (CLF
male), which is interpreted to mean ‘husband’ in appropriate contexts (e.g. kuv tus txiv
(1SG CLF man) ‘my husband’).
40 Chapter 1
(51) Generic
plasCS zoo-li [tus miv]CC
owl be(come).like CLF cat
‘Owls are like the cat.’ (IWH 12:7)
In this example, both the first and the second NP are interpreted as generic.
The first—plas ‘owl’—occurs without a classifier, while the second—tus mis—
‘the cat’ occurs with one.
1.5 The Locative Phrase and Spatial Deictics
Locative phrases are most commonly composed of a Noun Phrase preceded

by a ‘Spatial Deictic’. The small, closed class of morphemes here termed
‘Spatial Deictics’ (following Ratliff (1990)) has elsewhere been referred to as
‘prepositions of place’ (‘prepositions de lieu’) (Mottin, 1978, p. 70), ‘prepositions’
(Fuller, 1985, p. 39), ‘relative location nouns’ or ‘relator nouns’ (Clark, 1989),
and ‘denominal prepositions’ (Ratliff, 1992, pp. 104–112). The class includes the
following members: tom ‘there-distant’; ntawm ‘there-nearby’; pem ‘there-up
(a slope)’; nram ‘there-down (a slope)’; tim ‘there-across’; saum ‘there-above’;
hauv ‘there-inside’; nraum ‘there-outside’.27
All members of this word class, as delineated in this analysis, indicate some-
thing about the spatial properties of the place they refer to. Three of them
indicate a salient part or aspect of that place: saum ‘there-above/top’, hauv
‘there-inside’, and nraum ‘there-outside/other side’. Three others, namely pem
‘there-up (a slope)’, nram ‘there-down (a slope)’, and tim ‘there-across’ indi-
cate the location of the place referred to with respect to some topological (or
27 Ratliff (1992, p. 106) includes txhaim ‘outside’ (a form that is recorded by Bertrais-Charrier
(1964) but that may have fallen out of use) in her category of ‘denominal prepositions’.
Ratliff also adds thaum ‘time, when’ and zaum ‘time’ into this category (which thus con-
tains both spatial and temporal words) but does not include hauv ‘there-inside’. This is
because her focus is on describing “an important tonally defined form class” (p. 104). In
her analysis, it makes sense to include zaum ‘time’ as, like the spatial words in this group,
it is clearly an ‘m-tone’ word that is related to a cognate noun with a different tone: zaus ‘a
time, a turn, an occasion’. Here, however, the focus is on describing a syntactically rather
than a tonally defined class. As Ratliff (1992) shows, all these spatial and temporal words
can precede a demonstrative noun such as no ‘this’. However, the syntactic environments
in which zaum and thaum occur are otherwise very different from those described here
for the Spatial Deictics. For this reason I have not included them in this word class.
architectural etc.) feature (e.g. mountain, valley, stairs).28 In addition to their

reference to topology, these three Spatial Deictics, as well nraum (see above)—
when it means ‘there-other side’—also express deixis. That is, the location
or point of view of the speaker can clearly be relevant to their usage. Only two
of the Spatial Deictics, however, are purely deictic: ntawm ‘there-nearby’ and
tom ‘there-distant’; these indicate the relative distance of the place referred to
with respect to the location of the speaker, or to the speaker’s adopted point of
view; they make no reference to any part of that place itself, nor to any topo-
logical feature.
Spatial Deictics are directly followed by Noun Phrases (including pronouns,
as in (53), and proper nouns, as in (54)) that can be taken to express Locative
meaning (Location, Source, or Goal):
(52) Location
nwsS ua.si [hauv [lub vaj]]
3SG play in CLF garden
‘She is playing in the garden.’
(53) Source
kuvA qiv [kaum doslas]O [ntawm nws]E
1SG borrow ten dollar nearby 3SG
‘I borrowed $10 from him.’
(54) Goal
lawvA mus [nram Canberra]O lawm
3PL go down PN PRF
‘They have gone down to Canberra.’
It is possible for more than one Spatial Deictic to occur before a Noun Phrase,
as in the following example:
(55) zaum [pem hauv ntej]

sit up in front
‘Sit up in front.’ (WHD 19:18)
28 Ratliff (1992, p. 105) notes that pem ‘there-up’ is historically related to the noun pes ‘moun-
tain’, nram ‘there-down’ to the noun nras ‘plain’, and nraum ‘there-outside’ to the noun
nrau ‘place beyond’.
42 Chapter 1
A Locative Phrase can also consist of a Spatial Deictic alone. When a Spatial
Deictic with tone ‘m’ constitutes the locative phrase alone, it undergoes
the morphotonemic change to tone ‘d’, in keeping with the rules described
in §1.2.3.5.b above. For example:
(56) lawvA mus nradO lawm

3PL go place.down PRF
‘They have gone down there.’
Even though these words have the characteristics of conveying spatial infor-
mation and appearing pre-nominally—characteristics often associate with
words in the class of prepositions in many languages—they are not regarded
as prepositions in this study. This is because they do not fulfil another criteria
generally associated with prepositions: the function of conveying information
about the semantic role of the NP that follows them. A single Spatial Deictic
can occur in Locative Phrases that express a variety of semantic roles, and dif-
ferent Spatial Deictics can be used in Locative Phrases that express the same
semantic role. Noun Phrases referring to locatives of all types—Location,
Source, and Goal—co-occur with ntawm in examples (57)–(59):
(57) Location
nwsS khiav [ntawm khw]
3SG run nearby market
‘He ran (about) nearby at the market.’
(58) Goal
nwsA mus [ntawm khw]O
3SG go nearby market
‘He went nearby to market.’
(59) Source
nwsA/S tuaj [ntawm [lub zos]]O tuaj
3SG come nearby CLF village come
‘He came from nearby at the village.’
As long as the spatial reference is appropriate, other Spatial Deictics can be

substituted for ntawm in the examples above without affecting the interpreta-
tion of the semantic role (Location, Source, or Goal) of the following NP in the
sentence:
(60) Location (Compare (57) above)

nwsS khiav [nram khw]
3SG go down market
‘He ran (about) down at the market.’
(61) Goal (Compare (58) above)

nwsA mus [tim khw]O
3SG go across market
‘He went across to the market.’
(62) Source (Compare (59) above)

nwsA/S tuaj [pem [lub zos]]O tuaj
3SG come up CLF village come
‘He came from up at the village.’
If the Spatial Deictics in the sentences above do not convey information

regarding the semantic role of the NP that follows them, what element in the
sentence is responsible for conveying this information? This question will be
addressed in §1.6.5.3 below.
1.6 Simple Clause Types
In this section the basic structure of some common simple clause types in
White Hmong will be described. The ‘simple clause’ is an independent, declar-
ative, affirmative, active clause (Givón, 1984, p. 85).
1.6.1 Existential Clauses

Both in ‘bare’ existential clauses, which simply postulate the existence of some
entity or entities, and in presentative existential clauses, which serve to intro-
duce entities with indefinite specific reference into the discourse, the verb
muaj ‘have’ is utilised. The occurrence of the verb meaning ‘have’ not only in
possessive clauses but also in existential ones is quite common in other lan-
guages of the geographical area in which Hmong is spoken, including Thai,
Vietnamese, Khmer, and Mandarin (Clark, 1989, pp. 206–208).
The single argument of a ‘bare’ existential clause, that is the generic NP
whose existence is asserted/postulated, appears after the transitive verb
muaj ‘have’ in the position of grammatical object. The preverbal, grammati-
cal subject position is filled with a non-referential third person pronoun nws,
44 Chapter 1
similar to the so-called ‘dummy’ subject there in the English translations of

(63)–(65) below.
(63) nwsA muaj [tib neej zoo]O, nwsA muaj

3SG have human.being be(come).good 3SG have
[tib neej tsis zoo]O thiab
human.being NEG be(come).good also
‘There are good people and there are bad people too.’
(64) nwsA yeej muaj [Huab-tais ntuj]O

3SG HAB have Lord sky
‘There is a God.’
(65) nwsA yeej yuav-tsum muaj rogO

3SG HAB must have war
‘There must (always) be wars.’
While the preverbal grammatical subject slot is filled with the non-referential
3SG pronoun nws in ‘bare’ existential clauses, it remains empty in the presenta-
tive existential construction, as shown in examples (66)–(68) below:
(66) thaum ub muaj [ib tug tsov]S . . .

time yonder have one CLF tiger
‘Once upon a time there was a tiger.’ (LPV 1)
(67) muaj [ib hnub]S thaum sawvdawsS mus <ua-teb>PURP

have one day time everyone go make-field
tag lawm, . . .
finish PRF
‘One day (lit: there was a day), when everyone had gone off to work in the
fields, . . . (IWH 29:7)
(68) [nram kwj-deg nrad] muaj [ib tug niag

down gulley-water(t.c.) down(t.c.) have one CLF great
maum-zaj-laug]S . . .
female-dragon-elder
‘Down in the gulley down there, there was a great big old female
dragon . . .’ (DNH 138:69)
Presentative clauses can also be introduced by certain intransitive verbs that

describe the way in which a new referent arrives on the scene:
(69) tuaj [peb tug nyuag hluas-nkauj]S

come three CLF little young-unmarried.woman
‘. . . (there will) come three young girls . . .’ (DNH 130:8)
(70) tom-ntej ces yuav daug [cuaj tug dab <noj nyoog>REL]S;
at.first TOP IRR hatch nine CLF(t.c.) spirit eat be.raw
tom-qab yuav daug [yim tus dab <haus txias>REL]S . . .
afterwards IRR hatch eight CLF spirit drink be.cold
‘At first (there) will hatch nine spirits (who) eat (meat) raw; then (there)
will hatch eight spirits (who) drink (blood) cold.’ (DNH 26:7)
In this usage, the intransitive verb precedes the NP in S function, so the normal
constituent order of the intransitive clause is reversed. The effect, as in the
English translations, is rather poetic.
1.6.2 Nominal Subject Complement Clauses

Nominal subject complements most commonly follow the copula yog ‘be the
case’, as in examples (71) and (72):
(71) plasCS yog noogCC

owl COP bird
‘Owls are birds.’ (IWH 12:7)
(72) nwsCS yog [ib tug xibfwb]CC

3SG COP one CLF(t.c.) teacher
‘She is a teacher.’
When the referent of the subject complement is indefinite, as in the examples

above, it describes some attribute (generally permanent or semi-permanent)
of the subject (‘attributive’). However, when it has definite reference it serves
to identify the subject (‘referential’). For example:
(73) nwsCS yog [tus xib fwb (<uas pab kuv>REL)]CC

3SG COP CLF teacher REL help 1SG
‘He is the teacher (who helped me).’
46 Chapter 1
(74) nwsCS yog [Huab-tais tus ntxhais]CC

3SG COP Lord CLF daughter
‘She is the King’s daughter.’
1.6.3 Intransitive Clauses

1.6.3.1 Stative Intransitive Clauses
Clauses in which the predicate is an intransitive stative verb describing some
attribute of the subject are semantically similar to those with indefinite nomi-
nal subject complements (see §1.6.2 above). However, in this case, no copula
verb is required; the stative verb simply follows the subject NP, as in any other
example of the intransitive clause type. There is thus no need for a word class
‘Adjective’, separate from that of ‘Verb’, in this language. The states expressed
may be permanent, as in (75) below, or temporary, as in (76).
(75) [lub nkoj no]S loj

CLF boat this be.big
‘This boat is big.’
(76) nwsS mob heev

3SG be.sick very
‘He is very sick.’
Most of these verbs can be used not only to express a state, but also a change of
state. The interpretation is determined by the context:
(77) [koj cov plaub-hau]S ntev me-ntsis lawm

2SG CLF.PL hair be.long a.bit PRF
‘Your hair has got a bit longer.’
It is extremely common for verbs of this type to undergo reduplication when

used as predicates. In fact, a few of them, for example me ‘be small’, hardly ever
seem to appear except in their reduplicated form in this position.
(78) . . . [nws cov plaub]S dub~dub

3SG CLF.PL fur RDUP~be.black
‘. . . his fur was very very black.’ (IWH 29:7)
As noted in §1.3.3.3 above, if the predicate is composed of two syllables (e.g.

txom-nyem ‘be poor/destitute’) it is the first that is reduplicated:
(79) Txiv Nraug NtsuagS txom~txom-nyem

the.Orphan.Boy RDUP~be.poor
‘. . . the Orphan Boy was very very poor.’ (TNN p. 1)
1.6.3.2 Non-stative Intransitive Clauses

Intransitive clauses with non-stative predicates are of the same basic type as
those with stative predicates described above, the subject NP simply preceding
the verb. Three common kinds of intransitive predicates, change-of-state verbs,
motion verbs and verbs describing bodily functions, are illustrated below:
(80) [koj tus me-nyuam]S hlob zoo~zoo

2SG CLF child grow RDUP~be.good
‘Your child is growing well.’ (Heimbach, 1979, p. 64)
(81) nwsS txawm khiav~khiav . . .

3SG then RDUP~run
‘Then he ran and ran . . .’ (KNH p. 3)
(82) nwsS hnoos tas-hnub-tas-hmo

3SG cough continually
‘He coughed continually.’
Reduplication of the verb is quite common in these non-stative intransi-

tive clauses, although not as common as with stative intransitive verbs used
predicatively.
1.6.4 Reciprocal Clauses

Many predicates involving inherent reciprocity can occur either in the transi-
tive clause type, whereby one of the co-agents is expressed as the subject and
the other as the object, or in the intransitive clause type, whereby the two co-
agents are expressed as a plural or conjoined subject. In the latter case White
Hmong employs the morpheme sib before the verb. For example:
(83) a. nwsA ncaim [nws tus phoojywg]O

3SG separate 3SG CLF friend
‘She parted from her friend.’
(84) b. [ob tug phoojywg]S sib ncaim

two CLF(t.c.) friend RECP separate
‘Two friends parted from each other.’
48 Chapter 1
Many basically transitive verbs that do not involve inherent reciprocity can
also be de-transitivized with the morpheme sib to express reciprocal action.
The subject is either a plural or conjoined NP. For example:
(85) [kuv thiab [kuv tus txiv-hlob]]S sib pab

1SG and 1SG CLF uncle RECP help
‘My uncle and I help each other.’
Finally, the morpheme sib can be used in a non-reciprocal sense with a transi-
tive verb, indicating some linkage of the objects rather than the subjects in the
action. (Note, however, that the sib+V predicate does not take a direct object
even in this case. Rather a type of serial verb construction (referred to here
as the Disposal SVC) is used, in which the objects are expressed as the direct
object of the verb muab ‘take’ (see Chapter 3, §1.4.6)).
(86) [cov me-nyuam no]A/A/S tau muab [lawv cov

CLF.PL child this ACHV take 3PL CLF.PL
khoom-ua.si]O/O sib xyaws ua.si
things-play RECP mix play
‘These children got to share their toys.’ (Lit: ‘got to take their toys and mix
(them) together to play.’)
(87) kuvA/A muab [ob txog hlua]O/O sib khi ua.ke

1SG take two CLF rope RECP tie together
‘I tied two ropes together.’
1.6.5 Transitive Clauses

The transitive clause type involves a preverbal subject NP and a postverbal
object NP. Clauses containing archetypal transitive verbs, with agentive sub-
jects and affected or effected (i.e. created) objects, can occur in this pattern.
For example:
(88) kuvA dua [daim ntawv no]O lawm

1SG tear CLF paper this PRF
‘I have torn this piece of paper.’
(89) . . . kuvA thiaj txua [ib rab hneev]O

. . . 1SG so construct one CLF crossbow
‘. . . so I made a crossbow . . .’ (KNH p. 1)
There are many more clauses containing verbs with subjects that are non-
agentive and/or objects registering little or no effect as a result of the event or
state described, which also conform syntactically to the basic transitive clause
type. A few are described below.
1.6.5.1 Experiencer Subjects

Clauses with subjects that refer to Experiencers, such as those with predicates
of perception, cognition, and preference, conform to the canonical transitive
structure:
(90) wbA hnov xob-nrooO

1DU hear thunder
‘We heard thunder.’
(91) nwsA paub [kuv tus txiv]O

3SG know 1SG CLF male
‘He knows my husband.’
(92) nkawdA hlub [nkawd cov me-nyuam]O

3DU be.fond.of 3DU CLF.PL child
‘Those two are fond of their children.’
1.6.5.2 Possessive Clauses

Possessive clauses also pattern transitively, in spite of the fact that their objects
are not affected in any way.
(93) kuvA muaj [ob tug me-nyuam]O

1SG have two CLF(t.c.) child
‘I have two children.’
Givón (1984, p. 103) makes the following observation about the use of the tran-
sitive clause type to express stative possession in many languages:
Most commonly, a ‘have’ verb arises out of the semantic bleaching

of active possession verbs such as ‘get’, ‘grab’, ‘seize’, ‘take’, ‘obtain’ etc.,
whereby the sense of ‘acting to take possession’ has been bleached, leav-
ing behind only its implied result of ‘having possession’. This historical
process explains why the syntactic coding of such a verb conforms to the
transitive pattern of its source.
50 Chapter 1
In the light of these comments it is interesting to note the historical relation-

ship in White Hmong between the stative possessive verb muaj ‘have’ and the
active possessive verb muab ‘grasp with the hand’/‘take’ (David Strecker, per-
sonal communication).
1.6.5.3 Clauses Containing Locative Arguments

The term ‘Locative argument’ is used broadly in the discussion below, to refer
not only to arguments referring to Location, but also to those referring to
Source, Path, and Goal of Motion (i.e. origin, route, and destination), to Source
and Goal of Transfer (e.g. donor and recipient) and to Goal of Action (i.e.
beneficiary).
Certain verb types in White Hmong that pattern transitively may be
described as ‘inherent’ Location, Source, Path, or Goal Verbs: a single Noun
Phrase or Locative Phrase (depending on the verb type—see below) following
such a verb is automatically interpreted as the Location, Source, Path, or Goal
of the action or event described.
a Inherent Path Verbs

A Locative argument of a Path Verb is simply expressed by a Noun Phrase as O
argument, describing the Path of the motion of the subject. For example:
(94) nwsA nce ntaivO

3SG climb stair
‘She climbed the stairs.’
(95) lawvA hla [lub pas-dej]O

3PL pass CLF pond
‘They passed the pond.’
b Inherent Location Verbs

Verbs of the Stance/Rest type take core inner locative Location arguments.
A Locative Phrase directly following a verb of stance/rest is interpreted as
expressing the Location of the subject:
(96) nwsA zaum [saum [lub rooj]]O

3SG sit top CLF table
‘He is sitting on the table.’
(97) nkawdA nyob [nram hav-dej]O

3DU be.located down valley-water
‘They are down in the river valley.’
Transitive verbs that describe actions in which the location of the object can
be thought of as independent of the location of the action as a whole can also
take inner locative Location arguments (although these are not core inner loc-
atives as they are for verbs of the Stance/Rest type):
(98) pebA rauv tawsO [hauv [qhov-cub]]

1PL burn wood inside stove
‘We burn wood in a stove.’
(99) yusA puab ncuavO [hauv [nplooj tsawb]]

INDF wrap rice.cake inside leaf banana
‘One wraps rice cakes in a banana leaf.’
(100) NtxawmA nqa [Ntxawm tus me-nyuam]O [ntawm

PN carry PN CLF child nearby
xub-ntiag] xwb
breast only
‘Nzue only carried her child on her breast.’ (DNH 382:34)
c Inherent Source Verbs

There do not seem to be any Motion verbs in White Hmong that can be
thought of as ‘inherent’ Source verbs, taking an inner locative Source argu-
ment. It is certainly true that telic motion verbs like tawm ‘leave’, sawv ‘rise, get
up’, dim ‘escape’, and thim ‘retreat’ occur quite frequently with a following loca-
tive argument that is interpreted as Source. However, this interpretation only
seems to occur if the locative argument is then followed by a verb or phrase
expressing Goal (often in a serial verb construction), or if a Goal is otherwise
clear from the context.
(101) nwsA/S tawm [nram liaj]O mus

3SG leave down paddy.field go
‘He left the paddy field.’
(102) a. nwsA/S sawv [hauv [nws lub txaj]O los

3SG rise inside 3SG CLF bed come
‘She got up from her bed.’
52 Chapter 1
b. nwsA sawv [hauv [nws lub txaj]O

3SG rise inside 3SG CLF bed
‘She stood up on her bed.’
As is clear from example (102)b, in the absence of reference to Goal (provided

by the Goal verb los ‘come’ in (102)a), the O argument of sawv ‘rise, get up’ is
interpreted as a Location rather than a Source.
However, while Motion verbs do not seem to take inner locative Source argu-
ments, some Transfer verbs do. Transfer verbs that describe an action in which
the subject is responsible for transferring the location of the object (physically
or metaphorically), though not moving him/herself, can be inherent Source
Verbs. The kind of Transfer verb that belongs to this category is one in which
the subject is thought of as the Goal of the transfer, and so no other refer-
ence to a Goal is appropriate. Such verbs include qiv ‘borrow’, txeeb ‘snatch’,
muas ‘buy’, txais ‘receive’, yuav ‘obtain’, thov ‘request’, and rho ‘extract’. A single
Locative Phrase after one of these verbs is interpreted as Source:
(103) [tus tub-sab]A txeeb [kuv cov nyiaj]O [ntawm

CLF robber snatch 1SG CLF.PL money nearby
[kuv tes]]
1SG hand
‘The robber snatched my money from my hand.’
(104) kuvA thov nyiajO [ntawm koj]

1SG request money nearby 2SG
‘May I have some money?’ (Lit: ‘I request money (from) you.’)29
d Inherent Goal Verbs

The class of inherent Goal verbs includes the class of Deictic Motion verbs in
White Hmong: the verbs mus ‘go’ and tuaj ‘come’ and los ‘come/return home’.
29 Looking at examples (103) and (104) it will be clear why the Spatial Deictic ntawm ‘nearby’
has been misinterpreted by some scholars of the White Hmong language (e.g. Mottin
(1978, p. 73)) as a morpheme which itself contributes the notion of Source. In fact it is
simply the case that ntawm ‘nearby’ is the Spatial Deictic which occurs most frequently in
Source phrases following Transfer verbs because it is the least semantically marked of all
the members of this class. It will be clear from the examples of clauses with Motion verbs
above (examples (101) and (102)) that any of the Spatial Deictics can occur in a Locative
Phrase expressing Source, provided their deictic and spatial reference is appropriate.
A single Locative Phrase following one of these verbs is understood to refer to

the Goal of the motion:
(105) nwsA mus/tuaj [nram hav-dej]O

3SG go/come down valley-water
‘She went/came down to the river valley.’
(106) lawvA los [tom tsev]O

3PL come over.there house
‘They came home.’
e Locative Phrases with Other Verb Types

Following all other verb types a single Locative Phrase will be interpreted as
indicating the Location of the action or event described; that is, it is inter-
preted as an outer locative Location.
(107) [cov me-nyuam]S ua.si [hauv vaj]

CLF.PL child play inside garden
‘The children are playing in the garden.’
(108) [cov tub]A tua ntshuabO [nram hav-dej]

CLF.PL boy shoot otter down valley-water
‘The boys shoot otters down at the river valley.’
Even following some motion verbs, namely atelic motion verbs, a Locative
Phrase is interpreted as an outer locative Location, rather than as a Source or
Goal. For example:
(109) kuvS khiav [hauv khw]

1SG run inside market
‘I ran (about) in the market.’
(110) [tus poj-niam]A ev [nws tus me-nyuam]O [tom teb]

CLF woman carry 3SG CLF child over.there field
‘The woman carried her child about, over there at the field.’
1.6.6 Ditransitive Clauses

Ditransitive verbs involve verbs that can take three core arguments—the
subject and two objects. One of these objects is a Locative (Source or Goal
54 Chapter 1
of the transfer) and the other is a Theme (the thing transferred.) The kind of
verbs most likely to be used ditransitively are those that describe transactions.
As noted above, White Hmong Transfer verbs describing actions in which
the subject is thought of as the Goal of the transfer (i.e. the Recipient) are com-
monly used in clauses in which the Theme is expressed by a Noun Phrase, and
the Source expressed by a Locative Phrase:
(111) [tus tub-sab]A txeeb [kuv cov nyiaj]O [ntawm

CLF robber snatch 1SG CLF.PL money nearby
[kuv tes]]E
1SG hand
‘The robber snatched my money from my hand.’
(112) kuvA thov nyiajO [ntawm koj]E

1SG request money nearby 2SG
‘May I have some money?’ (Lit: ‘I request money (from) you.’)
There is another common way of expressing actions of this type that appears
to involve a ditransitive structure with two NP objects: the first expressing the
Source, and the second the Theme. Observe the following example, from Clark
(1980b, p. 31), and the translation given:
(113) kuv txais koj ib tawb ntsev coj mus noj

1SG borrow 2SG one basket salt carry go eat
“I’m borrowing from you a basket of salt to take for eating.”
(Bertrais-Charrier (1964) txais; translation Clark’s)
Note, however, that an alternative, and more accurate, interpretation is avail-

able for sentences like these, in which the pronoun koj (2SG) is understood not
as a separate argument (the Source of the transfer), but rather as the owner
of the thing transferred: ‘I’m borrowing your basket of salt to take to eat.’ In
fact, White Hmong consultants for this study will not accept examples like the
one above if the post-verbal NP is not understood to refer to the actual owner
of the thing transferred. This suggests that a ditransitive structure is probably
not involved at all. Hence the following sentence is acceptable if ‘my friend’ is
understood to be the original owner of the house, but not if s/he is simply act-
ing as an agent for someone else.
(114) kuvA muas [[kuv tus phoojywg] [ib lub tsev]]O

1SG buy 1SG CLF friend one CLF house
‘I bought my friend’s house.’
not: ‘I bought a house from my friend.’
Transfer verbs in which the subject is thought of as the Source of the trans-
fer and the Locative argument as the Goal (e.g. pub ‘give’, ‘send’ etc.) are not
used ditransitively in White Hmong. Instead, the Theme occurs as object and
the Goal is introduced by the morpheme rau. Loosely glossed here as ‘to’ (or
‘for’), the morpheme rau originates from a verb meaning ‘put in, insert’; see
Chapter 4, §1.2.3.4).
(115) a. nwsA muag [ib lub tsheb]O rau kuv

3SG sell one CLF car to 1SG
‘He sold a car to me.’
b. * nws muag kuv [ib lub tsheb]

3SG sell 1SG one CLF car
Intended meaning: ‘He sold me a car.’
(116) a. kuvA xa [ib tsab ntawv]O rau [kuv niam]

1SG send one CLF writing to 1SG mother
‘I sent a letter to my mother.’
b. * kuv xa [kuv niam] [ib tsab ntawv]

1SG send 1SG mother one CLF writing
Intended meaning: ‘I sent my mother a letter.’
However, in an appropriate context, the object of such verbs, including the

Transfer verb pub ‘give’ and the Speech Act verb hais ‘say’, can sometimes be
interpreted as a Goal rather than a Theme. If both Theme and Goal appear,
however, the Goal must be introduced by rau ‘to’.
(117) a. wbA pub nwsO

1DU give 3SG
‘We gave [it] (to) her.’
b. (i) wbA pub [ib daig tiab]O rau nws

1DU give one CLF skirt to 3SG
‘We gave a skirt to her.’
56 Chapter 1
b. (ii) * wb pub [ib daig tiab] nws

1DU give one CLF skirt 3SG
b. (iii) * wb pub nws [ib daig tiab]30

1DL give 3SG one CLF skirt
(118) a. yog kojS yuav mus, hais kuvO tso

COP 2SG IRR go speak 1SG first
‘If you are going, tell me first.’
b. (i) hais [lus Hmoob]O rau kuv

speak word Hmong to me
‘Speak Hmong to me.’
b. (ii) * hais [lus Hmoob] kuv

speak word Hmong 1SG
b. (iii) * hais kuv [lus Hmoob]

speak 1SG word Hmong
The restriction seems to be that no verb can simultaneously introduce a Theme

and a Goal into the clause, rather than that no verb can govern both Theme
and Goal arguments. See Enfield (2007, pp. 355–382) for a thorough discussion
of similar constraints in Lao.
1.7 Operations on the Simple Clause: Mood, Modality, and Aspect
1.7.1 Imperative
The most common form of the imperative differs little from a basic declara-
tive utterance in structure. It generally, though not invariably, ends with an
illocutionary particle, such as nawb, mas, es, or os, the effect of which is to
give the imperative utterance the force of an invitation or suggestion rather
than a command. The inclusion of the second person pronoun subject is
optional.
30 This structure is marginally acceptable to some speakers. Its acceptability may be con-
fined to those who are fluent in English, and is possibly influenced by the English ‘dative
shift’ construction.
(119) (koj)A/A nrog pebO noj movO mas

2SG be.with 1PL eat rice IP
‘Eat with us.’ (WHD 16:12)
(120) (koj)S zaum [pem hauv ntej]

2SG sit up in front
‘(You) sit up in front.’ (WHD 19:18)
It is also quite common to repeat the second person pronoun after the verb in
an imperative utterance; again, its appearance preverbally is optional.
(121) . . . kojA muaj numO los kojS mus kojS es

2SG have work CONJ 2SG go 2SG IP
‘. . . you have work (to do) so you go . . .’ (DNH 274:34)
(122) nebS kawm nebS os!

2DU study 2DU IP
‘Study, you two!’
(123) mam mus kojS os!

therefore go 2SG IP
‘Off you go then!’ (WHD 9:4)
The negative imperative is formed by placing the morpheme txhob immedi-

ately before the verb. This negative morpheme occurs in all irrealis contexts; it is
optionally preceded by the ordinary negative morpheme tsis (see §1.7.3 below).
(124) koj (tsis) txhob ua!

2SG NEG IRR.NEG do
‘Don’t do that.’
1.7.2 Interrogative and Uncertainty with puas

The morpheme puas signals that the truth-value of the proposition conveyed
by the clause is unknown to the speaker. In a main clause its function is to
indicate interrogative mood:
(125) koj puas nqis los [hauv no] na?

2SG descend come inside here IP
‘Are you coming down here?’ (DNH 399:14)
58 Chapter 1
In a subordinate clause it functions as a status marker, indicating uncertainty:31
(126) kuvA tsis paub <tias nws puas nyiam kuv>COMP

1SG NEG know COMP 3SG like 1SG
‘I don’t know whether she likes me.’
(127) kuvS/A/A tuaj <xyuas kojO saib <kojA puas muaj

1SG come check 2SG find.out 2SG have
[paj-ntaub <muag>REL]O>COMP>PURP
needlework sell
‘I’ve come to check to see whether you have any needlework to sell.’
(WHD 18:16)
1.7.3 Negative with tsis and txhob

When the speaker wishes to indicate certainty concerning the falsity of the
proposition expressed by the clause, the negative morpheme tsis is used:
(128) thaum ub tsis muaj hnub tsis muaj hli . . .

time yonder NEG have sun NEG have moon
‘Long ago there was no sun (and) there was no moon . . .’ (DNH 3:1)
(129) kojA tsis muaj [chaw rau peb nyob]O . . .

2SG NEG have place for 1PL dwell
‘. . . you do not have a place for us to stay . . .’
The negative morpheme txhob, on the other hand, occurs when the falsity of
the proposition is not certain. For example, it may be the subject of specula-
tion, as in example (130) below, or hope, as in (131) below. This irrealis negative
txhob is optionally preceded by the regular negative tsis, with no apparent dif-
ference in meaning.
(130) yog kuvA (tsis) txhob lem, pebS twb tsis sib tsoo
COP 1SG NEG IRR.NEG turn 1PL sure NEG RECP smash
‘If I hadn’t turned, we wouldn’t have smashed (into each other).’
31 A number of morphemes other than puas that expresses the speaker’s assessment of the
degree of likelihood of the realization of the proposition (sometimes referred to as the
category of ‘status’ (Foley & Van Valin, 1984, pp. 213–220)) in White Hmong have other,
related functions. See discussion on yuav and yuav-tsum below.
(131) (tsis) txhob mus los mas

NEG IRR.NEG go IP IP
‘Don’t go.’ (WHD 9:4)
(132) ab, SaubA twb hais <tias kom tsis tshob

oh PN sure say COMP.that COMP.to NEG IRR.NEG
tsa [Suav vwm cawv no]O>COMP, . . .
set.upright Chinese be.crazy alcohol this
‘Oh, Shao definitely told me not to revive this crazy drunk stranger, . . .’
(DNH 387:5)
1.7.4 Certainty with twb

The status morpheme twb is used when the speaker wishes to indicate cer-
tainty concerning the truth of the proposition:
(133) twb yog [koj tus brake]S tsis ua haujlwm, kojS

sure COP 2SG CLF brake NEG do work 2SG
thiaj sib tsoo
so RECP smash
‘Doubtless it was that your brake didn’t work, so you had the accident.’
(134) kuvA twb tsis paub

1SG sure NEG know
‘I really don’t know.’
(135) kuvS twb zoo-siab yog nwsS yuav tau tuaj

1SG sure be.good-liver COP 3SG IRR ACHV come
‘I’ll certainly be happy if he manages to come.’
1.7.5 Future, Likelihood, and Intention with yuav

The verb yuav is used before an noun or noun phrase in White Hmong in two
closely related ways: it can be used with the meaning ‘obtain/buy/acquire’, or
with the meaning ‘want’ (Clark, 1982; Li, n.d., pp. 27–28).
When yuav is used grammatically before a verb, Clark (1982, p. 125) suggests
that it is an auxiliary verb indicating ‘anticipatory or expectative aspect’. Most
other scholars regard it simply as a future tense marker in this position (e.g.
Bertrais-Charrier (1964), Heimbach (1979), Mottin (1978), Jaisser (1984), Fuller
(1985), Li (n.d.)). There is no doubt that it is frequently used in propositions
that have future time reference.
60 Chapter 1
(136) kuvS yuav zoo-siab yog nwsS tau tuaj

1SG be.good-liver COP 3SG ACHV come
‘I will be happy if she manages to come.’
(137) pebA yuav paub <sau [ntawv Hmoob]O>COMP mus

1PL know write writing Hmong go
xyuas lawvO
check 3PL
‘We will know how to write Hmong to greet [check on / keep in
touch with (N. J.)] them.’ (Clark, 1982, p. 137)
However, the future time reference of sentences with yuav can be relative
rather than absolute:
(138) ces Txiv Nraug NtsuagS yuav rov qab . . .

CONJ the.Orphan.Boy return back
‘The Orphan Boy was about to go back . . .’ (DNH 166:37)
Furthermore, yuav can appear in conditional sentences with no sense of future

time reference at all:
(139) ib yam li qhov pebA hais <tias nwsA de tau

one kind as thing 1PL say COMP 3SG pick get
[ib lub taub]O>COMP yuav tsis meej npaum li
one CLF pumpkin NEG clear equal as
qhov <nwsA de tau [ib lub taub dag]O>COMP
thing 3SG pick get one CLF pumpkin yellow
‘For example, if we say ‘he picked a pumpkin’ it would not be as clear as
‘he picked a yellow pumpkin.’ (WHG 13)
In examples such as this the speaker seems to be asserting his/her belief in the
likelihood of the truth of the proposition, and yuav can be characterised as
functioning to indicate epistemic modality.
It is important to note that, in the overwhelming majority of sentences with
yuav, the subject is human (or, at least, anthropomorphised). In this case, in
addition to the notions of future time reference or likelihood, intention or
desire is also often indicated. (This is the component of its meaning which
is most clearly derived from the use of yuav as a verb meaning ‘obtain / want
(something)’.)
(140) . . . nebS yuav mus ces mus es kuvA yuav kaw qhov-roojO
2DU go CONJ go IP 1SG shut door
‘. . . (if) you intend to go then go and I’ll shut the door.’ (DNH 274:31)
(141) yusA muaj [ib tug tub <yuav muab nyabO>REL]O

INDF have one CLF(t.c.) boy take son’s.wife
‘(Suppose) one has a son, (who) intends to take a wife.’ (THW 102)
(142) . . . nwsS me~me nwsA tsis tau luagO,

3SG RDUP~be.small 3SG NEG get companion
es nwsA yuav nyob [saum qab nthab]O xwb
CONJ 3SG stay top under drying.platform only
‘. . . he [the corn] was small and had no friends, so he just wanted to
hang under the drying platform . . .’ (DNH 5:12)
There is, in fact, a certain type of question in which yuav appears, in which it
has no future time reference at all, but in which it does seem to retain some-
thing of its sense of intention. Observe the following:
(143) cas kojS tseem yuav txaj-muag?

how 2SG still be.ashamed
‘Why should you still be shy?’ (TNN 3)
(144) ua-cas yuav qhib [lub qhov-rooj]O tag . . .

why open CLF door finish
‘Why did you (go and) open the door . . .?’ (TNN 6)
(145) ua-cas kojS yuav ruam ua-luag, ais?!

why 2SG be.stupid EMPH IP
‘Oh, why must you be so stupid?!’ (TNN 5)
In examples of this type the role of yuav seems to be to suggest that the reali-
sation of the action/state described is under the control of the subject, who is
portrayed as intentionally and stubbornly carrying it through.
This variation in the usage of a morpheme, from a deontic modality sense
(e.g. yuav ‘intend’), through an epistemic modality sense (yuav ‘likely’), to use
to indicate future tense (yuav ‘will’), is parallel to the diachronic development
of the English modal will. Diachronic development such as this is extremely
well attested cross-linguistically. However, all three senses of yuav in White
62 Chapter 1
Hmong are synchronically viable, with no sense seeming to be archaic or fall-

ing out of use.
1.7.6 Likelihood and Obligation with yuav-tsum and yuav-tau

The meaning of the expression yuav-tsum is clearly related to the meanings of
its component morphemes: yuav ‘likely/will’ and tsum, which is used elsewhere
as a verb meaning ‘obtain/be effective/hit the target’. This combination of the
semantic components of ‘likelihood’ and ‘success’ result in an indication of a
greater degree of certainty than that expressed by yuav alone, roughly equiva-
lent to that expressed by the epistemic sense of the English modal should:
(146) yog nwsA tua nwsA yuav-tsum raug liabO

COP 3SG shoot 3SG should hit.the.mark monkey
‘If he shoots he should hit a monkey.’
Much more common than this epistemic (status) sense of yuav-tsum ‘should’
is its use to express deontic modality:
(147) yog li mas kojA yuav-tsum qhia rau kuv nawb

COP like TOP 2SG should tell to 1SG IP
‘If that’s the case, you should tell me, OK?’ (DNH 35:51)
(148) kuvA yuav-tsum kawm <hais [lus Askiv]O>COMP

1SG should study speak word English
‘I should learn to speak English.’
Example (148) with yuav-tsum ‘should’ expresses an obligation that the speaker
should fulfil, but may or may not choose to; consequently it can co-occur with
a contrary statement such as the following:
(149) . . . tab-sis kuvA tseem muaj haujlwmO

but 1SG still have work
‘. . . but I still have work to do.’ i.e. ‘I’m too busy.’
As an alternative to yuav-tsum, yuav-tau is sometimes used to express likeli-

hood and obligation. This variant involves similar component morphemes:
yuav ‘likely, will’ and tau, elsewhere used as a verb meaning ‘get, come to have’.
This may be a regional variation, or there may be a slight difference in mean-
ing, with yuav-tau used when there is a slightly stronger sense of obligation:
(150) yuav-tau rov mus <qhib [lub qhov-rooj]O dua>PURP

must return go open CLF door again
‘. . . must go back to open the door again.’
1.7.7 Deontic Modality with a Complementation Strategy

Deontic modality in White Hmong is expressed using a complementation
strategy (Jarkey, 2006), with verbs such as xav ‘want to’, kam ‘be willing to’, txawj
‘know how to’, pib ‘begin to’, sim ‘try doing’, maj ‘be in a hurry to/be busy doing’,
laj ‘be tired of doing’. When these verbs are used in a modality sense the second
verb follows them directly, without the intervention of a complementizer. For
this and other reasons, these sentences bear a close resemblance to serial verb
constructions. However, in spite of a number of important similarities, they
are regarded in this analysis as a distinct phenomenon involving a comple-
mentation strategy rather than serialization. (See Jarkey (2006) for justification
of this analysis, and Chapter 5, §5.4 for further discussion).
Examples (151)–(154) illustrate verbs expressing deontic modality:
(151) kuvA xav haus dejO

1SG want drink water
‘I want to drink some water.’
(152) kojA puas kam qhia kuvO?

2SG Q be.willing teach 1SG
‘Are you willing to teach me?’ (WHD 5:1)
(153) kojS/A sim rov mus <xyuas saib>PURP

2SG try return go check find.out
‘Try going back (to) check on (them) and see.’
(154) kojA puas txawj hais [lus Hmoob]O?

2SG Q know.how speak word Hmong
‘Do you know how to speak Hmong?’
1.7.8 Deontic Modality with Postverbal tau

While deontic modality is primarily expressed using the preverbal comple-
mentation strategy described in §1.7.7 above, it is also expressed postverbally
with the morpheme tau ‘can’. The morpheme has a range of modal senses
when it occurs in this position in the clause, including ability, potential, and
permission. (See Chapter 3, §3.3.7 for examples.)
64 Chapter 1
Like other grammatical uses of tau (see §1.7.9 below), this postverbal use has
grammaticalized from the verb tau. Enfield (2003, p. 38 et passim) explains that
this verb means ‘get’ in the non-agentive sense of ‘come to have something’:
(155) kuvA tau [peb tug tses]O

1SG get/come.to.have three CLF(t.c.) fish
‘I got three fish (e.g. while fishing).’ (Enfield, 2003, p. 170)
The unacceptability of this verb in an imperative context confirms that it does

not refer to an action that is fully under the control of the subject:
(156) * kojA tau phau-ntawvO xwb nawb!

2SG get/come.to.have CLF-paper only IP
Lit: ‘Come to have a book!’ (Enfield, 2003, p. 171)
In addition to the obvious difference in position, postverbal tau differs syntac-

tically in a range of ways from the preverbal modals. Li (n.d., p. 11) notes that, in
this position, tau can be preceded by the negative, which follows all the other
modals; it can occur alone with the interrogative puas sentence finally; and it
can serve as an answer to a question. The following examples illustrate these
points (see also Li (n.d., pp. 12–13)):
(157) . . . kojS nyob kojA txhob nce nthabO nawb;

2SG stay 2SG IRR.NEG ascend drying.platform IP
nce tsis tau
ascend NEG can
‘. . . you stay (here) (and) don’t go up onto the drying platform, OK?
(You’re) not allowed to go up.’ (DNH 449:6)
(158) a. Q: kojA hais [lus Hmoob]O puas tau?

2SG speak word Hmong Q can
‘Can you speak Hmong?’
b. A: tau xwb
can only
‘Sure can!’
The key reason for these syntactic differences between postverbal tau and
other modality morphemes in Hmong is that this morpheme originates
from the use of the verb tau as the second verb in an Attainment Serial Verb
Construction. Its syntactic properties when it occurs postverbally to express

the meaning ‘can’ are related to its original meaning and usage in this SVC type.
(See Chapter 3, §3.3.7 for discussion and justification of this analysis).
1.7.9 Achievement with Preverbal tau

Like the postverbal modal tau, discussed in §1.7.8 above, the preverbal modal-
ity/aspect marker tau has clearly grammaticalized from the main verb tau ‘get,
come to have’. When used preverbally, tau occurs directly before the verb it
modifies, after the negative, that is, in the same position as modal verbs that
function in the complementation strategy described in §1.7.7 above. This sets
it clearly apart from the aspectual adverbs introduced in §1.7.10 below, which
all appear before rather than after the negative.
As pointed out by Li (n.d., p. 2) preverbal tau is almost invariably referred to
as a past tense marker in early studies of Hmong (e.g. Bertrais-Charrier (1964),
Heimbach (1979), Mottin (1978), D. Yang (1980), Jaisser (1984), Fuller (1985).
However, observe the following example in which it clearly cannot be inter-
preted as referring to past tense:
(159) mus [hnub Vas.Xuv] los [hnub vas.Xaum], kuvA tseem

go day Friday or day Saturday 1SG still
tsis tau paub
NEG ACHV know
‘I don’t know yet whether I’ll go on Friday or Saturday.’ (WHD 15:8)
Clark (1982) describes preverbal tau as a perfective aspect marker, and Li

(n.d.), as a marker of ‘Attainment aspect’, which, he says, belongs with lawm
‘Completion aspect’ to the perfective category (p. 30). In my own earlier work
(Jarkey, 1991) I also classified it as a marker of perfective aspect. However, the
position of tau with the modal verbs, after the negative rather than before,
where aspectual adverbs typically occur, suggests that it may have elements of
modal meaning.
The use of tau preverbally is often well translated as ‘get to, manage to’:
(160) tej-zaum kuvA yuav tau mus <ntsib nkawdO>PURP tag-kis

maybe 1SG IRR ACHV go meet 3DU tomorrow
‘Maybe I’ll manage to go (to) meet those two tomorrow.’ (Clark, 1982,
p. 132)
Givón (1973, p. 898) points out that modal verbs like succeed and manage pre-
suppose the falsity of the complement proposition before the time referred
66 Chapter 1
to, and imply the truth of this proposition after the time referred to. This pre-
supposition and implication are clearly relevant to this use of tau in White
Hmong. The description Enfield (2007, p. 231) gives for the Lao achievement
marker daj∅ is equally appropriate for this use of tau in White Hmong:
This marker conveys the idea that some prior state of affairs has made it
possible for the predicate state of affairs to be the case. Common trans-
lations of a predicate V marked with daj∅ include ‘did V’, ‘managed to
V’, and ‘got to V’. . . . The state of affairs denoted by the marked predicate
arises (or becomes possible) as an outcome of some previous state of
affairs being the case.
Regarding the modals succeed and manage to, Givón (1973, p. 898) also notes
that these verbs often have two interpretations: one in which there is a presup-
position of active attempt and the other in which there is no such presupposi-
tion. Sentences in White Hmong in which tau means ‘get/manage to’ generally
predicate animate subjects, but often don’t presuppose active attempt. Observe
the following example:
(161) kuvA mam li xa [wb daim duab]O/S tuaj rau

1SG therefore send 1DU CLF image come for
kojA tau saib
2SG ACHV look.at
‘So I’ll send our photograph so that you’ll get to see (it).’ (Clark, 1982,
p. 132)
This sense of ‘getting to do something’ with no active attempt on the part of

the subject is in keeping with the meaning of the lexical verb tau ‘get, come to
have’, which is non-agentive (Enfield, 2003, p. 171).
In the less common cases in which preverbal tau appears without an ani-
mate subject, it has a more aspectual sense; in the example below it clearly
suggests nothing about attempt, or lack thereof, on the part of the subject, but
only to the successful realisation of the predicate.
(162) txij [hnub <kuvA yuav poj-niamO>REL] los txog [tav

extend day 1SG obtain wife come arrive time
no], tau muaj [ib xyoo]S
this ACHV have one year
‘From the day I got married to the present time, it’s been one year.’
(Clark, 1979a, p. 6)
In this usage, the same basic meaning that Enfield (2007, p. 231) described for
the Lao achievement marker daj∅ still applies: “The state of affairs denoted
by the marked predicate arises (or becomes possible) as an outcome of some
previous state of affairs being the case.”
1.7.10 Imperfective Aspect

Imperfective aspect may be overtly signalled in White Hmong in a number of
different ways. Firstly, verbal reduplication (see §1.3.3.3 above) is commonly
used for this purpose; in the examples below reduplication serves to indicate
habitual and durative aspect:
(163) [tus lw-zaj] mas nwsS da~da-dej tau [ob

CLF lake TOP 3SG RDUP~immerse-water get two
peb xyoo no], . . .
three year this
‘That lake, he’s been taking his bath (in it) for several years now, . . .
(DNH 36:58)
(164) Txiv Nraug NtsuagS pw~pw ib tsam ov, [ob niam-txiv

the.Orphan.Boy RDUP~rest little.while IP couple
Suav]A txawm tso yoov qaibO/S, . . . los <tom
Chinese then send gnat come bite
Txiv Nraug Ntsuag>PURP
the.Orphan.Boy
‘(After) the Orphan Boy had been sleeping for a while, then the Chinese
couple sent gnats (to) bite him.’ (DNH 454:32)
Imperfective aspect may also be expressed by one of a series of adverbs that

occur preverbally, after the subject but before the negative.32 These include:
tab tom (imperfective); tseem ‘still’; pheej ‘continually’; sij ‘repeatedly’; yeej
‘originally’ (habitual).
(165) nwsS tab tom mob thaum [nws kwv-tij]A

3SG IPFV be(come).sick time 3SG clan.brothers
tuaj <xyuas nwsO>
come check 3SG
‘He was sick when his cousins came to check (on) him.’
32 Not all of these are actually attested in combination with the negative (e.g. tab tom (Li,
n.d., p. 26)), but this is no reason to assume that they do not belong to this set.
68 Chapter 1
(166) Yob Nraug NtsuagS/S/A rov los txog tsev,

PN the.Orphan return come arrive home
NtxawmS tseem quaj~quaj 33
PN still RDUP~cry
‘When Yor the Orphan came home, Nzue was still crying and crying . . .’
(DNH 277: 51)
(167) nwsA pheej noj tshuajO pheej tsis zoo

3SG keep.on eat medicine keep.on NEG be.good
‘She keeps on taking medicine (but) still isn’t well.’ (Heimbach 1979:242)
(168) . . . nkawdA mus <ua tebO>PURP los nkawdA/S yeej

3DU go make field CONJ 3DU HAB
tsis coj mus li
NEG take go at.all
‘. . . (when) they went to work the fields they would never take (her)
along.’ (DNH 449:3)
1.7.11 The Perfect with lawm

While the achievement marker tau indicates that a new state of affairs arises
as an outcome of a previous one, it makes no reference to whether it continues
to have any further relevance at the time referred to. This is the function of
the perfect marker lawm. This morpheme may be historically related to the
Mandarin particle le, and is functionally quite similar to it, as well as to Thai
lɛ̂ɛw/láew and Lao lèèw.34
The reference time to which lawm relates may be prior to, at or after the time
of the speech event. In example (169) below, lawm functions to express the
relevance of a situation that has come about to the time of the speech event:
(169) puag-ta kojA mus dab tsiO lawm?

just.now 2SG go which PRF
‘Where have you just been?’
33 In this example the reduplication of the verb quaj ‘cry’ seems to signal intensity, as well as
imperfective aspect. The combination of reduplication with an adverb like pheej ‘continu-
ally’ or tseem ‘still’ seems quite common in the expression of imperfective aspect.
34 For a thorough explication of the functions of le in Mandarin see Li, Thompson, and
Thompson (1982). Although these authors follow the traditional classification of the
Perfect as a type of aspect, it may be better regarded as a category of tense. This fact helps
to explain the peripheral position of the Perfect morpheme in both White Hmong and
Mandarin (see also Li et al. (1982, p. 41), regarding this peripheral position).
In (170) and (171) the reference is to some time in the future:
(170) luagA/A/S/A muab coj mus txog tsevO lawm, ces

other take take.along go arrive home PRF CONJ
thiaj.li tso neegO/S tuaj [tshab xo]O rau niam-txiv
so send person come CLF message for parents
‘[When] they have taken [her] to their home, then they [will] send
someone [to] bring the news to [her] parents.’ (THW 106)
(171) . . . [[koj txiv] wb]]A mus <ua teb>PURP lawm; kojS

2SG father 1DU go make fields PRF 2SG
nyob kojA txhob nce nthabO nawb
stay 2SG IRR .NEG ascend storage.platform IP
‘. . . [When] your father’n me have gone [to] work the fields, you
stay [here] [and] don’t you climb up on the storage platform, OK?’
(DNH 449:5)
Finally, the morpheme lawm may also function to express the relevance of a
situation to some time in the past:
(172) thaum Muam Nkauj LiagS twb loj~loj lawm

time PN sure RDUP~be(come).big PRF
los, nkawdA hlub~hlub ces nkawdA mus <ua
CONJ 3DU RDUP~cherish CONJ 3DU go make
tebO>PURP los nkawdA/S yeej tsis coj mus li
field CONJ 3DU HAB NEG take.along go at.all
‘When Mua Ngao Lia had really grown up, they loved (her) dearly and
(when) they went to work the fields they would never take (her) along.’
(DNH 449:3)
1.8 Conclusion to Chapter One
This short sketch of White Hmong grammar goes no further than a very basic
description of simple sentences. Where they are relevant to the topic of serial
verbs, higher-level units, such as sentences involving complement clauses or
juxtaposed clauses, are discussed in detail in other sections of this study. For
further discussion of complementation in White Hmong, see also Jaisser (1984)
and Jarkey (2006).
CHAPTER 2
Introduction to Serial Verb Constructions
. . . a peculiarity of Ewe is that we often find a row of verbs one after the
other. The chief features of this are that all the verbs stand next to each
other without being connected, that all have the same tense or mood,
and that in the event of their having a common subject or object, these
stand with the first, the others remaining bare . . . In English these con-
secutive verbs are partly rendered by composite sentences. But very often
several Ewe verbs may be expressed by a single verb in English. The expla-
nation for this is that the Ewe people describe every detail of action or
happening from beginning to end, and each detail has to be expressed by
a special verb: they dissect every happening and present it in its several
parts, whereas in English we seize on the leading event and express it by a
verb, while subordinate events are either not considered or are rendered
by means of a preposition, adverb, conjunction, or a prefix on the verb.
Westermann (1930, p. 126)
This description of ‘consecutive verbs’ in Ewe (Kwa, Niger-Congo) is the

English version, published in 1930, of a passage that had appeared even earlier
in German (Westermann, 1907). Along with the work of Christaller (1875) on
another Kwa language of the Niger-Congo family (Twi (Akan)), this was prob-
ably one of the earliest descriptions of the phenomenon that has come to be
most commonly referred to as a ‘serial verb construction’ (SVC).
Since Christaller and Westermann first identified the phenomenon around
the turn of the twentieth century, similar ‘rows of verbs’ that are simply juxta-
posed without overt connection have been recognised to occur in a very wide
range of languages and language families, not only in Africa, but also in East
and Southeast Asia, North and South America, Oceania and New Guinea, and
in Creole languages wherever they are found. Westermann’s description could
well be applied to the phenomenon in all of these languages:
– they involve a series of verbs (or verb-phrase-like components);

– these verbs stand next to each other without being connected (by overt
form of linkage such as conjunctions or complementizers);
– they all have the same tense or mood; and
– they have a common subject or object.

Introduction to SVCs 71
Examples (1)–(7) below show sentences that clearly fit this description from a
range of languages and language families:
(1) Mandarin (Li & Thompson, 1974a, p. 266)

tā zài chúfańg-lǐ zùo jiǎozi
he be.at kitchen-in make dumping
‘He is making dumplings in the kitchen.’
(2) Vietnamese (L.C. Thompson (1965, p. 232), cited in Clark (1978, p. 100))
chúng ta mau lên . . .
1PL INCL be.fast go.up
‘We’d better hurry up . . .’
(3) Tariana (North Arawak) (Aikhenvald, 2006a, p. 188)

kawhi nu-iɾa-ka nu-sita
manioc.flour 1SG-drink-REC.PST.VIS 1SG-finish
‘I have drunk manioc flour (and there is none left).’
(4) Thai (Tai-Kadai) (Thepkanjana, 1986, p. 156)

Sùrii ʔaàn nǎnsʉʉ̌ pay con còp
PN read book go until end
‘Suri kept reading until she finished the book.’
(5) Akuapem Twi dialect of Akan (Kwa, Niger-Congo) (Christaller (1933,

p. 145), cited in Schachter (1974, p. 258))
mede aburow migu msum
I.take corn I.flow water.in
‘I pour corn into the water.’
(6) Paamese (Oceanic, Austronesian) (Crowley, 1987, p. 48)

kaik komuasinau nauva netan
(kaiko ko-muasi-nau nau-vaa netano)
2SG 2SG-REAL-hit-1SG 1SG-REAL-go down
‘You hit me down.’
(7) Eastern Kayah Li (Karen, Tibeto-Burman) (Solnit, 2006, p. 150)

ʔa ʔō mɯ thʌ̄ʔíphrè
3 drink drunk whiskey
‘S/he got drunk on whiskey.’
72 CHAPTER 2
It is widely recognised that Westermann’s outline of the key features of ‘con-

secutive verbs’ in Ewe serves equally well to describe similar phenomena from
many languages, such as those given as examples (1)–(7) above. However,
while it may well constitute an adequate description of such examples, such
a list does not serve to delimit any particular construction in any particu-
lar language, nor to identify any cross-linguistically valid construction type.
Westermann’s description is actually applicable to a wide variety of signifi-
cantly different constructions in many languages that, for a range of reasons,
some scholars have chosen not to incorporate under the same heading. These
constructions differ, not in the basic arrangement of their components, but
rather in the syntactic and semantic relationships between these components,
and in the functions they perform.
2.1 The Term ‘Serial Verb Construction’
Many authors, commenting on the occurrence of the simple juxtaposition

of verbs in a considerable variety of languages and language families, have
observed that not all instances of such juxtaposition can be regarded as
realisations of the same phenomenon. A primary distinction is often drawn
between juxtaposed verbs contributing to the description of actions or situa-
tions that are clearly distinct from one another, and those that can be thought
of as together expressing a single action or situation.
Even before the turn of the twentieth century, Christaller (1875) had observed
a clear difference between ‘combinations’ of verbs in Twi (Akan, Kwa, Niger-
Congo), dividing them into ‘accidental’ and ‘essential’ combination types. In
the ‘accidental’ type:
. . . two or more predicates (verbs with, or without, complements or

adjuncts), expressing different successive actions, or a state simultane-
ous with another state or action, but having the same subject, are merely
joined together without conjunction and without repeating the subject.
In this case two (or more) sentences are thrown or contracted into one,
and the verbs are coordinate in sense as well as in form.
In the ‘essential’ type of verb combination, on the other hand, “. . . the actions
expressed by both verbs are simultaneous and in an internal or inseparable
relation or connection.” One verb is the ‘principal’ verb, and the other is either
‘auxiliary’ or ‘supplemental’. It is “coordinate only in form, but subordinate in
sense . . . It is the inherent semantic features of individual verbs which actually

determine the structural descriptions of concatenations.” (Christaller, 1875, p. 69)
A similar distinction is drawn by Matisoff (1969) in early work on the phe-
nomenon of juxtaposed verbs in Southeast Asian languages. He uses the term
‘concatenation’ rather than Christaller’s ‘combination’ or Westermann’s ‘con-
secutive verbs’ to describe constructions in Lahu (Tibeto-Burman) in which
“two or more verbs may be strung together or concatenated by simple juxta-
position to form complex verb phrases” (p. 69). Matisoff establishes a distinc-
tion between what he refers to as ‘true concatenation’, and cases of juxtaposed
verbs representing a series of temporally consecutive actions. He treats such
consecutive actions as belonging to two separate VPs (and hence two sepa-
rate underlying clauses in his analysis) that have come into ‘fortuitous concat-
enation’ (by the transformational deletion of the ‘suspensive’ particle). In the
case of ‘true concatenation’, on the other hand, Matisoff argues that, although
apparently involving no more than simple juxtaposition, the grammar of these
constructions is actually far from simple:
Several interlocking systems of semantic and syntactic constraints to-

gether determine the order, the membership, the constituent structure,
the meaning and the length of . . . concatenations. (p. 71)
One of the verbs in each true concatenation is the ‘logical head’ and the
others—the ‘versatile’ verbs—stand in some sort of subordinate relationship
to it. Regarding function, Lahu verb concatenation does the job assigned in
English to a “disparate array of subordinating devices: complementary infini-
tives, -ing complements, modal auxiliaries, adverbs, prepositional phrases,
even whole subordinate clauses.” Thus, like Christaller, Matisoff establishes a
terminological distinction between juxtaposed verbs that are basically coor-
dinate in meaning, and those that he sees as involving some kind of semantic
subordination of the second verb to the first. Furthermore, within this latter
group of ‘true concatenations’, he identifies a considerable range of different
sub-types.
By the 1960s, terms such as ‘serial verbal constructions’ (Stewart, 1963) and
‘serial verbs’ (Ansre, 1966) had come into use among linguists investigating
the languages of the Niger-Congo family, to describe cases of juxtaposed verbs
similar to many of those that Christaller (1875) had earlier identified as verb
‘combinations’ in Twi. By the early 1970s, the term ‘serial verb construction’
had migrated ‘out of Africa’, and was being used in descriptions of completely
unrelated languages, for example by Li and Thompson (1973) for Mandarin.
74 CHAPTER 2
Amongst the numerous scholars who have since employed the term ‘serial
verb construction’, the need to distinguish between the various types of ‘simple’
juxtaposition of verbs has continued to be recognised. There have been widely
differing opinions, however, regarding which cases of juxtaposed verbs the
term ‘SVC’ should be applied to and how the syntactic relationship between
the component verbs should be characterised. It is clear that the majority of
authors, either implicitly or explicitly, see the need to draw some kind of dis-
tinction between juxtaposed verbs representing a series of distinct actions or
events, juxtaposed verbs involving various types of syntactic subordination,
and, finally, juxtaposed verbs that work together to describe a single event,
but that do not involve subordination. What is often at issue is the question of
whether there are adequate reasons to include such a disparate array under the
one terminological umbrella, simply because they all exhibit verb concatena-
tion and, if separated terminologically, which of these types are to be referred
to as ‘SVCs’.
2.1.1 The Term ‘SVC’ Used for Distinct Actions

In their initial transformational analysis of serial verbs in Mandarin, Li and
Thompson (1973) delimit the class of serial verb sentences to those of the form
represented in (8), which contain two predicates, describe two distinct actions,
and are derived in their analysis from two underlying sentences.
(8) NP V(NP) V(NP)

Subject Predicate 1 Predicate 2
Li and Thompson distinguish these serial verb sentences from what they call
‘co-verb sentences’, which describe a single action, contain one predicate and
one ‘co-verb’, and for which, they argue, no such derivation is necessary. In spite
of the fact that ‘co-verbs’ are generally homophonous with verbs, these authors
argue that ‘co-verbs’ should be regarded not as verbs but as prepositions,
their function being to introduce phrases that could be labelled Benefactive,
Locative, Instrumental and the like.
Serial verb sentences are regarded by Li and Thompson as being semanti-
cally ambiguous, each having four possible interpretations of the relationship
between the two actions or events described:
i) They occur consecutively;

ii) They occur simultaneously;
iii) They are temporally alternating;
iv) The second describes the purpose of the first.
Although these authors do propose a basic difference in the syntactic structure

underlying sentences with a purpose interpretation (subordination) from that
of the other three types (coordination), they indicate that choosing the correct
interpretation depends on pragmatic factors such as the context of the utter-
ance and the real world knowledge of the addressee.1
In a later description of serialization in Mandarin, Li and Thompson (1981)
include under the term serial verb constructions many further types of multi-
verb sentences involving some form of subordination; these include sentences
with subject and object complements, so-called ‘control constructions’, and
descriptive clauses that modify object NPs. In spite of this significant extension
of the use of the term SVC in their later analysis, these authors still explicitly
exclude cases of juxtaposed verbs describing a single action of the ‘co-verb’
type.
2.1.2 The Term ‘SVC’ Used for a Single Action/Event

In the same year that Li and Thompson published their first description of
‘serial verb constructions’ in Mandarin, Lord’s influential article on the Kwa
languages of West Africa, ‘Serial verbs in transition’ (Lord, 1973), also appeared.
While Li and Thompson used the term ‘SVC’ to refer to distinct actions only,
Lord was one of the first to claim explicitly that “in the serial construction the
verb phrases necessarily refer to sub-parts or aspects of a single overall event”
(p. 269). She suggests that it is precisely because serial verb constructions rep-
resent “a single overall event” that some verbs in them, such as those express-
ing “semantic case relationships” like location, become prepositions.
Lord made her decision on this use the term ‘SVC’ in the midst of a vig-
orous debate in the field of West African linguistics regarding the appropri-
ate syntactic analysis of sequences of juxtaposed verbs (Awobuluyi (1973);
Bamgboṣe (1973); Schachter (1974); Stahlke (1970)). This debate focused on
questions of whether all or only some of these sequences should be accounted
for by a mono-clausal analysis, rather than the underlying multi-sentence
analysis that had been previously assumed. While Lord was clearly aware of
Li and Thompson’s use of the term ‘SVC’ for multi-event sequences in Chinese
(derived in their analysis from multiple underlying sentences), in the context
of the study of West African languages, she chose to use the term in precisely
1 See Paul (2008, pp. 4–5) for an explanation of the need for different parsing for each of these
interpretations, and (p. 10) for discussion regarding the need for a slight pause or explicit
marking by time adverbs to achieve the ‘alternating’ or ‘simultaneous’ interpretations. Paul
also argues strongly against Li and Thompson’s use of the term ‘construction’ to refer to these
highly varied syntactic structures.
76 CHAPTER 2
the opposite way, to refer to mono-clausal structures that function to represent

single events.
Scholars working in the field of Chinese linguistics, particularly those
studying Mandarin, have generally continued to use the term ‘SVC’ as Li and
Thompson used it, as a cover term for an extremely wide range of constructions,
including those that refer to distinct actions (Paul, 2008). Li and Thompson’s
use of the term also seems to have influenced some linguists working on
the languages of Mainland Southeast Asia, for example Clark (1978); Riddle
(1989); Thepkanjana (1986). This is certainly not to say that these linguists have
followed Li and Thompson’s use of the term in every respect, but that each
derives some elements from it in terms of what they include or exclude from
the category of ‘SVC’ in the languages they describe.
More generally, however, it has been the notion of a serial verb construction
portraying a single event, proposed by Lord (1973), that has been adopted in
influential typological studies (e.g. Aikhenvald & Dixon, 2006; Bisang, 2009;
Durie, 1997; Foley & Olson, 1985; Givón, 1991) and has come to be most widely
accepted. The most common use of the term ‘serial verb construction’, at least
since the mid-1980s, and with some notable exceptions, has thus been to refer
primarily to constructions containing juxtaposed verbs that together portray
what is construed a single action or event.
This is not to say that the notion of a ‘single action or event’ is by any means
a straightforward one. On the contrary, this notion is one that requires care-
ful consideration and has been widely discussed. This will be the topic of §2.2
below.
2.2 The Notion of a Single Event
As explained in §2.1.2 above, serial verb constructions are most commonly said
to involve a single clause, and to express a single event. However, by defini-
tion they consist of more than one predicate. How is it that a multi-predicate
sequence can be reconciled with a monoclausal structure, apparently repre-
senting a single event?
Pawley (1987, 1993) suggests that different languages use different logic
for reporting events. In English, for example, several conceptual situations/
events can be fused into a single clause by reducing certain situations/events
to the status of peripheral or backgrounded elements in the clause, expressed
as arguments of the verb (e.g. Instrument, Beneficiary, Direction, Location).
Furthermore, a single verb (such as an English verb like hunt or fetch) can itself
be used to stand for what is, in the real world, a fairly complex series of actions.
In Kalam (Papuan), on the other hand, Pawley says that each conceptual event
and each part of a complex series of actions is expressed by a separate predi-
cate, often in a serial verb construction.
Thus, Pawley claims, what is dealt with as a single unitary event in a lan-
guage without serial constructions, such as English, may actually be treated as
a series of sub-events in a language with serial constructions, like Kalam: each
sub-event represented by a separate predicate. According to this approach, it
would thus be the viewpoint of the onlooker (the linguist), rather than the
structure and organisation of the relevant language itself, that characterises
serial verb constructions as expressing a single event.
However, Givón (1987) claims that
Pawley’s argument harkens back to Aristotle’s notion of the verb (‘predi-

cate’) as a code element that forms the core of a single proposition. The
proposition, in turn, codes a ‘state’ or ‘event’ in the real world. Multi-verb
sequences thus code multi-event or multi-state sequences. (1987, p. 1)
He goes on to argue:
The temporal-physical distance between chunks of linguistically coded

information correlates directly to the conceptual distance between
them . . . It can thus be reasoned that if serial verb constructions indeed
reflect something about the cognitive segmentation of ‘events’, then
pauses of the type that characteristically appear . . . at the boundaries of
main/finite verbal clauses will also appear in verb-serializing languages
at serial-verb clause boundaries. (1987, p. 2)
To test this hypothesis Givón chose three serializing languages of New

Guinea and compared the length of pauses at the boundaries of ‘serial verb
clauses’ with those at the boundaries of main/finite clauses, and with those
between lexical words within the clause. The languages chosen, Tok Pisin
(Neo-Melanesian Creole), Tairora (Papuan) and Kalam (Papuan), differ with
regard to both the density (frequency) and type of serialization involved, as
well as differing typologically in other ways. However, in all three cases, Givón
concludes that:
Serial verb constructions display pause probabilities that fall within

the range characteristic of lexical words within the clause. They diverge
78 CHAPTER 2
rather dramatically from pause probabilities characteristic of inter-clausal

transition.2 Serial verb constructions, as a typological phenomenon, do
not seem to represent a different way of segmenting reality . . . Rather
they represent a different way of coding event segments that are chunked
in a roughly equivalent way cross-linguistically, and thus presumably also
cross-culturally . . . serial-verb constructions should be viewed within the
context of the typology of lexico-syntactic coding rather than the typol-
ogy of cultural-cognitive event segmentation. (1987, p. 33) (see also 1991,
p. 119)
While directly contradicting the Aristotelian notion of correspondence

between the event and the predicate, Givón’s findings regarding serial verbs
allow the maintenance of the traditional notion of correspondence between
the event and the clause.
Before further discussion of the way in which serial verbs are packaged
within the clause (§2.2.2 below), it will be useful to consider in more detail
what it might mean to construe a real-world situation in terms of the notion
of ‘event’.
2.2.1 Construal as a Single Event

In order to understand the idea of construal as a single event, we need to con-
sider further the notion of a ‘conceptual event’. In this study, a conceptual
event is regarded as both a cognitive and a cultural construct (Jarkey, 2010,
pp. 111–115).3
A conceptual event, as discussed in Jarkey (2010, p. 112), is a cognitive con-
struct in the sense that we treat our temporal experiences as a metaphor of
our physical experiences. That is, just as we tend to classify the physical world
into discrete entities, so we tend to classify the temporal world into discrete
events. These ‘discrete’ conceptual events, however, are by no means com-
pletely fixed elements in our mental representation of experience, even if we
see them as typical event types (see below). Conceptual events have mereo-
logical structures. That is, as pointed out by Bohnemeyer et al. (2007), parts of
events and combinations of events can themselves be considered instances
of events (see also Zacks and Tversky (2001)). What is, in one context, con-
ceived and portrayed as a single event (We hunted all morning) might easily, in
2 For similar observations see Bruce (1988, p. 25).

3 Please note that the term ‘conceptual event’ is not used here with the same meaning as it is
by Pawley (1987). Pawley relates a ‘conceptual event’ to what is expressed by a single verb in
a particular language.
another context (or even in the very next instant), be conceived and portrayed
as a series of multiple events (We saddled up at 8:00am, spotted our first fox
around 10:00, but didn’t manage to run one down until nearly midday).
A conceptual event is here regarded not only as a cognitive construct, but
also as a cultural one. While some event types may well be common across cul-
tures, other event types may be typical and salient in one culture, but atypical
and non-salient in another (Bruce, 1988; Durie, 1997, p. 322; Enfield, In press,
pp. 102–128). The relevance of culture to event typicality and salience is most
obvious when we think about the fact that different cultures often lexicalize
different events. As a certain event comes to have a particular salience in a
particular culture (such as, for example, the action of consulting an internet
search engine), a verb (such as the verb to google) may be introduced into the
language to refer to it.
We can think of serial verb constructions as involving two or more
‘lexicalized’ events, juxtaposed to express a single ‘conceptual’ event (remem-
bering, of course, that those ‘lexicalized’ events are themselves also instances of
‘conceptual events’). Of course there are likely to be certain, commonly occur-
ring event combinations, such as those involving different components of
a motion event, that will occur in many languages that utilise the strategy
of serialization. It is also to be expected, however, that different combina-
tions of lexical verbs may well occur in serial verb constructions in different
languages, to represent different salient conceptual events in the respective
cultures. It is only when such common or culturally significant association
occurs that what might otherwise be fragmented as two events or situations
are actually perceived, and treated, as one. Durie (1997) agrees, citing evidence
from White Hmong from Jarkey (1991) to support this claim (see Chapter 3
§3.1.2.3).
Bruce (1988) sees serialization as constituting a bridge between the syn-
tax and the lexicon. He points out that, just like serial verb constructions, it
is possible to analyse many lexical items into combinations of more simple
concepts (e.g. fetch = go, get, bring back). Unlike a serial verb construction,
however, a lexical item is usually thought of as an indivisible whole rather than
a series of component parts. Bruce maintains, following Langacker (1983), that
lexical items are semantic units by virtue of the fact that their processing is
fully automatic: there is no need to focus attention on component parts or
on how they are put together. While verbs in series must express convention-
ally or culturally associated ideas, the constructions are still analysable and
generally productive, and the component parts transparent. It is in this way
that Bruce sees serialization as constituting an interface between the lexicon
and syntax.
80 CHAPTER 2
2.2.2 Packaging as Single Event

This brings us to one of the most fundamental questions about serial verb con-
structions: in the complete absence of any form of conjunction or complemen-
tizer to link the two verbs, how is the relationship between them conveyed?
When predicates are joined by a conjunction or a complementizer, it is these
words that provide much of the information about the semantic relationship
between them. In a serial construction, on the other hand, no such information
is provided; there is no connecting morpheme that serves to convey the sense
in which the predicates are related—that is, the fact that they are intended to
be interpreted ‘as one event’. Where, then, does this information come from?
One answer to this question is that the grammatical construction itself can
contribute to the meaning. Grammar, or at least syntax, is often regarded as
simply a vehicle to carry the ‘meaning bearing’ elements of language: the lex-
emes. However, one way in which syntax itself can convey meaning is through
iconicity (Givón, 1985; Haiman, 1980, 1985). An iconic relationship may exist
between a grammatical construction used to describe a particular phenom-
enon, and features of the phenomenon itself. Haiman (1980, p. 516) explains:
The clearest example of such iconicity is that of sequence. Other things

being equal, the order of statements in a narrative description corre-
sponds to the order of the events they describe.
Serial verb constructions can be seen as iconic in several ways. Firstly, the very
absence of a conjunction or complementizer between verbs in series means
that they are not separated at all; they simply occur ‘back-to-back’, or with only
an object NP intervening. A second feature of serial verb constructions that
can be thought of as an iconic sign of their unity is the fact that, as discussed
above, there are no pauses between the elements, which occur under a single
intonation contour. Thirdly, the fact that the two predicates convey a single
event is reflected in their occurrence within a single clause. In these three ways
the semantic ‘oneness’ of verbs in series is conveyed by both their syntactic
and prosodic unity.
The occurrence of serial verbs within a single clause means that they are
parts of the one whole by virtue of what they share. One thing that is shared
by all verbs in a serial construction is temporal operations. Bohnemeyer et al.
(2007) suggest that a speaker’s construal of the temporal properties of extra-
linguistic stimuli are, in fact, key to their ‘packaging’ into discrete events. These
temporal properties include their beginning, end and duration, as well as their
occurrence relative to other ‘events’ (p. 502). Bohnemeyer and his co-authors
propose the ‘macro-event property’ (MEP) as a universal semantic test for

identifying a single event. The MEP is:
. . . a property of constructions that assesses the event construal they

convey—specifically, the “tightness of packaging” of subevents in the
construction. A construction has the MEP if temporal operations such as
time adverbials, temporal clauses, and tenses necessarily have scope over
all subevents encoded by the construction. (p. 497)
Serial verb constructions certainly have the MEP. In fact, they exhibit even
tighter ‘packaging’. In addition to shared temporal operations, they share at
least one core argument, often two when transitive verbs are involved, and
they invariably share all peripheral arguments (not only those expressing time
but also those expressing location). They also share all operators (or grammati-
cal categories) that relate to the core and periphery of the clause, such as deon-
tic and epistemic modality, evidentiality, and illocutionary force, in addition,
of course, to tense. It is as if they are linked together by their shared arguments,
and bound together by their common operators.
The iconic properties and shared elements described above are precisely
the kinds of features generally listed as characteristic of serial verb construc-
tions. Bisang (2009, p. 805) claims that the following factors identified in the
comprehensive typological overview by Aikhenvald (2006b) can all ultimately
be derived from the fact that serial verb constructions represent a single event:
– shared grammatical categories

– shared arguments
– monoclausality
– intonational properties [of a monoverbal clause]
– contiguity
– wordhood
– marking of grammatical categories [either ‘concordant’ or ‘single’ marking
(Aikhenvald, 2006b, pp. 39–40)].
As pointed out by Aikhenvald (2006b, pp. 37–39 and personal communica-

tion), the features of contiguity and wordhood are optional, and so are only
marginally connected with the issue of eventhood conveyed by the strategy
of serialization more generally. The features related to the shared grammati-
cal categories and shared arguments, as well as the syntactic and intona-
tional features of monoclausality, are clearly not only derived from but also
82 CHAPTER 2
fundamentally connected to the fact that serial verb constructions represent

a single event.
2.3 The Notion of Joint Predication
In addition to the importance of single eventhood in serial verb constructions

(§2.2 above), another issue that is widely recognised as significant is that of
joint predication (Aikhenvald, 2006b, pp. 1, 4; Bisang, 2009, p. 794; Durie, 1997,
p. 290). In fact, Durie (1997, pp. 289–290) goes so far as to say:
The archetypal serial verb construction consists of a sequence of two or

more verbs which in various (rather strong) senses, together act like a
single verb. [Emphasis added—NJ]
If we expect that the two or more verbs in a serial verb construction will act
as joint predicators, ‘like a single verb’, what kind of constructions might be
included, or excluded, as possible SVC types as a result?
The issue of joint predication relates closely in some cases to the notion of
the proposition—an important but under-explored area in the overall picture
of serialization. If a construction has one verb that is primarily responsible for
expressing the proposition, and one that functions as part of a complement
and is thus sub-propositional, can the verbs involved still be seen as acting as
joint predicators?
In many Mainland Southeast Asian languages there are a range of comple-
mentation types and purpose constructions that occur without a complemen-
tizer and that are structurally very similar to serial verb constructions. Many
scholars of these languages include at least some of these in their description
of SVC types. Observe the following examples:
(9) Eastern Kayah Li (Karen, Tibeto-Burman) (Solnit, 2006, p. 153)

vɛ̄ khɑ ʔírɛ dɯ ʌ́
1SG promise work own.accord NS
‘I promise to work myself.’
(10) Cantonese (Sinitic) (Matthews, 2006, p. 75)

ngo5 ceng2 keoi5dei6 sik6 faan6
I invite 3PL eat rice
‘I’m inviting them for dinner.’
(11) Cantonese (Matthews, 2006, p. 84)

ngo5 faan1 lai4 sik5-faan6
1SG return come eat-rice
‘I’m coming back to eat.’
Examples (9)–(11) above are regarded as serial verb constructions in the analy-
sis in which they are presented.
Similar constructions occur in White Hmong (see Chapter 5 for examples
and discussion), in which the second junct functions as a complement of the
first verb and is thus sub-propositional. The truth-value of the second junct
is independent of that of first, resulting in its defeasibility. I therefore do not
regard these types of verb concatenation in White Hmong as serial verb con-
structions; the two verbs do not seem to be functioning as joint predicators in
the expression of the proposition as a whole.
In other examples from the languages of Mainland Southeast Asia, also
involving complementation without a complementizer, the two verbs do have
the same truth-value. The examples below are again treated as serial verb con-
structions in the respective analyses:
(12) Thai (Thepkanjana, 1986, p. 21)

Sùrii tham kææ̂ w tææk
PN do/make glass break
‘Suri broke a glass.’
(13) Cantonese (Matthews, 2006, p. 75)

ngo5 zing2 keoi5 dit3
older.brother make 3SG fall
‘I made him fall.’
(14) Vietnamese (Mon-Khmer) (Bisang, 1992, p. 320, cited from Emeneau,

1951, p. 49)
anh5 xem2 tôi5 nha̓y3
older.brother watch 1SG jump
‘Look at how I jump; look at me jumping.’
Examples (12) and (13) above involve complementation with a causative

meaning. The fact that the first verb in both cases is implicative determines
that the second verb could not be felicitously denied if the first is asserted,
so the truth-value is shared. Example (14) involves a completely different
type of complement: an Activity complement (Dixon, 2006). In this case, the
84 CHAPTER 2
function of this construction—to focus on an unfolding action or situation—

means that, here again, the two juncts could not possibly have independent
truth-value.
Once again, very similar constructions occur in White Hmong and once
again, they are not regarded as serial verb constructions in this analysis
(Chapter 5). In these types of examples, there is no shared argument. In (12),
Suri did not ‘make the glass’, but rather she ‘made the glass break’. In (14), the
addressee is not asked just to ‘look at me’, but to ‘look at me jumping’. These are
not cases of joint predication, but of the second junct functioning as a comple-
ment of the first verb. The second verb is thus sub-propositional, and cannot
be thought of as working with the first to jointly express the proposition as a
whole.
The notion of two or more verbs working as joint predicators, neither one
functioning as a complement of the other and both contributing to the expres-
sion of a single proposition, will be seen in Chapter 3 to contribute much to our
understanding of the semantics of serial verb constructions in White Hmong.
This notion is useful in circumscribing those constructions that are included
here as SVC types in White Hmong, and in identifying those that are super-
ficially similar. In Chapter 5, other syntactic differences between the various
SVC types and these and other superficially similar constructions will be exam-
ined in detail.
2.4 Syntactic Framework for the Analysis of SVC Types in White

Hmong
The study of serial verbs cross-linguistically has shown that it is not possible
to develop a unified syntactic analysis of this phenomenon as it is realised in
every language in which it occurs. Furthermore, with the exception of lan-
guages that have quite limited verb serialization, it may not even be possible
to propose a single syntactic analysis for serial verbs within a single language.
For this reason, as explained above, the singular term ‘serial verb construction’
would not be appropriate cross-linguistically, and only rarely within a single
language (i.e. only in a language that has just one type of verb serialization). As
Aikhenvald (2006b, p. 2) observes, “SVCs are a grammatical technique cover-
ing a wide variety of meanings and functions. They do not constitute a single
grammatical category.”
However, in many languages it is clearly possible to identify a range of
‘constructions’—that is a number of construction types—each one of which
involves the simple juxtaposition of verbs that work together to express a single
event within a single clause. These are what are referred to in this analysis as
serial verb constructions (SVCs). Each SVC type identified in White Hmong
is thus given a distinct name, capitalized to emphasise the fact the term is
specific to White Hmong grammar, rather than intended to imply any a priori
or cross-linguistic category (following Bybee (1985, pp. 47, 141); Comrie (1976,
p. 10); Haspelmath (2011, p. 125)):
– Cotemporal SVCs, with sub-types:

– Cotemporal Motion
– Cotemporal Action and Motion
– Cotemporal Action and Stance
– Cotemporal Action
– Cause-Effect SVCs
– Attainment SVCs
– Disposal SVCs.
Along with a thorough investigation of the components and functions of

each of these SVC types in White Hmong, it is important to provide a precise
account of the ways in which the predicates in each one of these constructions
can be related to one another syntactically.
2.4.1 Key Notions for the Analysis

A framework for analysis that has proved extremely helpful in understand-
ing serial verb types in White Hmong is that initially presented by Foley and
Van Valin (1984) and Foley and Olson (1985). These authors utilised (the then-
developing) Role and Reference Grammar theories concerning, firstly, the lay-
ered structure of the clause and, secondly, the possible types of ‘interclausal’
relations, to describe the syntax of serial verbs.4
The particular theories used by these authors in the works cited above have
subsequently undergone some important modifications within the framework
of Role and Reference Grammar. These revisions have come about as a result of
further research and application of the model not related to the study of verb
serialization (see Van Valin (1987, 2005); Van Valin and LaPolla (1997)). With
some further modification, these most recent Role and Reference Grammar
theories will be used as a basis to present a framework for analysing the syn-
tactic structures of serial verb constructions in White Hmong.
4 The term ‘interclausal’, which was used in these early RRG works, is not really suitable in all
cases to which it was applied; some of these cases (including that of serial verb construc-
tions) do not involve the linking of complete clauses, but rather, the linking of components,
or ‘layers’, within a single clause.
86 CHAPTER 2
The purpose of this section is to give a brief and general outline, firstly, of
the theory of the layered structure of the clause and, secondly, of the theories
concerning how these layers can be linked together: the theories of ‘juncture’
and ‘nexus’. Where they are relevant to the analysis of serial verb construc-
tions, certain features of the revised RRG model will be compared with their
counterparts in the original model used by Foley and Van Valin (1984) and
Foley and Olson (1985).
2.4.1.1 The Layered Structure of the Clause and its Components

In Role and Reference Grammar, the clause is regarded as involving three
layers: the ‘nucleus’, the ‘core’, and the ‘clause’ itself. The innermost layer, the
‘nucleus’, contains the predicate, most commonly a verb (Van Valin, 2005,
p. 28). The next layer, the ‘core’, consists of the nucleus plus the (core) argu-
ments of the predicate. Just as the nucleus is a part of the core, so the core
is a part of the clause, hence the term ‘layers’. Adjuncts such as time phrases,
‘outer’ locatives, and beneficiaries (often called ‘peripheral arguments’ in other
frameworks) are part of the ‘periphery’ to the core.
The structure of a simple clause can thus be represented diagrammatically
as in Figure 4.
It is important to notice that not all clausal components are also clausal
layers. A relationship of inclusion holds between the clause and the core and,
in turn, between the core and the nucleus; the nucleus and the core are both
components of the clause, and also layers within the clause. A different kind
of relationship holds between the core and its periphery. The periphery acts as
a kind of modifier to the core (Van Valin, 1987, p. 2); it is thus a clausal compo-
nent but is not a layer.
CLAUSE
NUCLEUS
CORE
figure 4
PERIPHERY TO CORE Diagrammatic representation of a
simple clause5
5 The vertical alignment of the boxes representing the periphery and the core is not signifi-
cant; it simply emphasises the fact that this diagram is not meant to imply any particular
The nature of this peripheral component is one of the most significant differ-
ences between the most recent version of the RRG theory of clause structure
and earlier versions. In the earliest works each layer was seen as a part of the
next: the nucleus was a part of the core, and the core a part of the periphery.
Thus the periphery itself was the clause; it consisted of the core plus all non-
core arguments (Foley & Van Valin, 1984). As the theory developed further, the
periphery was identified as modifying the core rather than enclosing it, with
the core plus its periphery being regarded as together making up the clause
(Van Valin, 1987). Most recently, the notion of a periphery for each layer—not
only the core, but also the nucleus and the clause—has been added (Van Valin,
2005, p. 21). All three layers can be modified by different types of adverbs, for
example, and these occur in the relevant periphery. Only the periphery to the
core is illustrated in Figure 4 above.
2.4.1.2 Operators
Each layer of the clause is associated with its own set of semantic operators:
grammatical categories such as tense, aspect and mood. For example, aspec-
tual operators have scope only over the nucleus of the clause while modal
operators (that is, those signalling deontic modality: obligation, desire, cona-
tion, ability, and the like) are associated with the core. Operators such as tense,
‘status’ (epistemic modality) and illocutionary force, on the other hand, have
scope over the entire clause including the periphery.
2.4.1.3 Juncture
According to the RRG theory of ‘juncture’, there are three main ‘levels’ at which
the layers of the clause can be joined: nuclear, core, and clausal. (Sentential
juncture can also occur, but is not relevant to this analysis.)
Nuclear juncture involves the linking of two nuclei within a single core. This
is represented in Figure 5. Not only adjuncts but also all core arguments are
shared, and no arguments can intervene between the two predicates.
In nuclear juncture, each ‘junct’ consists only of a verb—the nucleus of a
clause. In both of the higher levels of juncture—core and clausal—the two
‘juncts’ may consist of complete clauses. That is, the core and the periphery are
not in an inclusive relationship, and are seen as tending to operate as a single
unit with regard to juncture. While nuclear juncture involves the linking of
two nuclei within a single core, with no possibility of independent arguments,
juncts linked at both the core and clausal layers may exhibit independent argu-
ments of any kind—core or peripheral.
configuration of arguments. It is only intended to show the relationships between the differ-
ent components of the clause.
88 CHAPTER 2
NUCL NUCL CORE
PERIPHERY figure 5
Diagrammatic representation of nuclear
juncture
As noted above, this is an aspect of the RRG theory of ‘interclausal’ syntax that
has been significantly revised. In earlier works core juncture was seen as paral-
lel to nuclear juncture, in the sense of involving incorporation of the linked
units within the next highest layer:
In core-level junctures two cores, each with its own nucleus and core
arguments, are joined together to form a larger complex core. The periph-
eral arguments must be shared by both cores, as they [the cores] form a
single complex unit within the peripheral layer. (Foley & Van Valin, 1984,
p. 188)
This notion of core juncture taking place within a single periphery was a
direct consequence of the original concentric notion of the structure of the
clause, with the core enclosing the nucleus and the periphery enclosing
the core (see 4.1.1 above).
Core juncture differs from nuclear juncture in that either predicate may have
arguments that are not shared, and in that arguments may intervene between
the two predicates. The defining feature of core layer juncture is that there
must be some ‘intersection’ of the two cores. This involves either the sharing of
at least one core argument by both juncts, or the inclusion (embedding) of one
junct as a core argument of the other.
Juncture at the core level does not imply a union between two cores within
a single periphery. Rather it implies the linking of one clause to another
(nucleus, core and periphery), with the point of linkage being within the
core. Of course there may be partial, or even total, intersection of peripheral
arguments in a core level juncture, but this is not necessarily the case; there is
always the option for completely independent peripheral arguments. This is
illustrated in Figure 6 and Figure 7.
Clausal juncture differs from core juncture in that it involves no ‘intersec-
tion’ of core arguments. If ‘intersection’ occurs at all, then it is in the periphery,
PERIPHERY
SHARED
NUCL CORE NUCL
CORE NP CORE
PERIPHERY
figure 6 Diagrammatic representation of core juncture in which the

point of intersection is a shared core argument
NUCL
NUCL CORE
CORE PERIPHERY
PERIPHERY
figure 7 Diagrammatic representation of core juncture in

which one junct is a subset of the core of the other
NUCL
CORE
PERIPHERY PERIPHERY
NUCL
CORE
figure 8 Diagrammatic representation of clausal juncture in

which the intersection of the two peripheries may be
null, partial, or total
90 CHAPTER 2
NUCL CORE
N CORE
PERIPHERY
PERIPHERY
figure 9 Diagrammatic representation of clausal juncture

in which one junct is a subset of the periphery of the
other
with either the sharing of peripheral arguments (Figure 8) or the embedding of

one junct as a peripheral argument of the other (Figure 9).
While sharing of core arguments does not take place in clausal junctures,
coreference between core arguments is common. The distinction is important.
“ ‘Sharing’ arguments involves the mapping of two identical arguments from
two logical structures into the same syntactic position in a clause” (Van Valin,
1987, p. 2). This occurs in both nuclear and core level junctures, but in clausal
junctures the two cores are independent; no core arguments are shared.
‘Coreference’, on the other hand, is a feature of clausal junctures (p. 3) and
simply means that the arguments in two (or more) distinct syntactic positions
refer to the same entity in the real world. According to Van Valin (p. 2):
An important diagnostic of the difference in English and many other lan-

guages is that a pronoun cannot appear in the dependent junct where
sharing is involved, whereas an overt pronoun can occur in cases of
coreference, . . .
To summarise: nuclear juncture involves the linking of two nuclei within a sin-
gle core, and the sharing of all core and peripheral arguments; core juncture
involves the linking of two clauses, with the point of juncture being at the core
level (intersection of core arguments); and clausal juncture involves the link-
ing of two clauses, with no intersection of core arguments.
2.4.1.4 Nexus
The term ‘nexus’ refers to the type of syntactic relationship that exists between
two juncts. There are three distinct nexus types: ‘coordinate’, ‘subordinate’ and
‘cosubordinate’.
When two juncts are in a coordinate relationship, neither one is embedded

in the other. Furthermore, the two juncts are independent of each other with
regard to the scope of operators at the level of juncture. In other words, coor-
dinate juncts exhibit no ‘grammatical category’ dependency (Van Valin, 1987).
This means that predicates joined at the nuclear level are in a coordinate
relationship if they can be independently modified by a nuclear level operator
such as aspect. Clauses joined at the core level are in a coordinate relationship
if there is no embedding involved and if a core level (deontic modal) operator
in the first junct does not have scope over the second junct as well. Finally,
clauses joined at the clausal level are in a coordinate relationship if neither
is embedded in the other and clause-layer operators such as tense, status and
illocutionary force have scope over only one junct, not both.
The defining feature of subordinate nexus is the embedding of one junct
within the other. That is, a junct that is subordinate to another functions as
either a core argument or an adjunct (peripheral argument) of that other
junct. Consequently nuclear juncture does not occur with the subordinate
nexus type; no embedding can occur within a nucleus.
Like coordinate juncts, subordinate juncts do not exhibit ‘grammatical
category’ dependency: they are independent of the superordinate junct with
regard to scope of operators at the level of juncture. However, there is another
sense in which subordinate juncts are often dependent. In English and many
other languages, a subordinate junct cannot be independently specified for
illocutionary force, but ‘must have the neutral, unmarked form of a statement’
(Foley & Van Valin, 1984, p. 239). If illocutionary force cannot be specified in a
subordinate junct, then it could never stand alone as a complete utterance; in
this sense it is dependent on the superordinate junct. Van Valin (1986) terms
this type of dependence ‘distributional’ dependence.
The term cosubordinate means that neither junct is embedded in the other,
but that the juncts exhibit ‘grammatical category’ dependency: all operators
can be shared at the level of juncture (Van Valin, 2007, p. 80).6 Juncts in a nuclear
cosubordinate relationship could therefore share aspectual operators, and
those in core cosubordinate relationship could share all deontic modal opera-
tors. Clause-level cosubordinate juncture requires juncts that are linked at the
6 Van Valin (2007, p. 80) proposes that, while cosubordination has generally been thought to
involve obligatory sharing of operators at the level of juncture, there are some languages
in which operator sharing is not obligatory. He cites data from Bickel 2003, from Belhare
(Tibeto-Burman) and Nepali (Indo-European). Even in these cases, he says, it is always pos-
sible that operators will be shared, a feature that clearly distinguishes cosubordination from
coordination and subordination.
92 CHAPTER 2
clause level, with no embedding, to allow clause-level operators like epistemic

modality to have scope over both juncts.
In summary, subordination can be easily distinguished from the other two
nexus types because it is the only type that involves the embedding of one
junct as an argument of the other. The two non-embedded nexus types—
coordination and cosubordination—are separated by the feature of ‘grammat-
ical category’ dependency; while juncts in a cosubordinate relationship can be
dependent at the level of juncture, those in a coordinate relationship must be
independent at that level.7
2.4.2 The Syntactic Analysis of White Hmong SVCs Proposed

2.4.2.1 Summary of Syntactic and Semantic Features
As noted at the beginning of this chapter (§2.1), a key challenge in the study
of languages that utilise the strategy of serialization is to determine which of
the sentence types containing concatenated verbs in that language warrant
the special title ‘serial verb construction’. Are there any particular syntactic
and semantic properties shared by some of these sentence types that set them
apart from the others? On the basis of evidence from White Hmong, particu-
larly that described in Chapter Three below, the following set of features are
proposed as the defining structural features of a serial verb construction in
this language:
7 In Role and Reference Grammar, the tripartite division of nexus types is actually expressed
in terms of binary features: coordination = [-embedded] [-dependent]; subordination =
[+embedded] [+dependent]; cosubordination = [-embedded] [+dependent]. However, as
Van Valin himself points out (1987), there is an important distinction between the type of
dependency exhibited by subordinate juncts—‘distributional’ dependency—and that exhib-
ited by cosubordinate juncts ‘grammatical category’ dependency. If the feature [+depen-
dent] is used to refer to both types of dependency, then it would surely have to be said that
coordinate juncts linked at the core level are also [+dependent]. Like subordinate juncts, the
second junct in a core coordinate construction cannot be independently specified for illocu-
tionary force, and so cannot stand alone as an complete utterance:
(i) *Fred got Mary to wash his car, doesn’t/didn’t she?
One reason why illocutionary force cannot be independently specified is because of the sub-
propositional nature of the junct in question; illocutionary force operates only over entire
propositions. The point is, however, that coordinate juncts are considered to be independent
in the RRG model, because they are independent with regard to operators at the level of junc-
ture. For this reason, if binary features are to be used at all, it would seem far better to propose
a primary division using the feature [±embedded], to distinguish subordination ([+embed-
ded]) from the other two nexus types, and a secondary division using the feature [±gram-
matical category dependency] to distinguish coordination ([-embedded] [-GCdependent])
from cosubordination ([-embedded] [+GCdependent]).
i) A series of two or more concatenated verbs (with or without intervening

core arguments);
ii) Each verb can occur on it own as the only predicate in the clause;
iii) No overt form of linkage;
iv) No pause between the juncts;
v) At least one shared core argument;
vi) Shared core operators (i.e. deontic modality);
vii) Shared clausal operators (e.g. tense, epistemic modality, illocutionary
force);
viii) A single set of peripheral adjuncts; the verbs belong to a single clause;
ix) No junct is embedded as an argument of another.
All of these structural features relate to semantic features that
x) The verbs are joint predicators within a single proposition, and express a
single event.
One purpose of this study is to show how this set of features serves to define
a particular group of concatenated verb sentences in White Hmong that can
justifiably be set apart from all other sentences involving concatenated verbs
in this language. With this aim in mind, it will be helpful to discuss how serial
verb constructions fit in to the syntactic framework the theory of juncture and
nexus introduced above.
2.4.2.2 Juncture and Nexus in Serial Verb Constructions

The analysis of serial verb constructions proposed by Foley and Van Valin
(1984) and Foley and Olson (1985) was based on the original Role and Reference
Grammar model of clause structure (see §2.4.1.1 above). In this model, each
layer of the clause was seen as enclosed within the next (nucleus within core,
and core within periphery), and juncts seen as linked to other juncts of the
same type, again within the next highest layer of the clause. Their analysis of
serial verb constructions was a monoclausal one: serial verb constructions
were seen as consisting of two nuclei within a single core, or two cores within a
single periphery. However, the revision of the RRG notion of the layered struc-
ture of the clause, and of the nature of the three levels of juncture (outlined in
Van Valin (1987)), has significant implications for the description of some types
of serial verb constructions (specifically core layer serial verb constructions).
These implications will be described below.
In the following two sections (a. Nuclear Layer SVCs and b. Core Layer SVCs)
the juncture and nexus types occurring with serial verbs will be illustrated.
94 CHAPTER 2
Specific evidence to justify the classification of the examples from languages

other than White Hmong below is not discussed here, although some of it may
be clear in the examples themselves. The relevant evidence can be found in the
works cited. Those types that occur with serial verbs in White Hmong are, of
course, discussed in detail in Chapter 3.
a Nuclear Layer SVCs

A serial verb construction consisting of two nuclei within one core would cer-
tainly be regarded as a monoclausal construction in the revised framework of
RRG, just as it was in the original model. Nuclear juncture was illustrated in
Figure 5 above.
There are two possible nexus relationships between nuclear juncts in a
serial verb construction: nuclear coordinate and nuclear cosubordinate. Serial
verb constructions in which the verbs are in a nuclear coordinate relationship
would allow independent nuclear layer operators for each junct. The repre-
sentation of nuclear juncture originally give as Figure 5 actually serves best to
represent this kind of nuclear layer SVC, as the two nuclei are shown as inde-
pendent of each other. This diagram is repeated here as Figure 10:
NUCL NUCL CORE
PERIPHERY figure 10
Diagrammatic representation of SVCs
involving nuclear coordination
NUCL CORE
NUCL NUCL
PERIPHERY figure 11
Diagrammatic representation of SVCs
involving nuclear cosubordination
Serial verb constructions in which the verbs are in a nuclear cosubordinate

relationship, on the other hand, would allow only one set of nuclear layer
operators, which would have scope over both juncts. This relationship is better
illustrated by a structure in which the two nuclei are linked together within a
single nucleus, as shown in Figure 11.
These two types of nuclear layer serialization—nuclear cosubordination
and nuclear coordination—are exemplified below:
i. Nuclear Cosubordination:
(15) Mandarin (Hansell, 1987, p. 40)
a. wo3A/A kan4 dong3 le [nei4 ben3 shu1]O/O
I read understand ASP that CLF book
‘I understand (by way of reading) that book.’
b. ta1A/A qiao1 po4 le [yi2 ge fan4wan3]O/O

he hit breakINTR ASP one CLFg ricebowl
‘He broke (by hitting) one rice bowl.’
(16) White Hmong (DNH 137: 67)

Niam Nkauj Zuag PajS/A twb los dhau [peb lub thoob]O
PN sure come pass three CLF bucket
lawm lau
PRF INTNS
‘Nia Ngao Zhua Pa had already come right past three buckets (of water).’
(17) Kuman (Papuan) (Piau, 1985, p. 133)

yeA/A mokonaO/O gat ne kodugwa
ye mokona gagl ne kod-uw-ka-a
3 greens cook eat finish-3SG-REALLY-DECL
‘She cooked (and) ate the greens.’
ii. Nuclear Coordination:

(18) Kuman (Papuan) (Piau, 1985, p. 133)
yeA/A mokonaO/O gagl kodo ne moglkwa
ye mokona gagl kod ne mogl-uw-ka-a
3 greens cook finish eat stay-3SG-REALLY-DECL
‘She finished cooking the greens and is eating them.’
All the examples of nuclear layer serialization given above consist of two jux-
taposed nuclei, sharing a single set of core arguments. In both the original
96 CHAPTER 2
and the revised RRG frameworks, such structures would clearly be regarded as
monoclausal.
In some examples of nuclear layer serialization, such as the Mandarin and
White Hmong examples (15) and (16) above, it is quite easy to see how the
serialized predicates can be thought of as representing different facets of a
single conceptual event. However, in other cases, such as the examples given
above from Kuman ((17) and (18) above), the action of ‘cooking (and) eating
the greens’ may appear to represent a series of events rather than a single
one. However, recall the observations made by Bruce (1988) and Durie (1997)
(§2.2.1) that commonly or culturally associated actions can be perceived
within the relevant speech community as single events, and that this percep-
tion is reflected in their grammatical treatment. While the particular aspects
of the preparation of food and its subsequent consumption, expressed in
these examples, may seem more like a series of distinct events for an English
speaker, they must represent such commonly associated actions that they can
be regarded as components of a single event in the Kuman-speaking commu-
nity. This is reflected in the fact that these actions can be represented together
by serial verbs. In (17), the two verbs are jointly under the scope of the same
aspect marker (kod ‘finish’), so the nexus relationship is cosubordinate. In (18),
this aspect marker appears between the verbs and has scope only over the first
of the two, gagl ‘cook’, so the nexus is coordinate.
b Core Layer SVCs

A core layer serial verb construction simply could not occur in the revised RRG
framework articulated by Van Valin (1987, 2005) and Van Valin and LaPolla
(1997). In the revised version of the theory, any kind of core juncture is seen
as involving the linking not of two cores within a single periphery, but of
two clauses, core plus periphery (the point of intersection being within the
core). A structure in which the point of intersection is a shared core argument
(rather than one junct being a subset of the core of the other) was illustrated
as Figure 6 in §2.4.1.2 above, and is repeated here.
Such a structure could hardly be said to be ‘monoclausal’. The fact that
this revised RRG view of core juncture always allows juncts linked at the core
level to have independent adjuncts (peripheral arguments) is at odds with
the fact that independent adjuncts cannot occur for each junct when a serial
verb construction is involved. While this structure is perfectly appropriate for
many kinds of core juncture, it is not useful for serial verb constructions. By
definition, a serial verb construction must involve a single clause with a single
periphery.
PERIPHERY
SHARED
NUCL CORE NUCL
CORE NP CORE
PERIPHERY
figure 6 Diagrammatic representation core juncture in which the point

of intersection is a shared core argument
As pointed out in Jarkey (1991), for this reason we need to expand the possible
juncture/nexus relationship types specifically in order to accommodate core
juncture structures that actually disallow independent peripheries. There are
two ways in which we could do this. The first of the two possibilities would
involve cosubordinate juncture at the core level, and the second, coordinate
juncture at the core level. In both cases, a single periphery would be associated
with the linked cores, resulting in structures similar to the original RRG model
for these types of core juncture.
In the first case, in which the juncts are in a core cosubordinate relationship,
the two cores would need to be linked within a higher core in order to ensure a
single set of core layer operators. This is represented in Figure 12.
CORE
NUCL
SHARED NUCL
CORE NP
CORE CORE
PERIPHERY
figure 12 Diagrammatic representation of SVCs involving core

cosubordination
98 CHAPTER 2
Linkage of this type at the core level in a cosubordinate nexus relationship is,
in fact, an extremely common kind of serial verb construction:

khǎwS/S/A/A/S wìn yɔɔ́ n khaâm saphaanO khâw baânO pay
he run reverse cross bridge enter house go
‘He ran back across the bridge into the house (away from the speaker’s
center of attention).’
(20) Yoruba (Benue-Congo) (Stahlke, 1970, p. 61)

moA/A mú ìwéO wá iléO
I took book come house
‘I brought a book home.’

nqa [ib lub yeeb.thooj]O haus puj.pauv tuaj
carry one CLF pipe inhale puff.puff come
‘. . . bringing a pipe along and puffing on it.’
(22) Paamese (Oceanic) (Crowley, 1987, p. 43)

kailA amuas vuasO/S emat
3PL 3PL.REAL.hit pig 3SG.REAL.die
‘They hit the pig (and it) died.’

yuav-tau hlais noob.qesO/S tawm . . .
must slice testicles leaveINTR
‘. . . (they) had to slice (the hog’s) testicles out.’
Independent validation of the structure proposed in Figure 12 has come from

a suggestion by Bohnemeyer and Van Valin (2009), which brings together
the RRG notion of the juncture/nexus relationships with the notion of the
Macro-Event Property (MEP) (Bohnemeyer et al., 2007; Bohnemeyer, Enfield,
Essegbey, & Kita, 2011). (See §2.2.2 above for an explanation of the MEP and its
relevance to serial verb constructions.) This collaborative work acknowledges
that multi-core constructions across languages generally lack the MEP, because
each core generally has its own periphery. However Bohnemeyer and Van Valin
(2009) suggest that multi-core constructions can have the MEP if, and only if,
they are in a nexus relationship of cosubordination. These scholars present
evidence from English infinitival complements, Ewe serial verb constructions,

and Japanese converb constructions; they argue that these constructions all
involve core cosubordination under a core node and that “core cosubordina-
tion preserves the MEP” (p. 22).
The fact that structures of this type have the MEP relates directly to the fact
that they represent single events. The fact that the juncts are linked at the core
level in a cosubordinate relationship means they will also have a single set of
core-layer operators. That is, deontic modality operators expressing desire,
conation, intention, obligation, and so on will also be shared, in addition to
temporal operations, such as tense, time adverbs, and temporal clauses. This
multi-verb structure thus involves an even tighter linkage than one that has the
MEP alone. Evidence of shared deontic modality for serial verb constructions
in White Hmong will be given in the sections describing SVC types involving
core cosubordination in Chapter Three below.
As noted above, there are two ways in which we could expand the possible
juncture/nexus relationship types specifically in order to accommodate core
juncture structures that actually disallow independent peripheries. The first
of these would involve cosubordinate juncture at the core level, and has been
illustrated above in Figure 12. The second of the possible juncture/nexus types
that would disallow independent peripheries would involve core coordinate
nexus. In this case, the two cores would not appear together under a higher
core, and so would be independent with regard to the scope of operators at the
level of juncture. This structure is illustrated in Figure 13 below.
As noted, in this juncture nexus combination involving core coordination,
core level operators would have scope over one junct only. However, core coor-
dination does not seem to be a possible type of SVC.
SHARED
NUCL CORE NUCL
CORE NP CORE
PERIPHERY
figure 13 Diagrammatic representation of core-level coordinate

juncture in which two cores share a single periphery
100 CHAPTER 2
Core coordination certainly does occur in some languages between juncts

that are merely juxtaposed, with no overt form of conjunction. However, the
available evidence suggests that these structures refer to more than one event.
Observe the following example from Barai (Papuan):
(24) Barai (Olson (1981), cited in Foley and Van Valin (1984, p. 246))
na [e ije] k-ea bu-me va-e
1SG person DEF say-3PL.ᴜ 3PL-CASUAL go-PST
‘I spoke to the people, (and) they just went.’
The two verbs in this sentence have a shared core argument, namely e ije
‘the people’, and the second junct is independently modified by what Olson
describes as the core layer ‘mode particle’ me ‘CASUAL’. Independent modifica-
tion by clause-level operators is not allowed. Olson concludes that the nexus/
juncture relationship is one of core coordination. However, he also “notes
that there is a significant pause between the two juncts in sentences such as
these; there is no such pause with core-level cosubordinate nexus” (Foley &
Van Valin, 1984, p. 246). The occurrence of such a pause suggests that these
juncts are not in a serial relationship, and the possibility of independent modal
operators emphasises the fact that they represent two distinct, though closely
related events.
Sebba (1987) gives examples from Sranan (Carribean Creole), which he
refers to as ‘Coordinating SVCs’, such as (25) below:
(25) Kofi naki Amba kiri en

PN hit PN kill 3SG
‘Kofi struck Ambai (and) killed heri .’
Sebba’s reasons for suggesting that these juncts are in a coordinate relationship
are convincing (pp. 109–110). However, such sentences would not be regarded
as a type of SVC in this analysis for two reasons. Firstly, Sebba notes that con-
ceptually the two verbs represent two events. Secondly, as in the example of
core coordination from Barai above, it is possible for a pause to occur between
the two juncts.
There is thus no clear evidence that the core coordinated type of junc-
ture/nexus relationship can occur between serialized verbs. This is the loos-
est form of juncture that the diagrams (specifically Figure 13) could be taken
to illustrate, and so it is quite likely that core serialization of this type simply
does not occur. This conclusion concurs with the proposal that serial verbs
must have a single set of core operators. It also supports the proposal made by
Bohnemeyer and Van Valin (2009) that core cosubordination has a special sta-
tus amongst core juncture types in that it can exhibit the MEP.
2.4.2.3 Summary of Juncture/Nexus Types Occurring with Serial Verbs

In their typological study, Foley and Van Valin (1984, p. 263) note verb seri-
alization occurring in the following juncture/nexus combinations: nuclear
coordinate; nuclear cosubordinate; core cosubordinate. The first two of these
combinations can be represented by Figure 10 and Figure 11 above, illustrat-
ing nuclear coordination and nuclear cosubordination respectively. The third
combination is represented by Figure 12, which illustrates core cosubordina-
tion. As will be shown in Chapter 3, in White Hmong serial verbs occur only in
the cosubordinate nexus type, but can involve either nuclear or core juncture.
2.5 Conclusion to Chapter Two
In this chapter, we have seen that we can account for the ‘part-whole’ relation-
ship of verbs in series in terms of the fact that they work together to express
both a single event and a single proposition, the prosodic and grammatical
structure of the construction, and the ability of such a structure to be inter-
preted iconically. However, we have not yet accounted for the fact that, within
the limits imposed by this ‘part-whole’ relationship, verbs in series can still
be seen as related to one another in quite different ways. The kinds of rela-
tionships that occur between serial verbs include, for example, temporal,
aspectual, and causal relations. Again, we cannot appeal to the presence of a
particular conjunction or complementizer as the element that is responsible
for the exact interpretation of the relationship in a particular SVC. What are
the factors that determine this interpretation?
It will be shown in Chapter Three that it is the order and arrangement of
components (including the level of juncture involved), as well as factors related
to the individual lexical items in the constructions—such as their transitivity,
the semantic roles they assign to their arguments, and their inherent aspectual
properties—that are responsible for the precise interpretation of the relation-
ship between the two facets of the one event expressed in serial verb construc-
tions in White Hmong.
CHAPTER 3
Productive Serial Verb Constructions in

White Hmong
In her article on serial verbs and propositions in White Hmong, Riddle (1989)
maintains that serial constructions in this language can represent either a
series of events or a single event:
. . . the interpretation of the number of propositions expressed by a serial

construction ultimately depends on lexical and pragmatic factors . . . in
many such concatenations, each verb can contribute its prototypical ver-
bal sense and thus be interpreted as representing a separate proposition,
depending on the larger linguistic and situational contexts in which it
occurs. (p. 2)
She gives the following example (cited from Johnson (1981, p. 19)), with two
alternative translations (Riddle, 1989, p. 10):
(1) nws txiav nrov pov tseg

3sɢ cut weed throw leave
‘S/he cut down the weeds.’
‘S/he cut down the weeds and threw them away.’ (HFT 19)
Riddle claims:
This sentence can be interpreted as expressing one proposition or two

depending solely on the pragmatic context. If the weeds in question were
cut and allowed simply to lie where they fell, like grass, then the serial
construction has the first interpretation given above and it represents a
single proposition. However, if the weeds were cut down and then picked
up and discarded somewhere else, then it has the second interpretation,
which represents two separate propositions.
It is actually not strictly correct to say that the interpretation of the sentence
above is entirely dependent on pragmatic factors. If the speaker intends the
sentence to express two distinct propositions, then this sense can be conveyed
clearly by the insertion of a distinct pause before the second of the two verbs.

productive svcs 103
No such pause would be possible if the ‘single event’ sense were intended.
Nevertheless, the important point to be made here is that, while Riddle does
observe that juxtaposed verbs in White Hmong can serve to express both a
single event and a series of distinct events, she does not choose to draw a ter-
minological distinction; Riddle classifies sentences with both of these mean-
ings as involving ‘serial constructions’.
As outlined in Chapter 2, the term ‘serial verb construction’ is used in quite
a different way in this study. Here, verb serialization in general is regarded as
a strategy for portraying a single proposition within a single clause using more
than one verb. When it comes to interpreting the meaning of a sentence that
makes use of such a strategy, there is no doubt that pragmatic context is cer-
tainly important. In addition, however, a full account of the meaning and use
of a particular SVC within a particular language also requires an understand-
ing of the structure, components, and function of that particular SVC type.
Four major types of productive serial verb constructions are identified
in this analysis of White Hmong. They are referred to here by the following
names: Attainment SVCs, Cause-Effect SVCs, Disposal SVCs and Cotemporal
SVCs. These four types of SVC vary with regard to:
– the number of verbs that can appear (two or more than two);
– the types of verbs involved (related to factors such as transitivity, effective-
ness and aktionsart);
– the nature of the units linked (level of juncture) (although the nexus rela-
tionship between the juncts is that of cosubordination in all cases);
– the semantic relationships between the verbs;
– the function of the shared arguments in relation to the verbs;
– the focus of the construction in terms of process (what the subject does)
and/or outcome (what happens to the object), and the function(s) of the
construction as a whole.
Table 5 is a simplified representation of each of these four major SVC types

identified in White Hmong. The shared arguments are highlighted in the table.
table 5 Productive SVC types—verbs and argument sharing
Attainment S/A V1 V2 (O)

Cause-Effect A V1 O→S V2
Disposal A V1 O V2 (V3) . . .
Cotemporal S/A V1 (O) V2 (O) (V3) . . .
104 CHAPTER 3
It will be clear simply by looking at Table 5 that what are here all identified
as types of SVC in White Hmong actually involve four quite distinct construc-
tions. These constructions differ in regard to
– whether two or more than two verbs can appear;

– how the components are basically ordered; and
– what the function of the shared arguments is in relation to each verb: A, S,
or O.
At the same time, of course, these four constructions have much in common
in terms of their structure—they are all ‘packaged’ as single events in the ways
typical of serial verb constructions, as described in Chapter 2, §2.2.2.2 and
§2.2.4.2.
In addition to these basic structural similarities, what else do these four con-
structions in White Hmong have in common? Bruce (1988, p. 32) claims that
serial verb constructions seem to fulfil a specific pragmatic function that is
distinct from that of other combinations of verbs. Is there anything that would
warrant classifying and describing these four constructions together as types
of SVC in terms of their function? Furthermore, how do they differ functionally
from other constructions in the language that also involve the simple juxtapo-
sition of verbs, but that are not regarded here as types of SVC?
There is, in fact, a key functional element found in all four constructions
identified as major SVC types in this study of White Hmong. That common ele-
ment is the fact that these constructions all facilitate the elaboration of detail
in portraying a single event, within the bounds of a single clause. This elabo-
ration of detail may relate to either an intransitive or a transitive process or
action. In some SVC types, the focus and elaboration of detail relates not only
to a process or action, but also to its outcome. In others, it relates entirely to
an action itself.
More precisely, the following types occur:
– an intransitive process and its outcome for the S argument (Attainment

SVCs with intransitive verbs)
– a transitive action and its outcome in relation to the O argument (Attainment
SVCs with transitive verbs; Cause-Effect SVCs; Disposal SVCs)
– an intransitive or a transitive action where the focus is entirely on the actor
and the detail of the action itself, rather than on any outcome of the action
at all (Cotemporal SVCs).
productive svcs 105
In all of these cases, the basic structure of multiple verbs within a single clause
facilitates precisely the rich and intricate portrayal of events that Westermann
(1930, p. 126) observed in the Ewe. Recall his observation that:
. . . the Ewe people describe every detail of action or happening from

beginning to end, and each detail has to be expressed by a special verb:
they dissect every happening and present it in its several parts,. . .
Westermann’s point that “each detail has to be expressed by a special verb”

in Ewe is an important one to note. It is certainly true that the structure of
the clause—the arrangement of the multiple verbs and their arguments in
it—are fundamental elements in establishing the function of each SVC type.
Nevertheless, the structure by itself conveys nothing at all without the lexi-
cal items that give it flesh. As this chapter will show, each one of these four
constructions in White Hmong is characterised by certain types of predicates,
which work together with each other and with the structure they occur in, to
create this description of “every detail of action or happening from beginning
to end”.
For this reason, as a starting point in the investigation of these construc-
tions, it will be helpful to briefly discuss the issue of ‘situation aspect’/aktion-
sart and the related classification of verbs and other predicates proposed by
Vendler (1957), and subsequently refined by a number of linguists, including
Dowty (1979), Foley and Van Valin (1984), Van Valin and LaPolla (1997), Smith
(1997) and Van Valin (2005).
Vendler divides verbs into four categories, which can be distinguished on
both semantic and syntactic grounds. These categories and some examples
which Vendler gives to illustrate them, along with their key, identifying fea-
tures noted by Smith (1997) are:
i) states (static, durative): want, love, dominate;

ii) activities (dynamic, durative, atelic): walk, run, push a cart;
iii) accomplishments (dynamic, durative, telic): run a mile, paint a picture,
grow up;
iv) achievements (dynamic, instantaneous, telic): recognise, find, win the
race, die.
To Vender’s four categories, Smith (1997, p. 3) adds a fifth, to cater for instanta-
neous actions:
v) semelfactive (dynamic, instantaneous, atelic): tap, knock, cough.

106 CHAPTER 3
Van Valin and LaPolla (1997) and Van Valin (2005) further distinguish between
active accomplishments, which are the telic uses of activities (run a mile, drink
a beer, paint a picture) and (other) non-dynamic accomplishments (grow up,
melt, learn Chinese).
It is widely recognised that these categories actually refer, not to the inher-
ent aspectual properties of individual verbs, but to properties of entire predi-
cations or clauses. Mourelatos (1981, p. 199) lists six factors which are involved
in determining the classification of a predication into these categories:
(a) the verb’s inherent meaning; (b) the nature of the verb’s arguments,
that is, of the subject and of the object(s), if any; (c) adverbials, if any;
(d) aspect [in the sense of ‘viewpoint’ aspect (Smith, 1997)—N.J.];
(e) tense as phase (e.g. the perfect); (f) tense as time reference to past,
present, or future.
It will be clear from the following discussion that, although not all predicates
in White Hmong can be slotted neatly into these categories, this system of clas-
sification does provide a useful basis for understanding the aspectual prop-
erties of the various predicate types that appear in White Hmong serial verb
constructions. The aspectual properties of the predicates themselves, and
the ways in which each one relates to other elements in the clause, are cru-
cial to the meaning and function of each construction. The properties of the
predicates involved will be seen to work together with the basic structure and
arrangement of the SVC types described in this chapter, to create the detailed
portrayal of events in each case.
3.1 Cotemporal Serial Verb Constructions
Serial verb constructions in White Hmong in which the relationship between

the verbs is cotemporal can be divided into two main types: those that refer
only to cotemporal motion, and those that incorporate reference to action
(including motion and action, stance and action, and cotemporal actions).
3.1.1 Cotemporal Motion Serial Verb Constructions

3.1.1.1 Introduction
Nowhere is it clearer than in the case of Cotemporal Motion SVCs that verb
serialization is a device for expressing the details of a single situation in terms
of its various components. It is well known that where a language without
productive svcs 107
serial verb constructions may use a single lexical item such as take or bring,
a serializing language will ‘break up’ such an action into components such as
carry go or carry come respectively. Furthermore, where a non-serializing lan-
guage may use a directional adposition or adverb to indicate the Path, Source,
Goal or direction of a motion event (e.g. up, over, back, from, to, away), a serial-
izing one will add another verb (e.g. ascend, cross, return, leave, go). Such differ-
ences are illustrated in the following examples from White Hmong, and their
English translations:
(2) [cov tub.rog]S/A/S/A khiav tawm [ntawm [lub

CLF.PL soldier run leave nearby CLF
kwj.ha]]O rov los tsevO
gully-valley(t.c.) return come home
‘The soldiers fled from the ravine and came back home.’
(3) [ib tug niag poj-dab]A/S taug hav-dejO

one CLF(t.c.) great female-spirit follow valley-water
hnyuj.hnyo los
slowly and steadily come
‘. . . a big ol’ female spirit came slowly and steadily along the river valley.’
(DNH 319:4)
(4) nqa [ib lub pob.zeb]O nce mus lawm [saum

carry one CLF stone ascend go depart top
[lub roob siab~siab saud]]O
CLF mountain RDUP~be.tall top
‘. . . take a stone up to the top of that tall mountain up there.’ (DNH 116: 8)
Various aspects of the motion event such as these are conceptually separable.
In the real world, however, they are not clearly divisible. They simply constitute
different facets of the one event and, as such, are ideal candidates for expres-
sion via serial verb constructions.
3.1.1.2 The Elements in a Cotemporal Motion SVC: Order and Function

There are a number of types, or semantic classes, of motion verbs that can
appear in a Cotemporal Motion SVC in White Hmong. The different classes
of verbs are strictly ordered, and each one expresses a different facet of the
motion event. A Cotemporal Motion SVC may be composed of two or more
verbs selected from any of the classes, provided they appear in the correct
108 CHAPTER 3
order. It is also possible to utilise more than one verb from a single class. The
list below shows the order in which the classes appear, and gives examples of
verbs from each one.
a Manner of Motion
i Transport Verbs
Verbs of this type are transitive and describe the way in which a subject trans-
ports an object through space: e.g. nqa ‘carry’, ris ‘carry on the back’, kwv ‘carry
on the shoulders’, lawv ‘drive animals’, coj ‘take (someone) along’.1
ii Locomotion Verbs
These verbs are intransitive and simply express the way in which the subject
moves through space: e.g. ya ‘fly’, dhia ‘jump’, nkag ‘crawl’.
b Motion with Respect to GROUND

In a motion situation, the FIGURE—the moving object—is perceived to be
in motion in relation to another object. This second object is termed the
GROUND (Talmy, 1975, p. 181). The GROUND at the origin of the motion event
is the Source; the GROUND relevant for the extent of the motion event is
the Path; the GROUND at the destination of the motion event is the Goal.2 A
1 It is important to distinguish between these ‘Transport’ verbs and ‘Transfer’ verbs. In the case
of ‘Transport’ verbs, both the subject and the object constitute the FIGURE (Talmy, 1975);
that is, both the subject and object move together. In the case of ‘Transfer’ verbs, on the other
hand, the FIGURE is the object alone. This is especially clear in the case of those Transfer
verbs that involve actual motion (e.g. xa ‘send (something)’, tso ‘send (someone)’, dov ‘roll’,
pov ‘throw’), as opposed to those in which the FIGURE (the object) only moves metaphori-
cally (e.g. pub ‘give’). The subject of Transfer verbs involving true motion is understood to
remain stationary, while only the object moves. Thus, while a Motion SVC is appropriate for
a Transport verb, where both the subject and the object move together, a Cause-Effect SVC
(see Chapter 3, §3.2 below) is used in the case of a Transfer verb:
(i) kuvA xa [ib tsab ntawv]O/S mus rau nws
1SG send one CLF writing go to 3SG
‘I sent a letter to him.’

(ii) tso dov [koj lub]O/S nqis lawm [ib phab] . . .
send roll 2SG CLF descend leave one side
‘Roll yours down one side . . .’ (DNH 116: 8)
2 What is here referred to as ‘Motion with respect to GROUND’ is termed PATH in Talmy’s
description of the motion situation: “The respect in which one object is considered as mov-
ing or located to another object is (functions as) the PATH . . . of the motion situation . . .”
(Talmy, 1975, p. 181). The term PATH is not used with this meaning here because it is more
productive svcs 109
previous motion event is also seen here as a potential type of GROUND, with
respect to which the FIGURE can be thought of as moving (see b iii. below).
Following verbs that describe Manner of Motion (see a i, ii above), the next
four types of verb in a Motion SVC all refer to the way in which the motion
interacts with the GROUND.
i Path Verbs
Examples of Path verbs in White Hmong include: hla ‘cross, encircle, pass’,
taug ‘follow (a stationary object e.g. a road)’, lawv ‘follow (a moving object)’.
Path verbs that also refer to vertical orientation—nce ‘ascend’, nqis ‘descend’—
appear after other Path verbs. These verbs are transitive, and the O argument is
a Noun Phrase referring to the Path of the motion.
(5) taug~taug dejO nqis havO mus

RDUP~follow water descend valley go
‘. . . followed the river down along the valley away.’ (DNH 183: 3)
ii Source Verbs
Examples include: tawm ‘leave’, dim ‘escape’, sawv ‘get up/rise’, khiav ‘flee’, thim
‘retreat’. The O argument of these verbs indicates the Source of the motion.
Unlike Path verbs, whose O argument refers to an entity and is expressed as a
Noun Phrase (e.g. nce ntaiv ‘climb stairs’, hla pas-dej ‘pass a pond’), the O argu-
ment of a Source verb refers to a place and is expressed as either a Locative
Phrase (Spatial Deictic plus Noun Phrase—see Chapter 1, §1.5) or by a proper
noun referring to a place:
(6) nwsA tawm [nram liaj]O lawm

3SG leave down paddy.field PRF
‘She has left the paddy field.’
(7) [cov Hmoob]A/S dim [(hauv) Nplog.teb]O los

CLF.PL Hmong escape inside Laos come
‘The Hmong escaped from Laos.’
useful to employ this word in the Fillmorean sense, referring to one type of GROUND
(i.e. Path as opposed to Source or Goal). The expression ‘Motion with respect to GROUND’
refers to Motion with respect to a Path, a Source or a Goal. The terms ‘Source’, ‘Path’, and
‘Goal’ are not used in Talmy’s analysis.
110 CHAPTER 3
(8) thaum nwsA khiav [Suav rog]O mas tsis yog

time 3SG flee China war TOP NEG COP
[rog loj]CC
war be(come).big
‘When he fled the Chinese war, it wasn’t a big war.’ (Fuller, 1985, p. 116)
iii Motion with Respect to Previous Motion Event

There seems to be only one verb that occurs in this slot: the intransitive verb
rov ‘return’. Its appearance is extremely common in Motion SVCs, and it is
often followed by the adverb qab ‘back’.
(9) nwsS/S/A khiav rov qab mus tsevO

3SG run return back go home
‘He ran back home.’
(10) thov nce rov tuaj . . .

request run return back
‘Please come back up . . .’ (DNH 25: 3)
iv Goal Verbs
Examples include: mus ‘go’, tuaj ‘come’, los ‘come/return home’. As with Source
verbs, the O argument of these Goal verbs refers to a place, and so is expressed
either as a Locative Phrase (Spatial Deictic plus Noun Phrase) or as a place
name. The O argument is interpreted as the Goal of the motion.
(11) a. niaj hnub [cov me-nyuam]A mus [nram pas-dej]O

every day CLF.PL child go down pond
b. * . . . mus pas-dej
go pond
‘Every day the children go down to the pond.’
(12) pebA tuaj AustraliaO thaum xyoo 1979

3PL come Australia time year
‘We came to Australia in 1979.’
(13) nwsA mus [tim Asmeslivkas]O lawm

3SG go across America PRF
‘He has gone to America.’
productive svcs 111
The Goal verbs mus ‘go’, tuaj ‘come’, and los ‘come/return home’ have the addi-
tional function of conveying deictic information. That is, these verbs serve to
indicate not only the Goal of the motion, but also the direction of the motion
event with respect to the point of view of the speaker. If the motion is away
from the speaker, or from some place that is chosen as the point of reference,
then the deictic verb mus ‘go’ is used. Alternatively, for motion towards the
speaker or the chosen point of reference, tuaj ‘come’ is appropriate. However,
when the point of reference is the home of the subject, then motion towards
it is expressed by los ‘come/return home’, rather than simply by tuaj ‘come’.
Not surprisingly, the home is very frequently adopted as a point of reference
in White Hmong, which may explain the lexicalization of this particular Goal.
It is extremely rare to find a Motion SVC in White Hmong that does not
include one of these three verbs; deictic information is practically indispensible
in the expression of motion in this language. In cases in which no Noun Phrase or
Locative Phrase follows these verbs, then their sole function is to indicate deixis.
table 6 Verb types and order in Cotemporal SVCs
a. Manner of Motion b. Motion with respect to Ground
Atelic Atelic & Telic Telic
i. Transport ii. Locomotion i. Path ii. Source iii. ‘Return’ iv. Goal
(Transitive) (Intransitive) (Transitive (Transitive (Intransitive) (A=S
and A=S and S=O Ambitransitive)
Ambitransitive) Ambitransitive)
coj ‘take (s.o.) ya ‘fly’ raws ‘chase, pursue’ tawm ‘leave’ rov ‘return’ mus ‘go’
along’ dhia ‘jump/ taug ‘follow dim ‘escape’ tuaj ‘come’
cab ‘lead’ dance’ s.t. stationary’ sawv ‘get up/ los ‘come home’
thauj ‘transport/ nkag ‘crawl’ lawv ‘follow s.t. rise’
give a ride’ khiav ‘run’ moving’ thim ‘retreat’
nqa ‘carry’ mus kev ‘walk’ hla ‘move across’ poob ‘fall’
ris ‘carry on nce ‘ascend’
back’ nqis ‘descend’
kwv ‘carry on ncig ‘encircle/
shoulders’ move around’
(dej) nyab ‘(water)
rise’
112 CHAPTER 3
3.1.1.3 Summary and Illustration: Order of Verbs in a Motion SVC

The following examples of serial verb constructions expressing motion serve
to illustrate the way in which verbs from these various classes can be com-
bined according to the order shown above. Only the verbs in bold belong to the
Motion SVC. Other verbs, such as txog ‘arrive’ in (15), are part of different SVC
types, and will be explained in subsequent chapters.
(14) nwsA/S/S coj me-nyuamO khiav~khiav tuaj

3SG take.along child RDUP~run come
‘She fled, bringing her children along.’
(15) ces plisS/S dhia loo tuaj txog

CONJ wildcat jump without.hesitation come arrive
‘. . . and then the wildcat came leaping out of the forest.’ (DNH 436: 17)
(16) thov <nce rov tuaj . . .>COMP

request ascend return come
‘Please come back up . . .’ (DNH 25: 3)
(17) ces [lub nruas ntawd]S/S/S thiaj maj.mam poob rov los . . .
CONJ CLF drum nearby so slowly fall return come
‘And so then that drum fell gently back down to earth.’ (DNH 115: 5)
3.1.1.4 Principles Governing the Order of Verbs in a Motion SVC

Mention has already been made of the clear iconic relationship between the
occurrence of more than one verb within a single clause in a Motion SVC, and
the way in which these verbs serve to describe different facets of a single motion
event. For example, the first verb may describe the manner of the motion,
while the second indicates something about the way in which the motion is
related to the GROUND—the Path, Source, Goal, or previous motion event—or
the way in the motion is related to the point of view of the speaker—the deixis
and direction.
The iconicity principle can also be invoked to explain, in part, the order in
which verbs appear in a serial verb construction of this type. The occurrence
of the motion event is first established with the indication of the manner of
motion (Type a Motion verbs), and only then can the extent or direction of this
motion be indicated (Type b Motion verbs).
However, it must be pointed out once again that a serial verb construction
in White Hmong does not indicate a sequence of events at all, but rather serves
to convey different aspects of a single motion event that occur cotemporally.
This is shown explicitly in the examples (18)–(20) below:
productive svcs 113
(18) hnub peb nas.tsuagS/S fawb hnyuj.hnyo mus txog

day three rat search.around steadily go arrive
‘On the third day, a rat went steadily searching around all the way there.’
(DNH 124: 14)
not: . . . searched around steadily and then went on . . .
(19) lawvA/A coj [cov me-nyuam]O los tsevO

3PL take.along CLF.PL child come home
‘They brought the children home.’
not: ‘. . . took the children (somewhere) and then came home.’
(20) nwsA/S taug [ib sab ntug dej]O tuaj

3SG follow one side edge water come
‘He’s following the river bank (hither).’
not: ‘. . . follows the river bank and then comes here.’
Furthermore, there is one aspect of the order of verbs in a Motion SVC in White
Hmong that is clearly not determined iconically. In a real-world motion event,
the FIGURE must begin at the Source and then proceed along the Path towards
the Goal. In the report of such an event in White Hmong, however, the order
of these elements is: first Path, then Source, and then Goal. The occurrence of
predicates expressing Path before those expressing Source completely over-
rides iconic principles of ordering. However, this ordering only seems aber-
rant when we expect the elements in the Motion SVC to express an ordered
sequence of events.
(21) [cov Hmoob]A/A/A hla [dej Na.Koom]O dim [hauv

CLF.PL Hmong cross water Mekong escape inside
Nplog.teb]O mus Thai.tebO
Laos go Thailand
‘The Hmong crossed the Mekong River, escaping from Laos and going to
Thailand.’
not: ‘. . . crossed the Mekong River and then escaped from Laos and then
went to Thailand.’
In example (21) above, the reference to the Path (hla dej Na Koom ‘cross the
Mekong River’) precedes the reference to the Source (dim hauv Nplog teb
‘escape (from) Laos’). The interpretation is cotemporal: crossing the river is
what constitutes the escape from Laos; neither one occurs before the other in
time. The predicates referring to the Source (dim hauv Nplog teb ‘escape (from)
Laos’) and the Goal (mus Thai teb ‘go to Thailand’), on the other hand, are actu-
114 CHAPTER 3
ally ordered in a way that could be interpreted as iconic. However, clearly these
predicates still represent two different aspects of the same event: one indicat-
ing the point of origin of the motion, and the other the destination. When they
occur in a Motion SVC, the temporal relationship between these predicates
should also be interpreted as co-temporal—‘escaping from Laos and going to
Thailand’—rather than as sequential—‘they escaped from Laos and then went
to Thailand’.
In fact, the ordering of predicates expressing Path before those expressing
Source is determined by the following basic principle: relative degree of telic-
ity determines the order of verbs in a Cotemporal Motion SVC. Thus the atelic
Manner of Motion verbs, such as ya ‘fly’ and nqa ‘carry’, invariably precede the
telic motion verbs expressing Source and Goal and direction, such as mus ‘go’
and tuaj ‘come’.
If the order is reversed, and a telic motion verb precedes an atelic one,
then the sequence cannot be interpreted as a Motion SVC, expressing a sin-
gle motion event. Instead, the verbs are understood to describe two distinct
events, the second indicating the purpose of the first (see Chapter 5, §5.5). This
is illustrated in the example (22):
(22) [tus tub]S mus <da-dej [nram hav-dej]>PURP

CLF boy go immerse-water down valley-water
‘The boy went to bathe down in the river valley.’
It is for this reason that predicates expressing Path precede those expressing
Source in a Motion SVC. Source verbs in White Hmong are inherently telic, and
a Purposive interpretation would arise if a Source verb were to appear before
a Path verb.
It must be acknowledged that not all Path verbs exhibit the same degree of
telicity. The inherent atelicity of Path verbs such as taug ‘follow (a stationary
object)’ and lawv ‘follow (a moving object)’ is quite apparent. However, Path
verbs like nce ‘ascend’, nqis ‘descend’ and hla ‘cross, encircle, pass’ are actually
telic, in the sense that they do have an intrinsic goal. The examples below dem-
onstrate that these Path verbs are, in fact, quite low in telicity. The attainment
of their goal is defeasible:
(23) kuvA tau nce roobO, tab.sis kuvA tsis txog

1SG ACHV ascend mountain but 1SG NEG arrive
[hauv saud]O
inside top
‘I managed to climb the mountain, but I didn’t reach the top.’
productive svcs 115
(24) nwsA tau hla [tus dej]O, tab.sis nwsA tsis txog
3SG ACHV cross CLF water but 3SG NEG arrive
[sab tid]O
side beyond
‘She did cross the river, but she didn’t reach the other side.’
Note the grammaticality of the White Hmong example (24) above, in spite of
the oddness of the English translation. Perhaps a more accurate gloss for the
atelic verb hla would be something like ‘proceed across’.
This issue of the defeasibility of many accomplishment predicates in White
Hmong is discussed further in regard to Attainment SVCs (§3.3 below). The
point to be made here is that relative degree of telicity seems to be the key fac-
tor in the order of verbs in a Cotemporal Motion SVC.
3.1.2 Cotemporal SVCs Involving Actions

3.1.2.1 Action and Motion
In addition to those involving two or more motion verbs, Cotemporal SVCs
may also consist of an atelic action verb accompanied by a motion verb (or by
a series of motion verbs). The action is understood to take place at the same
time as the motion. For example:
(25) nqa [ib lub yeeb.thooj]O haus puj.pauv tuaj

carry one CLF pipe inhale puff.puff come
‘. . . bringing a pipe along and puffing on it.’ (DNH 124: 9)
(26) ces [nplej thiab pob.kws]S/S qw zom.zaws los

CONJ rice and corn shout together come.home
‘And then the rice and the corn came home, all shouting at once.’
(DNH 126: 28)
(27) hnub peb [nas tsuag]S/S fawb hnyuj hnyo mus txog
day three rat search.about steadily go arrive
‘On the third day a rat went searching steadily about all the way there.’
(DNH 124: 14)
3.1.2.2 Action and Stance

Instead of a verb of motion, a predicate expressing stance/posture may also
appear with an atelic motion verb, in a Cotemporal SVC. In this case the action
verb most commonly appears as the second verb in the construction.
116 CHAPTER 3
(28) nwsA/A ib [ntawm lub phab.ntsa]O haus yuam.yeebO

lean lean nearby CLF wall inhale tobacco
‘S/he leaned against the wall nearby, smoking.’
(29) kuvA/A nyiam pw [saum txaj]O nyeem ntawvO

1SG like lie top bed read writing
‘I like lying in bed reading.’
(30) lawvS/S zaum-tsaws tham ib hmo nkaus

3PL sit chat one night INTNS
‘They sat chatting all night long.’
(31) nwsS/A sawv-ntsug noj movO

3SG stand eat rice
‘He ate standing up.’
In example (30) above, the word zaum-tsaws ‘sit’ is actually lexicalized from
an Attainment SVC (see §3.3 below) composed of zaum ‘sit down’ and tsaws
‘alight’. Likewise, in (31) above, sawv ntsug ‘stand’ (‘be in a standing position’) is
lexicalized from the Attainment SVC made up from sawv ‘stand up’ and ntsug
‘be upright’. These SVCs refer to stable states and are thus atelic: zaum tsaws ‘be
in a sitting position’, sawv ntsug ‘be in a standing position’.
The verbs zaum and sawv by themselves, however, are actually Change of
Stance verbs meaning ‘sit down’ and ‘stand up’ respectively. As such they are
clearly telic, and a sequence in which they occur alone as the initial verb would
have a purposive rather than a cotemporal interpretation (see Chapter 5, §5.5
for details):
(32) lawvS zaum <tham>PURP

3PL sit chat
‘They sat down to chat.’
(33) nwsS sawv <noj movO>PURP

3SG stand eat rice
‘He stood up to eat.’
3.1.2.3 Cotemporal Actions

When an action occurs cotemporally with a motion or with a posture, as
illustrated above, a speaker of White Hmong can choose to portray the two
as separate facets of a single event, and use a serial construction to do so.
productive svcs 117
However, when two action verbs are involved, it is far more likely that they will
be perceived, and thus portrayed, as two distinct events. In this case a serial
construction would not be appropriate.
There are, however, two patterns in which cotemporal actions are presented
as constituting two facets of a single event. The first of these is illustrate in
examples (34) and (35) below:
(34) kuvA/A mam zov ntsia <kojS mus>COMP

1SG so.then guard watch 2SG go
‘So then I’ll watch over you going.’
(35) [cov poj-niam]A/A npaj zaubO npaj movO

CLF.PL woman prepare vegetables prepare rice
‘The women prepare the food.’
In this type of Cotemporal Action SVC, the two juncts are close in meaning.
There is quite often even repetition of a single verb with different objects, as
in example (35). This produces what is known as an ‘Elaborate Expression’, a
euphonious pairing with an ABAC pattern, described in detail in Chapter 5,
§5.1. The effect of this similarity in meaning and partial repetition relates to
one of the key functions of Cotemporal SVCs in general: to focus on the action
of the subject and portray it in a lively and vivid way.
The same motivation can be seen in the second type of Cotemporal Action
SVCs, shown in examples (36) and (37):
(36) ces NtxawmS dhia loo los nres [hauv

and PN jump INTNS CONJ stand.upright inside
plawv]O quaj cem quaj cem <hais.tias . . .
middle cry scold cry scold COMP
‘And Nzeu leapt up and stood in the middle, crying [and] scolding [and]
crying [and] scolding, saying . . .’
(37) nwsS/A dhia tshov qeejO

3SG dance blow bamboo.pipes
‘He dances (while) playing the pipes.’
Although these examples might both seem as if they involve two quite dis-
tinct actions, they actually represent what is clearly perceived as a single event.
In (36), the young woman Nzue scolds as she cries and cries as she scolds,
her emotions tumbling out uncontrollably. In (37), the playing of the qeej
118 CHAPTER 3
‘bamboo pipes’ and the dancing together constitute the key elements of the
performance: not two actions but one. Whenever the qeej is played, the per-
former’s feet and body move and sway in time to the music. The assertion that
this is thought of as a single action is supported by the fact that the verb phrase
dhia qeej ‘dance the pipes’ can be used.
When two cotemporal actions are seen as more clearly distinct, the use of a
serial construction is completely unacceptable:
(38) a. nwsS dhia thiab mloog nkaujO

3SG dance and listen song
‘He dances (and) listens to music.’
b. * nwsS dhia mloog nkaujO

3SG dance listen song
The unacceptability of example (38)b illustrates the fact that this type of
Cotemporal SVC is quite limited in productivity. While still subject to syntac-
tic analysis, constructions like these are not dissimilar from lexicalized com-
pounds. Only actions that are commonly and closely associated in the context
of Hmong culture, those that are culturally ‘recognisable event types’, yield
acceptable combinations in this construction (Aikhenvald, 2006b, pp. 11–12;
Bruce, 1988; Durie, 1997, p. 322; Jarkey, 2010, pp. 128–131).
3.1.3 Nexus and Level of Juncture in Cotemporal SVCs

3.1.3.1 Level of Juncture
As described above (Chapter 2, §2.4), the clause can be thought of as consisting
of three basic ‘layers’: the nucleus (the predicate), the core (the nucleus plus
the core arguments of the predicate), and the periphery (all other arguments
such as time phrases, ‘outer’ locatives and beneficiaries). Nuclear juncture
occurs when two nuclei are linked together, sharing all core and peripheral
arguments. Core juncture generally involves the linking of two clauses, with
the point of intersection being within the core. Clausal juncture also involves
the linking of two whole clauses. If intersection occurs at all at this level, it is
between peripheral arguments only.
Verb serialization occurs at either the nuclear or core level. When it occurs
at the core level, a special case of core juncture is involved: instead of linkage
between two clauses (core and periphery), core level serialization involves the
linkage of two cores; all peripheral arguments must be shared.
What level of juncture is involved in the Cotemporal Serial Verb
Construction? It is quite clear from the examples given above that the level of
productive svcs 119
juncture in this type of SVC is not nuclear. In a nuclear juncture the two verbs
would appear together, with all arguments preceding and/or following. In a
Cotemporal SVC in White Hmong, however, a NP expressing the object can
appear immediately after one verb, intervening between it and the next.
(39) nwsA/S kwv [lub raj-dej]O los

3SG carry.on.shoulder CLF tube-water come.home
‘He carried the water pipe home.’
(40) taug taug dejO nqis havO mus

follow follow water descend valley go
‘. . . follow the river down the valley (away).’ (DNH 183: 3)
Core locative arguments, specifying Location and Source, can also appear
between the verbs in a Cotemporal SVC:
(41) nwsa/A sawv-ntsug [ntawm qhov-cub]o ua [mov <noj>REL]O

3SG stand nearby kitchen make rice eat
‘She stood in the kitchen making rice to eat.’
(42) [cov tub.rog]A/S tim [ntawm [lub kuj.ha]]O los

CLF.PL soldier retreat nearby CLF valley come.home
‘The soldiers retreated homeward [from] the valley.’
The appearance of arguments of the first verb between the two juncts in
Cotemporal SVCs demonstrates that the level of juncture in these construc-
tions is not nuclear.
The fact that the two juncts share a common core argument—the subject—
indicates that the level of juncture in these constructions must be core. This
sharing of just one core argument is typical of core layer juncture; in nuclear
juncture all core arguments are shared; in clausal juncture, none.
To prove that the A/S argument is really shared by the two juncts, rather
than simply coreferential between them (with zero anaphora in the second
junct), it is necessary to show that a coreferential argument (e.g. an overt
pronoun) cannot be inserted preverbally in the second junct (see Chapter 2,
§2.4.1.3). Examples (43) below and (44) below, in which a pronoun has been
inserted, are unacceptable (unless there is a significant pause between the two
juncts (see below)).
120 CHAPTER 3
(43) ?? nwsA kwv [lub raj-dej]O nwsS los

3SG carry.on.shoulder CLF tube-water 3SG come.home
thiab
also
for: ‘He carried the water pipe home.’ (Compare example (39))
(44) ?? ces pliss dhia loo nwsS/S tuaj txog

CONJ wildcat jump without.hesitation SG come arrive
for: ‘. . . and then the wildcat came leaping out of the forest.’ (Compare
example (15))
Note that O arguments need not be shared. As shown above, a NP or LP argu-

ment expressing the O argument of the first verb may intervene between the
two verbs. However, as we would expect, peripheral arguments—such as tem-
poral and ‘outer locative’ NPs—must be shared; they can not appear between
the verbs in the Cotemporal Serial Construction:
(45) a. nag.hmo [cov tub.rog]A/S tim los

yesterday CLF.PL soldier retreat come.home
‘Yesterday the soldiers retreated homeward.’
b. * [cov tub.rog]A/S tim nag.hmo los

CLF.PL soldier retreat yesterday come.home
The (b) example above would only be acceptable if a distinct pause were to
occur after the first verb, indicating that the two juncts were serving to express
distinct events: ‘The soldiers retreated, and yesterday (they) came home’. This
would be an example not of the Cotemporal Serial Construction, but of juxta-
posed clauses (see Chapter 5, §5.2). The Cotemporal Serial Construction, on
the other hand, involves not simply core juncture, but core juncture of a spe-
cial type: the linkage of two cores within a single periphery.
3.1.3.2 Nexus
The term ‘nexus’ refers to the type of relationship that exists between two
juncts. The concept is explained fully in Chapter 2, (§2.4.1.4), and briefly reiter-
ated here. A junct that is ‘subordinate’ to another is embedded as either a core
or a peripheral argument of that other junct. When two juncts are in a ‘coordi-
nate’ relationship, neither one is embedded in the other, and they are indepen-
dent with regard to operators at the level of juncture. The term ‘cosubordinate’
means that neither junct is embedded in the other, but that they are mutually
dependent: all operators are shared at the level of juncture.
productive svcs 121
Fuller (1985, pp. 107–116) shows that subordinate elements, such as NP

arguments, as well as subordinate location and time phrases and clauses, can
appear sentence initially as topics in White Hmong. Here are some examples
she gives:
(46) [cov tub.rog] mas yog Nyab.LajCC coob xwb

CLF.PL soldier TOP COP Vietnam be(c0me).many only
‘The soldiers, (they) were mostly Vietnamese.’ (Fuller, 1985, p. 109)
(47) thaum ntawd mas lawv ua-kam rau Suav xwb nawb
time nearby TOP PL do-deal to Chinese only IP
‘At that time, they did business with the Chinese only.’ (Fuller, 1985, p. 110)
(48) yav <kuvS hlob kiag>REL mas [peb lub zos]

time.period 1SG grow completely TOP 1PL CLF village
mas yog [peb tib pawg kwv-tij ]CC xwb
TOP COP 1PL single bunch clan.brothers only
‘When I grew up, our village, it was just us cousins.’3 (Fuller, 1985, p. 116)
The second junct of a Cotemporal Serial Construction, on the other hand, can-
not appear sentence initially as topic with the same meaning as in the SVC:
(49) a. nwsA/A nqa [nws rab hneev]O los tsevO

3SG carry 3SG CLF crossbow come home
‘He carried his crossbow home.’
b. * los tsev mas, nws nqa [nws rab hneev]

come home TOP 3SG carry 3SG CLF crossbow
(50) a. lawvS/A khiav mus [nram hav-dej]O

3PL run go down valley-water
‘They ran down to the river valley.’
b. mus [nram hav-dej]O mas, lawvS khiav

go down valley-water TOP 3PL run
‘(They) went down to the river valley, (and then) they ran.’
not: ‘Going down to the river valley, they were running.’
3 This example has two topics: first the temporal adverbial clause yav kuv hlob kiag ‘the time I
grew up’, and next the NP peb lub zos ‘our village’. This latter NP topic is not an argument in
the clause that follows.
122 CHAPTER 3
The second example above shows that such a sequence is sometimes gram-
matical, but that the interpretation is sequential rather than Cotemporal, and
the structure involves juxtaposed clauses. It is quite clear that the Cotemporal
Serial Constructions does not involve subordinate nexus.
The placement of certain adverbial expressions provides some further evi-
dence for the assertion that Cotemporal SVCs do not involve subordination,
at least not NP-type subordination. (The particular relevance of this evidence
will be seen when Cotemporal SVCs are compared in this respect with the
Purpose Construction (see Chapter 5, §5.5.1.1). A verb used adverbially, like
ceev ‘be fast’, is able to appear, among other places, directly after a verb and its
object, as in the first example below, but not between the verb and its object,
as in the second:
(51) a. fawb nrojO ceev~ceev mus . . .

search.around vegetation RDUP~ be.fast go
‘. . . go searching quickly around the grass . . .’
b. * fawb ceev~ceev nroj mus . . .

search.around RDUP~ be.fast vegetation go
Such adverbial expressions can readily appear between the verbs of a

Cotemporal SVC, indicating that the second does not function like an NP argu-
ment of the first:
(52) nce ceev~ceev mus . . .

ascend RDUP~ be.fast go
‘. . . climb up quickly (away from the speaker) . . .’
As the level of juncture in SVCs of this type has been shown to be core, then it
is the scope of the core-layer modal operators that will determine which of the
two non-embedded nexus types is involved: whether the juncts are dependent
at this level, and therefore cosubordinate, or independent, and therefore coor-
dinate. Observe the following examples:
(53) yuav-tau rov mus <qhib [lub qhov-rooj]O dua>PURP

must return go open CLF door again
‘. . . must go back to open the door again.’
(54) nwsS/A txawj dhia tshov qeejO

3SG know.how dance blow bamboo.pipes
‘He can dance [while] playing the pipes.’
productive svcs 123
(55) [cov Hmoob]A/A/A xav hla [tus dej]O dim [hauv

CLF Hmong want cross CLF water escape inside
Nplog.teb]O mus [tim Thaib.teb]O
Laos go across Thailand
‘The Hmong wanted to cross the river, escaping from Laos and going to
Thailand.’
In the three examples above, the modals yuav-tau (obligation), txawj (ability),
and xav (desire) all have scope over the entire SVC, not just over the predi-
cate that immediately follows them. It would be self-contradictory to make the
assertion in, say example (55) above, and then continue:
(56) tab.sis lawvA tsis xav mus Thaib.tebO

but 3PL NEG want go Thailand
‘. . . but they didn’t want to go to Thailand.’
The fact that the modal operators have scope over all cores in a Cotemporal
SVC indicates that these juncts are dependent at the level of juncture. The rela-
tionship is thus one of core cosubordination.
Although they cannot be independently modified by core layer modal oper-
ators, juncts in a core cosubordinate relationship can, and frequently are, inde-
pendently modified by nuclear layer modifiers, such as onomatopoeic adverbs:
(57) hnub peb nas.tsuagS/S/S fawb hnyuj.hnyo mus txog

day three rat search.around steadily go arrive
‘On the third day a rat went searching steadily around all the way there.’
(DNH 124: 14)
(58) ces [niag tsov]A/S txawm nce kig.kuag~kig.kuag mus

CONJ great tiger then ascend RDUP~grrr go
‘Then that big ol’ tiger went up (the ladder) growling.’ (DNH 399: 16)
(59) nkag qhawv.qho lawv qab

crawl INTS follow behind
‘to follow (someone) unwillingly’ (Heimbach 1979, p. 151)
(60) dejS/S nyab zuj.zuj tuaj

water rise gradually come
‘The water gradually rose up.’ (DNH 115: 2)
124 CHAPTER 3
The appearance of adverbs that modify the nuclear layer, such as those illus-
trated above, between the verbs in a Cotemporal SVC, is quite in keeping with
the conclusion that these verbs are in a core cosubordinate relationship.
3.2 Cause-Effect Serial Verb Constructions
The majority of the discussion in the literature about serial verb constructions
concerns those in which the shared argument is interpreted as either A or S of
all the verbs in the series (where A refers to the subject of a transitive clause
and S refers to the single participant in an intransitive clause). All examples of
the Cotemporal Serial Construction in White Hmong, discussed in §3.1 above,
are of this type. In the examples of Cotemporal SVCs below, the functions of
the shared argument in relation to each of the serial verbs is shown in paren-
theses after the translation:
(61) [tus noog]S/S ya mus lawm

CLF bird fly go PRF
‘The bird has flown away.’ (S=S)
(62) nwsA/S nqa [nws rab hneev]O los

3SG carry 3SG CLF crossbow come.home
‘He brought his crossbow home.’ (A=S)
(63) nwsA/A ib [ntawm lub phab.ntsa]O haus yuam.yeebO

3SG lean nearby CLF wall inhale tobacco
‘S/he leaned against the wall nearby, smoking.’ (A=A)
However, in some languages there are serial verb constructions in which the
shared argument is interpreted as the O argument of the first verb (where O
refers to the object in a transitive clause), but the S argument of the second.
Below are some examples.
(64) Yoruba (Benue-Congo) (Bamgboṣe, 1974, p. 23)

OlúA ti [ọmọ náà]O/S subú
PN pushed child the fall
‘Olu pushed the child (so that the child) fell.’ (O=S)
productive svcs 125

lomA phát [lǎŋkhaa baân]O/S pliw
wind blow roof house blow.away
‘The wind blew the house roof away.’ (O=S)
(66) Paamese (Oceanic) (Crowley, 1987, p. 43)

kailA amuas vuasO/S emat
3PL 3PL-REAL-hit pig 3SG-REAL-die
‘They hit the pig (so that it) died.’ (O=S)
(67) Lao (Tai-Kadai) (Bohnemeyer et al., 2011, p. 55)

kuu3A thup1 còòk5O/S tèèk5
1SG smash glass break
‘I smashed the glass (so that it) broke.’

yog kojA ntaus [lub zeb-zag no]O/S tawg no.ces
COP 2SG hit CLF stone-dragon this shatterINTR CONJ
‘If you hit this dragon-stone (so that it) shatters, then . . .’ (O=S)
In all of the examples given above, although the shared argument functions as
the O argument of the verb that precedes it and the S argument of the one that
follows, it is in the role of undergoer to both of these verbs. This is the case in
most examples of this type of serialization in the literature. However, in White
Hmong there are sentences of this type in which the shared NP is interpreted
as the undergoer of the verb that precedes it but as the actor of the verb that
follows:
(69) nwsA ntaus [tus dev]O/S khiav kiag

3SG hit CLF dog flee completely
‘He hit the dog (so that it) ran away.’4
4 Some authors give sentences such as the following as examples of serial verb constructions
with “causative” meaning:
(i) Akan (Schachter, 1974, p. 261)
Amma frɛɛ Kofi baae
Amma called Kofi came
‘Amma called Kofi (to come) in.’
(Also: ‘Amma called Kofi and came in’, i.e. S/A=S/A interpretation also possible.)
This sentence is apparently very similar to the White Hmong examples given above. Judging
from the translations Schachter gives for this type of example, however, the difference seems
to be that, while the White Hmong structure is truly causative, the Akan sentence is actually
126 CHAPTER 3
(70) Nraug.NpliamA rov tib chua Muam.Nkauj.LugO/S

PN return suddenly yank PN
dhau plaws rau tom tag
cross.over suddenly to across finish
‘Nrau Nplia again suddenly yanked Mua Nkau Lu across there.’
Lit.: ‘. . . (so that she) crossed over there).’ (DNH 429: 46)
Note, however, that V2 in these examples is still an intransitive verb; the shared
argument can still be characterised as ‘O=S’, even when S—the single argu-
ment of that intransitive verb—is an actor rather than an undergoer.
Foley and Olson (1985) claim that, in some languages, some serial verb con-
structions may be ambiguous between a variety of readings, including cause-
effect, as in this example from Thai:
(71) John khàp rót chon khwaay taay

PN drive car collide buffalo die
‘John drove the car, collided with a buffalo, and
i) it died.’ ii) the car stalled.’ iii) he died.’ (Foley & Olson, 1985, p. 26)
Clearly the ambiguity in examples like this depends on a number of factors,

such as whether the first NP as well as the shared NP can, in fact, felicitously
be predicated by the final verb in the construction, and whether the language
involved allows A=S serial verb constructions (or juxtaposed clauses) with an
appropriate reading—a sequential reading in the case of the examples above.
White Hmong does not allow a simple sequential reading for serial verb
constructions and, as a result, ambiguity of the type illustrated above for Thai
does not occur in this language. Thus, in the White Hmong example below, the
only possible in-terpretation is the cause-effect interpretation.
(72) kuvA xa [ib tsab ntawv]O/S mus rau Sydney

1SG send one CLF writing go to Sydney
‘I sent a letter to Sydney.’
not: ‘I sent a letter and (then) (I) went to Sydney.’
simply jussive: Amma called Kofi to come, but we do not know whether Kofi actually came as
a result, or not. Thus the calling and the coming in this example are not subject to the same
truth-value, and cannot be regarded as aspects of a single proposition. (See Chapter 5, §5.3.1
below for detailed discussion of similar sentences in White Hmong.)
productive svcs 127
While the sharing of A and S arguments occurs in all languages that exhibit
verb serialization, the sharing of O and S arguments—i.e. cause-effect
serialization—occurs in only some (Aikhenvald, 2006b, p. 14; Foley & Olson,
1985, p. 42). Some authors maintain that, in those languages in which it does
occur, this type of SVC constitutes a significant proportion of all serial verb
constructions in the language (Crowley, 1987, p. 47; Foley & Olson, 1985, p. 25).
However, others suggest that the sharing of O and S arguments occurs in only
exceptional cases (Jansen, Koopman, and Muysken (1978, p. 155), cited in
Sebba (1987, p. 86)). Where they are mentioned at all, the literature is divided
as to how sentences such as these should be analysed, and as to whether or not
they should be regarded as involving verb serialization at all.
3.2.1 Constraints on the Verbs Appearing in Cause-Effect SVCs

3.2.1.1 Constraints Described in the Literature
Most authors who address the question at all agree that the interpretation
of Cause-Effect SVCs is lexically triggered in some sense. That is, a serial verb
construction may be given a cause-effect reading if, and only if, certain lexical
items occur. However, different authors have put forward quite different ideas
as to whether the constraints of occurrence are actually on V1, or on V2, or on
both verbs in the construction.
Schachter (1974) regards the first verb in the construction as the one that
is responsible for the cause-effect interpretation. However, he does not make
any attempt to generalise as to exactly what features of a verb in this position
give rise to this interpretation. In fact, Schachter suggests that the possible
‘cause-effect’ interpretation of a verb should simply be indicated as part of its
lexical entry.
To account for the interpretation of these and other serial verb construc-
tions, Schachter suggests that there be a general rule whereby the first NP in a
serial construction is interpreted as the ‘semantic subject’ of all the verbs, but
that some individual verbs are marked as disallowing this rule. These verbs
carry certain syntactic and semantic lexical features indicating that their
object be understood as the ‘semantic subject’ of a verb that follows serially.
Still other verbs would be marked as allowing both possible interpretations.
Thepkanjana (1986) regards a small number of verbs in Thai as having an
idiomatic resultative or causative interpretation when occurring initially in
(what she considers to be) serial verb constructions, and thus requiring their
resultative/causative interpretation to be indicated in the lexicon. The relevant
verbs are the resultative verb tham ‘make/do’ and the causative verbs hây ‘give’
and tham hây (tham ‘make/do’ + hây ‘give’). These verbs, Thepkanjana argues,
128 CHAPTER 3
predicate not just the following NP but the entire complement that follows
them.5 For example (p. 21):
(73) Sùrii tham kææ̂ w tææk

PN do/make glass break
‘Suri broke a glass.’
This type of construction would be termed ‘complement clause serialization’

in the careful distinction of serialization types outlined by Aikhenvald (2006b,
pp. 14–20). All other ‘Resultative’ SVCs (i.e. similar to what are here termed
Cause-Effect SVCs) in Thai, according to Thepkanjana, simply require the
first verb to be transitive, and any further restrictions are on the final verb (see
below).
Amongst those authors who refer to constraints on the final verb in the con-
struction, there is almost universal agreement that this verb must be intransi-
tive (Bruce, 1988; Durie, 1988; Foley & Olson, 1985; Sebba, 1987; Thepkanjana,
1986).6 Foley and Olson further specify that it must be ‘stative or process intran-
sitive’. Thepkanjana argues for Thai that the final verb in a tham Resultative
construction (and presumably in all other ‘Resultative’ (here Cause-Effect)
SVCs) must be an intransitive verb that indicates either a change in status or
a coming into existence. It may be stative (e.g. ‘be flat’), inchoative (e.g. ‘die’,
‘open’) or activity (e.g. ‘blow away’),7 as long as it is a predicate that can be used
to indicate some kind of change from the original state of the shared NP.
3.2.1.2 Constraints on V1 and V2 in Cause-Effect SVCs in White Hmong

It is neither necessary nor revealing to mark all the transitive verbs that can
appear as V1 in a Cause-Effect SVC as ‘Causative verbs’ in the lexicon, as
Schachter (1974) suggests. It is quite possible to identify precisely which kinds
of transitive verbs are candidates for this position. However, the identification
5 For precisely this reason, constructions similar to this in White Hmong would not be regarded
as serial verb constructions in this analysis. See below (Chapter 5, §5.3.2).
6 (Solnit, 2006, p. 151) gives an example from Eastern Kayah Li (Karen, Tibeto-Burman) that he
refers to as ‘cause-effect (Directive)’, in which the second verb transitive. Because a ‘directive’
verb is involved, this type may be more similar to the White Hmong pattern discussed in
Chapter 5, §5.3.1.
7 Thepkanjana uses the term ‘activity’, presumably in the sense of Vendler (1967) (see the
Introduction to Chapter 3), but the example she gives, pliw ‘blow away’, appears to be more
like an achievement verb.
productive svcs 129
of these verbs alone will not be adequate: it is not the case that a Cause-Effect
SVC will always result when one of these verbs occurs as V1 in a series of verbs.
In the two examples below the same V1 occurs, but only the first example is a
Cause-Effect SVC:
(74) lawvA ua [lub tsev]O/S siab~siab

3PL make CLF house RDUP~be(come).tall
‘They built the house tall.’
(75) lawvA/A ua tau [lub tsev]O/O lawm

3PL make get CLF house PRF
‘They have built the house.’
In a similar way, it is clearly not adequate to refer only to the second verb in a
Cause-Effect SVC as the factor that determines the cause-effect interpretation.
In examples (76) and (77) the second verb is the same, but only example (76)
can be interpreted as involving a Cause-Effect Serial Construction.
(76) yusA/A cub movO/S siav

INDF steam rice be(come).cooked
‘One steams rice (so that it is) cooked.’ (KNH 16)
(77) lawvA noj nqaijO nyoos, tab.sis pebA noj nqaijO siav
3PL eat meat be.raw but 1PL eat meat be.cooked
The conclusion to be drawn from examples such as these is that it is a combi-

nation of the semantic properties of both V1 and V2 that results in the cause-
effect interpretation.
The constraints on V2 in a Cause-Effect SVC in White Hmong seem to concur
roughly with those given for Thai by Thepkanjana (1986). That is, any intransi-
tive verb can occur, as long as it can be used to indicate some kind of change
in the object of V1: change of state, change of location, or a kind of change
not included by Thepkanjana—change of action. Thus both state verbs and
activity verbs receive an inchoative interpretation when they occur as V2 in
this construction: state verbs signal a change of state, and activity verbs signal
the inception of a new activity. State verbs like nyoos ‘be fresh/raw’ and tshiab
‘be new’ do not appear in SVCs with cause-effect meaning because, in the real
world, states such as these do not occur as the result of change.
Actually it is not strictly true that V2 in a Cause-Effect SVC must be intransi-
tive in White Hmong. In fact, it may be structurally transitive. That is, motion
130 CHAPTER 3
verbs can appear that take Source, Path, or Goal arguments as objects (see
above Chapter 1, §1.6.5.3). In most of the examples below, these verbs do
appear as intransitive verbs, either with no object at all, or with something like
a ‘prepositional’ object (e.g. as the object of rau ‘to/for’, see Chapter 4 §4.2.3.4).
However, it is certainly possible to include the direct object of V2 in this con-
struction, as shown in examples (78) and (79) below:
(78) [tus Tsov tus tw]A tib xyob [tus Qav]O/A dhau
CLF tiger CLF tail sharply flick CLF frog pass.over
plaws [tus Tsov]O rau [pem hauv ntej] lawm
suddenly CLF tiger to up in front PRF
‘. . . Tiger’s tail had sharply flicked Frog suddenly past Tiger (and) up
ahead.’ (LPV 2)
(79) nwsA xyeeb [nws tus muam]O/A poob roojO

3SG push 3SG CLF sister fall table
‘He pushed his sister (so that she) fell (off) the table.’
In example (78), the NP tus Tsov ‘Tiger’ is the direct object of the verb dhau
‘pass over’, and represents the Path. In (79), rooj ‘table’ is the direct object of the
verb poob ‘fall’, and represents the Source.
Examples such as these indicate that the constraint noted by many authors
(see above), that the second verb in a Cause-Effect SVC must be intransitive,
needs to be modified in the case of White Hmong. If a transitive structure does
appear, however, it is important to note that it is always one of a marked kind.
The actor of a verb like dhau ‘pass over’ or poob ‘fall’, far from being a typical
actor, such as an Agent, actually has the semantic role of Theme, a semantic
role which is actually more likely to occur as an undergoer than as an Actor
((Foley & Van Valin, 1984, p. 59; Van Valin, 2005, pp. 61–62; Van Valin & LaPolla,
1997, p. 146). In other words, the two arguments of a transitive verb of this type
are certainly not in the roles typically assigned to the arguments of a transi-
tive predicate (Hopper & Thompson, 1980). Note, however, that these verbs
do convey the notion of change (change of location in this case)—the crucial
property of V2 in a Cause-Effect SVC.
As for the first verb in Cause-Effect SVCs in White Hmong, it is not true
that simply any transitive verb can occur, as Thepkanjana claims for the
‘Resultative’ SVC in Thai. The transitive verbs that do occur as V1 in this con-
struction in White Hmong are archetypal transitive verbs: their A arguments
are Agents, their O arguments, Patients or Themes, and they express some kind
of impingement of A on O—some kind of action that will cause a change. The
productive svcs 131
verbs that occur in this V1 slot can be classified into three types: Affective verbs,
Effective verbs, and Transfer verbs. Let us describe the kinds of Cause-Effect
SVC that occur with each of these types of predicate as V1.
a Affective Verbs as V1
An Affective verb describes an action, performed by an Agent, which brings
about some physical change in (i.e. ‘affects’) a Patient. Actions of this kind
clearly involve direct causation: the change in the Patient is a direct result of
the action of the Agent.
(80) yuav-tau hlais noob.qesO/S tawm . . .

must slice testicles leaveINTR
‘. . . (they) had to slice (its) [the hog’s] testicles out.’ (DNH 9: 46)
(81) pebA vau <tsuam xwb tuag>PURP lawm lauv

1PL fall.down crush only die RF IP
‘. . . we will have fallen down (to) just crush (him) to death.’ (DNH 125: 16)
(82) muab pob.kwsO/O/S ntaus~ntaus dam tag

take corn RDUP~ hit break.offINTR finish
‘. . . beat the corn (so that it) breaks off.’ (DNH 123: 4)
(83) HmoobA/A txawm muab nrojO txiav caj.dabO/S/S poob cuag

Hmong then take weed slice neck fall reach
li cas
like how
‘The Hmong then took the weeds (and) sliced their necks (so that they)
fell right down like anything.’ (DNH 126: 26)
(84) nwsA ntaus [tus dev]O/S khiav/quaj kiag

3SG hit CLF dog flee/cry completely
‘He hit the dog (so that it) ran away/whined.’
It is interesting to note that even archetypal transitive verbs in White Hmong,

such as those in the V1 slot in the examples above, do not actually fully incor-
porate the notion of ‘affect’. This notion is only unambiguously conveyed by
the entire SVCs above. Thus, for example, there is no exact equivalent in White
Hmong for an English verb like kill, which implies not only that the subject
does something ‘murderous’, but also that the object dies as a result. The White
Hmong verb that is semantically closest to English kill is tua. This verb would,
132 CHAPTER 3
indeed, be best translated as kill in certain contexts, such as in the following

example:
(85) [[kuv txiv] lawv]A twb coj tau [cuaj tug

1SG father 3PL REAL take.along get nine CLF(t.c.)
npua]O tuaj <tua>PURP
pig come
‘My father’n them succeeded in bringing along nine pigs to kill . . .’ (TNN 4)
In fact, the verb tua actually means something like ‘shoot at/stab at/perform
some murderous deed towards’. It is only well translated by kill in the example
above because it is reasonable to assume, in this context, that the deed will be
successful. However, the successful outcome is defeasible; it is not a part of the
semantic content of the verb tua itself, as shown by the following examples:
(86) nwsA tua liabO tsis raug

3SG monkey NEG hit.the.mark
‘He shot at monkeys (but) didn’t hit any.’
(87) lawvA tua nwsO tab.sis nwsS tsis tuag

3PL 3SG but 3SG NEG die
‘They shot him, but he didn’t die.’
The successful outcome of such an act can, however, be incontrovertibly con-

veyed by the use of a Cause-Effect SVC:8
(88) lawvA tua nwsO/S tuag

3PL 3SG die
‘They shot him dead.’
The kind of verbs that appear as V2, in a Cause-Effect SVC with an ‘Affective’
V1, are verbs describing change of location (e.g. tawm ‘leave’, poob ‘fall’, khaiv
‘flee’) change of state (e.g. tuag ‘die’, dam ‘break off’ siav be/become cooked’),
or inception of activity (e.g. quaj ‘cry’ khiav ‘run’). A verb of this type serves to
indicate the change or the new activity that comes about as a direct result of
the action described by the first verb in the construction.
8 See also §3.3.1, example (122) for an example of an Attainment SVC also serving to convey a
successful outcome for the action described by the verb tua.
productive svcs 133
b Effective Verbs as V1
The second kind of verb to appear as V1 in Cause-Effect SVCs in White Hmong
are ‘Effective’ verbs: verbs that indicate that the referent of the A argument cre-
ates that of the O argument, or brings it into being.

3PL make CLF house RDUP~be(come).tall
‘They built the house high.’
(90) nwsA khawb [lub qhov]O/S tob~tob

3SG dig CLF hole RDUP~be(come).deep
‘She dug the hole deep.’
The second verb in Cause-Effect SVCs of this kind is one that can indicate
either a State or a Change of State, such as siab ‘be/become tall’ and tob ‘be/
become deep’ in the examples above. These verbs convey more than just the
state that the Patient happens to be in after it is created; they indicate the state
that the Agent intended to bring about as a result of the action described by
the first verb. Again, the relationship between V1 and V2 in these sentences
clearly involves direct causation.
c Transfer Verbs as V1
A Transfer verb is one that describes the way in which an Agent ‘transfers’ a
Theme through space, without actually moving him/herself. In this case, just
as in the other two, direct causation is involved: the Agent acts directly on the
Theme to cause the change described by V2.
ahead.’ (LPV 2)
(92) nwsA xa [ib tsab ntawv]O/S mus rau [nws niam]

3SG send one CLF writing go to 3SG mother
‘She sent a letter to her mother.’
(93) pov pobO/S los rau kuv

throw ball come to 1SG
‘Throw a ball to me.’
134 CHAPTER 3
(94) qaum.ntujA txawm tso [hmuv tooj hmuv hlau]O/S

heaven then release spear copper spear iron
los <nkauj ntiaj.tebO>
come pierce earth
‘Then the heavens sent copper and iron spears down (to) pierce the
earth.’ (DNH 115: 5)
3.2.2 The Type of Causation Described by Cause-Effect SVCs in

White Hmong
3.2.2.1 Intentional or Unintentional Causation?
The Cause-Effect Serial Verb Construction can be used to express both inten-
tional and unintentional causation. A Cause-Effect SVC is generally used when
the A argument of the first verb performs the action described with the inten-
tion of bringing about the situation described by the second verb, as in the
following example:
(95) kuvA yuav xyeeb kojO/S poob ces quaj xwb!

1SG IRR push you fall CONJ cry only
‘I’ll push you down and then (you’ll) cry all right!’
However, a serial verb construction of this type can also be used to describe
unintentional causation:
ahead.’ (LPV 2)
Out of context, this example could be interpreted to mean either that the
action was performed deliberately or accidentally. However, in the context of
the story from which this excerpt was taken, it is quite clear that it was an acci-
dent. The story goes that Tiger and Frog were having a competition to see who
could jump further. Every time Tiger was about to jump, Frog would hang on
to his tail. As Tiger jumped, he would unintentionally flick Frog up ahead. The
Cause-Effect SVC is used here to express an unintentional result, rather than
something deliberately caused.
productive svcs 135
3.2.2.2 Direct or Indirect Causation?

No other construction in White Hmong signals a more direct form of causation
than the Cause-Effect SVC. Regardless of whether V1 is an Affective, Effective,
or Transfer verb this construction expresses a transitive action and its out-
come: the causer acts directly on the causee and, in doing so, brings about the
new situation described by V2.
The fact that the ‘cause’ and the ‘caused’ are expressed in a single clause in
these SVCs is clearly related to the fact that they indicate a direct form of cau-
sation. An event involving direct causation can be seen as a single event, hence
the propensity in many languages for this type of causation to be expressed by
lexical, rather than morphological or grammatical causatives (Shibatani, 1976;
Shibatani & Pardeshi, 2002).
White Hmong is a language, like many others in Mainland Southeast Asia, in
which verbs are fairly low in lexical complexity in certain domains. As we have
seen in the preceding section, the notions of direction and deixis are never
incorporated into verbs of locomotion and transport. Notions such as these
are introduced by a serial verb: a separate lexical item, but one that appears in
the same clause. In a similar way, in the absence of lexically complex causative
verbs, a serial verb construction is an ideal vehicle for conveying direct causa-
tion: two facets of a single event; two predicates within a single clause.
Mediated or indirect causation, on the other hand, requires more than one
event: the causer does something (not necessarily to the causee) and this leads
indirectly to the result. It is not surprising, then, that situations involving indi-
rect causation are expressed in White Hmong by constructions involving more
than one clause (see Chapter 5, §5.3 below).
The fact that the Cause-Effect SVC expresses a single event, a transitive
action and its outcome, means that, unlike in the case of the Purpose Con-
struction (see Chapter 5, §5.5), the second junct cannot be felicitously denied
when the proposition as a whole is asserted. Observe the following ungram-
matical examples:
(97) nwsA tsoo [lub khob]O/S tawg . . . * tab.sis [lub khob]S

3SG smash CLF cup break but CLF cup
tsis tawg
NEG break
‘She smashed the cup (so that it) broke . . . *but the cup didn’t break.’
(98) lawvA ua [lub tsev]O/S siab~siab . . . * tab.sis [lub

3PL make CLF house RECP~be(come).tall but CLF
tsev]S tsis siab
house NEG be(come).tall
136 CHAPTER 3
‘They built the house tall . . . *but the house wasn’t tall.’
Karttunen (1971) and Givón (1973, 1980, 1985, 2001b) use the term ‘implicative’
to describe complement-taking verbs for which the truth of the main clause
implies the truth of the complement clause. For example, the verb make in
English is an implicative verb, because:
(99) Jill made Jack fall down.
implies that
(100) Jack fell down.
Although they do not involve complement clauses, the term ‘implicative’ can
also be used with reference to Cause-Effect SVCs in White Hmong, because the
truth of the proposition as a whole implies the truth of the second junct.
3.2.3 Nexus and Level of Juncture in Cause-Effect SVCs

The two juncts in a Cause-Effect SVC are linked at the core level. This is because
the single shared argument in these constructions is a core argument: NP2 is
simultaneously the O argument of V1 and the S (or very occasionally the A)
argument of V2.
The fact that this shared argument can intervene between the verbs in
a Cause-Effect SVC rules out the possibility that nuclear level juncture is
involved. The fact that NP2 is indeed a shared argument, and not a coreferen-
tial one, can be demonstrated by the ungrammaticality of the occurrence of a
coreferential NP in the same clause. Observe the following:
(101) * nwsA ntaus [tus me-nyuam]O nwsS quaj kiag

3SG hit CLF child 3SG cry completely
‘He hit the child (so that) she cried.’
(102) nwsA ntaus [tus me-nyuam]O kom (nwsS) quaj kiag

3SG hit CLF child COMP 3SG cry completely
‘He hit the child so that she would cry.’
A noun phrase coreferential with the O argument of ntaus ‘hit’ can only appear
in the second junct if this second junct represents a separate clause. In exam-
productive svcs 137
ple (102) above, the complementizer kom introduces a new clause, and the
third person singular pronoun nws can optionally appear. However, the first
example involves only one clause, and the coreferential pronoun is ungram-
matical. This ungrammaticality demonstrates that NP2 is definitely a shared
argument, rather than a coreferential argument that undergoes zero pronomi-
nalization in the second junct. There is thus no possibility that clausal level
juncture is involved.
The fact that the juncts are linked in a serial relationship at the core level is
further corroborated by the fact that it is impossible for either junct to exhibit
independent peripheral arguments:
(103) * kuv xa [ib tsab ntawv] yav.sawv.ntxov mus rau

1SG send one CLF writing morning go to
[kuv niam]
1SG mother
‘I sent a letter in the morning to my mother.’
3.2.3.2 Nexus Relationship in Cause-Effect SVCs

Like the second junct in a Cotemporal SVC (see §3.1.3.2), the second junct in a
Cause-Effect SVC cannot appear sentence initially as a topic.
(104) * tawg mas, nws tsoo [lub khob]

break TOP 3SG smash CLF cup
‘(So that it) broke, he smashed the cup.’
(105) * [lub khob] tawg mas, nws tsoo

CLF cup break TOP 3SG smash
‘(So that) the cup broke, he smashed it.’9
(106) a. * siab mas, lawv ua [lub tsev]

be(come).tall TOP 3PL make CLF house
‘High, they built the house.’
9 This would be grammatical with an interpretation in which tawg ‘break’ is interpreted as an

attribute of lub khob ‘the cup’:
(i) [lub khob <tawg>REL] mas, nws
CLF cup break TOP 3SG
‘The broken cup, he smashed (it).’
138 CHAPTER 3
b. * [lub tsev] siab mas, lawv ua

CLF house be(c0me).tall TOP 3PL make
‘The house to be high, they built (it).’10
Recall the evidence presented by Fuller (1985, pp. 107–116) showing that sub-
ordinate elements, such as NP arguments, as well as subordinate location and
time phrases and clauses, can appear sentence initially as topics in White
Hmong (see §3.1.3.2 above). The fact that this cannot occur in the case of Cause-
Effect SVCs can be interpreted to mean that subordination is not involved in
this SVC type.
The scope of the core level, deontic modality, operators can be used as
a test to determine which of the two non-embedded nexus types—coordina-
tion or cosubordination—is involved in this core level juncture. It is possible
for sentences ostensibly similar to those above to have independent deontic
modality operators for each verb, as shown by the examples below. However,
the meaning of sentences such as these is quite different from that of the
Cause-Effect Serial Construction.
(107) nwsA txawj cub movO, txawj siav thiab

3SG know.how steam rice know.how cook also
‘He knows how to steam rice, (and) knows how to cook (it) too.’
The example above represents, not a single clause with a cause-effect relation-
ship between the first and the second verb, but rather a series of juxtaposed
clauses, with a semantic relationship of balance between them (see Chapter 5,
§5.2). In this example, the O argument of the first verb—mov ‘rice’—is not
interpreted as the S argument of the second. The second verb siav is under-
stood to be a transitive verb ‘cook’, rather than the intransitive verb ‘be
cooked’, and its A argument is coreferential with that of the preceding junct.
The fact that coreferentiality rather than sharing of arguments is involved
can be demonstrated by the fact that an overt pronoun can appear in the
second junct:
10 This would be grammatical with an interpretation in which siab ‘be(come) tall’ is inter-
preted as an attribute of lub tsev ‘the house’:
(i) [lub tsev <siab>REL] mas, lawv ua
CLF house be(c0me).tall TOP 3PL make
‘The tall house, they built it.’
productive svcs 139
(108) nwsA txawj cub movO, nwsA txawj siav thiab

3SG know.how steam rice 3SG know.how cook also
‘He knows how to steam rice, (and) he knows how to cook (it) too.’
It is not possible to include an independent deontic modality operator in the

second junct when a genuine Cause-Effect SVC is involved.
When an operator of this type appears before the first verb in a Cause-Effect
Serial Construction, it may not be immediately apparent that this operator has
scope over both verbs. In the example below, the A argument nws ‘he’ clearly
‘knows how to’ cub mov ‘steam rice’, but neither ‘he’ nor the ‘rice’ ‘know how to’
siav ‘be cooked’:
(109) nwsA txawj cub movO/S siav

3SG know.how steam rice be(come).cooked
‘He knows how to steam rice (so that it is) cooked.’
In fact, the deontic modality operator in this example does have scope over
both juncts in the construction: it is not just that ‘he knows how to steam rice’,
but that ‘he knows how to steam rice (so that it is) cooked’. Likewise, in the
example below, the meaning is not just that you should throw the ball, but that
you should throw it to me:
(110) kojA yuav-tsum pov pobO/S los rau kuv

2SG should throw ball come to 1SG
‘You should throw (the) ball (so that it) comes to me.’
The fact that the two juncts in a Cause-Effect SVC cannot be independently
modified by the core level deontic modality operators, and the fact that such
an operator occurring before the first verb has scope over both verbs in the
construction, indicates that the juncts are in a core cosubordinate relationship.
3.2.4 Cause-Effect SVCs: One Subject or Two?

A number of scholars, including Li and Thompson (1973), Nylander (1981)
(cited in Sebba (1987)), and Clark (1992) do not include what are here termed
‘Cause-Effect SVCs’ in the set of serial verb constructions at all. The definition
of a serial construction used by these scholars includes the condition that they
contain only one subject.
It is perfectly appropriate to require that all serial verb constructions have a
single grammatical subject. However, it is not actually necessary to discard the
requirement that serial verb constructions have a single grammatical subject
140 CHAPTER 3
in order to include sentences of this type in the class. What it is necessary to

discard is the requirement that the subject must be predicated by all verbs in
the series.
In fact, these sentences display the same degree of unity as other SVC types;
the difference is that the shared argument is the second NP (a non-subject NP)
rather than the first (in an AOV language). The unity of the construction comes,
as we would expect, from the way in which the two verbs contribute to the one
proposition: the expression of a transitive action and its direct and immediate
outcome. In fact, disallowing sentences of the type described in this section
from the class of serial verb constructions means ignoring the considerable
number of significant similarities, both syntactic and functional, between
these and other types of SVC (see §3.2.3 above regarding syntactic similari-
ties). It also means losing some important generalisations concerning serial
verb constructions, which serve to delineate serial verb constructions from
other constructions involving verbal concatenation.
As we have seen above, the interpretation of these sentences can be
accounted for by lexical factors rather than by syntactic rules of government,
referring to factors such as the transitivity of the verbs, their expression of
impingement and their expression of change.
3.2.4.1 The Question of Subject in Other Languages

Most authors who discuss serial constructions of this type describe the
shared argument as being simultaneously the ‘object’ of V1 and the ‘subject’
of V2 (Bamgboṣe, 1974; Foley & Olson, 1985; Crowley, 1987; Hansell, 1987). This
description may simply be an informal use of the term ‘subject’ in some cases.
Alternatively, it might be taken to imply that the serial verb constructions
exemplified above would, in fact, have two subjects: NP1, which is subject of
V1, and NP2, which is subject of V2 (in an AOV language). This notion of more
than one subject would be perfectly acceptable in a multiclausal analysis (e.g.
Bamgboṣe (1974, p. 22)): both NP1 and NP2 being the subject in their respec-
tive underlying clauses; the second instance of NP2 (the ‘subject’ of V2) being
deleted under coreference with the first (the ‘object’ of V1), leaving a single
‘surface’ subject (NP1). However, the notion of more than one subject is not
quite so straightforward in a monoclausal analysis.
Schachter (1974), one of the earliest scholars to present a monoclausal anal-
ysis of serial verb constructions, claims that there is always one, and only one,
‘syntactic subject’ in a serial verb construction. Schachter argues that, while
the shared NP in sentences of this type performs as the ‘semantic subject’ of
the second verb, it is still only the first NP that behaves as the ‘syntactic subject’.
Sebba (1987, pp. 86–87) draws the same distinction, maintaining that serial
productive svcs 141
verb constructions have only one ‘syntactic’ subject, but may have more than
one ‘semantic’ subject.
Schachter cites evidence (found in Christaller (1933, p. 145) from Akuapem,
a dialect of Akan (Kwa), in which subject concord prefixes on all verbs show
agreement with the first, rather than any subsequent NP, regardless of the
semantic interpretation:
(111) mede aburow migu msum

I-take corn I-flow water-in
‘I pour corn into the water.’
Schachter explains:
There is no doubt that aburow ‘corn’ is the semantic subject of gu ‘flow’

(as well as being both the syntactic and semantic object of de ‘take’), for
clearly it is the corn, and not the speaker, that is being poured into the
water . . . However, there is also no doubt that, as far as the syntactic rule
of subject concord is concerned, the subject of gu, like that of de, is a first-
person-singular pronoun. (pp. 258–9).
Schachter uses this evidence from Akuapem as syntactic justification for his
proposal that the underlying structure of all serial verb constructions involves
a single subject NP (and a sequence of concatenated VPs within a single
clause). As noted (§3.2.1.1 above), he argues that the semantic interpretation
(i.e. of which NP is ‘semantic subject’ of which verb) in serial verb construc-
tions should be accounted for by lexical rules (see below)).
The ‘single subject’ analysis does seem to be suitable for Akuapem, for the
reasons outlined by Schachter. However, Crowley (1987) presents evidence
from Paamese, an Oceanic language of Vanuatu, showing that the pronomi-
nal affix on the second verb in a Cause-Effect SVC agrees with the shared
argument (O argument of V1 = S argument of V2), and not with the initial NP
(A argument of V1):11
11 Crowley actually gives examples of Cause-Effect Serialization from two other Oceanic
languages, which he also refers to as ‘switch-subject’ serialization. However, the morpho-
logical evidence in these examples does not, in fact, point to the shared argument being
the subject of the final verb. In the first example he gives, from Gitua, the final verb has
no cross-reference marker. This is interpreted in Crowley’s gloss as cross-referencing a
3rd person singular subject (i.e. zero cross-reference marker). However, according to Bill
Foley (personal communication) it should probably be interpreted as no cross-reference
142 CHAPTER 3
(112) kaik komuasinau nauva netan

(kaiko ko-muasi-nau nau-vaa netano)12
2SG 2SG-REAL-hit-1SG 1SG-REAL-go down
‘You hit me down.’ (Crowley, 1987, p. 48)
(113) kail amuas vuas emat

(kaile a-muasi vuasi emate)
3PL 3PL-REAL-hit pig 3SG-REAL-die
‘They killed the pig by hitting it.’ (Lit: ‘They hit the pig (so that it) died.’)
(Crowley, 1987, p. 43)
According to Crowley, the non-initial verb in a serial verb construction of this

type in Paamese is marked “with the subject features that agree with the initial
object” (p. 47). That is, he regards both of the preverbal pronominal affixes in
the first example above, ko- ‘2SG’ and nau- ‘1SG’, to be subject concord markers.
This, together with his terminology ‘switch subject’ serial verb constructions,
clearly indicates that Crowley regards serial verb constructions of this type as
having two subjects.13 However, in Jarkey (1991, pp. 202–207) I discuss Crowley’s
evidence and find it somewhat inconclusive. In the absence of any further
evidence regarding the subject properties of the shared NPs in the Paamese
examples above, it is not possible to decide conclusively whether they should
be regarded as subjects or not.
The analysis of serial verb constructions presented in this study actually
does not have to make any reference to the notion of subject; the analysis
concentrates instead on the occurrence of a ‘shared argument’ in all types of
marker at all. More data from Gitua would be needed to confirm this. In the second exam-
ple he gives, from Ulithian, the postverbal 2nd person singular cross-reference marker
on V2 is actually for object (see Durie (1988)). That is, the shared argument in Ulithian is
overtly marked as an object rather than a subject, indicating that Causative Serialization
in this language definitely does have a single syntactic subject: the initial NP.
12 Note that it is not the shared NP ‘1SG pronoun’ that is repeated in this example; a shared
NP, by definition, could not be repeated. In fact, this referent does not actually appear as
an NP at all in this example. Rather, it is the 1SG cross-reference marker on the verb that
occurs twice: once postverbally to indicate, according to Crowley, that 1SG is the object of
muasi ‘hit’, and once preverbally to indicate that 1SG is the subject of vaa ‘go’.
13 The term ‘switch subject’, rather than ‘different subject’, is presumably used by Crowley
because this term at least hints at one of the most important features of these construc-
tions, namely that, in his terms, the ‘subject’ of V2 is, in fact, the ‘object’ of V1, not an
entirely new participant in the event.
productive svcs 143
serial verb construction. This shared argument is one factor that, together with
shared core operators and a single periphery, is responsible for linking the
verbs (and their respective cores when core level juncture is involved) together
in a single clause. The rules as to which argument is understood to be shared
in any particular SVC, in any particular language, appear to be lexically based,
as described above (§3.2.1).
3.2.4.2 The Question of Subject in White Hmong

If we are interested in determining whether the shared NP in Cause-Effect
SVCs in White Hmong exhibits any subject properties it is necessary to briefly
discuss the general question of the properties of subjects in White Hmong.
Keenan and Li (1975) list three categories of properties that are relevant
to determining subject-hood: coding properties, semantic properties, and
behaviour and control properties. Fuller (1985, pp. 71–85) examines NPs in
White Hmong with regard to properties from all three of these categories, and
concludes:
. . . the noun phrases that one would want to call subjects in Hmong have
limited but discernible subject properties. Coding properties include ini-
tial (pre-verbal) position only; . . . Semantic properties for these pre-
verbal NPs include a variety of semantic roles, a selectional relationship
with the verb, and a strong tendency toward referentiality; post-verbal
NPs may be non-referential and do not include agents (in the same
clause). Behaviour and control properties include several syntactic con-
structions which make reference to subjects, namely Equi constructions,
Conjunction Reduction constructions, zero Imperative Subjects and
Serial Verbs, in which only subjects are omitted, and Raising to Subject
constructions, and Passive constructions, in which the raised NP occurs
only in subject position. Hmong was found not to have Subject to Object
Raising constructions, and the evidence from Reflexivization and Recip-
rocal Formation does not bear conclusively on subject properties.
There is clearly one of Fuller’s subject properties that will not be used in this
exercise, and that is the one that states that only subjects are omitted in serial
verb constructions. Judging by examples of serial verb constructions given
here and elsewhere in her work (pp. 31–2), it is evident that Fuller does not
include sentences of the type under discussion in the set of serial verb con-
structions. It is also clear that most of the other behaviour and control proper-
ties mentioned, such as ‘Equi-NP deletion’, ‘Imperative Subject deletion’, and
‘Raising constructions’ are simply not relevant to the NPs in question. Let us
144 CHAPTER 3
examine the shared NP in the Cause-Effect SVC in White Hmong, with a view
to determining whether it exhibits any of the other properties mentioned.
a Coding Properties
The shared NP occurs both before and after verbs of which it is an argument;
it is syntactically coded as both the object of the first verb and as the subject
of the second.
b Semantic Properties
i Possible Semantic Roles of the Shared Argument
Fuller (p. 74) shows that preverbal NPs in White Hmong occur with a wide
variety of semantic roles. In the terminology used in this study, these roles
include Agent, Patient, Instrument, Experiencer, and Locative. However, Fuller
also points out the strong association between the preverbal position and the
role of Agent. If an Agent occurs in a sentence, it occurs in subject position of
a main or subordinate clause. Even in a passive-like sentence, an Agent can-
not occur postverbally in the same clause in White Hmong (see Fuller (1985,
pp. 136–156)).
It will be clear from the examples of the construction given elsewhere in
this section that the shared NP in a Cause-Effect SVC in White Hmong occurs
in an extremely limited set of semantic roles. With respect to the verb that
precedes it, it is always either Patient or Theme. With respect to the following
verb, it may also be Patient or Theme. This NP is never understood to be an
Agent. The only Agent in sentences of this type occurs initially, before the first
verb in the clause. We may conclude that, with respect to its possible semantic
roles, the shared NP does not exhibit subject properties.
ii Selectional Relationship with Verb

Although it is generally true that the shared NP in a Cause-Effect SVC exhibits
a selectional relationship with the following verb, this is not always the case.
In the SVC below, the NP ib tsab ntawv ‘a letter’ occurs before the verb mus ‘go’.
However, this NP would not occur as the subject of this verb in a simple clause:
(114) kuvA xa [ib tsab ntawv]O/S mus rau [kuv niam]

1SG send one CLF writing go to 1SG mother
‘I sent a letter to my mother.’
(115) ?? [ib tsab ntawv]S mus rau [kuv niam]

one CLF writing go to 1SG mother
for: ‘A letter went to my mother.’
productive svcs 145
It seems that the subject property of having a selectional relationship with the
following verb is not always exhibited by the shared NP in a Cause-Effect SVC.
This may be related to a certain degree of grammaticalization that is underway
in this and similar examples (see Chapter 4, §4.2.3.2).
However, it must be noted that this argument cuts both ways: while in a few
examples of the Cause-Effect SVC the shared NP does not have a selectional
relationship with V2, the initial NP never has a selectional relationship with
this verb. Neither NP has a clear claim to subject-hood on these grounds.
iii Referentiality
Fuller presents some evidence that suggests that non-referential NPs tend not
to occur as subjects in White Hmong. However, a non-referential NP can occur
as the shared NP in a Cause-Effect SVC. Observe the following example:
(116) yusA cub movO/S siav

‘One steams rice (so that it is) cooked.’
Notice, however, that a non-referential noun also occurs as subject of the first
verb in this example, showing that this property is, indeed, a tendency only,
and may not be a very useful test for subject-hood.
c Behaviour and Control Properties

As shown by Fuller, White Hmong subjects do not exhibit a very large number
of behaviour and control properties. Of these only one, namely ‘Conjunction
Reduction’, would be relevant to the NP under consideration. However, it
seems to be the case that this is not actually a valid test for subject-hood after
all. Although Fuller does show, in her discussion of subjects (p. 80), that a zero
NP in a conjoined clause can be coreferential with the subject of the preceding
clause, she does not prove that such a Noun Phrase cannot be coreferential
with a preceding non-subject NP. In fact, it seems that, in the right pragmatic
environment, those speakers who allow zero pronominalization to occur at
all in conjoined clauses of this type allow it to occur equally for subjects and
direct objects. This shows that zero pronominalization in conjoined clauses
cannot be used as a test for subject-hood in White Hmong. Observe the follow-
ing examples:
(117) NpliaA cem YeebO thiab hais <tias ‘Txhob ua li!’>COMP

PN scold PN and say COMP NEG.IRR do like
‘Mblia scolded Yeng and said “Don’t do that!” ’
146 CHAPTER 3
(118) NpliaA cem YeebO thiab quaj~quaj

PN scold PN and RDUP~cry
‘Mblia scolded Yeng and (Yeng) cried (and) cried.’
In the first example above, the zero NP in the second junct is taken to refer
to Mblia, the subject of the preceding junct. In the second example, however,
the zero NP is understood to refer to Yeeb, the direct object of the preceding
junct. In fact, the referent of a zero NP in a conjoined clause is grammatically
ambiguous; it is clearly pragmatics that determines the interpretation.
As we would expect, the same situation holds when the first junct in a series
of conjoined clauses is a Cause-Effect SVC.
(119) nwsA xyeeb [nws tus muam]O/S poob ces tawm mus xwb
3SG push 3SG CLF sister fall CONJ leave go only
‘He pushed his sister down and then just walked away.’
(120) kuvA yuav xyeeb kojO/S poob ces quaj xwb!

1SG IRR push you fall CONJ cry only
‘I’ll push you down and then (you’ll) cry all right!’
In the first example it is clearly the first NP in the Cause-Effect SVC, nws 3SG,
that is interpreted to be the coreferent of the zero NP in the second junct; in the
second example, however, it is the shared NP in the first junct, koj 2SG, which
is understood as the coreferent of the following zero NP. Again, the interpreta-
tion is quite clearly determined by pragmatic rather than grammatical factors.
The fact that some speakers simply do not accept conjoined sentences such
as the ones above with zero NPs in the second junct could be due to the ambi-
guity of coreference. This would account for the insistence of these speakers
that the sentences are not clear, and that the conjoined juncts should con-
tain overt nouns or pronouns. The fact that these sentences are grammatically
ambiguous to those speakers who do accept them does not throw any light
onto the question of whether the shared NP in a Cause-Effect SVC exhibits any
subject properties or not.
d Object Properties of the Shared NP

The shared NP in a Cause-Effect Serial Construction exhibits at least one prop-
erty that indicates that it is, in fact, an object rather than a subject. This prop-
erty is the ability to occur, under certain circumstances, in a topical position as
the object of the verb muab ‘take’ (see §3.4.6 for further discussion regarding
objects of this type):
productive svcs 147
(121) muab pob.kwsO/S ntaus~ntaus dam tag

take corn RDUP~hit break.offINTR finish
‘Beat the corn (so that it) breaks off.’ (DNH 123: 4)
e Conclusion
The weight of the evidence seems to point to the conclusion that the shared
NP in a Cause-Effect SVC in White Hmong is not a subject. This NP does not
exhibit a wide variety of semantic roles, it cannot be an Agent, and need not be
referential. Furthermore, it can appear in the topical sentence initial position
as the object of muab ‘take’, suggesting that it is an object. It is therefore reason-
able to conclude that all serial verb constructions in White Hmong, including
Cause-Effect SVCs, consist of a single clause with a single grammatical subject.
3.3 Attainment Serial Verb Constructions
3.3.1 Introduction
In a significant class of SVCs in White Hmong—called Attainment SVCs in this
analysis—the second of two verbs in a series serves to describe the successful
attainment of the goal of the action or event depicted by the first verb.14 The
two verbs in this type of SVC occur back-to-back, with no intervening NP, even
when the first verb in the construction is transitive. Observe the highlighted
verbs in the examples below:
(122) nwsA/A tua raug liabO/O

3SG shoot.at hit.the.mark monkey
‘He shot some monkeys.’
(123) yog [pab twg]A/A xub nrhiav pom [txoj

COP group which(ever) first search.for see CLF
ncauj-kev no]O/O ces . . .
mouth-path this CONJ
‘If one of the groups found the entrance to the path first, then . . .’
(SLY 110: 2)
14 Note that Enfield (2003) uses the term ‘attainment’ in a different way, to refer to the “actu-
alisation in reality of some predication” (p. 41).
148 CHAPTER 3
(124) kuvA/A raws cuag lawvO/O lawm

1SG pursue reach 3PL PRF
‘I have caught up with them.’
(125) thaum yusA/A puab tau [ib lub ncuav]O/O lawm

time INDF mould get one CLF rice.cake PRF
‘When one has moulded one rice cake . . .’ (Adapted from: KNH 16)
(126) xyoo no nwsA/A xaws tau [peb daig tiab]O/O lawm

year this 3SG sew get three CLF skirt PRF
‘This year she has made three skirts.’
(127) nwsA/A mus txog [tom khw]O/O lawm

3SG go arrive over.there market PRF
‘She has arrived at the market.’
(128) [hluav.taws kub~kub]S/S tuaj ti nkaus

fire RDUP~be.hot come be(come).close INTNS
‘. . . a very hot fire came very close . . .’
(129) dejS/S nyab zuj-zus tuaj ces txawm nyab~nyab

water rise gradually come CONJ then RDUP~rise
nto plaws ntujO
reach suddenly sky
‘The water came up gradually and then suddenly rose and rose right up
to the sky.’ (DNH 115: 2)
3.3.2 Types of Predicate in Attainment SVCs

3.3.2.1 The Second Verb
The second verb in an Attainment SVC is the simpler one to describe. If it were
to occur as the only verb in clauses such as those illustrated above, the predica-
tion would be an achievement (e.g. raug (liab) ‘hit (some monkeys)’, pom (txoj
ncauj-kev no) ‘see (the path)’, cuag (lawv) ‘catch (them)’ tau (ib lub ncuav) ‘get
(a rice cake)’, txog (tom khw) ‘arrive ([at the] market)’, ti ‘be(come) close’, nto
‘reach (the sky)’. However, in White Hmong achievement verbs rarely appear
alone as the only predicate in the clause. They are almost always a part of a
serial verb construction, frequently an Attainment SVC.
productive svcs 149
3.3.2.2 The First Verb

The first verb in each of the examples above, if appearing as the only verb in
the clause, would be translated as either an activity or an accomplishment
in English. Those that correspond to activities in English include tua (liab)
‘shoot at (monkeys)’, nrhiav (txoj ncauj-kev no) ‘search for (the path)’, raws
(lawv) ‘pursue (them)’. Those that would be translated as accomplishments in
English are of two kinds: first there are the Effective verbs, e.g. puab (lub ncuav)
‘mould (the rice cakes)’, xaws (daim tiab) ‘sew (the skirts)’; secondly there are
telic Motion verbs, e.g. mus (tom khw) ‘go (to market)’ tuaj ‘come’. However,
these types of predicates—activities and accomplishments—seem to pattern
somewhat differently in White Hmong (as in a number of other Asian lan-
guages) from the classification introduced by Vendler (1957), outlined in the
Introduction to this chapter.
a Goal-Oriented Activities as V1
Let us begin by looking more closely at examples (122)–(124) above. The first
verb in these sentences, if it were to occur alone, would be part of an activ-
ity predication. Activities are well known to co-occur with adverbial phrases
expressing duration, such as ‘for time-t’, with the interpretation that the action
is in progress for the duration of time-t. Observe the following examples:
(130) nwsA/A tua liabO/O ntev li ib teev

3SG shoot.at monkey long as one hour
‘He shot at monkeys for an hour.’
(131) hnub no, kuvA nrhiav [kuv nti nplhaib]O ib

day this 1SG search.for 1SG CLF ring one
hnub nkaus
day single
‘Today, I searched for my ring all day long.’
(132) kuvA raws lawvO ntev li ib teev

1SG pursue 3PL long as one hour
‘I pursued them for one hour.’
The predicates in examples (130)–(132) above are somewhat atypical of the

general notion of activities, in that each one seems to involve a transitive verb
oriented towards some kind of goal related to ‘attaining’ the object: one ‘shoots
at monkeys’ in order to ‘kill monkeys’; one ‘searches for a ring’ with a view
to ‘finding the ring’; one ‘pursues them’ in order to ‘catch them’. Activities,
150 CHAPTER 3
however, are supposed to “involve no culmination or anticipated result”

(Mourelatos, 1981, p. 192), as opposed to accomplishments, which are often said
to be ‘goal-oriented’.
The difference between accomplishments and activities is not, as is often
assumed, that accomplishments have a goal and activities do not. Instead,
accomplishments differ from activities in that they have an intrinsic goal;
they are ‘telic’ in the true sense of the word: “directed toward attaining a goal
or limit at which the action exhausts itself and passes into something else”
(Andersson (1972), cited in Dahl (1981)). If you persevere with ‘building a house’
or ‘eating a cake’ for long enough, ultimately you will have ‘built the house’ or
‘eaten the cake’, and in doing so you will have attained your goal. A goal of this
kind is often appropriately referred to as a ‘terminal point’. Activities, on the
other hand, can indeed have a goal, but it must be an extrinsic one. You can
persevere for an indefinite period of time ‘shooting at monkeys’, ‘searching for
your ring’, or ‘pursuing them’ but in the end you may have done no more than
‘shot at monkeys’, ‘searched for your ring’, or ‘pursued them’; you may never
attain your goal of ‘finding the ring’, ‘killing some monkeys’ or ‘catching them’.
A goal of this kind is something distinct from the action described, rather than
being a natural ‘terminal point’ of that action.15
In White Hmong, in order to indicate that the extrinsic goal of these activi-
ties is actually reached—that the object is ‘attained’—an Attainment SVC
must be created with the addition of an achievement predicate. The resulting
construction can co-occur with adverbial expressions indicating limit of time.
Adverbials indicating duration of time are not acceptable in most cases:
(133) a. nwsA/A siv sim.haum li ib teev tua

3SG use time as one hour shoot.at
raug liabO/O
hit.the.mark monkey
‘He took an hour to shoot some monkeys.’
15 Enfield (2003, 2007) refers to verbs of this type as ‘projected accomplishment verbs’. He
says that “unlike genuine resultative verbs they do not entail the resultant state. Instead,
they entail that the main subject carries out an activity with the intention of bringing about
that state” (2003, p. 230). In White Hmong, the intention of the subject is not necessarily
involved (as seen in example (129), in which the goal-oriented activity verb nyab ‘(water)
rises’ appears as V1 in an Attainment SVC with an inanimate subject). The term ‘goal-
oriented activity’ is used here, emphasising the fact that the events described by these
verbs are a sub-type of activity.
productive svcs 151
b. * nwsA/A tua raug liabO/O ntev li ib teev

3SG shoot.at hit.the.mark monkey long as one hour
Lit: ‘He shot some monkeys for an hour.’
(134) a. thaum ua.ntej peb teev kuvA/A nrhiav tau kuv

time before three hour 1SG search.for get 1SG
nti nplhaib]O/O
CLF ring
‘I found my ring by three o’clock.’
b. * kuvA/A nrhiav tau [kuv nti nplhaib]O/O

1SG search.for get 1SG CLF ring
kom.txog peb teev
until three hour
Lit: ‘I found my ring until three o’clock.’
Note, however, that when a change of location (or change of state) is involved,
as in example (135) below, a duration-of-time expression can occur (135b).
However, the meaning is not that the motion leading up to that change of loca-
tion takes place for that duration. Rather it is that the new location (or new
state) is maintained for that duration.
(135) a. ib hnub tom.qab kuvA/A mam mus raws

one day after 1SG then go pursue
cuag lawvO/O
catch 3PL
‘I caught up with them in a day.’
b. kuvA/A mus raws cuag lawvO/O tau ntev li

1SG go pursue catch 3PL get long as
ib hnub
one day
‘I caught up with them for a day.’
The examples of the Attainment Serial Verb Construction given above, in

which the first predicate expresses a goal-oriented activity, serve to focus the
attention of the listener on the attainment of the goal of that activity, rather
than on the activity itself. This conclusion is supported by examples (133)–(135)
above, which show that these SVCs either cannot co-occur with expressions
152 CHAPTER 3
indicating duration of time, or that time expressions of this sort refer to the
time after the goal has been attained rather than to the process leading up to
that goal.
Nevertheless, there is a difference between an Attainment SVC and an
achievement. The latter focuses exclusively on the moment of the achieve-
ment. It may well be assumed that there is some process leading up to it but, in
many languages, this process may not even be mentioned. In the Attainment
Serial Construction in White Hmong, on the other hand, even though the pri-
mary focus is definitely on the attainment of the goal, reference is still made to
the goal-oriented process leading up to that attainment. It is even possible to
draw considerable attention to this process, for example by reduplicating the
first verb:
(136) dejS/S nyab zuj-zus tuaj ces txawm nyab~nyab

water rise gradually come CONJ then RDUP~rise
nto plaws ntujO
reach suddenly sky
‘The water came up gradually and then suddenly rose and rose right up
to the sky.’ (DNH 115: 2)
In example (136), the duration of the process as the water gradually rises
is initially introduced in the first clause with a Cotemporal Motion SVC,
dej nyab zuj-zus tuaj ‘the water came up gradually’. This sense of the water
swelling relentless upward is reinforced in the second clause by the reduplica-
tion of the verb nyab ‘rise’ in the first part of an Attainment SVC, which ends
with the inevitable outcome nto plaws ntuj ‘reach suddenly (up) to the sky’. It
is because of this inclusion of interest in the process, as well as focus on the
goal, that the term ‘attainment’ rather than ‘achievement’ is used here to refer
to these SVCs.16
16 The distinction between ‘goal-oriented’ activities and corresponding accomplishments

or achievements is made in many languages, using different strategies. In English there
are restricted sets of minimal pairs, in which a verb with a prepositional object indi-
cates an activity, and the corresponding verb with a direct object indicates the related
achievement: e.g. shoot at a monkey versus shoot a monkey; kick at someone versus kick
someone. In Swedish, the distinction is more systematic (Dahl, 1981, p. 87), e.g. stickade
på en tröja ‘knitted at a sweater’ vs. stickade en tröja ‘knitted a sweater’. In some non-
Germanic European languages a similar distinction is made morphologically, by employ-
ing different aspectual forms of the same verb, e.g. Russian lečit’ (imperfective) ‘treat’,
productive svcs 153
b Effective Verbs as V1
Having described those examples of the Attainment Serial Verb Construction
in which the first verb, when appearing as the only predicate in the clause,
would be a goal-oriented activity, let us now turn to those in which it would
be translated as an accomplishment in English. Once again, all these verbs are
transitive, and in this case they express ‘attaining’ the object in the sense of
creating it.
In each of the examples below, the (a) sentence illustrates an Attainment
SVC; the (b) sentence illustrates a corresponding effective accomplishment
alone:
(137) a. thaum yusA/A puab tau [ib lub ncuav]O/O lawm

time INDF mould get one CLF rice.cake PRF
‘When one has moulded one rice cake . . .’ (Adapted from: KNH 16)
b. thaum yusA puab [ib lub ncuav]O

time INDF mould one CLF rice.cake
‘When one moulds a rice cake . . .’
(138) a. xyoo no nwsA/A xaws tau [peb daig tiab]O/O lawm

year this 3SG sew get three CLF skirt PRF
‘This year she has made three skirts.’
b. nwsA xaws [ib daig tiab]O rau kuv lawm

3SG sew one CLF skirt for 1SG PRF
‘She has sewn a skirt for me.’
Predicates like puab (ib lub ncuav) ‘mould (a rice cake)’ and xaws (ib daig
tiab) ‘sew (a skirt)’ differ from the goal-oriented activities discussed above
(e.g. nrhiav (kuv nti nplhaib) ‘search for my ring’) in that they have an intrinsic
rather than an extrinsic goal. That is, these predicates are truly telic in that
each one expresses an action which ultimately “exhausts itself and passes into
something else.” Dahl (1981, p. 81ff) explains that Slavic aspectologists observe
not one but two distinctions in relation to telicity. He gives the following exam-
ples to illustrate:
vylečit’ (perfective) ‘cure’ (Mourelatos, 1981, pp. 194–195); Ancient Greek peithein (present)
‘persuade/try to persuade’, peisai (aorist) ‘convince’ (Armstrong, 1981, p. 11).
154 CHAPTER 3
(a) I was writing.

(b) I was writing a letter.
(c) I wrote a letter (taken to imply ‘I finished it’).
The first distinction is between predicates that do not lead up to a terminal

point ((a) above) and those that do ((b) and (c)). Dahl coins the term ‘T prop-
erty’ to refer to those predicates that lead up to a terminal point. The second
distinction is between predicates that simply lead up to a terminal point
((b) above), and those for which the terminal point is actually reached ((c)).
This latter property he terms the ‘P property’.
It is clear from the evidence given below that a predicate in White Hmong
like xaws (ib daig tiab) ‘sew (a skirt)’, by itself, has the ‘T property’ but not nec-
essarily the ‘P property’; that is, it expresses an action that leads up to a termi-
nal point, but for which the terminal point has not necessarily been reached.
The focus of attention is on the action, rather than the terminal point, and the
attainment of the terminal point is implied but not entailed. The Attainment
SVC xaws tau (ib daig tiab) ‘make (a skirt)’, on the other hand, is shown below
to exhibit the ‘P property’. That is, although the action leading up to the termi-
nal point is referred to (by the first verb), the attainment of the terminal point
is explicitly entailed (indicated by the second verb), and the focus of attention
is on that attainment.17
This difference in focus is responsible for the fact that these verbs tend to be
used alone when reporting actions, and in serial constructions when reporting
results. Compare the following examples:
(139) a. xyoo no nwsA ua [ob daig teb]O

year this 3SG make two CLF field
‘This year he’s working two fields.’
b. xyoo no nwsA/A ua tau [ob nthab nplej]O/O

year this 3SG make get two platform rice
‘This year he grew two platforms(-worth) of rice.’
17 This feature of Accomplishments is common in the languages of Asia and South Asia,
including Chinese (Smith (1997); Tai (1984)), Hindi (Singh (1991)), Japanese (Ikegami
(1985); Tsujimura (2003)), Korean and Thai (Wechsler (2003)), Tamil (Talmy (1991)).
(See §3.3.4 for further discussion on Chinese.) Enfield’s analysis of the entailments of
accomplishments in Lao (2007, p. 416) suggests that this feature does not apply in that
language.
productive svcs 155
(140) kojS rau~rau-siab mus <txiav [taws ntau~ntau]O>PURP;

2SG RDUP-put.in-liver go chop firewood RDUP~much
es thaum kojA/A txiav tau ntau~ntau lawm tiag
CONJ when 2SG chop get RDUP~much PRF really
ces kojA rhaub [ib yiag dej kub~kub, npau~npau]O
CONJ 2SG heat one pot water RDUP~hot RDUP~boil
‘Do your best to go and chop a lot of firewood; and when you have
really chopped a lot, heat a pot of water boiling hot.’ (DNH 364: 59)
In the (a) example in (139), with the single verb ua ‘make’, the speaker is focus-
ing on an action: s/he may be actually watching the farmer at work in the
fields. In the (b) example, with the Attainment SVC ua tau ‘make get’, the focus
is on the product or result of the action, rather than on the action itself: the
speaker may be looking at the harvest laid out to dry. In a similar way, the single
verb txiav ‘chop’ in example (140), focuses on the process of the action that
the addressee is exhorted to perform—‘to (go and) chop a lot of fire-wood’—
while the Serial Verb Construction txiav tau ‘chop get’ describes the outcome
of that action. This outcome serves as background, as the speaker’s attention
turns to the next action that the addressee must perform: ‘heating a pot of
water boiling hot’.
Further evidence of this difference between accomplishment verbs alone
and the corresponding Attainment SVC is provided by the possibilities of co-
occurrence with certain time expressions:
(141) nwsA xaws [daim tiab no]O kom.txog peb teev

3SG sew CLF skirt this until three hour
‘She was sewing this skirt until three o’clock.’
(142) * nwsA/A xaws tau [daim tiab no]O/O kom.txog peb teev
3SG sew get CLF skirt this until three hour
Lit: ‘She made this skirt until three o’clock.’
Example (141) above shows that an adverbial expression indicating duration

of time is acceptable with an Effective verb alone, which focuses on the action
rather than its completion. It is not acceptable, however, with the correspond-
ing Attainment SVC in example (142), which focuses on the attainment of the
goal of the action, not on the process of the action itself. This contrast in focus
is further illustrated by the slightly different interpretations of the time expres-
sions in examples (143) and (144) below:
156 CHAPTER 3
(143) nwsA siv sij.haum peb teev los xaws [ib daig tiab]O
3SG use time three hour CONJ sew one CLF skirt
‘She spent three hours sewing a skirt.’
(144) nwsA/A siv sij.haum peb teev los xaws tau [ib
3SG use time three hour CONJ sew get one
daig tiab]O/O
CLF skirt
‘She took three hours to sew a skirt.’
With the verb xaws ‘sew’ alone, the time expression siv sij haum peb teev (lit.
‘use three hours’) is interpreted as referring to a duration: ‘spend three hours’;
there is no indication in this sentence as to whether the skirt was finished
or not. With the corresponding Attainment SVC, the same time expression
is interpreted as referring to a limit of time: ‘take three hours’; the sentence
unambiguously indicates that the skirt was completed.
It is entirely appropriate for a sentence with an Effective verb alone to be
used in a context where the goal is understood to have been attained, as long
as there is no particular attention directed toward that attainment:
(145) nwsA xaws [ib daig tiab]O rau kuv lawm

3SG sew one CLF skirt for 1SG PRF
‘She has sewn a skirt for me.’
While a sentence with a simple predicate like xaws (ib daig tiab) ‘sew (a skirt’)
is normally understood to imply that the goal is attained, that implication is
by no means unavoidable. The corresponding Attainment SVC, on the other
hand, can only be taken to mean that the goal is attained:
(146) nwsA xaws [daig tiab no]O lawm, tseem tsis tau tag
1SG sew CLF skirt this PRF still NEG get finish
‘She has been sewing this skirt, [but] she hasn’t finished it yet.’
(147) * nwsA/A xaws tau [daig tiab no]O/O lawm, tseem

3SG sew get CLF skirt this PRF still
tsis tau tag
NEG get finish
Lit: ‘She has made this skirt, [but] she hasn’t finished (it) yet.’
productive svcs 157
c Telic Motion Verbs as V1

In the previous section, Attainment SVCs in which the first verb is an Effective
verb were described. Another kind of accomplishment predicate that can
occur as the first verb in an Attainment SVC is one that expresses telic motion,
with verbs such as mus ‘go’, tuaj ‘come’, or los ‘come home’.
(148) nwsA/A mus txog [tom khw]O/O lawm

3SG go arrive over.there market PRF
‘She has arrived at the market.’
(149) kuvA/A tuaj txog kiag [ntawm nej lub taw-rooj]O/O

1SG come arrive INTNS nearby 2PL CLF foot-door
kiag . . .
INTNS
‘I really will come right to your doorstep . . .’ (DNH 124: 6)
(150) [hluav.taws kub kub]S/S tuaj ti nkaus . . .

fire hot hot come be(come).close INTNS
‘A very hot fire came very close . . .’ (HFT 1, p. 2)
(151) nkawdS/S/A tau poob los ti [npoo av]O

3DU ACHV fall come be(come).close edge earth
lawm
PRF
‘. . . they have managed to fall close to the edge of the earth.’
(DNH 2: 11)18
Clauses containing the single verb mus ‘go’, tuaj ‘come’, or los ‘come, return
home’ plus a Locative Phrase indicating a goal are like those with the Effective
verbs puab (ib lub ncuav) ‘mould (a rice cake)’ and xaws (ib daig tiab) ‘sew (a
skirt)’ in that they have an intrinsic rather than an extrinsic goal. Once again,
predicates of this kind express an action that ultimately “exhausts itself and
passes into something else”:
18 The three verbs in this construction poob los ti ‘fall come close’ actually represent a combi-
nation of two SVCs: the first, poob los ‘fall come’ is a Cotemporal SVC, which is then itself
serialized with ti ‘close’ to form an Attainment SVC. This phenomenon of linking SVCs is
discussed and exemplified in more detail below (§3.5.2).
158 CHAPTER 3
(152) nwsA mus [tom khw]O lawm

3SG go over.there market PRF
‘She has gone to market.’
However, as noted in §a above, with predicates such as this conveying change

of location (or change of state) the interpretation of a duration-of-time expres-
sion is different from that with other types of accomplishments:
(153) nwsA mus [tom khw]O ntev li ib teev

3SG go over.there market be.long as one hour
‘She went to market for one hour.’
Instead of referring to the duration of the action leading up to the terminal

point, the adverbial expression ‘for an hour’ refers to the duration of the main-
tenance of the change of location (i.e. ‘She went to the market (and stayed at
the market) for one hour’).
With the corresponding Attainment SVC, as with all other Attainment SVCs
illustrated above, only adverbial expressions indicating limit of time can occur;
those indicating duration of time are unacceptable:
(154) nwsA/A mus txog [tom khw]O/O tom.qab ib teev

3SG go arrive over.there market back one hour
‘She got to the market before one o’clock.’
(155) * nwsA/A mus txog [tom khw]O/O ntev li ib teev

3SG go arrive over.there market long as one hour
Lit: ‘She got to the market for one hour.’
Of all the different kinds of Attainment SVC discussed, those with change of
location verbs like this seem the closest to representing an achievement predi-
cation. They focus clearly on the moment at which the goal of V1 is achieved—
for example, the ‘arrival at the market’. Nevertheless, in this case as in all the
others, there is still mention made of the action leading up to that achieve-
ment, and this action clearly constitutes part of the meaning of the predica-
tion as a whole. As with the other types of Attainment SVC illustrated above,
the focus can be at least partially shared between the goal and the process, for
example by the reduplication of V1:
productive svcs 159
(156) Txiv.Nraug.NtsuagS/A mus~mus txog [nram qab pa-dej]O

the.Orphan.Boy RDUP~go arrive down back lake
‘The Orphan Boy continued walking [and] got down to the lower end
of the lake . . .’ (DNH 183: 3)
d Other Verb Types as V1

The three types of predicates described above—goal-oriented activities, effec-
tive accomplishments, and telic motion accomplishments—are actually not
the only ones that can occur as V1 in this type of Serial Verb Construction.
Observe the examples below:
(157) yog kojS/A dhia yeeb kuvO no ces cia li, . . .

COP 2SG jump overcome 1SG IP CONJ let as
‘If you jump [and] beat me [i.e. jump further than me] we’ll leave it at
that, . . .’ (LPV p. 1)
(158) kuvA/A puas hais raug [lo ntawd]O/O?

1SG Q speak hit.the.mark CLF that
‘Did I say it right?’
In these examples, the predicates dhia ‘jump’ and hais lo ntawd ‘say that’ are
both activities. Unlike goal-oriented activities such as nrhiav (nti nplhiab)
‘search for (the ring)’, these are ones that are not normally thought of as hav-
ing any goal at all. However, the use of the Attainment Serial Construction,
along with the context in which these examples occur, serves to portray these
activities as if they had a goal. In the first example, with the achievement verb
yeej ‘overcome’ as V2, what is achieved is not just jumping, but successfully
jumping further than the opponent. In the second example, with the V2 raug
‘hit the mark’, the activity of speaking is portrayed as having the goal of ‘speak-
ing correctly’. In these examples it is not V1, but rather V2—the achievement
verb—that, along with the context, is responsible for the notion that there is a
goal, and that it is successfully attained.
3.3.3 Components Common to All Attainment SVCs

In the discussion above we have seen that all instances of the Attainment SVC
function to indicate that the subject attains some kind of goal. In this type of
SVC there is a difference in the interpretation, not only depending on whether
V1 itself has a goal or not, but also depending on the type of goal involved.
When the first verb expresses an action that is directed toward the subject
‘attaining’ a goal, this means simply that the extrinsic or intrinsic goal of V1,
160 CHAPTER 3
expressed by V2, is achieved. In the case of goal-oriented activity predicates

(e.g. tua ‘shoot at’), effective accomplishment predicates (e.g. xaws ‘sew’), and
telic motion predicates (e.g. mus ‘go’), the goal is most commonly related in
some way to ‘attaining’ the object (e.g. ‘hitting’, ‘creating’, or ‘reaching’ the
object). In this case, the interpretation of the Attainment SVC is that the goal
is achieved: the target is hit, the new item is created, the destination is reached.
However, in the case of the activity predicates that do not have a goal, such
as dhia ‘jump’ and hais lo ntawd ‘say that’, the idea that there is a goal may be
introduced by the context along with the achievement verb (V2). In these cases
the goal is understood to be simply to perform a somewhat challenging action,
and it is the achievement verb as V2 that is responsible for the notion that the
subject attains this goal.
3.3.4 Comparison of Attainment SVCs with ‘Resultative Verb Compounds’

in Mandarin
Attainment SVCs in White Hmong are similar in many ways to some (though
certainly not all) examples of what are often referred to as ‘Resultative Verb
Compounds’ (Tai, 1984; Thompson, 1973) or ‘Complements of Result’ (Hansell,
1987, 1993) in Mandarin. Hansell (1987, p. 40) gives these examples:
(159) wo3 kan4 dong3 le nei4 ben3 shu1

I read understand ASP that CLF book
‘I understand (by way of reading) that book.’
(160) ta1 bing4 si3 le

he sick die ASP
‘He died of disease/ He got sick and died.’
Tai (1984) claims that Mandarin has no category of true Accomplishment

verbs. It seems that, as in White Hmong, an accomplishment interpretation
emerges in certain contexts with a simple verb only, but “to ensure the attain-
ment of goal, Chinese resorts to resultative verb compounds, of which the first
element indicates the action, the second the result.” (p. 290). This ability to
create a predication which unambiguously expresses attainment of a goal is
clearly an extremely important function of both the ‘Attainment’ SVC in White
Hmong and the ‘Complement of Result’ construction in Mandarin, both of
which involve juncture at the nuclear level.19
19 Hansell (1987, pp. 48–49) uses scope of operators tests to show that the relationship
between the verbs in a Mandarin Complement of Result construction is one of Nuclear
productive svcs 161
In spite of this important similarity between this construction in White

Hmong and some examples of the Complement of Result construction in
Mandarin, it is important to note that the two constructions are by no means
completely parallel. The most significant difference is with regard to the seman-
tic scope of the construction. Mandarin Complements of Result also include
examples that are semantically more similar to a completely different type of
SVC in White Hmong: the Cause-Effect SVC (see §3.2). In Mandarin, the two
verbs in a sentence with this cause-effect meaning appear immediately adja-
cent to one another, just as they do in the other examples of the Complement
of Result construction above:
(161) ta1 qiao1 po4 le yi2 ge fan4wan3

he hit breakINTR ASP one CLF ricebowl
‘He broke (by hitting) one ricebowl.’ (Hansell, 1987, p. 40)
However, in White Hmong, the shared NP in a Cause-Effect SVC appears

between the two verbs, in its ‘normal’ position in relation to both predicates; a
completely different type of juncture—core rather than nuclear—is involved:
(162) nwsA ntaus [ib lub ntim]O/S tawg rhe

3SG hit one CLF ricebowl break completely
‘He hit a ricebowl [and it] broke.’
3.3.5 Rules for Deriving Attainment SVCs: Syntactic or Lexical?

Both Chao (1968) and Thompson (1973) regard the Mandarin construction
described above as involving a type of compound verb that is created in the
lexicon. Thompson rejects the use of syntactic rules in the creation of these
‘Resultative Verb Compounds’, principally on the grounds that they cannot
be derived transformationally from two underlying sentences (pp. 362–8).
Hansell (1987), on the other hand, treats these constructions in Mandarin as a
type of SVC. He is able to adopt this analysis because he utilises the RRG notion
(similar to that utilised in this study) in which all verbs in a serial construction
are regarded as belonging to a single clause.
In the case of White Hmong it also seems reasonable, at first sight, to con-
sider sentences of the type illustrated above to involve compound verbs. The
fact that the second verb follows directly upon the first, without the possibil-
ity of even a shared object intervening, suggests that the two verbs could be
Cosubordination. See §3.3.6 below regarding juncture and nexus in White Hmong
Attainment SVCs.
162 CHAPTER 3
viewed as a single unit. However, there are a number of important reasons for
preferring an analysis involving syntactic rules in this case.
One reason for deriving Attainment SVCs in White Hmong by syntactic
rather than lexical rules is that, if we use an analysis of serial verb construc-
tions like the RRG one (as Hansell does for Mandarin), there already exists
in the syntax a mechanism for analysing verb-verb sequences such as this.
Additional rules in the lexicon would be simply unnecessary. Furthermore,
the derivation of sequences such as these by syntactic rather than lexical rules
does not by any means preclude the possibility of viewing the verbs involved
as co-predicators within the clause.
A second reason for deriving this construction by syntactic rather than lexi-
cal rules is that, as shown below, any verbs that can co-occur in an Attainment
SVC can also co-occur in expressions which involve more than one clause, but
in which they fulfil the same basic functions in relation to one another—the
expression of an action and its achievement. There is no doubt at all that these
multiple clause constructions involve entirely productive syntactic processes.
It would be uneconomical to try to account for the co-occurrence of these
verbs on some occasions by syntactic rules, but on others by lexical rules.
The expressions referred to are extremely common. They occur when an
operator intervenes between V1, the activity or accomplishment verb, and V2,
the achievement verb. For example:
(163) nwsA tua puas raug liabO?

3SG shoot.at Q hit.the.mark monkey
‘He shot (but) did (he) hit (any) monkeys?’
(164) nwsA xaws yuav tau [daim tiab]O

3SG sew IRR get CLF skirt
‘She’ll sew (and) will probably complete the skirt.’
(165) nwsA mus tsis tau txog [tom khw]O lawm

3SG go NEG ACHV arrive over.there market PRF
‘She has gone (but) hasn’t arrived at the market yet.’
The force of these expressions is to indicate the background information con-

cerning an action, using V1 (e.g. ‘we all know he’s been shooting’), and then to
make some further statement or question concerning the attainment of the
goal of that action (e.g. ‘did he hit any monkeys?’).
The semantic ‘separateness’ of the background information and the state-
ment or question concerning it is reflected by their syntactic separation into
productive svcs 163
two clauses. Instead of treating the action and its outcome as a single event as
in an Attainment SVC (e. g. tua raug ‘succeed in shooting’, xaws tau ‘complete
sewing’), the speaker breaks these two facets of the situation into two distinct
events: the shooting is expressed as distinct from the issue of hitting the tar-
get; the sewing is thought of as independent of the completion of the finished
product.
This semantic separation is reinforced in a number of ways. As we would
expect, the juncts can optionally be interrupted by a brief pause. The semantic
separation is also reinforced by syntactic separation—by the occurrence of the
operator between the two verbs. It can be further reinforced syntactically by
the placement of the object between the two predicates, a position in which it
can never appear in the Attainment SVC (see §3.3.6.1 below):
(166) nwsA tua liabO puas raug?

3SG shoot.at monkey Q hit.the.mark
‘Will he succeed in shooting any monkeys?’
(167) nwsA xaws [daim tiab]O yuav tau

3SG sew CLF skirt IRR get
‘She will complete sewing the skirt.’
Riddle (1989, pp. 3–4) notes the occurrence of sentences of this type in White
Hmong. She gives the following example:
(168) kojA hla puas yuav dhau?

2SG cross Q IRR pass.over
‘Can you get across (the river)?’ (Riddle, 1989, p. 4)
Riddle claims that the fact that sentences like this clearly express two separate
propositions (the act of crossing the river and the attainment of that goal, in the
example above) lends support to her argument that serial verb constructions in
this language can be multi-propositional. Her argument is based largely on the
assumption that, if two verbs can be used to express separate propositions in
any one construction, then they must be able to express separate propositions
in every construction in which they appear together. However, it is argued here
that when two verbs do express separate propositions they are not in a serial
relationship, but rather belong to separate clauses. However, when verbs occur
together in an Attainment SVC, or any other SVC type, then they are working
together to express a single proposition. Even in the absence of other syntactic
or lexical evidence of the ‘separateness’ of the clauses, a multi-propositional,
164 CHAPTER 3
multi-clausal structure will always allow a pause between the juncts. A serial
verb construction will not.
3.3.6 Nexus and Juncture in Attainment SVCs

The precise relationship between the verbs in this type of SVC in White Hmong
can be determined by examining the occurrence restrictions on arguments
and the scope of certain operators. As noted, no NP appears between the verbs
in an Attainment SVC, even if the first verb is transitive. A Noun Phrase or
Locative Phrase appearing after the final verb is always an argument of this
final verb. If the first verb is transitive, such a noun is also an argument of that
verb. The NP liab ‘monkey’, in the first example below, is the object of the first
verb tua ‘shoot at’, as well as of the final verb raug ‘hit the mark’. However, this
object NP cannot appear between the two verbs, as shown by the ungrammati-
cality of the (b) example below:
(169) a. nwsA/A tua raug liabO/O

3SG shoot.at hit.the.mark monkey
‘He shot a monkey.’
b. * nws tua liab raug

3SG shoot.at monkey hit.the.mark
This ungrammaticality cannot be attributed to any transitivity requirements of

the final verb, as this example is equally unacceptable even if a co-referential
NP appears as the object of raug ‘hit the mark’:
(170) * nws tua liab raug cov

3SG shoot.at monkey hit.the.mark CLF.PL
for: ‘He shot at some monkeys and hit some.’
The fact that no NP intervenes between the two verbs in this construction,
even if the first verb is transitive, suggests that the juncture relationship is at
the nuclear level. The fact that the verbs in this construction are joined at the
nuclear level explains how it is that, even though their relationship is described
in terms of syntactic rather than lexical rules, they so clearly act as joint predi-
cators within the clause. The two verbs are first linked to one another, and then
function as a single unit with respect to all arguments and higher-level opera-
tors within the clause.
productive svcs 165
This linkage of the verbs at the nuclear level is also related to the function
of this construction. As noted, an Attainment SVC serves to indicate the suc-
cessful attainment of the goal of the action described by the first verb. This is
clearly basically an aspectual function, performed in some other languages by
the use of aspectual morphemes such as the perfective. Aspect is, of course,
primarily a nuclear layer phenomenon, referring to the temporal constituency
of the action described by the predicate. So it is no surprise to find that a serial
verb construction that performs a basically aspectual function involves linkage
of the verbs at the nuclear level.
3.3.6.2 Nexus
In order to determine the nexus relationship between two verbs joined at the
nuclear level, it is normal to observe the use of nuclear layer operators such
as aspectual morphemes. Verbs joined at the nuclear level that can be jointly
modified by aspectual operators are in a cosubordinate relationship; those that
can be independently modified by such operators are in a coordinate relation-
ship. However, the relationship between verbs in an Attainment SVC is, itself,
a basically aspectual one, and further aspectual modification does not tend
to occur.
It is possible for certain adverbs to modify the first verb in an Attainment
SVC independently of the second, as in examples (171) and (172) below:
(171) Lwj NeemA thiaj.li los <ntaus nphawv-nphoov

PN so come.home hit bang-bong
tau [ib daig hauv-siab hlau]O/O <ua tsho-khuam
get one CLF chest iron make vest
kiag . . . >PURP>PURP
INTNS
‘So Leu Neng came home to bang out a sheet of iron to make a vest . . .’
(DNH 55: 173)
(172) kaj-ntug [tus tub]A/A nco dheev txog [nws

dawn CLF boy remember suddenly arrive 3SG
rab hneev]O/O
CLF crossbow
‘At dawn, the boy suddenly remembered his crossbow.’ (KNH 3)20
20 See Chapter 4, §4.2.3.5 for justification of this type of example as an Accomplishment

SVC.
166 CHAPTER 3
The fact that these adverbs have scope over only one of the two verbs in exam-
ples like these simply shows that they are nuclear-layer modifiers (that is, part
of the periphery to the nucleus in RRG terms (Van Valin, 2005, p. 21)). As these
adverbs are not operators, they provide no evidence regarding the type of
nexus involved.
In the absence of any clear evidence that a coordinate relationship is
involved, it seems reasonable to conclude that the verbs in an Attainment SVC
in White Hmong are in a nuclear cosubordinate relationship.
3.3.7 Grammaticalization of V2 tau to Express Deontic Modality ‘can’

In keeping with a parallel pattern in numerous languages across the region
(Enfield, 2003), in White Hmong the verb tau ‘get, come to have’ has grammati-
calized in the post-verbal position to express the deontic modality meaning
‘can’. This is illustrated in (173) below.
(173) kojA puas noj tau kua.txobO?

2SG Q eat can red.pepper
‘Can you eat red peppers?’
The syntactic patterning of this use of tau indicates that it has developed from
the use of an achievement verb as V2 in an Attainment SVC. It is particularly
reminiscent of examples such as those instances of the Attainment SVC illus-
trated in §3.3.2.2.d above, in which V2 indicates the successful performance of
the action described by V1, even when that action has no particular goal itself.
While V2 in an Attainment SVC must express a goal that is closely related to the
nature of V1, this grammaticalized, modal use of tau can be used with a wide
range of verb types and indicates simply that a goal is achieved, without speci-
fying what kind of goal it is. The only possible interpretation is that the ‘goal’
is the performance of the action itself: ‘Can you achieve eating red peppers?/
Can you eat red peppers?’
This grammaticalized postverbal use of tau to express the deontic modal-
ity meaning ‘can’ is extremely common in White Hmong. Enfield (2003, p. 193)
gives the examples below to show that it can be used in a range of senses,
including not only ability but also permission and possibility:
(174) kuvA hais tsis tau [lus Nyab-laj]O

1SG speak NEG can word Vietnamese
‘I can’t speak Vietnamese.’
productive svcs 167
a. . . . niam tsis pub <hais>

mother NEG give speak
‘. . . (my) mother won’t let (me) speak (it).’
b. . . . kuv tsis tau kawm

1SG NEG ACHV study
‘. . . I haven’t studied (it).’
c. . . . txaj-muag rau lawv

be.ashamed to 3SG
‘. . . I’m ashamed (to speak) (it) to/for them.’
A positive sentence with postverbal tau is often used to describe an actual state
of affairs, not just a potential one. Compare (175) and (176) below:
(175) lawvA txawj hais [lus Hmoob]O

3PL know.how speak word Hmong
‘They are able to speak Hmong.’
(176) lawvA hais tau [lus Hmoob]O

3PL speak can word Hmong
‘They can speak Hmong.’
Often understood as: ‘They speak Hmong.’
In example (175) above, txawj ‘know how to’ is a Modal verb expressing ability.
There is no indication in this sentence as to whether those referred to actually
do speak Hmong; we are only told that they know how to do so. The correspond-
ing example (176) with tau, on the other hand, is often taken to express not just
a potential, but an actual state-of-affairs: ‘They speak Hmong’. As Enfield (2003,
p. 40) explains,
An important aspect of ‘possibility’ is its logical and pragmatic relation-

ship with ‘attainment’, i.e. the realisation in fact of a predication.21 The
ideas ‘q can happen’ and ‘q happens’ are naturally close.
21 Note that Enfield’s use of the term ‘attainment’ to refer to “the realisation in fact of a
predication” (2003, pp. 40–41) is somewhat different from the way the term is used in this
study, to refer to the ‘attainment’ of a goal (either intrinsic or extrinsic) of a predicate.
168 CHAPTER 3
Enfield (2003, p. 194) goes on to note that, in this and many other ways with
respect to the grammaticalization of a word originally meaning ‘get, come to
have’, Hmong is similar to other languages of the region. However, it differs
from most with regard to the position of the grammaticalized element. That
is, the position of tau immediately after the verb and before the object, which
is the norm in Hmong, is unusual amongst languages of the area; in many of
the other languages Enfield discusses, including Lao, Khmer, and Kmhmu
Cwang, a morpheme with a similar meaning and usage appears after rather
than before the object of the verb.
Perhaps due to the influence of the pattern in these other languages, as well
as due to its grammaticalization, tau can actually appear in both positions in
White Hmong: not only directly after the verb (in the same position as V2 in
an Attainment SVC) but also outside the core, after the verb and its object.
This position is most likely when it is modified by a negative or interrogative.
Enfield (2003, p. 205 and 218) gives the following examples of the alternation
between the two possible orders, noting (p. 218) that the appearance of tau
closer to the verb is more idiomatic:
(177) a. kojA mus puas tau tajlajO

2SG go Q can market
‘Can you go to the market?’
b. kojA mus tajlajO puas tau

2SG go market Q can
‘Can you go to the market?’
(178) a. kuvA haus tsis tau cawvO

1SG drink NEG can alcohol
‘I can’t drink alcohol.’
b. kuvA haus cawvO tsis tau

1SG drink alcohol NEG can
‘I can’t drink alcohol.’
Enfield (2003, pp. 218–219) suggests that the preferred placement of tau imme-
diately after the verb in Hmong may be due to the construction being at a
more advanced stage of grammaticalization in this language in comparison
to its neighbours. He suggests that this is a “distinguishing behavioural feature
perhaps iconic of increasing semantic ‘proximity’ to the verb entailed by the
modal function ‘can’.”
productive svcs 169
In this analysis, however, the preference for the placement of the modal
use of tau immediately after the verb is attributed to its origin as V2 in the
Attainment SVC, and it is its optional appearance outside the core that is con-
sidered to be a possible result of the grammaticalization process. These two,
related issues are discussed in turn below.
While many of the other languages discussed by Enfield (2003) have a single
basic ‘resultative’ pattern (A V1 O V2), White Hmong has two distinct construc-
tions—the Cause-Effect SVC (A V1 O V2) and the Attainment SVC (A V1 V2 O).
Compare the examples below from Lao and White Hmong:
(179) Lao—Different-Subject Resultative (Enfield, 2003, p. 417)

man2A tii3 [pêt3 toø-nan4]O/S taaj3
3SG hit duck CLF-that die
‘He hit that duck dead.’
(180) White Hmong—Cause-Effect SVC

Yawg LwvA ntaus Yawg LiajO/S tuag tag . . .
PN hit PN die finish
‘After Old Leu had hit Old Lia dead . . .’ (DNH 304: 10)
(181) Lao—Projected Resultative (Enfield, 2003, p. 416)

man2A/A haa3 kacèè3O/O hên3 lèèw4
3SG seek key see PRF
‘He’s found the key.’
(182) White Hmong—Attainment SVC

nwsA/A nrhiav pom [nti nplhaib]O/O lawm
3SG seek see CLF ring PRF
‘He’s found the ring.’
Examples (179) above and (180) above, from Lao and White Hmong respec-
tively, are similar in that they both express a relationship of direct causation
between V1 and V2 and both exhibit the pattern A V1 O/S V2, with O/S the
shared argument. Enfield (2007) refers to the Lao example, (179), as a ‘differ-
ent-subject resultative’, and in this analysis I term the White Hmong equiva-
lent, (180), a Cause-Effect SVC. Examples (181) from Lao and (182) from White
Hmong, however, involve another type of relationship between the verbs: in
this case both A and O are shared and the second verb describes the successful
outcome of the goal of the first. In this case the Lao example, (181)—termed by
Enfield (2007) a ‘projected resultative’—exhibits the same basic pattern as the
170 CHAPTER 3
previous example, (179): A V1 O V2. The White Hmong pattern in (182), how-
ever, is different from that in the previous example, (180), with the two verbs
appearing immediately adjacent to one another and before the shared object
in this case: A V1 V2 O. This pattern is thus considered a completely different
construction, and referred to here as an Attainment SVC.
As discussed in §3.3.6.1 above, the function of V2 in this serial construction
in White Hmong is basically an aspectual one, involving nuclear-level juncture
and nuclear-level scope. As tau grammaticalized to developed a modal inter-
pretation with scope over not just the nucleus but the entire core, including
the object, its placement after the object presumably became more acceptable.
It is possible that the post-core placement of parallel morphemes in many of
the neighbouring languages may also have influenced the development of this
alternative pattern.
3.4 Disposal Serial Verb Constructions
3.4.1 Introduction
In the preceding section (§3.3 above), we saw that when two verbs in an
Attainment SVC are both transitive, both A and O arguments are shared. The
shared O argument appears after the second verb in the construction; it can-
not intervene between the two verbs. There is another type of SVC in White
Hmong—referred to here as a Disposal SVC—that also involves a sequence
of transitive verbs with both A and O again shared. In this type, however, the
shared O argument does intervene between the verbs. Observe the following
examples:
(183) nwsA/A pov [nws rab hneev]O/O tseg

3SG throw 3SG CLF crossbow leave
‘He threw his crossbow away.’ (KNH 2)
(184) kuvA nqis los es koj thiaj.li zoo tom

1SG descend come CONJ 2SG so be.good bite
kuvO/O noj
1SG eat
‘I will come down (the ladder) so you (can) gobble me right up (DNH
437: 24)
(185) a. kuvA/A tso nwsO/O tseg

1SG release 3SG leave
‘I abandoned him.’
productive svcs 171
(186) nwsA/A tsa [tus xyoob-tuam-tswm]O/O/ cia

3SG standTR CLF type.of.large.bamboo put.aside
‘He stood the bamboo pole to one side.’ (Adapted from DNH 30: 22)
(187) luagA/A muab nwsO/O tsuav ua-lwj-ua-liam tag

others take 3SG chop.to.pieces recklessly finish
‘Others took it [the ever-living spirit] and chopped it right up into all
sorts of pieces.’ (DNH 52: 149)
This phenomenon—the appearance of a shared O argument between tran-

sitive verbs—has been observed in a number of other serializing languages,
including Twi (Christaller, 1875; Stewart, 1963), Ewe (Westermann, 1930),
Yoruba (Lord, 1974), and Sranan (Sebba, 1987). For example:
(188) Sranan (Sebba, 1987)

a. denA/A fon owrukukuO/O kiri
they beat owl kill
‘They beat owl to death.’ (p. 43)
b. miA/A fringi [a batra]O/O broko

I throw the bottle break.TR
‘I threw and broke the bottle.’ (p. 43)
c. KofiA/A fringi [a nyan]O/O trowe

PN throw the food eject
‘Kofi threw the food away.’ (p. 113)
d. KofiA/A teki [den krosi]O/O kibri

PN take the.PL clothes hide
‘Kofi hid the clothes.’ (p. 131)
3.4.2 The Semantic Relationship Between the Verbs

With regard to the semantics of the construction, it would be far more remark-
able if two or more transitive verbs in a putative SVC were to have no shared O
argument. This does indeed occur in some SVC-like constructions in Sranan,
but these would not be regarded as involving serial verb constructions in this
analysis.
(189) Kofi sutu Amba kiri Kwaku

PN shoot PN kill PN
‘Kofi shot Amba (but) killed Kwaku.’ (Sebba, 1987, p. 109)
172 CHAPTER 3
The sharing of an O argument, on the other hand, is precisely what would be

expected if two (or more) transitive verbs were, indeed, being used together to
express a single event. The two actions described are conceptually separable,
but are performed under circumstances in which a ‘single event’ interpreta-
tion is not only possible, but actually most likely. These circumstances are pre-
cisely those illustrated in examples (183)–(187) above from White Hmong (as
well as (188)a–d from Sranan). In each case, the action/event described by the
first verb in the construction leads up to, and culminates in, that described
by the second. There is no disjuncture between the two: the first precipitates
the second, and the second is contingent upon the first. When more than two
actions are involved, as we will see below, this pattern of linked actions simply
continues.
Looking more closely at the kinds of transitive verbs that occur in the exam-
ples above, we see that, in all cases, the subject (A) is an Agent and the object
(O), a Patient or Theme. Table 7 shows examples of the kind of verbs that occur,
and the order in which they appear.
table 7 Verb types and order in Disposal SVCs
1. Manipulate 2. Impinge / Impact 3. Dispatch 4. Dispose/ Consume/Transfer

(Transitive) (Transitive) (Transitive) (Transitive)
muab ‘take’ tua ‘kill/shoot’ tso ‘release’ tseg ‘abandon’

tsa ‘stand’TR hlawv ‘burn’TR pov ‘throw’ cia ‘set aside’
ziab ‘dry’TR xa ‘send’ noj ‘eat’
txhoov ‘cut’ w ‘scatter’ pub ‘give’
txiav ‘chop’ rau ‘put.in’ (→ ‘to/for’)
tsuav ‘chop.to.pieces’
tsoob ‘smash’
hlais ‘slice’
faib ‘divide’
tom ‘bite’
nplawm ‘whip’
The first kind of verb that can appear in the Disposal SVC is one that involves
the subject manipulating the object, taking it up or putting it down. By far the
most common verb to appear in this position is the verb muab ‘take, pick up’.
A wide range of verbs appear in the second slot—verbs that describe impinge-
ment or impact, often quite destructive in nature. The verbs in the third slot
productive svcs 173
are characterised as expressing ‘dispatch’, following the term used by Enfield

(2007, pp. 368–372) for somewhat similar sequences in Lao. Finally, verbs in the
fourth position describe ‘disposing’ of the object in some way: relinquishing
it, consuming it, or transferring it to anther location. Combinations of two or
more verbs from any slot can appear, but always in this order.
The term ‘Disposal SVC’, which is used in this study to refer to SVCs of this
type, is a borrowed one; it is used by Chinese linguists to refer to the kinds of
propositions expressed in a construction that bears close resemblance to one
particular type of Disposal SVCs in White Hmong—those with the verb muab
‘take’ (see §3.4.6 below for discussion).
3.4.3 The Position of the Shared Object

The intersection of cores is the hallmark of the level of juncture represented
by the Disposal Serial Construction—core juncture (see §3.4.4.1 below). The
kind of core juncture occurring in serial verb constructions—that is, non-
embedded core juncture—involves the sharing of at least one core argument.
In the Disposal SVC, not one, but two, core arguments are shared: the sub-
ject (A) and the object (O).
As noted above (Chapter 2, §2.4.1.3), a shared argument, by definition,
cannot occur in more than one syntactic position; the sharing of arguments
“involves the mapping of two identical arguments from two logical structures
into the same syntactic position in a clause” (Van Valin, 1987, p. 2). The nature
of the juncture—core level juncture—dictates that the two verbs do not occur
back-to-back, as they do in nuclear juncture. The shared O argument appears
directly after V1, in its normal position in relation to this verb. However, the
fact that it is a shared argument means that it will not appear again after V2.
Observe the examples below:

b. * kuv tso tseg nws

1SG release leave 3SG
c. * kuv tso [kuv tus kwv] tseg nws

1SG release 1SG CLF younger.brother leave 3SG
The ungrammaticality of example (190)b above shows that the semantic rela-
tionship between verbs in a Disposal SVC cannot be expressed in a nuclear
juncture. The ungrammaticality of example (190)c reinforces the fact that the
174 CHAPTER 3
O argument, like the A argument, is truly shared; it is not simply a coreferential

argument that can optionally be deleted after the second verb.
3.4.4 Nexus and Juncture in Disposal SVCs

The fact that the shared O argument intervenes between the verbs in a Disposal
SVC indicates that juncture cannot be at the nuclear level. The fact that no
anaphor of this argument can appear after the second verb indicates that junc-
ture is not at the clausal level; it is truly a shared argument, not a coreferential
argument that optionally undergoes zero anaphora in the second junct. These
points have been illustrated in example (190) above.
Recall that the sharing of at least one core argument occurs in (non-embed-
ded) core level juncture (see Chapter 2, §2.4.2). Disposal SVCs exhibit two
shared arguments: A=A and O=O. This construction involves not just core level
juncture, it involves SVC-type core cosubordinate juncture. This implies that
the two juncts each consists of a single core and together constitute a linked
core within single clause; it is not possible for each junct to contain indepen-
dent peripheral arguments. This is confirmed by the obvious unacceptability
of examples like the (b) example below:
(191) a. nwsA/A pov [khob dej]O/O tseg

3SG throw cup water leave
‘He threw the cup of water away.’
b. * nws pov [khob dej] [hauv [lub tsev]] tseg

3SG throw cup water inside CLF house leave
[nram [lub vaj]]
down CLF garden
‘He threw the cup of water in the house (and) left it down in the
garden.’
3.4.4.2 Nexus
As noted before (§3.1.3.2), subordinate elements, both NPs and adjuncts, can
appear sentence-initially as topics in White Hmong. The second junct of a
Disposal SVC, however, cannot appear in this position, showing that subordi-
nated nexus is not involved in this construction:

productive svcs 175
b. * tseg (nws) mas, kuv tso nws

leave 3SG TOP 1SG release 3SG
(193) a. [tus tsov]A/A tom [tus noog]O/O noj

CLF tiger bite CLF bird eat
‘The tiger gobbled the bird up.’
b. * noj (tus noog) mas, [tus tsov] tom [tus noog]

eat CLF bird TOP CLF tiger bite CLF bird
The scope of core level, modal operators confirms that, of the two non-
embedded nexus types, this construction involves a cosubordinate rather than
a coordinate relationship. This is because these operators have scope over the
entire construction, and can modify neither junct independently. Observe the
examples below:
(194) a. [tus tsov]A/A pib tom [tus noog]O/O noj lawm

CLF tiger start bite CLF bird eat PRF
‘The tiger has started to gobble up the bird.’
b. * [tus tsov] tom [tus noog] pib noj lawm

CLF tiger bite CLF bird start eat PRF
Lit: ‘The tiger has bitten the bird (and) started to eat.’
We can conclude from the evidence presented above that, as in Cotemporal

and Cause-Effect SVCs, the juncts in a Disposal SVC are in a Core Cosubordinate
relationship. This is clearly by far the most common type of juncture/nexus
relationship found in serial verb constructions in White Hmong, the only excep-
tion being the Attainment SVC, which represents a Nuclear Cosubordinate
relationship.
3.4.5 Disposal SVCs Compared with Other SVC Types

3.4.5.1 Comparison with Attainment SVCs
This type of serialization can be seen to have similarities to, and differences
from, Attainment SVCs in White Hmong. In both types it can be said that V1
‘leads up to’ V2 in some way. In the Attainment SVC, V2 describes the attain-
ment of a goal expressed by V1 and, where V1 is a verb that expresses either
an intrinsic or extrinsic goal, that goal is related in some way to attaining the
object. In the Disposal SVC it is true to say that V1 is directed toward V2, and
indeed, in some cases (though certainly not all), that V2 is the goal of V1. For
176 CHAPTER 3
example, one may pov rab hneev ‘throw/cast away the crossbow’ with the goal
of tseg ‘leaving’ it behind, or xaws daim tiab ‘sew the skirt’ with the goal of pub
kuv ‘giving it to me’. However, it is quite clear that, even in these cases where
the second verb may be thought of as indicating a goal of the first, that goal is
not related in any way to attaining the object, as in the case of the Attainment
SVC. On the contrary, as noted above, the actions described actually result
instead in destroying, abandoning, or relinquishing the object.
3.4.5.2 Comparison with Cause-Effect SVCs

Disposal SVCs are syntactically similar to Cause-Effect SVCs in that a shared
argument intervenes between the verbs. However, the Disposal SVC differs
from the Cause-Effect, in that V2 governs, rather than predicates, this shared
argument. Like V1, V2 in a Disposal SVC predicates the subject. In the Cause-
Effect SVC, on the other hand, V2 predicates the object of V1, and has no predi-
cation relationship with the subject, as in the following example:
(195) pov pobO/S los rau kuv

throw ball come to 1SG
‘Throw a ball to me.’ (Cause-Effect SVC)
Compare:

3SG throw 3SG CLF crossbow leave.TR
‘He threw his crossbow away.’ (KNH 2) (Disposal SVC)
In the first example above—the Cause-Effect SVC—V2 is intransitive, and the

shared argument pob ‘ball’ is simultaneously the O argument of V1 and the S
argument of V2 (O=S). In the second example—the Disposal SVC—V2 is tran-
sitive, and nws rab hneev ‘his crossbow’ is the O argument of both verbs (O=O).
Furthermore, in Disposal SVCs, this shared argument is only one of two shared
arguments, the other being the subject of the verbs (A=A).
One respect in which Disposal SVCs resemble Cause-Effect SVCs is in the
strong propensity for the O argument to appear initially, as the object of the
verb muab ‘take’, as in example (187). Here is another example:
(197) nwsA hais <tias kuvA/A/A muab nwsO/O/O tso tseg>COMP

3SG say COMP 1SG take 3SG relinquish leave
‘He told me to abandon him.’ (Jaisser, 1989, p. 6)
productive svcs 177
The use of this construction is particularly common when a Disposal SVC

is composed of more than two transitive verbs. Example (198) shows four
verbs appearing in the order indicated in Table 7: manipulation (muab ‘take’),
impingement (hlawv ‘burn’), dispatch (pov ‘cast away’), and disposal (tseg
‘leave’).
(198) muab hlaws pov tseg

take burn cast.away leave
‘Burn it up.’ (Heimbach, 1979, p. 233)
The use of muab ‘take’ in Disposal SVCs in discussed in more detail in §3.4.6
below.
3.4.6 Disposal SVCs with muab ‘take’

In examples (187) and (198) above, muab appears with one or more other tran-
sitive verbs that are themselves clearly in a ‘disposal’ relationship. It is very
common for the verb muab ‘take’ to occur as V1 in a Disposal SVC of this type.
An further example is given here:
(199) ces txawm muab pov rau tshav.ntuj . . .

CONJ then take throw put.in sunlight
‘. . . so then (they) took (it) and threw (it) into the sunlight.’ (DNH 12: 61)
In the examples given above, the object is physically manipulated by being

literally picked up by the subject. However, there are many examples in which
muab ‘take’ clearly has a similar function, but in which it does not indicate that
the object was actually ‘picked up’. Observe the following:
(200) nwsA/A muab poj-niamO/O nrauj lawm

3SG take woman divorce PRF
‘He has divorced his wife.’ (Heimbach, 1979, p. 174)
Similar use, both literal and non-literal, of a verb meaning ‘take’ (or the like) has
been observed in many serializing languages. One of the most widely discussed
constructions of this type is the so-called ‘bǎ Construction’ in Mandarin. The
morpheme bǎ actually no longer occurs as a main verb in modern Mandarin.
However, its historical credentials as a verb meaning ‘take’ are well established
(Li & Thompson, 1974a).22
22 Unlike the morpheme bǎ in Mandarin, the morpheme muab ‘take’ is still used as a verb in
White Hmong. This is shown in the following example:
178 CHAPTER 3
The term ‘disposal’ was first used, and is still used, to refer to the nature
of the proposition expressed by the ‘bǎ Construction’ in Mandarin. In this usage
the term has far wider implications than those mentioned so far in the discus-
sion of Disposal SVCs in White Hmong. While ‘disposal’ may, indeed, refer to
sentences that indicate that the object is ‘disposed of’ in the sense described
above—of being destroyed, consumed, or relinquished—this degree of
impingement certainly need not occur in the ‘bǎ Construction’. The Chinese
grammarian Wang Li gives the following explanation (Li & Thompson, 1981,
p. 468):
The disposal form states how a person is handled, manipulated, or dealt

with; how something is disposed of; or how an affair is conducted.
Li and Thompson agree, adding: “roughly, disposal has to do with what happens
to the direct object” [emphasis in the original]. Thus the ‘bǎ Construction’ in
Mandarin is most likely to be used with action verbs rather than state verbs. It
is also most likely to be used when the object is definite or generic, and is thus
somewhat topical. (See Li and Thompson (1981, pp. 463–491) for full explana-
tion and exemplification, and for discussion of some interesting exceptions.)
Disposal SVCs with the verb muab ‘take’ in White Hmong are like sentences
with the ‘bǎ Construction’ in Mandarin in that they exhibit the property of
disposal in this wider sense. It is not necessarily the case that the object is
significantly impinged upon by the action described; the sentence simply
indicates what happens to the object. Thus, some muab sentences describe
something about the way in which a certain object is dealt with, without any
indication that it was affected or changed by the action in any significant way.
For example:
(201) muab lawvO/O cais raws.li [[lawv lub cev]

take 3PL separate according.to 3PL CLF body
thiab [lawv qhov-feem]]
and 3PL type
‘(You) classify them according to their shape and their type.’ (WHG 11)
(i) yusA muaj [ib tug tub]O, yuav muab nyabO

INDF have one CLF(t.c.) boy IRR take son’s.wife
‘(Suppose) one has a son, (and he) intends to take a wife.’ (THW 102)
productive svcs 179
(202) nkawmA/A/A muab hu ua Muam.Nkauj.Liag no

3DU take call make PN IP
‘. . . they named (their daughter) Mua Ngao Lia.’ (DNH 449: 1)23
(203) kuvA/A/A muab faib ua npaum li no hom . . .

1SG take divide make equal as this CLF(kinds)
‘I divided (them [the words]) into this many classes.’ (WHG 9)
The last two examples above also serve to illustrate the fact that, in many cases,
a Noun Phrase referring to the object does not actually appear in the sentence
with muab at all, having undergone zero anaphora as a result of its high degree
of topicality. As for object NPs that do appear, these do, indeed, tend to be
definite, as in all of the examples above. Personal pronouns commonly occur,
again emphasising the frequent topicality of this position.
However, as discussed in §3.4.7 below, a construction very similar to this one
is used to introduce both a theme and a goal argument into the clause and, in
this case at least, it is possible for an indefinite, non-specific NP to appear. This
is illustrated in example (204):
(204) ces mam muab nyiaj.txiagO/O tsab rau

CONJ so take money offer to
niam.ntxawm.txiv.ntxawm niam.hlob.txiv.hlob
junior.paternal.aunt/uncle senior.paternal.aunt/uncle
‘. . . and so (you) offer money to the paternal aunts and uncles.’
(Adapted from: THW 102)
3.4.7 V2 Introducing an Unshared Argument

In §3.4.3 above, we saw that the shared O argument in a Disposal SVC appears
directly after V1, and does not appear again after V2. It is possible for an object
to appear after V2 in a construction that seems to be derived from this type of
SVC. Such an object can only appear, however, if it is not a shared argument.
For example, a Noun Phrase expressing a Recipient can appear:
(205) nwsA/A xa [ib qho khoom]O/O pub wbO

3SG send one CLF goods give 1DU
‘She sent some things to us (as a present).’
23 The word nkawm 3DU ‘a pair’ actually appears as nkaws in the reference cited. This is
assumed to be a typographical error.
180 CHAPTER 3
(206) [kuv niam]A/A xaws [ib daim tiab]O/O pub kuvO

1SG mother sew one CLF skirt give me
‘My mother sewed a skirt (and) gave (it to) me.’
These examples are interesting in that they show how this type of construction
allows the second verb (pub ‘give’) to govern more than one object—a Theme
and a Recipient—without violating the condition that no verb can simultane-
ously introduce both a Theme and a Recipient into the clause in White Hmong
(see Chapter 1, §1.6.6). The Theme argument of pub ‘give’ is a shared argument,
introduced by V1 and so not repeated after V2. Only the Recipient argument
appears after V2, as it is unshared.
These examples are also interesting in that they show how this construction
can facilitate the expression of a Recipient for verbs like xa ‘send’, which and
cannot, by themselves, introduce both a theme and a Recipient argument into
the clause. The construction can also facilitate the expression of a Recipient
for verbs like xaws ‘sew’, which can neither govern a Recipient argument, nor
appear in a true Disposal SVC at all. Whereas a Beneficiary or Recipient can be
expressed as a direct object of many action verbs in the so-called ‘Dative Shift’
construction in English, this is not possible in White Hmong:
(207) * nws xa kuv [qho khoom]

3SG send 1SG CLF goods
Intended meaning: ‘She sent me (some) things.’24
(208) * [kuv niam] xaws kuv tiab

1SG mother sew 1SG skirt
Intended meaning: ‘My mother sewed me a skirt.’
Although the verbs xa ‘send’ and xaws ‘sew’ cannot themselves introduce
Recipient NPs into these clauses, an argument with that semantic role can be
brought into the clause by another verb, pub ‘give’. A similar construction is
24 Example (207) is grammatical with an alternative parsing:

(i) nwsA xa [kuv qho khoom]O
3SG send 1SG CLF goods
‘She sent my goods.’
This alternative parsing is not possible for example (208); the sequence kuv tiab (1SG skirt)
cannot be interpreted to mean ‘my skirt’ without a classifier: kuv daim tiab (1SG CLF skirt)
‘my skirt’.
productive svcs 181
used in many other serializing languages with the same function. The example
below is again from Sranan (Sebba, 1987, p. 113):
(209) KofiA/A tyari [a nyan]O/O gi [en mama]O

PN carry the food give his mother
‘Kofi brought his mother the food.’
See Chapter Four, particularly Chapter 4 §4.2.3.4, for further discussion.
3.5 Conclusion to Chapter Three
3.5.1 SVCs Express Various Facets of a Single Event

One of the main purposes of this chapter has been to show how each of the
different types of serial construction serves, in its own way, to portray the vari-
ous facets of a single event. It may well be the case that the individual predi-
cates involved could be used to describe more than one event, provided they
occurred in separate clauses; however, when they occur in a single clause, in
a serial construction, they are portrayed as constituting a single event. This
issue has been discussed in some detail with regard to each of the SVC types
described above; it will be helpful to summarise that discussion here.
3.5.1.1 Cotemporal SVCs

Cotemporal SVCs in White Hmong (see §3.1 above) can be divided into two
main types: those which refer only to motion, and those which refer to action:
motion and action, stance and action, or cotemporal actions.
Those which refer only to motion distinguish the following aspects of the
motion event, each of which is indicated by a separate verb in the series, and
any number of which may be expressed in a single series: manner of transport;
manner of locomotion; Path of motion; Source of motion; ‘returning’ to a pre-
vious location; deixis/Goal of motion. While these facets of the motion event
can be conceptually separated, they clearly constitute elements of a single
motion event, rather than a series of distinct actions.
Those Cotemporal SVCs that refer to both motion and action, or stance and
action, simply serve to describe the motion or posture that the subject adopts
while performing the relevant action. The subject is not portrayed as doing two
things, but rather doing one thing while adopting a certain posture, or moving
in a certain way or in a certain direction.
It seems that serial verb constructions in White Hmong cannot usually
be used to refer to cotemporal actions: doing two things at once. The only
182 CHAPTER 3
e xceptions are when the actions occur together so commonly, perhaps univer-
sally or perhaps particularly in the Hmong cultural milieu, they are thought of
as practically inseparable. Actions of this type include ‘preparing vegetables
and preparing rice’, ‘crying and scolding’, ‘swaying the body (dancing) while
playing the bamboo pipes’ (see §3.1.2.3 above). Apart from these kinds of
examples, neither simple cotemporal nor simple sequential actions occur in
serial verb constructions in White Hmong.
3.5.1.2 Cause-Effect SVCs

Indirect causation can be thought of as involving two distinct, though closely
related, events: the ‘causing event’ and the ‘caused event’. However, Cause-
Effect SVCs in White Hmong (see §3.2 above) were shown to convey an
extremely direct form of causation: V1 expresses a transitive action, and V2,
its outcome. Less direct causation involving two distinct events, on the other
hand, is conveyed in White Hmong by various constructions in which the verbs
are not as tightly integrated.
The first verb in the Cause-Effect SVC is always an archetypal transitive verb:
an Affective, Effective, or Transfer verb. These actions result directly in some
change in the object, described by the second verb: change of state, change
of location, or inception of a new activity. This second verb is almost always
intransitive. Even in the few cases in which it is structurally transitive, it is a
transitive verb of a marked kind, with a Theme as subject.
3.5.1.3 Attainment SVCs

The Attainment SVC in White Hmong (see §3.3 above) functions to indicate
that the subject attains a goal. The attainment of the goal is expressed by V2,
an achievement verb, while the goal itself is expressed by V1 (along with the
object argument). This first verb may be of the kind that expresses an action
that is directed toward the subject ‘attaining’ the object in some way: an activ-
ity predicate that has an extrinsic goal; a process-oriented, effective accom-
plishment predicate, in which the intrinsic goal is for the subject to create or
‘get’ something new; a motion predicate that has the intrinsic goal of reaching
a destination. An Attainment SVC with this type of V1 means simply that the
extrinsic or intrinsic goal of V1 is achieved: the object is ‘attained’. Finally, in a
small number of instances, V1 may be a verb that does not express an action
directed toward the subject attaining the object (e.g. an activity predicate that
has no goal). In these cases it is the achievement verb (V2) in the context of this
construction that is responsible for the notion that the subject attains a goal.
The interpretation is that this ‘goal’ is the performance of the action described
by V1 itself. It is this final function of the Attainment SVC that seems most
productive svcs 183
likely to have led to the grammaticalization of tau ‘get, come to have’ to express
the deontic modality meaning of ‘can’.
There is no way in which the two verbs in this type of SVC could possibly be
thought of as expressing separate events. The second verb in this construction
follows directly upon the first, and indicates either the attainment of its goal
or the attainment of its performance. It was noted, in fact, that a similar con-
struction in Mandarin has been treated as involving compounding of verbs,
rather than serialization. Even though this is dealt with as a syntactic, rather
than a lexical, phenomenon in the present analysis of White Hmong, the fact
that serialization is seen as taking place at the nuclear level in this construc-
tion means that the verbs are functioning as a single unit with respect to all
arguments and higher level operators within the clause. The two verbs clearly
represent two different facets of the same event.
3.5.1.4 Disposal SVCs

The verbs in a Disposal SVC (§3.4 above) are both transitive and share two argu-
ments: A=A and O=O. The event described by the first verb is understood to lead
up to, and culminate in, the event described by the second. In the majority of
examples of this construction, both verbs express a high degree of impingement
on their object, and the second indicates the way in which the object is ‘disposed
of’ by the subject—whether it be destroyed, consumed, or relinquished.
However, in Disposal SVCs in which the first verb is muab ‘take’, it is not
necessarily the case that the object is significantly affected by the action
described. These sentences may simply indicate something about the way in
which the object is dealt with by the subject. Disposal SVCs with muab ‘take’ in
White Hmong are similar to sentences with the ‘bǎ Construction’ in Mandarin,
and to serial verb constructions with a verb meaning ‘take’ in many other seri-
alizing languages, in that they function to present a topical object, and to say
what ‘happens’ to it. Once again, it is clear that the verbs that appear together
in this SVC type serve to express different facets of a single event.
3.5.2 Linked Serial Verb Constructions

It has been shown in the discussion in this chapter that, while some SVC types
in White Hmong are limited to just two verbs, others can involve a consider-
ably larger number of verbs in series. Cotemporal Motion SVCs, for example,
can contain both transitive and intransitive verbs, each of which expresses a
different facet of the motion event: manner of transport, manner of locomo-
tion, Path, Source, ‘Return’, and Goal (see §3.1.1.2 above). The Disposal SVC can
also consist of more than two verbs, all transitive, expressing meanings such as
manipulation, impingement, dispatch and disposal (see §3.4 above).
184 CHAPTER 3
The other two kinds of serial verb constructions described in this chapter
really do each have only two ‘facets’ to the events they convey. In the case of
Cause-Effect SVCs these are, first, a transitive action and, second, its outcome.
In Attainment SVCs they are, first, the action (activity or accomplishment) that
involves some kind of goal and, second, the attainment of that goal.
It is quite normal, however, for SVCs involving any one of these types to
be ‘interwoven’ or ‘linked’ together, sometimes resulting in serial verb con-
structions of considerable length and complexity. This mechanism is used,
particularly in narrative texts, to provide a multi-focal construal of the events
described. Linked serial verb constructions can facilitate focus on both the
process involved in an event and its outcome or result, within a single clause.
This is shown in example (210):
(210) Cotemporal plus Attainment

ces [lub nruas ntawd]S/S/S/A thiaj maj.mam poob
CONJ CLF drum that so slowly fall
rov qab los txog [hauv ntiaj-teb]O . . .
return back come arrive inside earth
‘And so then that drum came slowly falling back to earth, . . .’ (DNH 115: 5)
In this example, two SVCs share the same subject, lub nruas ntawd ‘that drum’.
A Cotemporal Motion SVC comes first—poob rov qab los ‘come falling back’—
and the final verb in this SVC—los ‘come (home)’—then becomes the first
verb in an Attainment SVC los txog hauv ntiaj-teb ‘arrive here on earth’, which
is added to express the ultimate outcome of the motion event. The shared sub-
ject—lub nruas ntawd ‘that drum’—functions as the S argument of the first
three verbs—poob ‘fall’, rov ‘return’, and los ‘come’—and as the A argument of
the final verb—txog ‘arrive’—which introduces the destination—hauv ntiaj-
teb ‘earth’—as its O argument.
As seen in example (211) below, all types of core argument—A, S, and O—
can be shared across linked serial verb constructions. The order in which
shared arguments appear will, of course, depend on the order and nature of
the component SVCs involved:
(211) Disposal plus Cause-Effect plus Attainment

hmoobA/A txawm muab nrojO txiav caj-dabO/S/S poob
Hmong then take weed chop neck fall
cuag li cas
reach like how
‘The Hmong [farmer] then chopped the weeds’ necks [so that they]
fell right down like anything! (DNH 125: 26)
productive svcs 185
Example (211) begins with a Disposal SVC, hmoob muab nroj txiav ‘the Hmong
farmer took the weeds and chopped (them)’, which focuses on the process in
which the shared A argument hmoob ‘the Hmong farmer’ deals with the spe-
cific, shared O argument, nroj ‘weed’ in a highly destructive manner. The sec-
ond transitive verb of this Disposal SVC, txiav ‘chop’, then becomes the first
verb in a Cause-Effect SVC, txiav caj-dab poob (chop neck fall). This Cause-
Effect SVC serves to pivot the focus away from the process initiated by the A
argument, hmoob ‘the Hmong farmer’, and direct it towards the outcome of the
event. This outcome relates to the shared argument in this SVC, caj-dab ‘neck’,
which functions as the O argument of txiav ‘chop’ and the S argument of poob
‘fall’. The verb poob ‘fall’ then becomes the first verb in an Attainment SVC poob
cuag ‘fall reach’, in which caj-dab ‘neck’ functions as the S argument of both
verbs. This final SVC serves to reinforce the outcome of the event for the object:
the ‘necks’ [of the weeds] do not just ‘fall’, but ‘fall right down’.
Because Attainment SVCs involve nuclear level juncture, they can be linked
either after another SVC type, as shown in examples (210) and (211) above, or
linked within another SVC type:
(212) Cotemporal plus Attainment

[kuv txiv lawv]A/A/S twb coj tau [cuaj
1SG father 3PL REAL take.along get nine
tug npua]O/O tuaj <tua>PURP
CLF(t.c.) pig come kill
‘My father ‘n them succeeded in bringing along nine pigs to kill . . .’
(TNN p. 4)
Example (212) begins with an Attainment SVC, coj tau ‘take.along get’, in which
both A and O arguments are shared. This Attainment SVC constitutes the ini-
tial junct of a Cotemporal SVC: [coj tau] tuaj ‘[bring get] come’, which shares
the subject kuv txiv lawv ‘my father ‘n them’, the A argument of the first junct
and the S argument of the second. The final, telic verb in this linked SVC, tuaj
‘come’ then introduces a Purpose construction: [[coj tau] tuaj] tua ‘succeed in
bringing along (in order to) kill’ (see Chapter 5, §5.5 below).
In spite of their detail and complexity, linked serial verb constructions in
White Hmong can still be seen to refer to a single event. They thus differ from
the phenomenon of juxtaposed serial verb constructions used to describe
rapid sequences of events in narrative, described by Bisang (1992, p. 7) (cited in
Bisang (2009, p. 807)) and exemplified from Khmer:
186 CHAPTER 3
(213) tɤ̀ːp [stùh tɤ̀] [deɲ cap] [yᴐ̀ːk mᴐ̀ːk ʔaop]

then jump.up go pursue catch take come embrace
‘Then [she] jumped up, caught [the duckling] and took it into her
arms.’
Bisang explains that this kind of juxtaposition of serial verb constructions is a

discourse phenomenon used for “expressing rapid sequences of events” (similar
to what Pawley (2009, pp. 121–122) describes as a ‘narrative SVC’). Examples like
(213) in Khmer thus do not meet the criterion of referring to a single event and
so cannot be analysed as together constituting a single serial verb construction.
Linked serial verb constructions in White Hmong, on the other hand, do
refer to a single event and do constitute a single clause. Each SVC junct shares
with the other junct(s) at least one core argument, all peripheral arguments,
and all core and clause-layer operators; there is neither a pause nor any overt
form of linkage between the juncts and the juncts can be thought of as repre-
senting different facets of a single event.
The multi-focal construal facilitated by linked serial verb constructions in
White Hmong works alongside other strategies, such as the use of elaborate
expressions, adverbials, repetition, and so on, to enhance the vividness of the
portrayal of an event in a narrative text and to give a strong sense of immediacy
and involvement.
3.5.3 Relationship Between Lexical Complexity and Serialization

It has been noted that White Hmong verbs have a fairly low degree of lexi-
cal complexity in some domains: the notion of deixis is never incorporated
into the meaning of a verb of motion; an effective outcome is not taken for
granted even with verbs that express a high degree of impingement; the notion
of attainment of the telic goal is not an integral part of an accomplishment
verb; no single predicate, even one that describes an event involving three par-
ticipants, can simultaneously introduce both a Theme and a Recipient into the
clause. Instead, White Hmong employs the strategy of verb serialization, by
which a series of semantically not-so-complex predicates can come together in
a single clause to build up the expression of more complex and detailed events.
It has been observed that, where each type of serial construction described
above is used in White Hmong, in at least some cases a single lexical item might
be employed in a non-serializing language. This is certainly not to say that
verbs in series in White Hmong are simply substitutes for semantically and/
or syntactically more complex lexical verbs in a language that does not use
the strategy of verb serialization. However, it does draw attention to the fact
that, although more than one lexical verb is employed in serial c onstructions,
productive svcs 187
they do exhibit certain properties that would often be associated with a single
lexical verb, most notably the property of expressing what is construed as a
single event.25
Most significant, however, is not the fact that some serial verb constructions
can be translated by single lexical items in some languages, on some occasions,
but the fact that all serial verb constructions express a single event. Although
this construction clearly violates the Aristotelian notion of the correspon-
dence between the event and the predicate, it has been demonstrated in this
chapter that it supports the maintenance of the traditional notion of the cor-
respondence between the event and the clause.
25 Again, this is not to say that single lexical verbs necessarily express ‘simple’ events; on
the contrary, they may well express quite complex events, which in some cases may be
thought of in terms of a sequence of ‘sub-events’. This is easy to appreciate in the case of
an English verb like fetch, which involves motion in two directions as well as action, or
breakTR, which expresses both an action and a result. However, because these ‘sub-events’
are expressed by a single lexical item, it is relatively easy to appreciate the fact that they
are conceptualised as single events, in spite of their semantic complexity. In the case of
serial verb constructions it may not be so easy to appreciate the fact that they are concep-
tualised as single events, hence the comparison with semantically complex lexical verbs
like these.
CHAPTER 4
Valency-Increasing SVCs and Grammaticalization
4.1 The Phenomenon of Valency-Increasing SVCs
4.1.1 Adposition or Verbs?

The valency-increasing function of serialization is achieved by introducing a
second verb into the core of the clause—a verb that shares at least one core
argument with the other verb, but that brings with it another core argument.
It is not the valency of either verb that is increased by this process, but the
valency of the entire clause.
Many authors have observed that, in such a situation, one component of the
event may be evaluated as the ‘main event’ and the rest as the ‘circumstances’
(Givón (1975); Lord (1973)). Lord (1973, p. 293) gives an example:
Because location is generally considered to be less significant than other

aspects of events, in the serial construction the Locative verb phrase is
not as important as the other verb semantically (the meaning of the sen-
tence is clear . . .)1
The verb that is less semantically prominent and that marks the circum-
stances, may lose certain semantic and syntactic characteristics, and be reana-
lysed as an adposition. Instead of being expressed by a series of two verbs, the
event would now be expressed by one verb and one adposition. While the verb
that undergoes the change to adposition is semantically depleted, the one that
retains its verbal status may become more complex.
There has been considerable debate concerning this issue. In a language
that employs serialization, should a morpheme that is homophonous with
a verb and that serves to introduce the circumstances of an event into the
1 This distinction between the ‘main event’ or ‘more semantically prominent’ verb, on the
one hand, and the ‘circumstances’ or ‘less semantically prominent’ verb on the other, over-
laps with but is not identical to the distinction between ‘major’ and ‘minor’ verbs drawn by
Aikhenvald (2006b, p. 22). This latter distinction relates to whether a verb is from an open
class (a ‘major’ verb) or a closed class (a ‘minor’ verb). In many cases, where there is a distinc-
tion between the ‘main event’ and the ‘circumstances’ in a serial verb construction, the verb
that marks the main event would be a ‘major’ verb, and the one that marks the circumstances
would be a ‘minor’ verb.

Valency-increasing Svcs 189
clause still actually be regarded as a true verb, or must it necessarily be con-

sidered an adposition? The difference is an important one. If it is a verb in a
serial construction, then it is introducing an argument that is part of the core,
even though an argument fulfilling a similar role in another language may not
appear as a core argument. If it is an adposition, on the other hand, it may
introduce either a core argument, which is part of the logical structure of the
predicate (e.g. an inner locative), or a peripheral argument, which is not (e.g.
an outer locative, a time phrase, etc.).
The term ‘coverb’ has been widely employed to refer to morphemes that are
homophonous with a verb in the language concerned (or at least in some stage
of the history of that language), but that seem to fulfil a function similar to that
of an adposition in many other languages. This term has been used in stud-
ies of Chinese, and also adopted by linguists and teachers working elsewhere,
particularly on Southeast Asian languages and some African languages.2 Other
terms, for example ‘Verbid’ (Ansre, 1966), ‘Prepositional-Verb’ (Pawley, 1973,
pp. 142–147), and ‘Verbal-Preposition’ (Durie, 1988), have also been coined. In
some cases it is not completely clear whether the use of terms such as these
is meant to identify a new word class, separate from both verbs and adposi-
tions, or simply designed to avoid any clear commitment regarding what part
of speech these words really belong to.
Some linguists, notably Li and Thompson (1973, 1974a) and Clark (1978,
1979a), have argued that all ‘coverbs’ in the languages they describe should,
in fact, simply be regarded as derived prepositions. The reason why the term
‘preposition’, rather than the more general ‘adposition’, is used in the discussion
of this issue by these and many other authors is twofold. Firstly, the syntactic
typology of the verb serializing languages they have investigated—for example
Mandarin, Vietnamese, Thai, Khmer, White Hmong and Niger-Congo African
languages—is one in which the verb generally precedes the direct object. An
adposition that develops from such a verb would, consequently, also precede
its object; hence the term ‘preposition’. In a serializing language in which the
object precedes the verb (such as Tibeto-Burman languages like Lahu and Lisu,
or many of the languages of Papua New Guinea), on the other hand, an erst-
while verb might, presumably, develop into a postposition. However, the devel-
opment of verbs into postpositions is not, in fact, widely attested, and this is
2 In studies of Chinese, the term ‘coverb’ has often been used for such morphemes not only
when they do have a corresponding verb in the modern language, but also in cases where
the corresponding verb is no longer used as a verb (e.g. bǎ ‘Archaic Chinese ‘take’) (Li &
Thompson, 1973, p. 257).
190 CHAPTER 4
presumably the second reason for the general association of the term ‘coverb’
and others like it with ‘preposition’ rather than simply ‘adposition’.3
4.1.2 Criteria for Part-of-Speech Classification

“The shift from verb to preposition usually involves the depletion of some
semantic material out of the erstwhile verb.” (Givón, 1975, p. 82, emphasis in
the original). Givón gives examples in which these ‘erstwhile verbs’ cannot be
interpreted literally, such as the use of a verb meaning ‘give’ as a benefactive
case-marker, or one meaning ‘take’ as an accusative case-marker.
Some scholars view semantic change as evidence for part-of-speech change.
Sebba (1987, p. 74) claims, regarding the Sranan coverb gi ‘give/to’ in example
(1) below, that “it is not possible to interpret gi as having anything to do with
‘giving’ . . . and thus there is no a priori case for treating gi here as a verb at all.”
(1) fu gowtu no fadon gi gron

for gold NEG fall ground
‘So that the gold would not fall to the ground.’
Li and Thompson (1974a, p. 268) also present evidence of semantic depletion as

an argument in favour of a ‘coverb’ being regarded as a preposition. They show
how the morpheme ná in Mandarin need not have its literal verbal meaning
‘take hold’, as in example (2) below.
(2) tā ná shǒu cā hàn

he hand wipe sweat
‘He wiped away the sweat with his hand.’
Li and Thompson maintain that serial verb sentences and ‘coverb’ sentences
differ semantically in that the former represent two separate actions (Chapter 2,
§2.1.1) while the latter represent only one—that described by the main verb
and modified by the ‘coverb’. Thus the semantic change that takes place when
a verb (at least a non-state verb) becomes a preposition involves the loss of the
feature of action. Clark (1979a, p. 5) builds on this notion, claiming:
3 One possible explanation for this is that verb-final languages tend to favour nuclear rather
than core junctures in serial verb constructions (Foley & Olson, 1985, pp. 45–47). The phe-
nomenon discussed here is a core juncture phenomenon: one verb serving to introduce an
argument for another (rather than two verbs linked together and sharing all arguments, as in
a nuclear juncture.)
that it is precisely at that point at which a lexical item loses the feature
[+actn] that the item is a P and not a V, and . . . that, regardless of external
form (such as inflection and word order), a lexical item can be regarded
as V or P according to the presence or absence of a feature with respect
to ‘action’.
However, Thepkanjana (1986) suggests that this notion of ‘semantic bleaching’

in these contexts is the result of a failure to recognise that verbs are simply
being used in a metaphoric or figurative sense: referring to abstract rather than
concrete entities. For example, she says that the Thai verb hây ‘give’ has a literal
meaning of ‘to hand something to a person’ whereas, when used as a benefac-
tive, it refers to the transfer of a favour or service. The only valid argument for
claiming that a verb has been reanalysed as a grammatical morpheme, accord-
ing to Thepkanjana, is a change in its morphological or syntactic properties.
Thepkanjana is not correct in her assertion that the ‘coverbs-are-
prepositions’ school fails to recognise the non-literal sense of coverbs. The
authors discussed above are only too ready to acknowledge this. In fact this
seems to be precisely what they mean when they use terms such as seman-
tic ‘depletion’ or ‘bleaching’. The real difference of opinion lies in the issue of
whether the loss of a literal or concrete interpretation actually constitutes one
criterion for assuming a change in part-of-speech classification; Thepkanjana
believes that it does not.
The view adopted in this study is that the meaning and function of a mor-
pheme and its grammatical behaviour are inextricably related, but that this
should not be taken to imply that meaning and function can be used alone
to determine part-of-speech classification. In the absence of formal evidence,
how can one determine what degree of semantic variation from some notion
of ‘core meaning’ is sufficient to comprise change in word class? When is varia-
tion simply abstraction or metaphor, and when is it a sign of something more?
It is axiomatic that different word classes exhibit different grammatical behav-
iour. It is only if some change in formal properties can be observed that change
in word class can be deemed to have taken place.
The assertion that formal properties are the proper criteria for ascertaining
part-of-speech classification, however, does not mean that semantic criteria
are not relevant when it comes to certain lexical items undergoing a change in
their word class. They are not only relevant; they are crucial. If certain items
come to function in a way that is fairly removed from the ‘core meaning’ of
their original part of speech (e.g. verb), then it is likely that their formal prop-
erties will eventually follow suit. Conversely, if a certain item does not exhibit
192 CHAPTER 4
some variation in function or meaning, then its formal properties are not likely
to change. Grammatical behaviour is by no means capricious.
4.1.3 Part-of-Speech Constraints vs. Environmental Constraints

The assertion that it is formal properties that are the proper criteria for ascer-
taining part-of-speech classification should not be taken to mean that, in any
given sentence, an item assigned to a particular word class must be able to
exhibit all the formal properties possible for the class as a whole. It is impor-
tant to draw a distinction between the syntactic and morphological constraints
imposed on the members of a particular word class—e.g. prepositions cannot
be marked for tense—and those imposed by the environment—e.g. the first
verb in a serial construction (in language X) cannot be marked for tense. Thus,
for example, a valency-increasing morpheme that cannot be marked for tense
cannot automatically be relegated to the class of preposition.
Several authors discussing part-of-speech classification for putative serial
verbs in general (that is, referring not only to those that function as valency-
increasers, but also those that function in a similar way to adverbs, aspect-
markers, and conjunctions etc.) have observed certain syntactic constraints
on the morphemes in question, and yet have still concluded that these mor-
phemes should be regarded as verbs.
Bamgboṣe (1974), describing serial verbs in Yoruba (Benue-Congo), claims
that “the difference in the behaviour of a verb varies according to whether
it occurs as the only verb or as one of a number of verbs in a sentence”
(p. 41). In what Bamgboṣe refers to as ‘modifying’ SVCs in Yoruba—that is,
those in which one of the verbs serves to modify the other semantically in
some way—the ‘modifying verbs’ do not meet certain criteria for verb-hood
such as occurrence in the frame # NP . . . (NP) #, possibility of topicalization by
reduplication, and ‘selection’ of subject. Nevertheless, Bamgboṣe still regards
these morphemes as true verbs, rather than as adverbs, prepositions or auxilia-
ries. He notes that even in a ‘linking’ SVC, in which neither verb is semantically
less prominent, some verbs lose certain normal verbal properties, such as the
ability to be topicalized.
Referring to serial verb constructions in Thai in which one verb serves to
mark the aspect of the other, Thepkanjana (1986, p. 88) observes:
The aspectual verbs which were originally transitive [e.g. sia ‘lose’, ʔaw
‘take’] lose some syntactic properties, namely, their argument struc-
tures or subcategorization requirements. That means that they do not
have an independent syntactic status as ‘full’ transitive verbs in the
string.
Like Bamgboṣe, Thepkanjana argues that words such as these should still
be regarded as verbs, which simply lose certain syntactic properties (e.g. sub-
categorization for object) along with semantic ones (literal interpretation)
when they occur in a serial construction.4
Both of the authors quoted above reach similar conclusions: that verbs in
serial constructions need not necessarily exhibit all the formal properties pos-
sible for verbs in the language concerned in order to be still regarded as true
verbs. Furthermore, it should be noted that different kinds of serial verb con-
structions will impose different kinds of syntactic constraints on the verbs that
can appear in them. It is necessary to determine what constraints are imposed
by the nature of the particular syntactic environment in which a morpheme
appears; it is then necessary to ascertain if the morpheme in question does
or does not possess all the formal properties expected of a verb in that envi-
ronment. If it does not, then a true change in part-of-speech classification
has presumably taken (or is taking) place. If, on the other hand, it is clearly
the environment that is responsible for any syntactic constraints on the mor-
pheme, then there is no justification for assuming a change in its word class.
The exact syntactic constraints imposed by a particular type of SVC should
properly be specified in the grammar, and are not relevant to the lexical entries
of individual verbs. However, in the case of certain types of ‘asymmetrical’
serial constructions that allow only small, closed classes of verbs to appear
(Aikhenvald, 2006b, pp. 21–28), it is obviously essential to give an indication
in the lexicon that the verbs in question belong to that particular class. When
it comes to environmentally determined semantic constraints, on the other
hand, (i.e. the non-literal usage of certain verbs in some types of SVC) then
it would be essential to give a precise description of these constraints in the
lexical entry of each relevant verb (see Enfield (2009, pp. 447–448), Foley and
Olson (1985, pp. 44–45, 50)). A broad statement in the grammar, indicating that
verbs in a particular construction take on, for example, a ‘prepositional’ mean-
ing or an ‘aspectual’ meaning, is clearly inadequate.
4.1.4 Part-of-Speech Change: Discrete or Gradual?

4.1.4.1 Change as a Process for Individual Morphemes
Givón (1975) claims that semantic reanalysis is likely to be the first change to
take place, but that verbs becoming prepositions may then gradually lose the
4 In this analysis these ‘aspectual’ verbs in Thai would probably be regarded as being in a
nuclear level serial verb construction, explaining their inability to take independent argu-
ments (see 2.4.1.3).
194 CHAPTER 4
ability to take verbal affixes such as modalities, subject agreement or object

pronouns. He suggests that, in a language undergoing such a change, many
words may have already become prepositions according to some criteria, but
still be verbs according to others. Claudi and Heine (1985) agree with this claim,
stating that: “decategorialization lags behind desemanticization” and that cer-
tain morphemes may exhibit “morphosyntax which is half-way between that
of a verb and a preposition.” (p. 43).
Li and Thompson (1973) also view this change of category from verb to
preposition as a process, by which a serial verb at first serves the function of a
preposition, and then gradually loses its formal verbal properties and is finally
actually reanalysed as a preposition. So, while they are quite unequivocal in
their claim that all coverbs in Mandarin should be regarded as prepositions,
they also (perhaps paradoxically) describe (p. 260) how, “in the transition from
verbs to prepositions, some morphemes have progressed farther than others.”
This ‘progress’ is reflected in both the syntax and the semantics of coverbs.
Some Mandarin coverbs exhibit the vestigial verbal property of optional co-
occurrence with the verbal suffix -zhe, thus maintaining a syntactic link to
their former status as verbs. Among those coverbs that have corresponding
verbs in the modern language, some retain a closer semantic connection to
these verbs than others. Those that are semantically more divergent and, of
course, those that have synchronically lost their verbal antecedent altogether,
are regarded as farther along the path of preposition-hood.
In terms of semantic characterisation, Marybeth Clark (personal communi-
cation) accepts this idea of some words having progressed farther than others
in the transition from verb to preposition. However, she claims that syntacti-
cally a word must belong to one category or the other; it cannot be something
in between. If a word functions as a preposition in a particular sentence, then
it is a preposition in that sentence, regardless of whether or not it functions as
a verb elsewhere (Clark, 1978, p. 181).
If, on the other hand, one assumes that formal and syntactic properties are
the only proper criteria for determining part-of-speech classification, then the
evidence from at least some languages does point to the conclusion that word
classes are not completely discrete, and that the transition of a lexical item
from one word class to another should be viewed as a process. Thus, some
items ‘retain’ certain formal properties of a particular category while losing
others. (Examples include: the Mandarin coverbs mentioned above, which
can optionally co-occur with the verbal suffix -zhe, but which cannot take the
verbal aspect suffix -le; the Sranan coverb gi ‘give/to’, which exhibits the prepo-
sitional property of being able to be fronted with its object, and the verbal-
property of being able to be stranded (Sebba, 1987, p. 73).
The fact that certain morphemes exhibit some of the formal or syntactic
properties of one word class and some of another does not mean that we
should establish yet another word class (e.g. ‘coverb’, ‘prepositional verb’) spe-
cifically to cater for these morphemes. It may well be that these ‘in between’
morphemes are not themselves completely homogeneous with regard to for-
mal properties, and we would end up having to establish a separate word class
for each one. The fact that part-of-speech change is seen as gradual rather than
discrete means that, like it or not, we are forced to conclude that some mor-
phemes, in some environments, simply cannot be unequivocally assigned to
any one word class.
4.1.4.2 Change as a Process in the Language

There is a sense in which Clark (1978) does view the presence of coverbs in a
language as a signal that a process is taking place. She argues that originally
each of the words involved occurs only as a verb. During the ‘coverb stage’
the preposition exists side-by-side with the verb and their semantic similar-
ity makes the derivational relationship clear. It becomes gradually less clear,
however, as semantic divergence occurs, or the verb comes to be used less
frequently, perhaps only in a vestigial way, until finally it is dropped entirely
and only the preposition remains. Clark (1979a) presents evidence of this deri-
vational process taking place in White Hmong: she says the coverb rau ‘to/
for’ is a preposition, derived from the verb rau ‘put’, which has become highly
restricted as a main verb. (See §4.2.3.4 below for further discussion.)
With regard to the speed at which this process might take place, Durie
(1988) notes,
To the extent that the oblique-coding verb is used independently, this

will inhibit semantic bleaching and subsequent reanalysis. This means
that if a very common verb is used in serialization to code an oblique
role, the verbal status of the serial usage will endure longer. (p. 5)
He goes on to observe,
To the extent that serialization is a common process in the language,

this will provide motivation for interpreting an oblique coding verb as a
proper verb, and will inhibit the drift to preposition. (p. 5)
Thus, grammaticalization is constrained by both “the frequency with which a

verb occurs independently in discourse” and “the productiveness of serializa-
tion in the language” (p. 20).
196 CHAPTER 4
Givón (1975) suggests that, in a language undergoing the reanalysis of verbs

to prepositions, some verbs would begin the process of change much later than
others, or progress towards preposition-hood at a different pace; as a result the
synchronic analysis of such a language may be “rather messy.”5
Hopper and Traugott (2003, p. 16) agree:
. . . language development is an ongoing process, and one that often

reveals itself as change that is only incompletely achieved at any given
stage of a language.
As Hopper and Traugott note, this understanding of language change as

a process presents a challenge to the notion of categories, not only cross-
linguistically (Haspelmath, 2011), but even within a single language. It certainly
reminds us of the fluid rather than discrete boundaries between word classes.
4.2 Valency-Increasing Serialization in White Hmong
4.2.1 Introduction
Clark (1979a, p. 8) claims that “a word fulfilling a prepositional function can not
be a verb.” She goes on to explain:
It seems only necessary to say that at the point when a word is borrowed
from the verb inventory to function as a preposition it becomes a prepo-
sition. Therefore, there was no need for it to function first as a verb with
another verb. It remains only to say that, in general, there is no such stage
as ‘verbs in series’ with respect to synchronically derived prepositions.
Clark (p. 1) rejects the term ‘coverb’, used elsewhere in the literature to describe
the phenomenon she is writing about, on the grounds that it is “misleading
5 Givón (1975) also notes that the change from serial verbs to prepositions may result in a lan-
guage of somewhat mixed syntactic typology. In an SVO language, where it is the first verb
in a series that is reanalysed, the new prepositional object now precedes the ‘main’ verb,
resulting in a change from SVO to SOV word order for these sentences. However, in the same
language, where it is the second verb in the series that becomes a preposition, the process
does not result in a change in the basic SVO word order.
Wheatley (1984) suggests that a mixed syntactic typology may arise in a language that
was exclusively SOV at an earlier stage: “It is the serializing process that separates goals from
patients (among other constituents) and sets the stage for the isolation of the former on the
right of the verb in the event that grammaticalization of the second verb takes place” (p. 355).
because it implies that these prepositions are some kind of verb or are a dis-
tinctive syntactic category, neither of which is the case.” Instead, she chooses
the term ‘synchronically derived preposition’ to refer to a “preposition [that]
has a corresponding verb which is homophonous and broadly synonymous
and from which the preposition is derived.”
Clark identifies three ‘synchronically derived prepositions’ in White Hmong:
nyob ‘at’ (derived from the verb nyob ‘be at’); nrog ‘with’ (derived from the verb
nrog ‘be with’); txog ‘reaching, up to’ (derived from the verb txog ‘reach, arrive
at’). She adds another ‘preposition’, rau ‘to’, which, she says, is in the process of
losing its synchronic-derivation relationship with its corresponding verb, rau
‘put [in]’, a verb which now has very restricted usage. Furthermore, according
to Clark, there are two homophonous ‘prepositions’, txij ‘reaching, up to’ and
txij ‘since’, that are underived prepositions, there being no corresponding verb
in the language. Finally, Clark (1980b, pp. 14–15) proposes that there are three
‘directional adverbs’ in White Hmong, all of which are derived from intransi-
tive Goal verbs: mus ‘away’ (from the verb mus ‘go’), tuaj ‘hither’ (from tuaj
‘come’), los ‘hither, back’ (from los ‘come, return [home]’).6
The view adopted in this study regarding the appropriate criteria for part-of-
speech classification has already been discussed (see §4.1 above). Briefly, while
function and meaning are intricately related to part-of-speech category, formal
and syntactic similarity are the appropriate criteria for ascertaining word class.
Furthermore, different parts of speech can be utilised for similar functions in
different languages. Therefore, if the lexical items identified by Clark as either
‘derived prepositions’ or ‘adverbs’ are to be regarded as such, some evidence
beyond that of their function would be required.
The purpose of this section is to examine the functions and syntactic prop-
erties of the White Hmong lexical items listed above, namely nyob, nrog, txog,
rau, txij, mus, tuaj, and los. The lexical items mus, tuaj and los, are also relevant
to the topic of ‘valency increasing’ because, in some sentences, they serve to
introduce a Goal argument into the clause (see below §4.2.3.2 below). The
investigation of the syntactic properties of these items will reveal, when they
appear with a(nother) verb in the same clause, whether there is any evidence
to suggest that they should be regarded either simply as verbs in serial verb
constructions, on the one hand, or as verbs undergoing or having undergone
grammaticalization to prepositions or ‘directional adverbs’, on the other.
6 The lexical items mus, tuaj and los, are also relevant to the topic of ‘valency increas-
ing’ because, in some sentences, they serve to introduce a Goal argument into the clause
(see §4.2.3.2).
198 CHAPTER 4
4.2.2 Criteria Relevant to Part-of-Speech Classification in White Hmong

Among the criteria that have been used by other linguists to determine
whether the relevant morphemes in the languages they describe should be
regarded as verbs or prepositions, a number of formal tests are proposed.
A few of these tests are applicable in White Hmong. However, because of the
typological characteristics of the language, and because of the particular syn-
tactic properties of serial verb constructions in White Hmong, the number and
type of tests that can be applied to determine the part-of-speech classification
of the morphemes in question in this language is extremely limited.
4.2.2.1 Frequency
There is no doubt that all of the lexical items identified by Clark as either
prepositions or adverbs do share one important property of grammatical mor-
phemes, namely frequency. As Durie (1988, pp. 4–5) notes:
When verb serialization is used consistently in a language to encode

oblique roles, . . . these verbs will tend to be more frequently serialized
than used as independent verbs, because the need to code basic oblique
roles in discourse tends to occur particularly frequently. A principal ‘cen-
ter of gravity’ of the verb’s use in discourse will be as an oblique coding
strategy.
High frequency is thus clearly a feature that would suggest that the process of
grammaticalization may well be under way, if not completed in some cases
(Bybee, 2003). However, this feature alone may not be sufficient evidence to
claim that grammaticalization has actually occurred, and that the morphemes
in question should thus be regarded as representing a different part of speech.
The feature of frequency is, after all, inextricably associated with function.
4.2.2.2 Morphosyntactic Properties of Verbs

One of the simplest tests for verb-hood in many languages is to observe the
presence or absence of verbal morphology on the items in question. The loss of
verbal properties such as morphological agreement is well documented as an
outcome of the process of grammaticalization in Verbs in several West African
languages, including Ewe and Twi (Ansre, 1966, pp. 29–32). White Hmong
verbs are not inflected in any way, although many are able to undergo the
morphological process of reduplication. There are, of course, also morphemes
that co-occur with verbs to mark polarity, aspect, modality etc., and the pos-
sibility of co-occurrence with such morphemes could be used as a diagnostic
test for verb-hood.
However, it is important to remember that verbs do not exhibit all their

potential syntactic properties in every environment in which they can appear
(see §4.1.3 above). Serial constructions themselves impose a considerable
number of constraints on the verbs that occur in them. If any putative multi-
verb sentences are, in fact, true serial verb constructions, then the verbs
involved would clearly be expected to conform to these constraints. They
would, at least, share all clause-layer operators, such as polarity and illocution-
ary force, as well as the core modality operators. Furthermore, even though
some languages allow independent aspectual operators over the juncts in
serial verb constructions, this is not the case with any of the SVC-types identi-
fied for White Hmong. Thus the ability to co-occur independently with verbal
operators is not a useful test for verb-hood in White Hmong as long as verb
serialization may be involved.
4.2.2.3 Predication of Subject

An important property of verbs in three of the four types of serialization
identified in White Hmong, namely Cotemporal SVCs, Attainment SVCs, and
Disposal SVCs, is that they must predicate the subject; the subject is, in fact,
a shared argument in these SVC types.7 If a sentence putatively involved one
of these three SVC types, a verb-like morpheme marking the ‘circumstances’
would have to exhibit a predication relationship with the subject of the clause,
in order to be considered a verb in a serial construction. The loss of this predi-
cation relationship between an erstwhile verb and the subject of the clause has
been observed in a range of languages (e.g. Chinese (Li & Thompson, 1974a,
p. 271), Vietnamese (Clark, 1978, p. 113)). Such a loss would mean that the sub-
ject was no longer a shared argument, and indicate that the morpheme was
therefore not a verb in a serial construction, but had become grammaticalized.
4.2.2.4 Comparison with Prepositions

There are two distinct types of words that could be considered ‘candidates’ for
preposition-hood in White Hmong. One is the class of morphemes referred
to in this study as ‘Spatial Deictics’, and termed ‘Prepositions’ by a number of
other scholars (Mottin (1978, p. 70), Fuller (1985, p. 39), Ratliff (1992, pp. 104–
112)). Included in this class are morphemes such as ntawm ‘there-nearby’, pem
‘there-up (a slope)’, tim ‘there-across’, hauv ‘there-inside’, saum ‘there-above’,
and others. As shown in Chapter 1, §1.5, these morphemes do not serve to
7 In the case of Cause-Effect SVCs, it was noted that the second verb in the series does not, in
fact, predicate the initial NP, even though the available evidence seems to point to the con-
clusion that this NP may be regarded as the subject of these sentences (see Chapter 3, §3.2.4).
200 CHAPTER 4
indicate anything about the role of the NP that follows them. NPs with various
semantic roles in the sentence can be preceded by an identical Spatial Deictic.
Instead, their function is to indicate something about the spatial properties of
the NP that follows them. This could be the location of that NP with respect to
the location of the speaker (or to the speaker’s adopted point of view), or with
respect to some topological feature (e.g. pem ‘there-up (a slope)’, tim ‘there-
across’). Alternatively, it could be the part or aspect of the referent of that NP
which is relevant to the action or event described (e.g. hauv ‘there-inside’, saum
‘there-above’). These morphemes are thus not regarded as prepositions in this
analysis.8
The other candidate for preposition-hood in White Hmong is, of course, the
very morphemes that are the focus of this chapter themselves: morphemes that
are homophonous with a verb in the language and that occur with a(nother)
verb within a single clause to perform a valency-increasing function. These are
regarded as prepositions by several scholars, including Mottin (1978) and Clark
(1979a, 1989).
As noted above, Clark (1979a) claims that there are actually two, homopho-
nous prepositions txij in White Hmong that should definitely not be included
in her category of ‘synchronically derived prepositions’ because, she says, they
do not have a corresponding verb. She gives the following examples:
8 It may seem rather contradictory to assert that Spatial Deictics should not be regarded as
prepositions on the grounds that they do not function like prepositions, when it has been
stated that formal criteria are the only proper criteria for part of speech classification.
However, the question in this case is not whether a small set of morphemes belongs to Class
X or to Class Y (as in the case of ‘coverbs’). It is, rather, whether the members of Class Z—a
class which can be clearly defined in formal terms—should be referred to as ‘prepositions’
or not. There is no doubt, on formal grounds, that these morphemes do constitute a distinct
class; but what should this class be called? When it comes to naming the parts of speech
in a language, it is clearly universal rather than language particular criteria that must be
called upon. If none of the members of a word class in a certain language exhibit the seman-
tic features common to prepositions in other languages, then it is clearly inappropriate to
refer to this class as the class of ‘prepositions’. Mottin has presumably chosen to call these
morphemes ‘prepositions de lieu’ because of their spatial reference and because the simplest
translations in French, as in English, are prepositions. Although spatial reference is a signifi-
cant feature in the notion of ‘preposition’ in many languages, it does not seem an adequate
basis for classification when the morphemes involved do not perform the more significant
function of indicating the role of the following NP in the clause.
(3) dej tob txij duav

water deep waist
‘The water is up to the waist.’ (Clark, 1979a, p. 6)
(4) txij hnub kuv yuav poj-niam los . . .

day 1SG obtain wife come
‘Since the day I got married . . .’ (Clark, 1979a, p. 6)
In example (3), Clark analyses txij as spatial preposition meaning ‘up to’, and in
(4), as a temporal preposition meaning ‘since’.
However, contrary to Clark’s claim, there is actually a verb txij meaning
‘extend to, reach (in the vertical dimension)’, as shown by the grammaticality
of examples such as (5) and (6):
(5) dejA txij duavO lawm

water waist PRF
‘The water has reached the waist.’9
(6) nwsA tsis txij kuvO

3SG NEG 1SG
‘He’s not as tall as me.’
For this reason, txij will be included in the discussion below.
4.2.2.5 Sentence-Initial Occurrence as Topic

Even though there is no absolutely clear class of preposition in White
Hmong, it is still assumed that, if some verbs have actually been grammati-
calized, they may now exhibit certain formal or syntactic properties that are
not characteristic of verbs in the language. One property that distinguishes
some erstwhile verbs from verbs in several languages is their ability to occur
9 (Clark, 1979a, p. 6) gives the following sentence, in which txij occurs as the only predicate in
the clause, as ungrammatical:
(i) * dejA txij duavO
water waist
Thus, according to Clark, txij can only be used as a preposition, and is not a verb at all.
However, the sentence above was probably judged to be ungrammatical because it is not suf-
ficiently ‘bounded’ (see (Li, n.d., pp. 8–10)). When the perfect morpheme lawm appears
finally, as in example (5), the sentence is perfectly acceptable.
202 CHAPTER 4
sentence initially with their object as either a topic or the focus of a question
(Lord, 1973, pp. 280–283; Claudi & Heine, 1985, p. 43; Sebba, 1987, p. 114). In the
case of White Hmong only sentence initial occurrence as topic would be rel-
evant, as the focus of a question does not appear sentence initially. A property
such as ability to occur sentence initially as (part of) a topic being associated
with the morphemes concerned when they function to introduce oblique roles
but not with verbs in series may indeed suggest that a change in word class has
taken place.
4.2.2.6 Peripheral Position in Clause

Givón (1975) notes that “when a verb is semantically reanalysed as a preposition
or conjunction, it quite often it remains, for a long time afterwards, at its origi-
nal serial-verb position.” (p. 84). This would clearly constitute good evidence to
establish the verbal origins of a morpheme that is clearly not a verb synchron-
ically. Although Givón does not mention its occurrence, the opposite situation
would also be relevant in cases in which the part of speech of a verb-like mor-
pheme were in question: If a verb-like morpheme appeared in a more periph-
eral position than would be expected in the SVC type from which the sentence
is putatively derived, then this would be evidence of grammaticalization.
4.2.2.7 Summary of Criteria for Part-of-Speech Classification

The following is a list of the tests which can be applied to the morphemes under
investigation in White Hmong to attempt to determine whether they should be
regarded simply as verbs in serial constructions, or as erstwhile verbs which, in
the relevant environment, should more properly be regarded as prepositions.
1. Does the morpheme predicate the subject of the clause?

2. Can the morpheme and its object appear sentence initially as topic?
3. Does the morpheme appear in a more peripheral position than would be
expected in the SVC type from which the sentence is putatively derived?
4.2.3 Examination of Valency-Increasing Serialization in Hmong

4.2.3.1 Nyob
The verb nyob means ‘be located/stay/dwell’, as illustrated in the following
examples:
(7) [cov me-nyuam]A nyob [hauv lub vaj]O

CLF.PL child inside CLF garden
‘The children are in the garden.’
(8) thaum kuvS los <nyob [lub teb.chaws no]O>PURP . . .

time 1SG come CLF country this
‘. . . the time I came to live in this country.’ (WHD 20: 20)
The morpheme nyob is also used with another verb to indicate either the loca-
tion of a state or the location of an action, and can introduce either an inner
or an outer locative:
(9) Location of state; inner locative

muaj [ib tsob ntoo] (nyob) [hauv lub vaj]
have one CLF tree inside CLF garden
‘There is a tree in the garden.’
(10) Location of action; outer locative

[cov me-nyuam] ua.si (nyob) [hauv lub vaj]
CLF.PL child play inside CLF garden
In both examples above the occurrence of nyob is optional. A Locative Phrase

alone is, of course, interpreted as an inner locative with a verb such as the verb
muaj ‘have’ in its existential use in example (9) (Chapter 1, §1.5). As shown in
example (10), a Locative Phrase alone (i.e. without nyob) can interpreted as an
outer locative Location with any verb that does not take a core locative phrase
(e.g. Location, Source, Path, Goal).
The sentences above in which nyob does appear and functions to introduce
the locative argument are clearly similar to examples of the Cotemporal Serial
Construction, particularly to those expressing Cotemporal Stance and Action.
However, they differ in that, as noted (Chapter 3, §3.1.2.2), it is most commonly
the stance expression that appears first in Cotemporal SVCs of this type. For
example:
(11) lawvS/S zawm-tsaws tham ib hmo nkaus

Example (12) below shows the verb nyob itself appearing in first position in a
Cotemporal SVC of this type:
204 CHAPTER 4
(12) no ces [tus kwv]A txawm nyob [ntawm ntug dej]O

this CONJ CLF younger.brother then stay nearby edge water
tos Txiv.Nraug.Ntsuag txawm mus <suav~suav ntsesO>PURP
wait the.Orphan.Boy then go RDUP~count fish
‘And so then the younger brother stayed by the edge of the stream wait-
ing; the Orphan Boy went to count the fish.’ (DNH 183: 2)
In this sentence nyob does not simply function to introduce the location of the
waiting; it has its full lexical meaning ‘stay’, which is contrasted with the verb
mus ‘go’ in the following juxtaposed clause.
When it functions to introduce a Locative Phrase, on the other hand, nyob
and its object cannot appear before the action or state verb it accompanies:
(13) a. [cov me-nyuam]S ua.si nyob [hauv lub vaj]O

CLF.PL child play inside CLF garden
b. [cov me-nyuam]A/S nyob [hauv lub vaj]O ua.si

CLF.PL child inside CLF garden play
Lit: ‘The children stayed in the garden playing.’10
not: ‘The children are playing in the garden.’
(14) a. lawvA muaj [ib tsob ntoo]O nyob [hauv lub vaj]O
3PL have one CLF tree in CLF garden
‘They have a tree in the garden.’
b. * lawvA/A nyob [hauv lub vaj]O muaj [ib tsob ntoo]O

3PL in CLF garden have one CLF tree
Lit: ‘They stay in the garden having a tree.’
not: ‘They have a tree in the garden.’
Example (14) above shows quite clearly that nyob appears finally to introduce a
Locative Phrase that expresses the Location of the event or situation described
by the preceding verb; it does not necessarily describe the Location of the sub-
ject. It is not lawv ‘they’, who are located ‘in the garden’ at all in this example.
10 An alternative interpretation involving a relative clause is also possible:

(i) [cov me-nyuam <nyob [hauv [lub vaj]]>REL]S ua.si
CLF.PL child inside CLF garden play
‘The children (who are) in the garden are playing.’
(See Li & Thompson (1974a, p. 271) for a similar example from Mandarin). In
a Cotemporal SVC, on the other hand, both verbs must predicate the subject.
A location marked by nyob can, however, occur in sentence-initial topic
position, both in sentences describing the location of an action and those
describing the location of a state:
(15) nyob [teb.chaws Nplog], pebA siv tsuO <cub mov>PURP

country Laos 1PL use steamer steam rice
‘In Laos, we used a steamer to steam rice.’ (IWH 17: 7)
(16) nyob [teb.chaws Sam.Tue] mas muaj tsovS heev

country PN TOP have tiger very
‘In the Sam Tue area, there were lots of tigers.’ (Fuller, 1985, p. 115)
Neither junct of a Cotemporal SVC, on the other hand, can occur as a topic:
(17) a. ?? noj mov mas, nws sawv-ntsug

eat rice TOP 3SG stand
Intended meaning: ‘Eating, he stands up.’
(Might be interpreted as: ‘He eats, (and then) he stands.’)
b. ?? sawv-ntsug mas, nws noj mov

stand TOP 3SG eat rice
Intended meaning: ‘Standing, he eats.’
(Might be interpreted as: ‘He stands, (and then) he eats.’)
Thus the use of nyob discussed here differs from the verb nyob appearing in a
Cotemporal SVC in at least three significant ways:
a. Along with its object, it can only appear in final position (unless topical-
ized); the verb and its object appear in first position, after the subject.
b. It functions to indicate the location of the action or state, and need not
predicate the subject; a verb in a Cotemporal SVC always predicates the
subject.
c. Along with its object, it can appear sentence initially as topic; a junct of a
Cotemporal SVC cannot.
Differences such as these suggest that in these cases nyob and its object are
not functioning as a joint predicator in a serial verb construction. Instead, the
syntactic evidence points to the conclusion that this use of nyob would best
206 CHAPTER 4
be analysed as a locative preposition meaning ‘in’. This preposition is derived

from the verb nyob ‘be located, stay’, but must be distinguished from it.
4.2.3.2 Mus, Tuaj, Los

The verbs mus ‘go’, tuaj ‘come’, and los ‘return home’ are the three Deictic
Motion verbs in White Hmong. When a single Locative Phrase follows one of
these Deictic Motion verbs it is invariably interpreted as indicating the Goal of
the motion.11
(18) nwsA mus [tim Asmeslivkas]O lawm

3SG go across America PRF
‘He has gone (across) to America.’
(19) niaj hnub [cov me-nyuam]A mus [nram pas-dej]O

every day CLF.PL child go down pond
‘Every day the children go down to the pond.’
(20) pebA tuaj AustraliaO thaum xyoo 1979

1PL come PN time year
‘We came to Australia in 1979.’
When a Locative Phrase directly follows a verb expressing Locomotion or

Transport, on the other hand, it can only be interpreted as an outer locative,
indicating Location:12
(21) [tus noog]S ya [saum tsob ntoo]

CLF bird fly top CLF tree
‘The bird flew about, above the tree.’
not: ‘The bird flew to the top of the tree.’
11 This is not the case when more than one Locative Phrase follows one of these verbs,
in which case the first is interpreted as Source and the second as Goal (see Chapter 1,
§1.6.5.3).
12 ‘Locomotion’ verbs are those that describe the movement of a subject through space (e.g.
ya ‘fly’, nkag ‘crawl’) and ‘transport’ verbs are those in which a subject transports an object
through space (e.g. nqa ‘carry’, lawv ‘drive’).
(22) lawvA lawv [cov npua]O [pem teb]

3PL drive CLF.PL pig up field
‘They drove the pigs about, up in the fields.’
not: ‘They drove the pigs up to the fields.’
(23) nwsA ev [nws tus me-nyuam]O [nram liaj]

3SG carry.on.back 3SG CLF child down paddy.field
‘She carries her child about down in the paddy fields.’
not: ‘She carries her child down to the paddy fields.’
To indicate the goal of a locomotion or transport verb, mus ‘go’, tuaj ‘come’, and
los ‘return home’ may be used, introducing the Locative Phrase:
(24) [tus noog]S/A ya mus [saum tsob ntoo]O

CLF bird fly go top CLF tree
‘The bird flew to the top of the tree.’
(25) lawvA/A lawv [cov npua]O tuaj [pem teb]O

3PL drive CLF.PL pig come up field
‘They drove the pigs up here to the field.’
(26) nwsA/A ev [nws tus me-nyuam]O mus [nram liaj]O

3SG carry.on.back 3SG CLF child go down paddy
‘She carries her child down to the paddy fields.’
Clark (1980b) also notes the general occurrence of mus, tuaj and los in sen-
tences such as those above. She regards these words as derived adverbs, the
function of which is to add deictic information only, maintaining that it is the
non-deictic locomotion or transport verb that introduces the goal argument
Clark (1980b, pp. 14–15):
Some Goal verbs, such as khiav ‘run’, do not have inherent features with
respect to center, that is, they do not specify the direction hither or
thither of the path of the patient. Therefore they frequently cooccur with
directional adverbs which have such a feature. These adverbs are derived
from intransitive Goal verbs which are marked with respect to center
[-cntr] mus ‘go’ >---> [-cntr] mus ‘away’, [+cntr] tuaj ‘come’ >---> [+cntr]
tuaj ‘hither’, and [+cntr] los ‘come, return’ >---> [+cntr] los ‘hither, back’.
The use of such an adverb is shown in [i] and many other sentences.
208 CHAPTER 4
[i] kuv tus ntxhais khiav los (rau) hauv tsev

[1SG CLF daughter run] ‘hither’ [to inside house]
‘My daughter came running into the house.’
. . . Transport verbs—verbs whose AGTs transport their PATs through

space (Clark, 1977, pp. 17, 20–25)—seem to require the presence of the
directional adverbs which specify the [+center] direction of the PAT’s
path. These adverbs follow the verb and its PAT and precede the LOC
when the LOC is present, as in [ii] and [iii].
[ii] nws ev nws tus me-nyuam mus rau nram liaj

[3SG back.carry 3SG CLF child] ‘away’ [to down paddy.field]
‘She back-carries her child down to the fields (when she goes to work).’
[iii] nws nqa ib lub kaus mus teb

[3SG carry one CLF umbrella] ‘away’ [field]
‘He carried an umbrella to the fields.’ (Heimbach 1969: 78)
An important function of mus, tuaj, and los in the sentences Clark gives is,
indeed, to convey deictic information. However, it seems from the evidence
given in examples (21)–(26) above that they are, in addition, responsible for
the Goal interpretation of the Locative Phrases in these sentences. Although
Clark does give at least one example of a Locomotion verb followed directly by
Locative Phrase that is translated as a Goal, native speakers of White Hmong
consulted for this study reject this interpretation:
(27) noog ya tom tsob ntoo

bird fly over.there CLF tree
Translation given:
‘The bird flew over there to the trees.’ (Clark, 1980a, p. 51)
White Hmong language consultants maintain that the only possible interpre-
tation for this sentence seems to be with an outer locative of Location: ‘A bird
flew about over at the trees’, similar to example (21) above.
Clark (1980b, p. 22) also notes an example from Smalley (1976, p. 120), in
which a Locative Phrase following a transport verb is translated as a Goal:
(28) thiaj.li rov qab tuaj <lawv npuaO [[puag tim] [qub
so return back come drive pig yonder across old
tsev] [tim no]]>PURP , lawv npuaO mus txog mas, . . .
house across here drive pig go arrive TOP
Translation given:
‘Then we returned to drive the pigs from over at the old village to over
here. (We) drove the pigs . . .’ (Smalley, 1976, p. 120)
Correct translation:
‘So we came back to drive (some) pigs from way across at the old place
across there, to drive (some) pigs there, . . .’
However, the phrase tim no ‘across here’ cannot, in fact, be interpreted as a Goal
in this sentence. Instead, it actually functions to modify the preceding NP, qub
tsev ‘(the) old place’, yielding qub tsev tim no ‘(the) old place across here’. Only
after this does the goal oriented expression lawv npua mus txog ‘drive (the)
pigs (and) go arrive there’ appear, allowing the preceding Locative Phrase puag
tim qub tsev tim no ‘way across at the old place across here’ to be interpreted
as Source.13 A correct translation for this sentence would thus be: ‘So we came
back to drive (some) pigs from way across at the old place across here, to drive
some pigs there, . . .’ The repetition of the transport verb lawv ‘drive’ before the
goal-oriented expression mus txog ‘go arrive’ is simply due to the heaviness of
the Locative Phrase indicating the Source. Repetition of some preceding mate-
rial after a heavy phrase is extremely common in White Hmong.
It can be concluded that a Locative Phrase following a locomotion or trans-
port verb cannot be interpreted as a Goal. One way of specifying the Goal of
verbs of this kind is to utilise a Cotemporal SVC with one of the Deictic Motion
verbs. The Deictic Motion verbs cannot be interpreted simply as directional
adverbs, because they introduce the Locative Phrases that follow them and are
responsible for their interpretation as Goals.
It is not, however, always the case that mus ‘go’, tuaj ‘come’ and los ‘come/
return home’ serve to introduce Goal arguments. In examples such as (29)–
(30) below their function really is to add deictic information only. It could be
argued that, in these contexts, these morphemes are derived adverbs rather
than verbs.
13 It is not at all uncommon for a Spatial Deictic to appear both before and after a Noun
Phrase (see Chapter 1, §1.5). Before the NP, a Spatial Deictic serves to locate the referent
topographically and/or relative to the speaker’s point of view. After the NP, it serves as a
restrictive modifier.
210 CHAPTER 4
(29) taug~taug dejO nqis havO mus

RDUP~follow water descend valley
‘(He) followed the river down the valley, away.’
(30) [cov tub-rog]A thim [ntawm lub kwj-ha]O los

CLF.PL soldier retreat nearby CLF valley
‘The soldiers retreated homeward from the valley.’
However, although it is possible to analyse mus and los in the examples above
as adverbs, it is both unnecessary and uneconomical to do so. The syntactic
mechanism for dealing with them as verbs, namely the Cotemporal Serial
Verb Construction, already exists and is independently justified. The fact that
they do not serve to introduce overt goal phrases in these cases does not mean
that they automatically forfeit their status as verbs. It simply means that these
are ambitransitive verbs of the A=S type. There is no doubt that, in these sen-
tences, just as in all those exemplified so far, the deictic motion verbs can be
interpreted as predicating the subject, and thus do have a shared argument
with the other verb in the core.
There are, however, some sentences in which mus ‘go’, tuaj ‘come’ and los
‘come/return home’ appear, in which they cannot be interpreted as predicat-
ing the subject. Observe example (31):
(31) kuvA yuav nyob [qhov no]O mus txog [thaum kuvA yuav
1SG IRR dwell place this time 1SG IRR
tau [kuv ib tsev]O] tso
get 1SG one house first
‘I will live here until I get my own house.’ (WHD 20: 20)
In this sentence, mus (with the assistance of txog (from the verb txog ‘arrive’;
see §4.2.3.5 below)) is serving to introduce a temporal rather than a physical
Goal. It is not the subject, kuv ‘I’, who will ‘go’, but rather the ‘living here’ that
will ‘go on to the time . . .’. In the absence of a predication relationship with
the subject, mus cannot be regarded as a verb. Together with txog (see §4.2.3.5
below) its only role here is to introduce the (temporal) Goal/end-point of the
state described.
It can thus be concluded that mus has been grammaticalized in exam-
ples such as (31) above. This conclusion is reinforced by the fact that, in this
and similar roles, a phrase introduced by mus ‘go’, tuaj ‘come’, or los ‘come
home’ can, and frequently does, appear sentence initially as, or as part of,
a topic:
(32) los txog [xyoo 1960], [cov neeg Hmoob]A thiaj tau
year CLF.PL person Hmong so ACHV
pib kawm los mus
begin study CONJ go
‘So by 1960, the Hmong people had started to learn (the writing system)
and went on (learning it).’ (WHG 6)
(33) txij li [thaum xyoo 1960] los, [me-nyuam Hmoob]A thiaj

as time year child Hmong so
tau kawm ntawvO coob txog li [60–70 paws.xees]
ACHV study writing many reach as percent
‘Since about 1960, as many as 60–70% of Hmong children have conse-
quently learnt to (read and) write.’ (WHG 6)
(34) [ntawm peb lub Muong.Sam] mus rau [nram Sam.Teu] mas
nearby 1PL CLF PN down PN TOP
mus [xya hnub] kev
go seven day way
‘From our Muong Sam down to Sam Teu, (you) walk seven days.’ (Fuller,
1985, p. 115)
In example (34) above, the phrase introduced by mus (followed by rau (see
§4.2.3.4 below)) expresses distance rather than duration. However, as in the
other examples, mus is not predicated on the subject; this is actually a sub-
jectless sentence. Instead it serves to express the distance of the action/
event described by the predicate mus . . . kev ‘make one’s way/walk’. In all the
examples of this type above, the morphemes mus and los can be regarded as
having grammaticalized. Even in this role, however, in addition to introduc-
ing a Goal argument they retain the semantic component of giving deictic
information.
4.2.3.3 Txij
The verb txij means ‘extend up to/ reach as far as (in the vertical dimension)’,
as in the following examples:
(35) dejA txij duavO lawm

water waist PRF
‘The water has reached the waist.’
212 CHAPTER 4
(36) nwsA tsis txij kuvO

3SG NEG 1SG
‘He’s not as tall as me.’
Txij can be used with another verb, with similar meaning. In the examples
below, txij is interpreted as marking the vertical extent of a change (change of
state or change of location) or the vertical extent of an action.
(37) dejS/A tob txij duav

water be(come).deep waist
‘The water is up to the waist/deepened to waist level.’ (Clark, 1979a, p. 6)
(38) [tus qav]S/A dhia txij [kuv xwb-pwg]!

CLF frog jump 1SG shoulder
‘The frog jumped as high as my shoulder!’
Sentences of this kind may seem to resemble examples of Attainment SVCs, in

which the final achievement verb indicates the successful attainment of the
goal expressed by the first verb. For example:
(39) [tus tub]A/A mus txog [nram hav-dej]O/O

‘The boy got down to the river valley.’ (KNH 2)
However, in fact, these sentences with txij are quite different. In an Attainment
SVC with transitive verbs, both A and O are shared, so the final NP is an argu-
ment of both verbs. A grammatical sentence would be possible, even if the
achievement verb did not appear at all:
(40) [tus tub]A mus [nram hav-dej]O

CLF boy go down valley-water
‘The boy went down to the river valley.’
Alternatively, the achievement verb may appear (with ellipsis of its object)
sentence finally, in a separate, juxtaposed clause (see Chapter 3, §3.3.4):
(41) [tus tub]A mus [nram hav-dej]O txog lawm

CLF boy go down valley-water arrive PRF
‘The boy went down to the river valley (and) (he) has arrived (there).’
This is not the case in a construction with txij, which is entirely responsible for
introducing the final NP:
(42) a. [kuv tus kwv]S/A siab txij kuvO

1SG CLF younger.brother be(come).tall 1SG
‘My younger brother is as tall as I [am].’ (Clark, 1979b, p. 12)
b. * [kuv tus kwv] siab kuv

1SG CLF younger.brother be(come).tall 1SG
Lit: ‘My younger brother is tall me.’
c. * [kuv tus kwv] siab kuv txij lawm

1SG CLF younger.brother be(come).tall 1SG PRF
Lit: ‘My younger brother is tall me (and) has extended to.’
In the examples above, txij actually functions in much the same way as one of
the Deictic Motion verbs (mus ‘go’, tuaj ‘come’, and los ‘come home’) appearing
finally in a Cotemporal SVC to introduce the Goal of a motion event. In this
SVC type, only the S/A argument is shared, and transitive verbs can introduce
their own O arguments.
The only difference between a Deictic Motion verb in a Cotemporal SVC and
txij in examples (37), (38) and (42) above is that txij introduces an argument
expressing vertical spatial extent, a parameter in which deixis is generally not
perceived as relevant. In these examples there seems to be no reason to regard
the txij as anything but a telic Goal verb, appearing finally in a Cotemporal
Serial Construction. Its meaning is clearly the same as when it appears as the
only predicate in the clause; it simply acts with the verb which precedes it as a
joint predicator; the subject is shared.
As with all examples of the Cotemporal SVC, the final verb and its object in
sentences like those above cannot appear sentence initially as a topic:
(43) * txij duav mas, dej tob

waist TOP water be(come).deep
‘Up to the waist, the water deepened.’
(44) * txij kuv mas, [kuv tus kwv] siab

1SG TOP 1SG CLF younger.brother be(come).tall
‘As I [am], my younger brother is/became tall.’
214 CHAPTER 4
In the example below, however, the expression introduced by txij does occur
sentence initially as part of a topic, suggesting that, in this usage, it has been
grammaticalized:
(45) txij [[lub zos] <peb nyob>REL] rau yav.pem.toj mas ua

CLF village 3PL live to mountain TOP make
tau [[teb pob.kws] thiab [teb yeeb]]O/O nkaus xwb
get field corn and field opium only only
‘From the village we lived in up to the mountains, (we) grew corn and
opium only.’ (Fuller, 1985, pp. 114–115)
The extent marked by txij in examples (37), (38) and (42) above, discussed pre-
viously, referred to the extent of the subject: ‘the water is deep, extending to
the waist’ / ‘he is tall, extending to my shoulder’ etc. The significant difference
with example (45) is that the extent marked by txij refers not to the extent of
the subject but to the extent of the predicate. That is, in this example, it is the
‘growing (of corn and opium)’ that extends ‘from the village’. This difference in
the meaning of txij is obviously being reflected in its grammatical behaviour.
Note that, in example (45), the expression with txij is not only interpreted as
marking extent, but also Source. However, this is a component of the meaning
that is not contributed by txij itself, but by its context. As with all expressions
indicating Source in White Hmong (except those constituting the inner loca-
tive argument of a Source Verb) (see Chapter 1, §1.5), the expression txij lub zos
peb nyob ‘extend to the village we lived in’ in example (45) above is only inter-
preted as Source by virtue of the goal-oriented expression which immediately
follows (rau yav pem toj ‘up to the mountains’).
A similar use of txij can also be used to mark a temporal, rather than spa-
tial extent. In these examples also, the expression with txij is interpreted as a
Source by virtue of the goal-oriented expression (los (txog) . . .) ‘come (arrive)’)
that follows.14
14 The fact that txij is interpreted as marking Source only when a goal-oriented expression
follows indicates that there is no need to propose, as Clark (1979a) does, two separate
morphemes: txij ‘up to’ and txij ‘since’ (see §4.2.1). It is only necessary to recognise that txij
can be used to express both spatial and temporal extent.
(46) txij [hnub <kuv yuav poj-niam>REL] los txog [tav no]
day 1SG obtain wife time this
tau muaj [ib xyoo]S
get have one year
‘From the day I got married to the present time, it’s been one year.’ (Clark,
1979a, p. 6)
(47) txij li [thaum xyoo 1960] los, [me-nyuam Hmoob]A thiaj

as time year child Hmong so
tau kawm ntawvO coob txog li 60–70 paws.xees
ACHV study writing many reach like percent
‘Since about 1960, as many as 60–70% of Hmong children consequently
learnt to read and write.’ (WHG 6)
Again, in these examples, the expression with txij appears sentence initially as
a topic. This is possible because it is serving to mark the extent of the predicate
(e.g. the ‘learning (to read and write)’ in example (47)) rather than that of the
subject (the ‘Hmong children’). It has lost its own role as a predicate within the
clause, and functions only as a preposition, to introduce the time expression.15
4.2.3.4 Rau
The verb rau is most commonly glossed in English as ‘put’ (e.g. (Heimbach,
1979, p. 275). However, as shown in example (48) below, while rau can be used
to introduce the Goal of the verb tso ‘put, place’, it cannot be used in place of
the verb tso:
(48) a. kuvA tso [phau ntawv ko]O rau [saum lub rooj]
1SG put CLF writing that top CLF table
‘I put that book on the table.’
b. * kuv rau [phau ntawv ko] rau [saum lub rooj]

1SG CLF writing that top CLF table
15 In this usage, txij often precedes a time word, such as the noun hnub ‘day’ in example
(46) or thaum ‘time’ in (47), and is accompanied by los ‘come’, which signals the goal,
thus allowing the interpretation of txij hnub/thaum as indicating the temporal source,
‘since’: txij hnub/thaum ntawd los ces (txij day/time that come CONJ) ‘since that day/time’.
Compare: txij no mus (txij now go) ‘from now on’.
216 CHAPTER 4
On the basis of example (49) below in which the verb rau does appear, a more
correct English gloss would seem to be ‘put in/insert’:
(49) rau piam-thajO los tsis rau

sugar or NEG
‘With or without sugar?’ (Lit: ‘Put in sugar or not put (it) in?’) (WHD 17: 13)
In example (49), the object of the verb rau—piam-thaj ‘sugar’—is a Theme;

this is probably the most common role of the object when rau is the only predi-
cate in the clause. Other examples of rau taking a Theme as its object include:
(50) rau nplejO

unhulled.rice
‘put unhulled rice into (a mill) / plant rice seedlings in (holes in the
ground)’
and possibly the idiomatic expression:
(51) rau~rau-siabO
RDUP~rau-liver
‘try very hard’ (‘put your heart into it’)
A more restricted usage of this verb is one in which the object of the verb rau
is an item of clothing, as in: rau khau ‘put on shoes’, rau kaus mom ‘put on a
hat’, rau hnab looj tes ‘put on gloves’. In this case the object is a Goal or des-
tination rather than a Theme: the relevant body part is ‘inserted into’ these
items of clothing. (These items differ from other kinds of clothing, such as
tiab ‘skirt’, ris ‘trousers’, tsho ‘jacket’, with which the verb hnav ‘put on’ is used.)
Although this use of the verb rau to introduce a Goal NP is quite restricted, it
does shed light on the extremely common use of rau to introduce Goals of vari-
ous kinds. In the following example rau functions to indicate first Theme and
then Goal:
(52) ua.ntej rau dejO/O rau [hauv laujkaub]O

first water inside pot
‘First you put some water into a pot.’ (Adapted from: WHD 17: 13)
This second use of rau in examples like (52) is more easily understood when its
true meaning as an ordinary verb—‘to put in’ rather than simply ‘to put’—is
revealed. It is also more easily understood when we note that, even as the only
predicate in the clause, it can function to introduce not only Themes, but also
Goals.
In some cases, as in example (52) above and examples (53) and (54) below,
the meaning of rau following another verb to introduce a Goal is identical, or
very close, to its meaning as a main verb introducing its own Goal.
(53) nwsA/A cog [yub nplej]O/O rau [tom liaj]O

3SG plant seedling rice over.there paddy.field
‘She planted rice seedlings in the paddy fields.’
(54) tsaws los hliv rau [hauv lub dab]O

lift.off CONJ pour in CLF trough
‘. . . lift (it) off and pour (it) into the trough.’ (KNH 16)
In these two examples, and also in example (52) above, the verb that precedes
rau is a Transfer verb (rau ‘insert’, cog ‘plant’, hliv ‘pour’)—a transitive verb that
describes an action in which the subject is responsible for the change of loca-
tion (physical or metaphorical) of the object, but in which the subject him/
herself does not move. Furthermore, all three of these Transfer verbs are ones
that involve the actual insertion of their object into a receptacle (the Goal),
and so the semantics of rau are ideally suited to introducing the Goal of these
actions (e.g. (53) ‘She planted the rice seedlings, inserting (them) into the
paddy field’; (54) ‘. . . pour (it), putting (it) into a trough’).
These sentences with rau are clearly examples of the construction intro-
duced in Chapter 3, §3.4.7, which is derived from the Disposal SVC. Both verbs
predicate the subject (an Agent), and both share the same direct object (a
Theme). There is also an unshared argument—the object that appears after
rau—that indicates the Goal. Thus rau has two object arguments (a Goal and
a Theme), but is responsible for introducing only one of them (the Goal) into
the clause.
When appearing with other types of Transfer Verbs, ones which do not
involve the actual insertion of their object into a receptacle, rau is used in a
far less literal sense. However, it still clearly serves to introduce a Goal of some
sort:
(55) wbA/A pub [ib daig tiab]O/O rau nwsO

‘We gave a skirt to her.’
218 CHAPTER 4
(56) nwsA/A muag [ib lub tsheb]O/O rau [nws tus phoojywg]
3SG sell one CLF car 3SG CLF friend
‘He sold a car to his friend.’
In sentences such as (55) and (56), although rau has lost a significant compo-
nent of its literal meaning (in that actions like and ‘giving’ and ‘selling’ do not
involve any actual or metaphorical ‘insertion’ into a receptacle), it retains the
same gross semantic force, describing the transfer of a Theme, by the Agent, to
a destination. Instead of meaning ‘put/insert (it) into’, in these examples rau
has the more general meaning “transfer (it) to” (‘We gave some money, “trans-
ferring” it to him’ in (55); ‘He sold a car, ‘transferring’ it to his friend’ in (56)).
The role of rau in sentences of this type is only properly appreciated in the
context of an understanding of the restrictions on the expression of verbal
valency in White Hmong. As mentioned before, White Hmong is a language in
which verbs have a low degree of lexical complexity in certain respects. One
important ramification of this is the fact that there are no Goal verbs in White
Hmong that can simultaneously introduce two object arguments into the
clause (see Chapter 1, §1.6.6). Thus there are no exact White Hmong equiva-
lents to English sentences like the following, in which the English verbs give
and sell introduce first a Goal and then a Theme:
(57) We gave her a skirt.
(58) He sold his friend a car.
The White Hmong verbs pub ‘give’ and muag ‘sell’ can each introduce only one
object in one clause. In the case of muag ‘sell’, as with most other Transfer
verbs, that object is interpreted as a Theme:
(59) a. nwsA muag [ib lub tsheb]O

3SG sell one CLF car
‘He sold a car.’
b. ?? nwsA muag [nws tus phoojywg]O

3SG sell 3SG CLF friend
Lit: ‘He sold his friend.’ (not: ‘He sold (it) to his friend.’)
c. * nwsA muag [nws tus phoojywg]O [ib lub tsheb]E

3SG sell 3SG CLF friend one CLF car
Intended meaning: ‘He sold his friend a car.’
In the case of pub ‘give’, a single object tends to be interpreted as a Goal. Both
Theme and Goal cannot appear together:
(60) a. wbA pub nwsO

1DU give 3SG
‘We gave (it) to her.’
b. ?? wb pub [ib daig tiab]

1DU give one CLF skirt
Intended meaning: ‘We gave a skirt (to someone).’
c. * wb pub [ib daig tiab] nws

Intended meaning: ‘We gave her a skirt.’
d. * wb pub nws [ib daig tiab]

1DU give 3SG one CLF skirt
Intended meaning: ‘We gave her a skirt.’
To simultaneously express both the Theme and the Goal of these actions, rau
is employed to increase the valency of the clause, as shown above. Even when
rau itself is used as a main verb, it must be used again in support of this verb
if the speaker wishes to express both Theme and Goal. This was illustrated in
example (52) above and is repeated here in (61) below:
(61) a. ua.ntej rau dejO/O rau [hauv laujkaub]O

‘First (you) put some water into a pot.’ (Adapted from: WHD 17: 13)
b. ?? ua.ntej rau dej [hauv laujkaub]

Lit: ‘First put some water in (i.e. into something) in a pot.’
The only possible interpretation for (61)b is that hauv laujkaub ‘inside pot’ is an
outer locative Location: ‘First you put some water in (something) (while you
are) in a pot’. Although exhibiting some divergence from, or ‘bleaching’ of, its
basic lexical meaning of ‘put into, insert’, the morpheme rau in the examples
given above retains the feature of predicating the subject—the Agent respon-
sible for the transfer—and of introducing an object which is the Goal of the
transfer.
220 CHAPTER 4
Another kind of Goal that can be introduced by rau is the Goal of a Speech
Act verb. If the Speech Act verb takes a nominal object, as in (62) below, this
object appears directly after the verb, before the NP introduced by rau; if the
Speech Act verb takes a complement clause, as in (63) and (64), this tends to
appear sentence finally because of its relative weight.
(62) nwsA cem kuvO rau lawv

3SG scold 1SG 3PL
‘He complained about me to them.’
(63) [tus tsov]A thiaj hais rau [tus qav] <tias . . . >COMP
CLF tiger so say CLF frog COMP
‘. . . so the tiger said to the frog . . .’
(64) kuvA thiaj qhia rau nws <hais.tias . . . >COMP

1SG so tell 3SG COMP
‘So I told him that . . .’
This use of rau is further still from its literal meaning of ‘put/insert into’,
although the semantic connection is clear. In all three examples given above,
rau retains the feature of predicating the subject. However, in these examples
there is no shared object. It is not kuv ‘I/me’ who is ‘transferred to them’ in
example (62), but rather the whole act of ‘scolding me’; likewise, in (63) and
(64), it is not ‘the fact that . . .’ that is transferred, but the telling of it. In this
usage, rau does not introduce an object that is the Goal of a transfer, but rather
an individual who is the Goal/Recipient of a speech act.
There is yet another usage of rau, in which it appears as an alternative (or in
addition) to the use of mus ‘go’, tuaj ‘come’, and los ‘return home’, to introduce
the Goal of a Locomotion or Transport verb (see examples (24)–(26) in §4.2.3.2
above). Although still serving to introduce a destination, this usage of rau
marks a significant departure from those discussed so far. This is because, in
examples of this type, rau is functioning to express the movement of the sub-
ject to a Goal; the subject does not ‘transfer’ something to the Goal, but rather,
moves to the Goal him/herself (only in some cases ‘transporting’ an object
with it).
(65) wbS rov qab dhia dua rau [sab tod] soj
1DU return back jump again side over.there IP
‘We two will go back and jump to that side again!’ (LPV 2)
(66) [lub tsheb]S khiav [ntawm [lub zos no] rau [lub
CLF vehicle run nearby CLF village this CLF
zos nrad]
village down.there
‘The bus runs from this village to the village down there.’
(67) wbA hla [tus dej]O rau Thaib.teb

1DU cross CLF water Thailand
‘We crossed the river to Thailand.’
(68) lawvA lawv [cov npua]O rau [pem teb]

3PL drive CLF.PL pig up field
‘They drove the pigs up to the fields.’
Even in the final example, (68) above, in which the subject is responsible for
the change of location of the object, there is still no actual ‘transfer’ involved.
Instead, the subject ‘transports’ the object to the destination, moving there
him/herself and taking the object along.
Obviously rau in these examples differs from mus ‘go’, tuaj ‘come’, and los
‘return home’ in that it lacks marking for deixis. It also differs in that it indi-
cates that the Goal has been or will be attained. This component of the mean-
ing of this grammaticalized use of rau is clearly related to its basic meaning
as a predicate ‘to put in, to insert’; rau really marks a Destination rather than
simply Goal. It is also related to the assumed origin of this usage of rau in the
Disposal SVC type. Recall that the Disposal SVC is one of the SVC types that
function to assert the effectiveness of the action of the A argument. When
mus ‘go’, tuaj ‘come’, and los ‘return home’ are used, on the other hand, they
function to introduce the final junct in a Cotemporal SVC, which focuses
on the action of the subject (rather than on the outcome in relation to any
object), and there is no assertion that the Goal is attained. While it is gener-
ally assumed that a Goal introduced by one of these Deictic Motion verbs
is, indeed, reached, this assumption can be contradicted. This applies both
when the Deictic Motion verb is the only verb in the clause, and when it is in a
Cotemporal SVC:
(69) a. nwsS mus tsis txogA nram hav-dej

3SG go NEG arrive down valley-water
‘He went down to(wards) the river valley, (but) didn’t get there.’
222 CHAPTER 4
b. nws khiavS/S mus tsis txogA nram hav-dej

3SG run go NEG arrive down valley-water
‘He ran down to(wards) the river valley, (but) didn’t get there.’
To specify both deixis and attainment of Goal, a Deictic Motion verb can be
used in combination with rau:
(70) nwsS mus rau Asmeslivkas lawm

3SG go America PRF
‘He has gone to America.’
(71) ces Muam.Nkauj.LiagA txawm lawvS/S qab mus rau hauv

CONJ PN then follow behind go inside
‘. . . and then Mua Nkau Lia followed (him) inside (away from the
speaker).’ (DNH 451: 11)
As noted above, this usage of rau differs from some of the others discussed
above, in that the subject does not ‘transfer’ something to the Goal, but
rather, moves there him/herself. Nevertheless, even when such a consider-
able portion of its original lexical meaning has been ‘bleached’, rau can still
be thought of as predicating the subject in these examples. It simply serves to
mark the Destination of the subject, rather than the Destination of the Theme.
Consequently, expressions of this type with rau are not completely acceptable
if they appear initially as part of a topic:
(72) ?? ntawm [lub zos no] rau [lub zos nrad] mas
nearby CLF village this CLF village down.there TOP
[lub tsheb]S khiav
CLF vehicle run
‘From this village to the village down there, the bus runs.’
There are, however, sentences that are quite similar in some ways, in which an
expression with rau can occur as part of a topic. Recall the example given as
(45) above and repeated here as (73):
(73) txij [[lub zos] <peb nyob>REL ] rau yav.pem.toj mas ua

CLF village 3PL live to mountain TOP make
tau [[teb pob.kws] thiab [teb yeeb]]O/O nkaus xwb
get field corn and field opium only only
‘From the village we lived in up to the mountains, (we) grew corn and
opium only.’ (Fuller, 1985, pp. 114–115)
The difference is that rau, like txij (§4.2.3.3 above), does not predicate the sub-
ject in this type of example. The expression with rau is a goal-oriented one
that refers to the extent of the predicate: the ‘growing (of corn and opium)’.
The ability of rau in this usage to occur as topic is precisely what we would
expect, on the basis of comparison with the behaviour of nyob, txij, mus, tuaj,
and los. It is only when the property of predicating the subject is lost, and rau
serves simply to modify the predicate, that it is grammaticalized sufficiently to
appear sentence initially as a topic.
4.2.3.5 Txog
The verb txog means ‘arrive / reach a destination’, as in example (74) below:
(74) txog novO xwb es kojS rov qab nawb

arrive here only CONJ 2SG return back IP
‘(You’ve) only just got here, and now you’re going back!’ (VLT 15)
However, as mentioned previously (Chapter 3, §3.3.2.1), achievement verbs like

txog do not occur very commonly as the only candidate for verb-hood in the
clause. It is far more common for txog to occur after a Motion verb, introducing
the Terminal Point of the motion:
(75) nkawdS mus~mus ib hnub yuav tsaus-ntuj, mus txog [ib

3DU RDUP~go one day IRR dark-sky go one
lub zos]O/O
CLF village
‘They walked and walked all day almost until nightfall, and arrived in a
village.’ (DNH 451: 11)
(76) [tus qav]S/A dhia txog [sab pas-dej tid]O

CLF frog jump side pond beyond(t.c.)
‘The frog jumped, landing on the other side of the pond.’
Sentences of this type are clearly examples of the Attainment Serial Construc
tion (Chapter 3, §3.3), in which the second verb serves to describe the success-
ful attainment of a goal. In keeping with the properties of all achievement
predicates that can appear as V2 in an Attainment SVC, the verb txog can also
appear in a separate, juxtaposed clause, in which it can be independently mod-
ified by operators such as aspect and negation (see Chapter 3, §3.3.5):
224 CHAPTER 4
(77) nwsS tseem tuaj tsis tau txog kuvO

3SG still come NEG ACHV 1SG
‘He hasn’t reached me yet.’
(78) Txiv.Nraug.NtsuagS mus tsis txog [niam.tais yawm.txiv

PN go NEG wife’s.mother wife’s.father
tsev]O
house
‘. . . the Orphan Boy could not get to his parents-in-law’s house (that day).’
(DNH 453: 30)
Another common use of txog is one in which its object is understood to be a

topic of speech or cognition. It is as if the subject mentally or verbally ‘arrives’
at the topic.
(79) peb tham txog [cov xov ntawd]

1PL chat CLF.PL news that
‘We discussed the news.’
(80) [tus tub] nco dheev txog [nws rab hneev]

CLF boy remember suddenly 3SG CLF crossbow
‘The boy suddenly remembered his crossbow.’
(81) peb thiaj.li to.taub txog [qhov <peb hais>REL]

1PL so understand thing 1PL say
‘. . . so we understand the things we are saying.’ (WHG 9)
Even in this figurative usage, txog can appear in a juxtaposed clause, with inde-
pendent operators such as aspect or negation:
(82) nws hais tsis txog [dab neeg Hmoob] (nws hais
3SG speak NEG CLF folk.tale Hmong 3SG speak
txog [dab neeg Nplog])
CLF folk.tale Lao
‘He didn’t talk about Hmong folk tales (he talked about Lao folk tales).’
(83) [nws tus xib.fwb] tseem paub tsis tau txog nws thiab
3SG CLF teacher still know NEG ACHV 3SG also
‘His teacher still didn’t know about him.’
In examples such as these, where txog appears in a separate, juxtaposed clause,

it is clearly the only candidate for verb-hood in that clause. Extension from its
literal meaning, therefore, does not preclude txog from being used as a verb in
these sentences. However, the positive sentences above, in which txog appears
in the same clause as the verb of speech or cognition, differ from Attainment
SVCs in that the NP introduced by txog is not shared. That is, the first verbs
in these sentences cannot themselves introduce a topic of speech or thought
without txog. This seems to be an extension of the use of txog in an Attainment
SVC, involving at least some level of grammaticalization. The semantic con-
nection is still clear: here txog serves to mark the attainment of the Goal (that
is, the topic) of speech or thought, the subject metaphorically ‘arriving’ at this
topic.
Whether txog is used to introduce a destination or a topic of speech or
thought, it cannot appear sentence initially in a sentential topic, as shown by
the ungrammaticality of the following examples:
(84) * txog [sab pas-dej tid] mas, tus qav dhia

side pond beyond(t.c.) TOP CLF frog jump
Lit: ‘Arriving at the other side of the pond, the frog jumped.’
(85) * txog [dab neeg Hmoob] mas, nws hais

CLF story Hmong TOP 3SG speak
Lit: ‘About Hmong, he spoke.’
This inability to appear in a topic is consistent with the fact that, in this usage,
txog maintains a predication relationship with the subject of the clause.
As shown in Chapter 3, §3.3.6.2, an adverb can intervene between the speech
or cognition verb and txog:
(86) kaj-ntug tus tub nco dheev txog nws rab hneev
dawn CLF boy remember suddenly arrive 3SG CLF crossbow
‘At dawn, the boy suddenly remembered his crossbow.’ (KNH 3)
This may be taken as a sign of some loosening of the juncture between txog
and the speech or cognition verb that precedes it. However, as shown by other
examples in Chapter 3, §3.3.6.2, adverbs do normally intervene between the
verbs in an Attainment SVC, and so this should not be taken as evidence that
grammaticalization may be underway in this usage of txog.
While txij is used with a time phrase to indicate the origin of an extent
of time (see §4.2.3.3 above), a time phrase introduced by txog marks the
conclusion—the time up to which the action/event described occurs.
226 CHAPTER 4
(87) kuvA yuav nyob [qhov no]O mus txog [thaum kuvA yuav
1SG IRR dwell place this go time 1SG IRR
tau [kuv ib tsev]O tso
get 1SG one house first
‘I will live here until I get my own house.’ (WHD 20: 20)
Predictably, this use of txog can be topicalized, as it does not involve predica-
tion of the subject:
(88) los txog xyoo 1960, [cov neeg Hmoob]A thiaj

come year CLF.PL person Hmong so
tau pib kawm los mus
ACHV begin study CONJ go
‘So by 1960, the Hmong people had started to learn (it) and went on
(learning).’ (WHG 6)
(89) txij [hnub <kuv yuav poj-niam>REL ] los txog [tav no],
day 1SG obtain wife come time this
tau muaj [ib xyoo]S
ACHV have one year
‘From the day I got married to the present time, it’s been one year.’
(Clark, 1979a, p. 6)
4.2.3.6 Nrog
Like many of the other morphemes discussed here, nrog rarely appears as the
only candidate for verb-hood in the clause. When it does, it means ‘be with’, as
in the following sentence:
(90) nrog [koj tus nus] nraim nawb

2SG CLF brother straight IP
‘Stay with your brother all the time, O.K.?’
Its meaning can be extended to indicate ‘be with’ in the sense of ‘support’:
(91) kuv nrog nws thiab

1SG 3SG also
‘I’m with him.’ (i.e. ‘I support him.’)
As pointed out by Clark (1980b, p. 12), the verb nrog refers to location rather
than motion; by itself, it does not mean ‘accompany’. It only has this interpreta-
tion when it appears with a Motion verb:
(92) nws nrog [nws tus phoojywg] mus kawm-ntawv

3SG 3SG CLF friend go study-writing
‘He goes to school with his friend. / He accompanies his friend to school.’
In the example above it is the verb mus ‘go’ which contributes the motion com-
ponent of the meaning. When nrog occurs with a verb that does not describe
motion, no such component is present:
(93) nag.hmo kuv nrog nkawd noj mov

yesterday 1SG 3DU eat rice
‘Yesterday I ate with them.’
The sentences in which nrog appears with another verb clearly resemble
Cotemporal SVCs expressing stance and action. The other verb in the sentence,
however, is not restricted to a predicate expressing action, but may also express
change of location (e.g. mus ‘go’, exemplified in (92) above) or change of state
(as in example (94) below).
(94) kojA puas paub <tias hnubS qis zus es

2SG Q know COMP sun be(come).low gradually CONJ
[kuv lub siab]S me zus nrog no>COMP nab?
1SG CLF liver be(come).small gradually this IP
‘Do you know that as the sun slowly sets my heart gradually shrinks with
it?’ (VLT 14)
Note that in this sentence nrog appears after the verb me ‘be/become small’,
unlike the usual order of verbs in a Cotemporal SVC, in which the stance verb
normally comes first. Again, in example (95) below, nrog appears finally:
(95) pebA muaj [kev <hloov [cov theem.suab <uas tsis sib
1PL have way change CLF.PL tone REL NEG RECP
haum, tsis zoo mloog nrog [lwm lub theem.suab]>REL
fit NEG be.good listen next CLF tone
‘We have a way of changing the tones that do not fit with each other,
which do not sound good with the next tone . . .’ (WHG 8)
228 CHAPTER 4
Even when it occurs with a verb expressing change of location, nrog can appear
finally:
(96) kuvA mus [kawm-ntawv]O nrog nws

1SG go study-writing 3SG
‘I go to school with her.’
The fact that nrog can appear finally in the examples above could be inter-
preted as a certain degree of preposition-like behaviour. However, it could also
be attributed to the fact that the order of verbs in a Cotemporal SVC expressing
stance and action is somewhat flexible. Although it is generally the case that
the stance verb appears first, the opposite order is also possible:
(97) a. nwsS/A sawv-ntsug noj mov

3SG stand eat rice
b. nwsA/S noj mov sawv-ntsug

3SG eat rice stand
‘He eats standing up.’
The clause-final appearance of nrog is certainly not as significant as that of

nyob (see §4.2.3.1 above), which can only appear in final position when intro-
ducing an oblique argument. The morpheme nrog, on the other hand, does
seem to favour initial position in this usage, but exhibits some flexibility.
Furthermore, in spite of varying degrees of divergence from its lexical mean-
ing, ‘be with’, it must be said that, unlike nyob, the morpheme nrog in all of
the examples above does seem to maintain a predication relationship with the
subject. Again, unlike nyob, nrog and its object cannot appear sentence ini-
tially as a topic:
(98) a. kuvA/A nrog [kuv tus phoojywg]O mus kawm-ntawv

1SG 1SG CLF friend go study-writing
‘I go to school with my friend.
b. * nrog [kuv tus phoojywg] mas, kuv mus nrog kawm-ntawv

1SG CLF friend TOP 1SG go study-writing
(99) a. nag.hmo kuvA/A nrog nkawdO noj movO

yesterday 1SG 3DU eat rice
‘Yesterday I ate with them.’
b. * nag.hmo nrog nkawd mas, kuv noj mov

yesterday 3DU TOP 1SG eat rice
In general, it seems reasonable to conclude that the behaviour of nrog in this

usage is basically that of a verb in a Cotemporal SVC of the Stance and Action
type.
The meaning of nrog ‘be with’ is often extended slightly in two important
ways. Firstly, it can be used to indicate not just that two parties do something
together, but that they help each other to do it. For example:
(100) kuv nrog koj rho kuv cov pej.xeem mus

1SG 2SG extract 1SG CLF.PL populace go
‘I’ll help you to take my people away.’ (Clark, 1980b, p. 32)
(101) lawv nrog nws saib [nws cov me-nyuam]

3PL 3SG watch 3SG CLF.PL child
‘They helped her look after her children.’
Secondly, it is often used when one party uses property belonging to another,
in which case it seems to imply that the property is being shared—that both
parties have access to it:
(102) kuv nrog hlob Yeeb siv nws tus ntaiv

1SG uncle PN use 3SG CLF ladder
‘I used Uncle Yeng’s ladder.’
(103) kuv nrog nws hnav nws lub tsho

1SG 3SG wear 3SG CLF jacket
‘I wore her jacket.’
The sentences above tend to be interpreted as meaning that the two parties
are really physically together at the time referred to; thus the basic meaning of
nrog ‘be with’ is not lost in sentences of this type; it is simply extended.
This second extension of the meaning of nrog is very commonly used in
polite requests such as the following:
(104) nrog koj siv koj cov tais?

2SG use 2SG CLF.PL bowl
‘May I use your bowls?’
230 CHAPTER 4
(105) nrog niam qiv khaub.ncaws?

mother borrow clothes
‘Mum, may I borrow some clothes?’
(106) kuv tuaj nrog koj thov nyiaj me.ntsis

1SG come 2SG request money a.little
‘I came to ask for a bit of money.’
Clark (1980b), discussing sentences similar to (105) and (106) above, claims
that nrog serves to introduce an animate Source argument of a Transfer verb,
such as qiv ‘borrow’ or thov ‘request’. However, nrog is clearly being used here
with exactly the same meaning as in sentences with other verb types, such
as siv ‘use’, hnav ‘wear’ above. This suggests that the interpretation of the NP
introduced by nrog as a Source is not strictly correct. It is true that the refer-
ent of this NP is, indeed, the Source of the transfer in the real world, but in
the examples above it is expressed as a comitative argument. Note that it can-
not be used when a transfer takes place between parties who are not actually
physically together:
(107) ?? kuv nrog kuv tus muam txais tau ib tsab ntawv
1SG 1SG CLF sister receive get one CLF writing
Intended meaning: ‘I received a letter from my sister.’
This example is given as grammatical by Clark (p. 33), but is not accepted by
White Hmong consultants for this study, with the meaning given. The only
possible interpretation that seems to be available is: ‘I received a letter with
my sister.’ (i.e. she was with me when I got it, or it was addressed to both of us).
As many verbs expressing transfer are inherent Source verbs in White Hmong,
a simple Locative Phrase will be interpreted as expressing Source with a verb
such as txais ‘receive’ (see 1.5.5.3):
(108) kuv txais tau ib tsab ntawv ntawm kuv tus muam
1SG receive get one CLF writing nearby 1SG CLF sister
‘I received a letter from my sister.’ (Clark, 1980b, p. 31)
4.3 Conclusion to Chapter Four
In the first section of this chapter it was asserted that the proper criteria for
determining part-of-speech classification are formal, rather than seman-
tic or functional. Nevertheless, this does not in any way imply that meaning
and function are isolated from grammatical behaviour. On the contrary, this
relationship, far from being arbitrary or idiosyncratic, can be shown to be thor-
oughly motivated.
It was noted that, in order to ascertain the verbal status or otherwise of a
verb-like morpheme, it is essential to determine what properties would be
expected of a verb appearing in the environment in which the morpheme
occurs. In particular, it has been necessary to determine the exact SVC type
from which the sentence is putatively derived.
In the case of White Hmong, syntactic tests reveal that some change in
grammatical behaviour does indeed take place when erstwhile verbs are used
to fulfil certain functions. Most notably, when a verb-like morpheme is used to
introduce the Location, Extent, or Goal/Destination of the action as a whole,
rather than that of the subject, its grammatical behaviour is less like that of a
verb in a serial construction and more like the kind of behaviour that might be
expected of a preposition. For example, it may appear with its object as, or as
part of, a sentence initial topic, and/or it may appear in a more peripheral posi-
tion in the clause than would the corresponding serial verb. In such cases the
morpheme concerned could be regarded as, at the very least, on the pathway
to grammaticalization as a preposition.
The same syntactic tests, however, have also revealed the fact that, in other
cases when verbs are used to introduce oblique arguments, there is generally
little or no change in grammatical behaviour from that expected in a serial verb
construction of the relevant type. This is usually the case when the verb main-
tains an unequivocal predication relationship with the subject of the clause,
and thus maintains at least one shared argument with the other predicate in
the clause. The sharing of the subject argument is what would be expected
in the Cotemporal, Attainment, and Disposal Serial Constructions which the
sentences concerned represent. In such cases the morpheme should still be
regarded as a verb.
The argument that whenever a verb is performing a ‘preposition-like’ func-
tion it should be expected to exhibit more ‘preposition-like’ behaviour, is
ostensibly championing the link between form and function. However, such
an argument ignores the function of serial verb constructions themselves: to
allow two or more predicates to work together to express different facets of a
single event. One way that this operates is for one predicate to introduce argu-
ments into the clause that are part of its logical structure, not shared with the
logical structure of the other predicate, but which can nevertheless be thought
of as involved in the same event. In this way the valency of the entire clause is
greater than that of a clause that contains either of the two predicates alone.
Thus the function of valency increasing is closely related to the grammar of the
construction utilised to express it: two verbs within one clause expressing two
232 CHAPTER 4
facets of one event. Verb-serializing languages do not always need prepositions

to fulfil the valency-increasing function, because they can utilise verbs.
However, a verb in White Hmong may extend this function of introducing
arguments into the core to that of serving to introduce peripheral arguments
such as Outer Locative Location, Extent etc. In such cases there is no shared
argument; the erstwhile verb clearly does not maintain a predication relation-
ship with the subject. It is no longer a case of two verbs working together within
the core, but of one verb that predicates the subject, and one preposition—the
sole function of which is to introduce a peripheral argument into the clause.
CHAPTER 5
Superficially Similar Constructions
There are a considerable number of constructions in White Hmong which

superficially resemble serial verb constructions, but which are not regarded as
involving verb serialization for a variety of reasons. In this section these con-
structions are described, and the reasons why they do not qualify as serial verb
constructions are briefly reviewed.
5.1 Elaborate Expressions
Elaborate expressions are an aesthetic device, generally involving four com-

ponents, and found in many of the languages of Southeast Asia, including
Thai (Haas, 1964, pp. xvii–xviii), Lahu (Matisoff, 1973, pp. 81–86, 297–301), and
Burmese (Wheatley, 1990, p. 850). These expressions utilise devices such as
repetition and similarity of sound and meaning, making the language sound
more beautiful and balanced, and less abrupt. Some types of elaborate expres-
sions in White Hmong are similar to Cotemporal SVCs, particularly those that
portray cotemporal actions. Both the similarities and the differences are wor-
thy of note.
In White Hmong, elaborate expressions are referred to as lus ua txwm (word
make pair) ‘paired words’. Johns and Strecker (1982, p. 162) explain:
Paired words typically occur in four-part expressions in which the first

and third parts are related in some way and the second and fourth parts
are related in some way.
They illustrate this with the examples below, in which the first and third words
are identical, while the second and fourth are semantically similar in the first
case, and almost synonymous in the second:
(1) khwv iab khwv daw

toil be.bitter toil be.salty
‘arduous toil’
(2) kav teb kav chaw

rule land rule place
‘rule a country’

234 Chapter 5
In example (1), each half of the expression is composed of a noun modified by

an intransitive stative verb (NVNV), the whole expression thus functioning as
a noun phrase, with the pattern ABAC. In (2), each half comprises a transitive
verb followed by a noun (VNVN), and the whole functions as a predicate, again
with the pattern ABAC. Mortensen (2003: 17–22) illustrates a wide variety of
other possible types of elaborate expression in White Hmong, including those
in which each half is made up of noun plus noun, intransitive verb plus noun,
verb plus verb, etc., along with the additional patterns ABCB and ABCD.
It is elaborate expressions like example (2) above that are most similar to
Cotemporal SVCs. In terms of arrangement of components, they are identi-
cal to Cotemporal SVCs with transitive verbs: the first and third elements are
transitive verbs and the second and fourth are NPs in O function in relation to
those verbs. Semantically, like Cotemporal SVCs, they convey a strong sense of
two parts of one whole—two related aspects of one event or situation.
As a result, elaborate expressions of this type tend most commonly to be
treated by speakers as belonging to a single clause, rather than two separate
clauses, as shown in the following examples:
(3) txij thaum nws mus kawm txuj kawm ci tag . . .

since time 3SG go study skill study art finish
‘Since he finished learning the art [of magic] . . .’ (DNH 29:20)
(4) yuav-tsum noj mov haus dej

should eat rice drink water
‘should eat and drink’
(5) tsis muaj plaub muaj ntug . . .

NEG have dispute have litigation
‘don’t have legal disputes . . .’ (DNH 118:19)
Each of the examples above shows a different operator with scope over the
entire elaborate expression: the aspectual operator tag ‘finish’ in (3), the core
modal operator yuav tsum ‘should’ in (4), and the negative tsis ‘not’ in (5). The
elaborate expressions in these examples are clearly being treated as belonging
to a single clause.
However, elaborate expressions can also involve a much looser form of junc-
ture, namely juxtaposed clauses (see §5.2 below), in which each junct can be
modified independently by operators at all levels. Examples (6) and (7) show
that both juncture possibilities can and do occur with the same expression:
Superficially Similar Constructions 235
(6) peb tsis muaj tes muaj taw

3PL NEG have hand have foot
‘We have neither hands nor feet.’ (DNH 123:5)
(7) tsis muaj tes tsis muaj taw li

NEG have hand NEG have foot at.all
‘. . . it didn’t have hands and didn’t have feet.’ (DNH 117:14)
In (6) the two parts of the expression are treated as belonging to a single clause,
like a serial verb construction, and the negative tsis has scope over both juncts;
in (7), however, the same operator is used to modify each part independently.
Elaborate expressions (of this VNVN type at least) are thus by no means all
inviolate compounds, but can be subject to the normal syntactic processes of
the language. Observe example (8):
(8) yog neb noj taus Siv.Yis qav haus taus Siv.Yis dej no
COP 2DU eat can PN food drink can PN water IP
‘. . . if you can eat Shi Yi’s food and drink Shi Yi’s drink . . .’ (DNH 31:23)
The basic components here are similar to those in example (4) above—noj mov
haus dej (eat rice drink water). In this case, the second and fourth e lements—
the nouns qav ‘food’ and dej ‘water’—are each modified by a proper noun
functioning as a possessive, Siv Yis. In addition, the two predicates in this
example are not simple verbs, but are both followed by a deontic modal (noj
taus . . . haus taus (eat can . . . drink can)), resulting in further separation of the
components of the expression.
Although some elaborate expressions, like the example above, are com-
pletely accessible to literal interpretation, and subject to all the normal syn-
tactic processes of the language, many examples are somewhat idiomatic. For
example:
(9) hais lus khawm lus

speak word embrace word
‘swear to a fact / take an oath’ (Johns & Strecker, 1982, p. 163)
Another indication of the unity of expressions such as this, in addition to their

idiomatic meaning, is that the pairs of related words that regularly appear
together in either first and third, or second and fourth slot in these expressions
tend to become frozen in different ways. First, they are fixed in their order. For
236 Chapter 5
example, yeeb ‘shamanism’ . . . yaig ‘divination’ . . . never appear in the opposite

order, as yaig . . . yeeb. . . . Mortensen (2003, pp. 7, 23) explains that the order
of the nominal elements is based on the historical tone categories. Secondly,
they are often frozen in terms of the partner they can appear with. The word
qoob ‘grains, crops’ appears with loo ‘dry field’ to mean ‘grains, crops’, as in the
expression ua qoob ua loo (make crops make fields) ‘to tend crops’ but qoob is
never paired with teb ‘land, field’ (*ua qoob ua teb). This is so even though the
archaic word loo ‘dry field’ seems to have been entirely replaced by teb ‘land,
field’ elsewhere in the modern language, surviving only in collocation with
qoob in elaborate expressions (Johns & Strecker, 1982, p. 110) and in the com-
pound qoob-loo ‘grain, crops, products’.
However, while word pairs often tend to be frozen, there is some flexibility
when it comes to combining word pairs with one another in different elabo-
rate expressions. Johns & Strecker (1982, pp. 163–164) give the example of the
word pair . . . teb ‘land’ . . . chaw ‘place’:
If we insert kav ‘rule’ we get kav teb kav chaw ‘rule your country.’ If we
insert hlub ‘cherish, help, take care of’ we get hlub teb hlub chaw ‘protect
and cherish your country’ and so forth.
Ultimately, the distinction between elaborate expressions and other strategies

in the language that can be similar in form, such as SVCs expressing Cotemporal
Actions, should not be thought of as entirely clear-cut. Of course, elaborate
expressions are not restricted to the VNVN type; as shown above, these expres-
sions also occur in arrangements such as NVNV and a variety of others. For
this reason, if no other, the description of this phenomenon must be indepen-
dent of that of verb serialization. Furthermore, those cases that do occur in
the VNVN pattern allow not only action verbs but also state verbs, as shown in
examples (5)–(7) above, further distinguishing them from Cotemporal Action
SVCs. Nevertheless, elaborate expressions of the VNVN type do seem to at least
partially intersect with the phenomenon of serialization. At one extreme are
the completely frozen or highly idiomatic examples of elaborate expressions,
and at the other are the entirely literal and productive serial verb construc-
tions. However, if an elaborate expression of the VNVN type is entirely acces-
sible to literal interpretation, and the two juncts are treated as belonging to a
single clause, then it could simply be regarded as a rather poetic and metri-
cally balanced kind of SVC. Similarly, if a Cotemporal Serial Construction is
utilised to express multiple actions as part of a single event, then a graceful
speaker of the language might choose to overlay some of the features of an
elaborate expression, incorporating elements such as rhyme and alliteration,
to emphasise the balance and symmetry between the actions concerned. In

such cases there would be little to distinguish this construction from an elabo-
rate expression.
5.2 Juxtaposed Clauses
Some languages that utilise the strategy of verb serialization, including White
Hmong, are highly paratactic in other ways, allowing clauses with a range of
syntactic and semantic relationships to be simply juxtaposed, with no overt
form of linkage. A number of authors (e.g. Li and Thompson (1973), Tai (1984),
Hansell (1987), Sebba (1987), Clark (1992)), include certain constructions con-
taining juxtaposed clauses in the class of serial verb constructions, some refer-
ring to them as ‘coordinate’ serial verb constructions. In this analysis, however,
sentences in White Hmong containing more than one clause are considered,
by definition, not to involve serialization. If the term ‘serialization’ is to have
any meaning beyond simply ‘juxtaposition’ or ‘concatenation’ in the analysis of
this language, then it seems necessary to exclude these multiclausal structures.
It is only by doing so that serialization can be presented as a cohesive phenom-
enon, exhibiting common features both syntactically and semantically.
It is important to note that the phenomenon of juxtaposed clauses in
White Hmong should certainly not be thought of as involving a single con-
struction type; a range of constructions in this language can involve the
simple juxtaposition of clauses. The purpose of this section is simply to out-
line their differences from serial verb constructions. Further investigation is
clearly warranted.
In some cases of juxtaposed clauses in White Hmong, the only obvious dif-
ference from a serial construction will be the distinct pause that can optionally
occur between the clauses. They can, however, be shown to differ from serial
verb constructions syntactically, by observing possibilities such as the option
for independent core operators and independent peripheral arguments for
each junct. Depending on the exact level of juncture, the juncts in a sequence
of juxtaposed clauses may even exhibit independent clause-layer operators
and have no shared core arguments at all.
Juxtaposed clauses differ from serial verb constructions semantically in
that each junct, being a separate clause, expresses a separate event. Various
semantic relationships can hold between such clauses, including contrast,
circumstance/condition, and consequence.
When juxtaposed clauses are in a contrastive relationship they may have
the same subject, as in example (10) below, or have different subjects, as in (11):
238 Chapter 5
(10) me Muam.Nkauj.Liag, kojA nyob [hauv tsev] kojA txhob

little PN 2SG stay inside house 2SG IRR.NEG
nce nthabO nawb
ascend platform IP
‘Dear Mua Ngao Lia, you stay at home, (but) don’t you go up to the attic
platform, OK?’ (DNH 449:3)
(11) ua ib hnub luagA muab [mov <txaus rau

do one day other give rice enough for
nws coj los <noj>PURP]O xwb
3SG take.along come.home eat only
‘(He) worked all day (but) they [Lit: the others—the Chinese] gave him
only enough rice to take home to eat’. (TNN 1)
(12) yuav noj neegO, yuav muaj mobO muaj tuagO

IRR eat person IRR have sickness have death
‘(The evil spirits) will eat people, (and) (the people of the world) will have
sickness and death.’ (DNH 27:8)
(13) no ces [tus kwv]A txawm nyob

this CONJ CLF younger.brother then stay
[ntawm ntug dej]O tos Txiv.Nraug.Ntsuag txawm mus
nearby edge water wait the.Orphan.Boy then go
<suav~suav ntsesO>PURP
RDUP~count fish
‘And so then the younger brother stayed by the edge of the stream wait-
ing; the Orphan Boy went to count the fish.’ (DNH 183:2)
Likewise, when the relationship is one of circumstance or condition, the

clauses may have the same subject, as in (14) and (15), or different subjects, as
in (16) and (17):
(14) yuav tawv~tawv tsis zoo noj

IRR RDUP~be.tough NEG good eat
‘(The meat) will be very tough (and so) not good to eat.’ (DNH 8:43)
(15) kojA tsis paub qhov-twgO kojA xa mus rau kuv

2SG NEG know something 2SG send go to 1SG
‘(If) there’s anything you don’t know, send (a letter) to me.’ (Clark, 1992,
p. 153)
(16) thiab nebA kho tag nyiajS npaum li cas?

and 2DU fix finish money equal as how
‘And having repaired (it), how much did it cost? (Clark, 1992, p. 153)
(17) kojS ua, kuvS zoo-siab heev

2SG do 1SG be.good-liver very
‘You do so (and) I’ll be very happy.’
In juxtaposed clauses that exhibit different subjects, no core arguments are

shared and juncture is clearly at the clausal level. When two juxtaposed clauses
have the same subject, this may be a shared argument, as in example (14) above,
in which case the juncture is at the core level. However, example (18) below dif-
fers in that, even though the subjects of the two clauses are the same, it allows
a coreferential subject pronoun in the second junct. Coreferential arguments
are not shared arguments (see Chapter 2, §2.4.1.3) so, in sentences like this in
which no core arguments are shared between the juxtaposed juncts, and we
can conclude that the level of juncture is at the clausal level.
(18) me Muam.Nkauj.Liag, kojA nyob [hauv tsev]O, kojA

little 2SG stay inside house 2SG
txhob nce nthabO nawb
NEG.IMP ascend platform IP
‘Dear Mua Ngao Lia, you stay at home, (but) don’t you go up to the attic
platform, OK?’ (DNH 449:3)
An important point of contrast between serial verb constructions and juxta-

posed clauses concerns the scope of certain operators. The two juncts in a serial
construction share all core and clause-level operators. Those juxtaposed clauses
that share a subject, and thus involve (non-serial) core level juncture, do tend
to share some clause-layer operators like epistemic modality and illocutionary
force. Note the shared epistemic modality operator in example (14) above. This
sharing of outer clause-layer operators, coupled with the fact that the subject is
shared, reinforces the fact that the second clause is not completely independent
of the first in some types of juncture with juxtaposed clauses.
However, when different subjects are involved, even illocutionary force may
be independent, showing that the type of juncture between juxtaposed clauses
may even be as loose as peripheral coordinate, each junct being able to stand
alone as a complete clause. This is illustrated in example (16) above, and in (19)
and (20) below:
240 Chapter 5
(19) nejS ua.ntej mus xya hnub nawb, kuvS tuaj nawb
2PL first go seven days IP 1SG come IP
‘You go first (and wait) for seven days O.K., (and) I will come, OK?’ (DNH
124:6)
(20) nyob os, los tsevO mas

stay IP return.home home IP
‘You stay, (and) I’ll go home.’ (WHD 3)
This is in complete contrast to serial verb constructions, in which the two

juncts share all clause-layer operators and arguments.
As shown above (Chapter 3, §3.1.3), Cotemporal SVCs are joined at the core
level and share a common periphery. Juxtaposed clauses may also be joined at
the core level, or at the clausal level. In either case, they differ from serial verb
constructions in that each junct in a series of juxtaposed clauses can have its own
peripheral arguments, such as outer locative and time phrases. This reinforces the
point that these constructions are multiclausal; verb serialization is not involved:
(21) mus ua Suav zog tsis tau kev pw, niab hnub
go do Chinese work NEG get way rest every day
pw ntawm tej tshav.puam
rest nearby somewhere open.ground
‘He went to work for the Chinese (but) had no place to rest; (so) every day
(he) rested in an open paddock somewhere nearby.’ (TNN p. l)
Example (21) above actually contains three juxtaposed juncts. The first mus
ua Suav zog ‘went to work for the Chinese’ is in a contrastive relationship to
the second tsis tau kev pw ‘had no place to rest’. The final junct expresses con-
sequence, and contains an independent time phrase: niab hnub pw ntawm tej
tshav puam ‘every day rested in an open paddock nearby’.
There is actually a joke that the White Hmong tell which simultane-
ously exploits both the structural similarity of the juxtaposed and the serial
constructions—that is the lack of an overt form of linkage between the two
juncts—and the difference between them with regard to the scope of the neg-
ative operator:
(22) a. kuvA tsis yug qaibO yug osO, kuvA tau qeO
1SG NEG raise chicken raise duck 1SG get egg
<noj>PURP . Yog vim li cas?
eat COP because as how
‘I don’t raise chickens/raise ducks, I get eggs to eat. How can this be?’
b. vim tias kuvA tsis yug qaibO tab.sis kuvA

because COMP 1SG NEG raise chicken but 1SG
yug osO
raise duck
‘It’s because I don’t raise chickens but I do raise ducks.’
In the first line, tsis yug qaib yug os ‘not raise chickens raise ducks’ is ambigu-
ous. If it is taken as involving juxtaposed clauses, then the negative tsis is
understood to have scope over only the first junct yug qaib ‘raise chickens’
and the correct interpretation can be made: ‘I don’t raise chickens (but) I
(do) raise ducks’. If, on the other hand, this sequence is taken as involving a
serial construction of the Cotemporal Action type, with the two juncts in a
core cosubordinate relationship, then the negative will have scope over both
juncts: ‘I don’t raise chickens (and I don’t) raise ducks’. This is clearly the
interpretation that the addressee is expected to make, and which results in
the apparent contradiction with the next statement: Kuv tau qe noj ‘I get eggs
to eat’.
What are the factors that lead to the serial interpretation being the more
likely of the two, even though it does not seem to make sense? Firstly, and
most obviously, the speaker wishing to promote the serial interpretation would
not pause between the two juncts as one might between two separate clauses.
Secondly, these two juncts have the form of Elaborate Expressions: in this
case an identity between the first and third elements and a close semantic tie
between the second and fourth. As noted above (§5.1), Elaborate Expressions
of the V NP V NP type exhibit similar semantic properties to those of serial
verb constructions (two facets of one whole) and, as a result, tend most often
to be interpreted as serial verb constructions, rather than as two independent
clauses. Thus the actual lexical items used contribute here to the interpreta-
tion of the sequence as one clause rather than two.
5.3 Complementation Resembling Cause-Effect SVCs
Some types of transitive verb occur as V1 in constructions that superfi-

cially resemble Cause-Effect SVCs, but which actually involve complement
clauses. In these cases, the superordinate and subordinate verbs are simply
juxtaposed, with no requirement for overt linkage in the form of a comple-
mentizer. Each of these different types of constructions, all of which are
superficially similar to the Cause-Effect Serial Construction, will be briefly
described in turn.
242 Chapter 5
5.3.1 Complements of Non-Implicative Manipulative Verbs that Appear

without a Complementizer
As shown by Jaisser (1984, pp. 63–77) and Jarkey (2006, pp. 125–126), non-
implicative Manipulative verbs, for example ntuas ‘advise/urge’, txwv ‘forbid’,
thov ‘beg/ask’, can all take complement clauses introduced by the complemen-
tizer kom.1 The manipulatee may appear as the object of the higher verb, or as
the subject of the lower verb. For example:
(23) a. kuvA ntuas LisO <kom mus>COMP

1SG advise PN COMP go
b. kuvA ntuas <kom LisS mus>COMP

1SG advise COMP PN go
‘I advised Lee to go.’ (Jaisser, 1984, p. 63)
These sentences are clearly not superficially very similar to any type of serial
verb construction, either syntactically or semantically. However, it will be help-
ful to understand a little more about this construction, in order to appreciate
the syntactic and semantic similarities to the Cause-Effect SVC that occur in
some cases when the complementizer is absent (see below).
Jaisser (1984, pp. 105–107) proposes that sentences with the complementizer
kom are only used when the action described in the complement clause is
intended, but as yet unrealised, at the time of the speech event. However, as in
example (23) above with kom, the complement clause is actually understood
to express something that was intended by the causer and which may or may
not have actually been realised at the time of the speech event. That is, the sen-
tence is non-implicative; the truth of the entire proposition does not imply the
truth of the complement clause. Because these sentences are non-implicative,
and so do not indicate that the desired outcome is actually realised, they are
best described as ‘intentional’ or ‘purposive’, rather than ‘causative’.
It has been shown above that the complements of non-implicative
Manipulative Verbs can take complement clauses introduced by the com-
plementizer kom. However, it is not uncommon for a verb of this kind to be
1 The morpheme kom can also function as a conjunction meaning ‘so that, in order to’:
(i) nwsA thawb [tus muam]O kom poob roojO
3SG push CLF sister CONJ fall table
‘He pushed his sister so that (she would) fall off the table.’
This use of kom and its use as a complementizer are both grammaticalized from the verb kom
‘to order’.
followed directly by its complement, without any complementizer. These

sentences look extremely similar to examples of the Cause-Effect Serial
Construction, especially to those in which V1 is a transfer verb (Chapter 3,
§3.2.1.2.c). Notice in examples (24)–(27) below that it is not obvious whether
the causee should be thought of as the O argument of the first verb or the
S argument of the second:
(24) koj mus hu [cov tub.mab.tub.qhe] los noj mov

2SG go call CLF.PL servant come eat rice
‘Go and call the servants to come and eat.’ (TNN 4)
(25) ces thiaj li tso [tus ntxhais] mus tsav nruab.hnub,

CONJ so as release CLF girl go drive daytime
tsav lub tsom kub
drive CLF (lamp).glass gold
‘And so (he) sent the girl (to) go and guide the daytime, (to) guide the
golden lamp [the sun].’ (DNH 10:55)
(26) tub, tso [koj tus poj-niam] tawm los rau kuv saib
son release 2SG CLF wife leave come for 1SG look.at
‘Son, let your wife come out for me to look at.’ (TNN 7)
(27) ces txawm cheem Muam.Nkauj.Liag pw ib hmos

CONJ then detain PN lie.down one night
‘And then (she) detained [i.e. invited] Mua Nkau Lia (to) stay for one
night.’ (DNH 451:14)
As seen in these examples, the verbs that appear in the V1 position in sentences
of this type may be Jussive verbs (e.g. hu ‘call’), or they may be verbs that indi-
cate manipulation without actually specifying that a speech act is performed
(e.g. tso ‘send, release’, cheem ‘detain’).
Sentences such as these differ from examples of the Cause-Effect Serial
Construction in that, just like the equivalent sentences with the complemen-
tizer kom, they are non-implicative; they involve manipulation rather than
causation. NP2 in these examples is always an Actor rather than an Undergoer
in relation to the second verb, and consequently NP1 is never entirely in con-
trol of the outcome. As a result of these factors, these sentences are defeasible.
The truth of the second junct is not necessarily implied by the truth of the
whole sentence; the entire proposition can be asserted, and the second junct
can still felicitously be denied:
244 Chapter 5
(28) nws hu [nws tus poj-niam] los tsev tab.sis [tus

3SG call 3SG CLF wife come home but CLF
poj-niam] tsis los
wife NEG come
‘He called his wife to come home but she didn’t come.’
(29) kuv txiv tso kuv mus ntawd tab.sis kuv tsis mus
1SG father release 1SG go nearby but 1SG NEG go
‘My father sent me there . . . but I didn’t go.’
However, even though these sentences are non-implicative, the clear assump-
tion tends to be that the causation is successful and that second junct does
occur. A denial of this occurrence, of the kind illustrated above, is rather unex-
pected. In other words, although V1 in these examples is a non-implicative
verb, the subject of this verb is still judged to have considerable control over
the outcome of the event, and the success of the manipulation is judged to be
likely. Why should this be the case?
Firstly, note that the person being manipulated in all the examples above is
always someone who would automatically be expected to fulfil the wishes of
the manipulator. In example (24) above, for example, the referent of this NP is
cov tub mab tub qhe ‘the servants’, and in example (27) it is young Mua Nkau
Lia, a potential daughter-in-law of the manipulator in the story from which
this example is taken. Wives, sons, and daughters are other typical referents of
the NP in this position in examples of this type. In all cases, although a human
being, this person is someone who, in the situation described, is not perceived
as having very much intention or volition him or herself.
Secondly, the action described by V2 in the construction without the com-
plementizer is invariably expressed by an intransitive verb (or SVC) indicating
change of location or stance (los ‘come’, mus ‘go’, tawm (los) ‘come out’, pw ‘lie
down’ in the examples above). The subject of these verbs is indeed an Actor,
rather than an Undergoer, but the Actor of verbs such as these is clearly not
highly agentive.
It is not at all surprising, then, that when one is trying to manipulate a more
powerful person, particularly to do something that involves a higher degree of
agentivity, the complementizer kom is used with a verb of this kind:
(30) nws huA <kom [niam thiab txiv]A quib qhov-roojO>COMP

3SG call COMP mother and father open door
‘He called (his) mother and father to open the door.’ (DNH 5:13)
These facts correlate with observations made in early work by Givón (1980,
1985), concerning the relationship between the semantic and syntactic proper-
ties of sentences involving complement clauses. Givón established a semantic
‘binding scale’ whereby:
The stronger the influence exerted over the agent of the complement
clause by the agent of the main-clause verb, by whatever means, the
higher is the main-clause verb on the binding scale the less is the agent
of its complement clause capable of acting independently . . . [and] the
more is the intended manipulation likely to succeed. (Givón, 1980, p. 335)
The syntactic consequences of this semantic dimension are as follows:
The higher a verb is on the binding scale, the less would its complement
tend to be syntactically coded as an independent/main clause. (Givón,
1980, p. 337)
The Manipulative sentences in which no complementizer appears are clearly

higher on the ‘binding scale’ than those in which a complementizer does
appear, even though both types are non-implicative. That is, in the type of sen-
tence in which no complementizer appears, the subject of the main verb has
stronger influence over the subject of the complement verb, who is, in turn, less
capable of acting independently. Consequently, the intended manipulation is
more likely to succeed. Givón’s observations concerning the syntactic conse-
quences of this semantic difference are borne out: the complement clause of
these sentences does lack an important signal of separate clause-hood, namely
a complementizer.
All other things being equal, the use of a subordinating morpheme which
neatly separates the main clause from its complement clause is a coding
acknowledgement that the two clauses are semantically still indepen-
dent of each other, at least to some extent. (Givón, 1980, p. 371)
So it seems that it is no accident that these sentences, in which V1 is a non-

implicative Manipulative Verb and V2 is introduced without a complementizer,
are similar both in form and meaning to the Cause-Effect Serial Construction.
Although we are dealing with one verb that expresses the proposition and one
that is sub-propositional in the case of the Manipulative sentences, the two
juncts in which they appear are structurally integrated in a way that parallels
their semantic ‘bondedness’.
246 Chapter 5
Even in the case of the Cause-Effect SVC itself, the integration of the
cause and the effect into a single clause could also be viewed as an exten-
sion of Givón’s observations concerning the correlations between the degree
of semantic ‘binding’ between two verbs and their degree of structural inte-
gration. There is no doubt that the two verbs in a Cause-Effect SVC in White
Hmong are strongly ‘bound’ semantically. Even when the shared argument is a
human being, s/he is not seen as acting independently in the event at all, the
referent of the initial NP having total influence over the outcome of the event.
Furthermore, as shown in Chapter 3, §3.2, these sentences involve an implica-
tive relationship: the truth of the second junct is implied by the truth of the
first. Although Givón is not referring to serialization in the following excerpt,
the relevance of his remarks to Cause-Effect SVCs is obvious:
In semantic terms, implicative verbs code a situation where one cannot

conceive of time separation between the event of the main clause and
that of the subordinate clause. Rather, they ‘hold’ at the very same point
in time. They are both equally asserted, they are both equally true. Their
syntactic/structural integration into a single clause/proposition form is
thus a most natural reflection of that semantic reality. [Emphasis in the
original.]
5.3.2 Complements of the Causative Verb ua and the Permissive Verbs

Like the non-implicative manipulative verbs described in §5.3.1 above, the
causative verb ua ‘make/do’ may be followed directly by a complement clause,
without the intervention of a complementizer:
(31) kuvA ua <kojS chim-siab>COMP

1SG make 2SG be(c0me).angry-liver
‘I made you angry.’
(32) nwsA ua <[tus muam]A poob rooj>COMP

3SG make CLF sister fall table
‘He caused his sister to fall off the table.’
Before discussing sentences in which this verb takes a complement clause, it

should be mentioned that the verb ua can actually occur in a Cause-Effect SVC,
in which case it is an Effective verb meaning ‘make/build something’:

3PL build CLF house RDUP~be(come).tall
‘They built the house high.’
In this sentence the NP lub tsev ‘the house’ really is a shared argument: it is
the O argument of ua ‘make/do’, as well as being the S argument of siav
‘be(come).tall’, the verb that follows it.
However, as seen in examples (31)–(32) above, ua can also be used as a more
general causative verb, in which case it no longer has its specific meaning of
‘make/build’. Like the Thai verbs tham ‘make/do’ and hây ‘give’, this causative
ua governs not just the NP that follows it, but rather the entire junct expressed
by the combination of this NP and the following predicate. For this reason sen-
tences with causative ua cannot be regarded as Cause-Effect SVCs. There is no
shared argument: it is not the case that the S argument of the final verb is the
O argument of the first.
Jaisser (1984) maintains that this type of causative sentence indicates
that the caused action is completed. This interpretation is the result of the
fact that these sentences, like those involving the Cause-Effect SVC, are
implicative: the truth of the proposition implies the truth of the complement
clause. Thus example (34) below is ungrammatical:
(34) nws ua <[tus muam] poob rooj>COMP * . . . tab.sis [tus

3SG make CLF sister fall table but CLF
muam] tsis poob rooj
sister NEG fall table
‘He caused his sister to fall off the table . . . but she didn’t fall off the table.’
However, these causatives differ semantically from the Cause-Effect SVC in a

number of ways. Firstly, whereas the latter may indicate intentional or uninten-
tional causation, causatives with ua always indicate unintentional causation.
Secondly, they do not indicate the exact nature of the causing event—only
that the causer did something which, without his or her intention, brought
about the caused event. The first verb in a Cause-Effect SVC, on the other hand,
always specifies the precise nature of the causing action. Finally, the relation-
ship between the cause and the result in a causative with ua may be quite
indirect; it certainly never involves the causer acting directly on the causee as
does the causer in a Cause-Effect SVC. Furthermore, as noted, the causee never
appears as the direct object of the superordinate clause at all.
248 Chapter 5
Like the Causative verb ua ‘make’, permissive verbs such as kam ‘be willing’,
kheev ‘be willing’, pub ‘give (permission)’, and cia ‘let’ can all take complement
clauses that are not introduced by a complementizer. Observe the following
examples:
(35) nws kam <[kuv txiv] ua>COMP

3SG willing 1SG father do
‘She is willing (for) my father (to) do (it).’
(36) tsis kheev <ua>COMP

NEG willing do
‘be unwilling (to) do (it)/ (for someone else to) do (it)’ (Heimbach
1979:97)
(37) [leej niam leej txi] [niaj hnub] mus ua teb es

CLF mother CLF father every day go make field CONJ
cia <Muam.Nkauj.Liag [niaj hnub] nyob [hauv tsev]>COMP xwb
let PN every day stay interior house only
‘Every day the mother and father went to work the fields and every day
just let Mua Ngao Lia stay at home.’ (DNH 449:2)
(38) wb pub <kuv mus>COMP

3SG give 1SG go
‘She gave me permission (to) go.’
Again, there is no shared argument in sentences of this type: the permis-

sive verb predicates not just the NP that immediately follows but the entire
complement.
5.3.3 Complements of Perception Verbs

Perception verbs, such as pom ‘see’ and hnov ‘hear’, also occur in sentences that
superficially resemble Cause-Effect SVCs:
(39) lawv hnov <hluav.taws kub>COMP

3PL hear fire burn
‘They saw a fire burning.’ (Johnson (1981), cited in Jaisser (1984))
With regard to this example and others involving complements of perception

verbs, Jaisser (1984, p. 96) claims that:
Semantically, the NP following the perception verb acts as both the

object of the perception verb and the subject of the verb of the comple-
ment . . . What did they hear . . .? They heard the fire. What was burning?
The fire was burning . . .
Jaisser is not actually suggesting that NP2, hluav taws ‘fire’ in this example is a
shared argument, but rather one that undergoes equi-NP deletion in her analy-
sis. That is, she proposes an underlying structure for sentences such as this
which involves treating perception verbs as Equi verbs, and deleting the ‘lower’
subject—the second occurrence of hluav taws ‘fire’ in the example given:
Analyzing perception verbs as Equi verbs has the advantage of capturing

both the object relationship of ‘fire’ to ‘hear’ and its subject relationship
to ‘burn’, . . . (Jaisser, 1984, p. 98)
However, an analysis such as this is probably not necessary for these sentences.
The reason is that the semantic interpretation on which it is based seems to be
incorrect. While it is certainly true that, in the situation described, ‘they’ did,
in fact, ‘hear a fire’, this is not information conveyed by the sentence. The first
verb (hnov ‘hear’) governs not only the NP that directly follows it (hluav taws
‘fire’), but the whole of the following junct (hluav taws kub ‘a fire burns’). What
was seen, according to this sentence, was not just the fire but the event of the
fire burning.
This observation is also made by Thepkanjana (1986, p. 123) concerning sim-
ilar sentences in Thai. Thepkanjana does still include sentences of this type in
the class of Thai serial verb constructions. However, they are not regarded as
serial constructions in this analysis. This is because, like sentences with the
causative verb ua, they do not conform to the constraint that the O argument
of the first verb functions as the S argument of the final one. Once again, there
is no shared argument in sentences such as this.
5.4 A Complementation Strategy with Modal Verbs
As shown above, complements of the causative verb ua ‘make’, permissive

verbs, and perception verbs are all types of complement clauses that can
appear without a complementizer. The resemblance of these sentences to
serial verb constructions, however, goes little further than the juxtaposition of
250 Chapter 5
the two verbs: the complement clauses do not share any arguments with the
main verb, apart from the fact, of course, that the entire complement clause is
itself an argument of this verb.
Constructions with Modal verbs, such as txawj ‘know how’ and sim ‘try’,
as well as xav ‘want’, kam and kheev ‘be willing’ with a modal interpretation
(‘want/like/be willing to do something oneself’), are also used without a com-
plementizer. These verbs differ from those discussed above (i.e. causative ua,
permissive verbs, and perception verbs), however, in that they all share an
argument—the subject—with the verb that follows them. In this respect they
are more like serial constructions. Of course, like the complement-taking verbs
discussed above, these Modal verbs also differ from verbs in series in that they
do not act as joint predicators with the other verb in the sentence; rather they
function as a complementation strategy (Jarkey, 2006).
(40) cov me-nyuam xav ua.si nram pas-dej

CLF.PL child want play down pond
‘The children want to play down at the pond.’
(41) koj puas kam mus tag.kis?

2SG Q willing go tomorrow
‘Are you willing to go tomorrow?’
In both of the examples above the peripheral arguments, nram pas-dej ‘down
at the pond’ and tag kis ‘tomorrow’ respectively, refer only to the location/time
of V2. They make no reference to the location/time setting of the modal verbs
in the sentence. However, some other Modal verb sentences are similar to
serial verb sentences, in that any temporal or spatial reference applies equally
to both the Modal verb and the verb that follows it. Observe these examples:
(42) nws sim tsav tsheb nram hav-dej

3SG try drive vehicle down valley-water
‘He tried to drive the car down in the river valley.’
(43) puag thaum ub, koj puas txawj hais lus Hmoob?
INTNS time yonder 2SG Q know.how speak word Hmong
‘Back in the old days, were you able to speak Hmong?’
(44) lawv pib sib tham hauv chaw.pw/ thaum peb teev
3PL start RECP chat in bedroom time three hour
‘They started chatting in the bedroom/ at three o’clock.’
If the scope of these temporal and spatial phrases over both juncts in examples
(42)–(44) above cannot be attributed to serialization, to what, then, can it be
attributed?
(Givón, 1973, p. 100) makes some observations that throw light on this ques-
tion. He points out that no Modal verbs, whether they are implicative or non-
implicative, allow the complement clause to have a time reference preceding
that of the main verb:
(45) *Yesterday John managed/wanted to read a book the day before.
However, non-implicative Modal verbs may exhibit the phenomenon of

‘future-gap’: the complement clause may have a time reference after that of
the main clause:
(46) Yesterday John wanted/decided/planned/hoped to read a book the follow-

ing day.
Implicative verbs, on the other hand, do not allow ‘future-gap’:
(47) *Yesterday John managed/started to read a book the following day.
In other words, implicative (and neg-implicative) verbs require simulta-

neity of time. And this is of course obvious, since the truth of the impli-
cative (or neg-implicative) verb at the designated time implies—at the
very time—the truth (or falsity respectively) of this complement verb. By
definition, a future-gap may not be opened. (Givón, 1973, p. 100)
Givón’s observations serve to explain why the verb following non-implicative

Modal verbs like xav ‘want’ and kam ‘be willing’ can have time reference inde-
pendent of that of the Modal verb, and why the complements of implicative
Modals like pib ‘start’ cannot. However, it is interesting to note that the Modal
verbs sim ‘try’ and txawj ‘know how’ are non-implicative, and yet, like the
implicative Modals, these verbs do not allow future-gap:
(48) * nag.hmo nws sim tsav tsheb tag.kis

yesterday 3SG try drive vehicle tomorrow
‘Yesterday he tried to drive the car tomorrow.’
(49) * tav ntej, kuv txawj hais lus Moob Ntsuab tav no
time first 1SG know.how speak word Mong Green time this
‘Previously, I was/am able go speak Mong Njua now.’
252 Chapter 5
Givón’s observations concerning the relationship between implicativity and

simultaneity of time reference need to be extended to account for these
discrepancies. In fact, they can be accounted for by utilising his own theory of
the ‘binding scale’.
The notion of ‘binding’ proposed by Givón (1980) has already been intro-
duced with regard to the semantic relationship between manipulative verbs
and their complements (see §5.5.3.1 above). The more tightly ‘bound’ the two
verbs are, the more is the subject of the main verb in control of the outcome,
the less independent is the subject of the second verb, and the more likely is
the intended outcome to be successful. The syntactic consequences of binding
are that the complement exhibits fewer features of a separate clause; it is more
integrated syntactically into the main clause.
The notion of binding can also be applied to Modal verbs, even though no
influence over another party occurs (Givón, 1980, p. 342). In this case also, the
semantic binding between the Modal verb and the verb that follows is again
related to the likelihood of the realisation of the action described by that sec-
ond verb. Thus, Modals such as xav ‘want’, which are non-implicative and sim-
ply express the ‘emotional attachment’ of the subject towards the realisation of
the second verb, are very loosely bound, while Modals such as pib ‘start’, which
are implicative and non-emotional, are tightly bound to the second verb.
One of the syntactic consequences of the degree of binding between Modal
verbs and the verbs that follow is the degree to which the second verb allows
independent peripheral arguments. Modal verbs such as sim ‘try’ and txawj
‘know how’ are not implicative, but they are, nevertheless, quite high on the
binding scale. This is because the subject of these verbs is not portrayed as hav-
ing an ‘emotional attachment’ towards the desired outcome; furthermore, in
the case of sim ‘try’, even though s/he may not be successful, the subject actu-
ally does something aimed at bringing the desired outcome about. As noted
above, like the implicative Modal verbs, the non-implicative sim ‘try’ and txawj
‘know how’ do not allow future-gap. Thus it seems that Modal verbs that do
not allow future-gap are not just the implicative Modals, such as manage/start,
but any Modal verb that is relatively high on the binding scale, whether it is
implicative or not.
Furthermore, Modal verbs high on the binding scale, again regardless of
whether they are implicative or non-implicative, are understood to be true at
the same location as the action described by the second verb:
(50) He started to prepare dinner in the kitchen. (implicative)
(51) He tried to drive the car in the paddock. (non-implicative)

However, in the case of Modal verbs lower on the binding scale, it is not neces-
sarily the case that they are true at the same location as their complements:
(52) The children wanted to play in the garden.
(53) I like reading in bed.
The fact that some Modal verb sentences, namely those high on the binding
scale, exhibit the property of a single set of peripheral arguments, can thus
be attributed to the phenomenon of binding. This explains the similarity to
serial verb constructions in White Hmong in this respect. However, it remains
true that the Modal verb has scope over the following verb and its core argu-
ments, a situation not compatible with serialization, in which neither verb has
scope over the other, but in which both verbs contribute jointly to the whole
proposition.
5.5 Purpose Constructions
Many types of clauses expressing purpose in White Hmong are introduced

by the complementizer kom ‘so that/in order to’ (see Jaisser (1984, pp. 52–58),
Jarkey (2006, pp. 124–127)). The subject of such a purpose clause may be the
same as, or different from, the subject of the main clause. However, some verbs
in White Hmong, including Deictic Motion (mus ‘go’, tuaj ‘come’, los ‘come/
return home’) and Change of Stance verbs (e.g. sawv ‘stand up’, zaum ‘sit down’,
txhos-caug ‘kneel down’) can introduce another junct expressing purpose,
without a complementizer (see Chapter 3, §3.1.2.2, Chapter 5, §5.5.1.1). In these
sentences the subjects of the verb expressing purpose and of the following
verb are always the same. For example:
(54) kuvS tuaj <kawm [lus Hmoob]O>PURP

1SG come study word Hmong
‘I’ve come to learn Hmong.’
(55) wbS txhos-caug <rub [cov nroj]O/S tawm>PURP

1DU kneel.down-knee pull CLF.PL weed leave
‘We two knelt down to pull out the weeds.’
It is not uncommon to have more than two verbs in a purpose construction.

In (56) below, both the Deictic Motion verb mus ‘go’ introduces the Change of
254 Chapter 5
Stance verb txhos-caug ‘kneel down’, which in turn introduces the action pe
‘bow in worship’:
(56) sawvdawsS tseem yuav mus <txhos-caug <pe . . .>PURP>PURP

everyone still IRR go kneel.down-knee bow.in.worship
‘Everyone will still go to kneel down to worship (him) . . .’ (SLY 177)
These Purpose Constructions have some important syntactic features in com-

mon with serial verb constructions; for example, the verbs are simply jux-
taposed and they have a common subject. It is thus not at all surprising to
find similar constructions included in the class of serial verb constructions in
descriptions of some languages (e.g. Thepkanjana (1986)). However, in the case
of White Hmong at least, these constructions differ from serial verb construc-
tions both syntactically and semantically, and with regard to the arrangement
and type of lexical items involved. These differences are discussed briefly below.
5.5.1 Semantic and Lexical Differences Between SVCs and Purpose

Constructions
5.5.1.1 Comparison with Cotemporal SVCs
As shown in the discussion of Cotemporal Motion SVCs (Chapter 3, §3.1.1), the
final verb of a serial verb construction of this type is commonly one of the
Deictic Motion verbs mus ‘go’, tuaj ‘come’, or los ‘come/return home’. Unlike
most other types of Motion verb in White Hmong these verbs are telic.2 They
can function in a Cotemporal Motion SVC not only to express deixis, but also
to introduce the Goal of the entire motion event. For example:
(57) nwsS/S/A khiav rov qab mus tsev

3SG flee return back go home
‘She fled back home.’
While these Deictic Motion verbs appear in the final position in a Cotemporal
SVC, they appear in the first position in a Purpose Construction. In this case
the next junct expresses an action or state that is understood to be the goal or
purpose of the deictic motion, as shown in the examples below:
(58) yog kuvS los <yuav NtsiasO>PURP os

COP 1SG come obtain PN IP
‘. . . the fact is, I’ve come to marry Ntsia.’ (DNH 451:13)
2 Source Verbs, such as tawm ‘leave’ and dim ‘escape’ are the only other kind of telic motion
verb in White Hmong.
(59) txij puag thaum kuvS tuaj <nyob [lub teb.chaws no]>PURP
since INTNS time 1SG come dwell CLF country this
‘Since I came to live in this country . . .’ (WHD p. 20)
The order of the verbs is crucial. This is shown clearly in the two examples
below, in which the same two verbs appear, but in the opposite order:
(60) neegA txawm kom <devS mus <nqa>PURP>COMP

human then cause dog go carry
‘Then (the) people got (the) dog to go to get (some seeds).’ (DNH 119:20)
(61) nwsA/S nqa [nws rab hneev]O mus

3SG carry 3SG CLF crossbow go
‘He carried his crossbow away.’ (KNH p. 4)
In the first example above, the first verb is the telic Motion verb, mus ‘go’,
and the second an atelic one, nqa ‘carry’; the interpretation is Purposive: ‘go
to get’. In the second example, the first verb is the atelic nqa ‘carry’, and the
second is the telic mus ‘go’; the interpretation in this case is Cotemporal:
‘carry away’.
Although the Deictic Motion verbs are by far the most common kind of
Motion verb used to introduce a Purpose Construction, there are other telic
Motion verbs that can also occur in this position. Verbs indicating change of
stance are examples of such verbs:
(62) lawvS zaum <tham>PURP

3PL sit.down chat
‘They sat down to chat.’
(63) tag.kis sawv tsees <noj mov>PURP tas . . .

next.day stand.up INTNS eat rice finish
‘The next day, after (he) had got up to eat . . .’ (TNN 1)
As noted previously (Chapter 3, §3.1.2.2), these Purpose Constructions can be

compared to another type of Cotemporal Serial Construction—Stance and
Action—in which the first junct refers to a stable stance (atelic) (rather than a
change of stance (telic)), and the second junct refers to an action. In (64) and
(65) below, an Attainment SVC is used to express a stable stance, and forms the
first junct of a Cotemporal Stance and Action SVC:
256 Chapter 5
(64) lawvS/S zaum-tsaws tham [ib hmo nkaus]

(65) nwsS/A sawv-ntsug noj mov

3SG stand eat rice
‘He ate standing up.’
In this case the basic order of the Change of Stance verb and the action verb
is the same in the two constructions, the former before the latter. The crucial
factor in determining the interpretation is the telicity of the first junct: the
examples above of the Cotemporal Serial Construction have an atelic stance
verb as V1 the examples of the Purpose Construction, (62) and (63) have a telic
Change of Stance verb as V1.
These examples of the Purpose Construction seem to be semantically very
close to the corresponding examples of the Cotemporal Serial Construction.
The Purpose sentences with Change of Stance verbs above could perfectly well
be used in place of their Cotemporal SVC counterparts, to describe identical
situations: lawv zaum tham ‘They sat down to chat’—‘They sat chatting’. The
reason for this is that the Purpose construction has what is here termed an
‘implication of result’. If one sits down to do something, then it is assumed that
one will, indeed, sit doing it as a result. Although these sentences do literally
express purpose, due to the telicity of V1, the interpretation would generally be
that the purpose was fulfilled, and that the action and the posture (the resul-
tant stable state) occurred cotemporally.
However, this interpretation is the result of inference only. The Purpose
Construction is non-implicative, and so the truth-value of the second junct is
independent of that of the first. This junct is defeasible: it can be felicitously
denied without affecting the truth of the entire proposition:
(66) a. lawvS zaum <tham>PURP

3PL sit chat
‘They sat down to chat . . .’
b. . . . tab.sis lawvS ho tsis tham

but 3PL actually NEG chat
‘. . . but actually they didn’t chat.’
Compare this with the Cotemporal Serial Construction, in which the truth-
values of the two juncts are inextricably linked:
(67) a. lawvS/S <zaum-tsaws tham [ib hmo nkaus] . . .

‘They sat chatting all night long . . .’
b. * . . . tab.sis lawvS ho tsis tham

but 3PL actually NEG chat
‘. . . but actually they didn’t chat.’
In the normal course of events, of course, the action described by the second junct
in a Purpose Construction is assumed to occur. As a result, some examples of the
Purpose Construction may appear to represent a simple sequence of events. The
most natural English translation may actually be ‘V1 and V2’ rather than ‘V1 to V2’,
especially when the action described by V1 is unrealised. For example:
(68) kojS mus <hu [cov tub.mab.tub.qhe]O los <noj mov>PURP>PURP

2SG go call CLF.PL servants come eat rice
‘Go and call the servants to come and eat.’
Lit.: ‘Go (to) call the servants (to) come (to) eat.’ (TNN 4)
The meaning of the construction is, however, definitely purpose rather than
simple sequence. The first junct is understood to constitute a necessary, but
not sufficient condition for the realisation of the second. Again, the truth-value
of the second junct is quite independent of that of the first:
(69) nwsS mus <hu lawv los>PURP , tab.sis ho tsis hu

3SG go call 3PL come but actually NEG call
‘She went (to) call them, but actually didn’t call.
(70) lawvS los <noj mov>PURP , tab.sis ho tsis noj

3SG come eat rice but actually NEG eat
‘They came (to) eat, but actually didn’t eat.’
The fact that the truth-values of the two juncts in a Purpose Construction are
independent underlines the fact that these sentences involve one proposi-
tional junct and one that is sub-propositional. The truth-values of the juncts
in a serial construction, on the other hand, could not possibly be independent,
given the fact that the verbs work together to express a proposition.
5.5.1.2 Comparison with Attainment SVCs

It has been shown above that a Cotemporal Serial Construction involving
one of the Deictic Motion verbs—mus ‘go’, tuaj ‘come’, or los ‘come/return
258 Chapter 5
home’—differs from a Purpose Construction involving one of these verbs

with regard to the order of the verbs. A sequence such as nqa mus ‘carry
go’, in which the Deictic Motion verb appears second, is interpreted as a
Cotemporal SVC; a sequence such as mus nqa ‘go carry’, in which the telic
verb appears first, is interpreted as a Purpose Construction.
In an Attainment SVC involving a deictic motion verb, however, the deictic
motion verb appears in first position, just as it does in a Purpose Construction:
(71) [tus tub ]A/A mus txog [nram hav-dej]O/O

‘The boy arrived down at the river valley.’ (Attainment)
(72) [tus tub]S mus <da.dej [nram hav-dej]>PURP

CLF boy go bathe down valley-water
‘The boy went to bathe down in the river valley.’ (Purpose)
As previously shown, when the second verb describes the purpose of the first,
as in example (72) above, the occurrence of the action described by this verb
can felicitously be denied, without affecting the truth of the entire proposition:
(73) . . . tab.sis nwsS ho tsis da-dej vim-tias dejS

. . . but 3SG actually NEG immerse-water because water
no~no kiag
RDUP~be(c0me).cold completely
‘. . . but actually he didn’t bathe because the water was so cold.’
If the second verb describes the attainment of the goal of the first, on the other
hand, as in the Attainment SVC in example (71) above, the action described
would be understood to have actually taken place. It would be incongruous to
first assert it and to then deny it:
(74) * . . . tab.sis nws ho tsis txog nram hav-dej

but 3SG actually NEG arrive down valley-water
‘. . . but actually he didn’t arrive down at the river valley.’
The kind of verb appearing in the second junct is clearly the crucial factor that
allows a sequence such as those above to be correctly interpreted as either
an Attainment SVC or a Purpose Construction. Following a Deictic Motion
verb, the only possible interpretation of an achievement predicate, such as
txog ‘arrive’ in the example above, is as the second verb in an Attainment SVC
interpretation: the achievement verb expresses the successful attainment of

the goal of the Motion verb which precedes it. It cannot be understood to
express the purpose of that motion. A verb which does not express achieve-
ment, (e.g. an activity predicate like da-dej ‘bathe’) following a Deictic Motion
verb, is understood to itself convey the goal, that is the Purpose, of the motion.
5.5.1.3 Purpose Constructions with muab ‘take’

It is very common in White Hmong, as in many other serializing languages
(Bamgboṣe, 1974, p. 46; Li & Thompson, 1974b, pp. 266–267; Foley & Olson, 1985,
passim; Sebba, 1987, p. 132), for a verb that means ‘take’, ‘hold’, ‘use’ or the like
to be used to introduce an argument expressing instrument into the clause.
Probably the verb used most frequently for this purpose in White Hmong is
muab ‘take in the hand/take hold of’, as in the following example:
(75) nwsA muab [rab riam]O <hlais nqaijO>PURP

3SG take CLF knife slice meat
‘She took the knife (to) slice (some) meat.’ (i.e. ‘She sliced meat with a
knife.’)
Other verbs which can also be used in this way are siv ‘use’ and xuas ‘pick up/
take hold of’:
(76) nyob [teb.chaws Nplog], pebA siv tsuO <cub mov>PURP

be.located country Laos 1PL use steamer steam rice
‘In Laos, we use a steamer (to) steam rice.’
(77) yog tsis muaj tsu mas pebA xuas laujkaubO <ncu>PURP
COP NEG have steamer TOP 1PL take pot simmer
‘If there isn’t a steamer, we take a pot (and) simmer (it).’
The semantic differences between similar sentences with these three verbs are
discussed by Riddle (1989, pp. 6–9). Her discussion shows clearly that these
verbs do not just serve, like prepositions, to introduce instrumental arguments,
but that their own semantic content is reflected in the interpretation of exactly
how the instrument is manipulated or used.
This construction in White Hmong has, not surprisingly, been assumed to
involve Verb Serialization (e.g. Riddle (1989), Owensby (1986, pp. 239–240)).
However, it is instead considered here to be another example of the Purpose
Construction. In this case, V1 is not a telic Motion or Change of Stance verb,
but rather a telic Action verb. Just as in all other examples of the Purpose
260 Chapter 5
Construction, even though the complement is generally assumed to be realised

if the proposition as a whole is asserted, in fact the truth-values of the two
juncts are independent; the construction is non-implicative:
(78) nwsA muab [rab riam]O <hlais nqaisO>PURP tab.sis nwsA ho tsis
3SG take CLF knife slice meat but 3SG actually NEG
hlais vim.tias [rab riam]S npub heev
slice because CLF knife be(come).blunt very
‘She took the knife to slice (some) meat . . . but actually she didn’t slice
(any) because the knife was very blunt.’
(79) nwsA siv zogO <ntov ntooO>PURP tab.sis nwsA ntov ntooO tsis tau
3SG use strength chop tree but 3SG chop tree NEG can
‘He used (all) his strength to chop down the tree . . . but he couldn’t chop
it down.’
Riddle (1989) points out a number of other ways in which the nature of sen-
tences such as these is revealed. She observes (p. 7) that the question words in
the sentences below apply only to the first junct, and are not relevant to the
second:
(80) kojA tau xuas/muab dab.tsiO <tua [tus noog]O>PURP?

2SG ACHV grasp/take what kill CLF bird
‘What did you kill the bird with?’
(81) a. vim.li.cas nwsA thiaj li tau siv phomO <tua [tus noog]O>PURP?
why 3SG so as ACHV use gun kill CLF bird
‘Why did s/he use a gun to kill the bird?’
b. vim.tias nwsA tsis muaj hneevO

because 3SG NEG have crossbow
‘Because s/he doesn’t have a crossbow.’
Riddle notes that the second example above is not interpreted as mean-
ing ‘Why did s/he shoot the bird?’ This is shown by the appropriateness of a
response referring only to the instrument.
Riddle also demonstrates the nature of these sentences with evidence of the
scope of the negative, as in the first example below, and the scope of certain
adverbials, as in the second:
(82) nwsA tsis tau siv phomO <tua noogO>PURP . NwsA ruab.hluas xwb
3SG NEG ACHV use gun kill bird 3SG snare only
‘S/he didn’t use a gun to kill the bird. S/he just snared it.’ (Riddle, 1989,
p. 7)
(83) nwsA maj.mam muab phomO <tua noogO>PURP

3SG carefully take gun kill bird
‘S/he carefully took a gun to kill the bird.’ (Riddle, 1989, p. 8)
These sentences are regarded by Riddle as serial verb constructions, and so she
uses the evidence presented above to support her argument that serial verb
constructions in White Hmong can be multi-propositional. The very fact that
these sentences involve one verb that is responsible for expressing the proposi-
tion and one that is sub-propositional is one reason why they are not regarded
as serial verb constructions in this analysis.
5.5.2 Syntactic Differences Between SVCs and Purpose Constructions

Li and Thompson (1973) regard examples of the Purpose Construction in
Mandarin as serial verb constructions, because they conform to the ‘surface’
structure NP V (NP) V (NP), and because these authors define serial verb con-
structions as involving two separate actions (see Chapter 2, §2.1.1). According
to Li and Thompson, while all other SVC types involve the coordination of two
underlying sentences, serial verb constructions with a purpose reading differ
syntactically in a number of ways, all of which point to the first verb being the
main verb, and the second subordinate to it.
The Purpose Construction in White Hmong exhibits some similar syntactic
features to its counterpart in Mandarin, suggesting that it also involves sub-
ordination of the second junct to the first. However, while Li and Thompson
(1973) treat the Purpose Construction in Mandarin as a serial verb construc-
tion, it is precisely because it involves a subordinate relationship between the
two juncts that the Purpose Construction is not regarded as a serial verb con-
struction in this analysis of White Hmong.
a The ‘Purpose’ Junct Can Appear as a Topic

Recall the fact that subordinate elements, such as NP arguments and subor-
dinate location and time phrases, can appear sentence initially as topics in
White Hmong. A junct expressing purpose can also appear sentence initially
as topic. Observe the alternatives in examples (84) below and (85) below, and
the authentic example in (86) below:
262 Chapter 5
(84) a. kuvS tuaj <kawm [lus Hmoob]O>PURP

1SG come study word Hmong
‘I’ve come to learn Hmong.’
b. <kawm [lus Hmoob]O>PURP mas, kuvS tuaj

study word Hmong TOP 1SG come
‘(It is) to learn Hmong (that) I’ve come.’
(85) a. kojS yuav-tsum zaum <noj mov>PURP os

2SG should sit.down eat rice IP
‘You should sit down to eat.’
b. <noj mov>PURP mas, kojS yuav-tsum zaum os

eat rice TOP 2SG should sit.down IP
‘To eat, you must sit down.’
(86) Pub.ThawjA <ua rog>PURP mas muab riam.ntxubO xwb

PN make war TOP take knife only
‘The Pu Taw, to make war, used only the knife.’ (Fuller, 1985, p. 122)
The second junct of a serial construction, on the other hand, cannot appear
sentence initially as topic with the same meaning as in the SVC:
(87) a. nwsA/A nqa [nws rab hneev]O los tsev

3SG carry 3SG CLF crossbow come home
‘He carried his crossbow home.’
b. * los tsev mas, nwsA nqa [nws rab hneev]O

come home TOP 3SG carry 3SG CLF crossbow
(88) a. lawvS/A yuav-tsum khiav mus Thai.teb

3PL should flee go Thailand
‘They had to flee to Thailand.’
b. * mus Thai.teb mas, lawvS yuav-tsum khiav

go Thailand TOP 3PL should flee
The fact that the second junct in a Purpose Construction can appear sentence
initially as a topic suggests that it must be a subordinate junct: a complement
of the other junct in the construction. The fact that the second junct in a serial
construction cannot appear initially as a topic indicates that serial verb con-
structions do not involve subordination.
b The ‘Purpose’ Junct Behaves Like an Argument of V1 with Regard

to the Placement of Adverbials
As noted previously (Chapter 3, §3.1.3.2), a verb used adverbially, like ceev
‘be(come) fast’, is able to appear, among other places, directly after a verb
and its object, as in the first example below, but not between the verb and its
object, as in the second:
(89) a. fawb nrojO ceev~ceev mus

search.around vegetation RDUP~be.fast go
‘go searching quickly around the grass’
b. * fawb ceev~ceev nrojO mus

search.around RDUP~be.fast vegetation go
Such adverbial expressions can readily appear between the verbs of a

Cotemporal SVC:
(90) nce ceev~ceev mus <xyuas [saum nthab]O>PURP . . .

ascend RDUP~be.fast go check top drying.platform
‘climb quickly up to check the platform’
In a Purpose Construction, however, such adverbial expressions cannot appear

between the verbs, but must appear after the second junct, indicating that this
second junct behaves like a Noun Phrase argument of the first, in this respect:
(91) a. * nce mus ceev~ceev <xyuas [saum nthab]>PURP . . .

climb go RDUP~be.fast check top drying.platform
b. nce mus <xyuas [saum nthab]>PURP ceev~ceev

climb go check top drying.platform RDUP~be.fast
‘climb up to check the platform quickly . . .’
In addition to the fact that Purpose Constructions differ syntactically from

serial verb constructions in that they involve subordination, they also differ in
that, to a limited extent at least, the purpose junct can have peripheral argu-
ments independent of the initial junct. This fact is in keeping with the status
of this junct as a sub-propositional clause. Observe the following examples:
264 Chapter 5
(92) nwsS mus <yuav zaubO [tom khw]>PURP lawm

3SG go obtain vegetable over.there market PRF
‘She has gone to buy vegetables at (the) market.’
The locative phrase tom khw ‘(the) market over there’ expresses the location
at which the buying takes place. The verb yuav ‘obtain, buy’ is not a verb that
takes an inner locative argument; tom khw ‘(the) market over there’ is a periph-
eral, outer locative, argument in this context. Furthermore, it clearly applies
only to the second junct in the construction; this peripheral locative argument
is not related in any way to the first verb mus ‘go’.3
5.5.3 Syntactic Similarities Between SVCs and Purpose Constructions

In spite of the fact that the Purpose Construction can be shown to differ syn-
tactically from the serial construction, it remains true that the two construc-
tions also exhibit some important syntactic similarities—similarities that
should not be ignored. It is not only the fact that, in both types of sentence,
two verbs with the same subject are simply juxtaposed. It is also that, although
the second junct in a Purpose Construction can take independent peripheral
arguments (as shown above), this ability is somewhat limited. In this respect,
this second junct is, in fact, less like an independent clause, and more syntacti-
cally integrated into the main clause.
3 A Locative Phrase appearing after the second junct in a Purpose Construction cannot be
interpreted as a Goal:
(i) nwsS mus <yuav zaubO [tom khw]>PURP lawm
3SG go obtain vegetable over.there market PRF
‘She has gone to buy vegetables at (the) market.’
not: ‘She has gone to market to buy vegetables.’
Nor can a Locative Phrase expressing Goal appear between the two juncts of a Purpose
Construction. That is, when V1 is the type of predicate that can take a locative Goal argument,
it seems that the speaker can choose to express either a locative-type Goal or a purpose-type
Goal, but not both.
(ii)* nwsA mus [tom khw] <yuav zaubO>PURP
3SG go over.there market buy vegetable
for: ‘She went to market to buy vegetables.’
A Locative Phrase can only appear between the juncts in a Purposive SVC if it is an inner
locative expressing Location:
(iii) nwsA sawv [ntawm qhov-cub] <noj movO>PURP
3SG stand.up nearby kitchen eat rice
‘S/he stood up in the kitchen to eat.’
The limitation referred to is with regard to peripheral arguments expressing

time. Although it would be logically possible in some situations for the event
expressed by the main clause to occur at a different time to that of the purpose
clause, such a scenario cannot be grammatically expressed by the Purpose
Construction in White Hmong; the purpose clause would have to be introduced
by a complementizer:
(93) a. * kuvS yuav-tsum mus tam.sis.no <ntsib nwsO thaum

1SG should go now meet 3SG time
ob teev>PURP
two o’clock
‘I should leave now (to) meet him at two o’clock.’
b. kuvS yuav-tsum mus tam.sis.no kom.xwv <kuvA thiaj ntsib

1SG should go now so.that 1SG so meet
nwsO thaum ob teev>COMP
3SG time two hour
‘I should go now so that I will meet him at two o’clock.’
The reason for such restrictions can again be explained by the notion of ‘bind-
ing’ (Givón, 1980, 1985, 2001b). Although the Purpose construction is quite
clearly non-implicative, as shown above, it is, nevertheless, quite high on the
binding scale. Like some modal verbs, this construction expresses intention
on the part of the subject. However, unlike sentences with modal verbs such
as xav ‘want’ (see §5.4 above), the Purpose Construction does not indicate any
‘emotional attachment’ on the part of the subject (precisely the feature that
results in sentences with the modal ‘want’ being so low on the binding hier-
archy). In this respect the Purpose Construction is more like the modal use of
yuav ‘intend’. However, the subject in a Purpose Construction not only intends
to perform the action described by V2, but, just as in the case of the modal verb
sim ‘try’ (see §5.4 above), s/he actually does something in order to realise it: the
subject ‘changes location to do it’ (e.g. mus V ‘go to V’), ‘changes stance to do it’
(e.g. sawv V ‘stand up to V’), or ‘takes something to do it (with) (e.g. muab NP V
NP ‘take NP V NP’).
Thus, although the Purpose Construction is not implicative, nevertheless,
like sentences with the modal verb sim ‘try’, it does involve a considerable
degree of binding between the two verbs. This fact explains why the second
junct is syntactically integrated into the main clause to the extent that it can-
not exhibit independent peripheral arguments expressing time. Nevertheless,
the binding is loose enough to allow independent expression of peripheral
266 Chapter 5
arguments expressing location, and this is consistent with the status of the
second junct as a complement clause.
5.6 State Verbs Used Attributively and Adverbially
As shown above (§5.3), there are a considerable number of sentences involv-

ing complement clauses that superficially resemble the Cause-Effect Serial Verb
Construction. Other kinds of sentences that resemble some examples of the
Cause-Effect SVC are those in which the final verb is a State verb; the State verbs
in these examples are actually functioning either attributively or adverbially.
5.6.1 State Verbs Used Attributively

When V2 is a State verb, the Cause-Effect Serial Construction in White Hmong
is often fairly accurately translated by a factitive construction in English:
(94) lawvA ncaws [lub qhov-rooj]O/S qhib

3PL kick CLF door be(come).open
‘They kicked the door open.’
(95) nwsA khawb [lub qhov]O/S tob~tob

3SG dig CLF hole RDUP~be(c0me).deep
‘She dug the hole deep.’
However, these White Hmong sentences have an alternative interpretation,

which is not available to their English counterparts. The sentence in example
(94) above could also be understood to mean ‘They kicked the open door’, and
(95), ‘She dug the deep hole’. That is, because of the head-modifier order of
White Hmong NPs, and because the State verbs in these examples can be inter-
preted as modifying the nouns that precede them, these sentences can have an
interpretation in which V2 functions like an attributive adjective. This alterna-
tive reading of these sentences can be represented as follows:
(96) lawvA ncaws [lub qhov-rooj qhib]O

3PL kick CLF door be(come).open
‘They kicked the open door.’
(97) nwsA khawb [lub qhov tob~tob]O

3SG dig CLF hole RDUP~be(c0me).deep
‘She dug the deep hole.’
One factor that will determine the accessibility of this ‘attributive’ interpreta-
tion of these sentences is the likelihood of a particular State verb acting as a
modifier to the preceding noun in the relevant context. For example, in the
first sentence above, this interpretation is relatively less accessible, because of
the obvious fact that there is more point in ‘kicking a door open’ than in ‘kick-
ing an open door’.
This factor, along with the potential ambiguity of such a sequence, is
exploited in the following White Hmong riddle:
(98) a. Q: koj ncaws ntxhw tuag, puas tau?

2SG kick elephant dead Q can
‘Can you kick an elephant dead?’
b. A: tau xwb! KojA ncaws tau [ntxhw uas tuag]O

can only 2SG kick can elephant REL be(come).dead
lawm!
PRF
‘Sure can! Of course you can kick an elephant that is dead.’
The question in this riddle is framed as a kind of challenge: ‘Can you achieve . . .?’
As a result, the causative interpretation, which obviously presents a consider-
able challenge, leaps to mind: ‘Can you kick an elephant (so that it is/becomes)
dead?’ The ‘attributive’ interpretation on the other hand—‘Can you kick a dead
elephant?’ would present no challenge whatsoever, rendering it relatively less
accessible. Nevertheless, this sentence, like many other examples of the Cause-
Effect Serial Construction in which V2 is State verb, is genuinely ambiguous.
5.6.2 State Verbs Used Adverbially/Depictively

Another function for State verbs in White Hmong, in addition to those dis-
cussed above, is an adverbial function:
(99) [koj tus me-nyuam]S hlob zoo~zoo

2SG CLF child grow RDUP~be.good
‘Your child is growing well.’ (Heimbach, 1979, p. 64)
In example (99) the final intransitive State verb zoo ‘be good’ is used as a man-
ner adverb, describing the way in which the child is growing: hlob zoo ‘grow well’.
If the first verb were transitive rather than intransitive, the sentence would
have the same basic pattern as the potentially ambiguous type discussed in
§5.6.1 above: NP VTR NP VSTATE. In this case, it is possible for a three-way
268 Chapter 5
ambiguity to arise between a Cause-Effect SVC reading, an Attributive reading,

and an Adverbial reading. This is illustrated in example (100) below:
(100) lawv ua lub tsheb ceev~ceev

3PL make CLF car RDUP~be.fast
a. lawvA ua [lub tsheb]O/S ceev~ceev

‘They built the car (to be) fast.’ (Cause-Effect SVC)
b. lawvA ua [lub tsheb]O ceev~ceev

‘They built the car quickly.’ (Adverbial)
c. lawvA ua [lub tsheb ceev~ceev]O

‘They built the fast car.’ (Attributive)
It is, of course, not very common for the combination of elements in a sen-
tence with this basic pattern to allow this kind of multiple ambiguity. In exam-
ple (101) below the pattern occurs, but the state expressed by the final verb kub
‘be hot’ cannot be understood to result from the action described by the first
verb haus ‘drink’ (one cannot ‘drink milk (so that it gets) hot’), ruling out the
Cause-Effect SVC interpretation. Only the other two interpretations are avail-
able in this case:
(101) haus khob mis kub

drink cup milk be(come).hot
a. ‘Drink (a) hot cup (of) milk.’ (Attributive interpretation of V2)
b. ‘Drink (a) cup (of) milk hot.’ (‘Depictive’ interpretation of V2)
The interpretation given as (101)(b) above is termed here ‘depictive’, following

Enfield (2007, pp. 397–410) who uses this term for similar subtype of adverbial
construction in Lao. Here the final State verb does not refer to the manner in
which the subject performs the action (as does ceev ‘be fast, quickly’ in (100)
above), but focuses instead on the object, describing the condition in which
the milk is drunk.
Not only semantic factors, but also syntactic ones, can influence which inter-
pretations are possible. In example (102) below, in which the object is fronted
as the object of muab ‘take’ in an initial Disposal SVC (Chapter 3, §3.4.6),
either the cause-effect nor the attributive interpretation is available, and only
n
the depictive interpretation is possible:
(102) ces txawm yuav muab [niag cuam]O noj nyoos ntag no
CONJ then IRR take great gibbon eat be.raw IP IP
‘. . . and then (he) was going to take that big ol’ gibbon (and) eat (it)
raw.’ (DNH 398:7)
Apart from the obvious semantic difference, sentences with the adverbial/
depictive interpretation also differ syntactically from the Cause-Effect Serial
Construction in that the two verbs do not share an argument. With regard to
the situation described by example (102) above, while it is true to say both that
‘he was going to eat the gibbon’ and that ‘the gibbon would be raw’, only the
first of these propositions is actually expressed explicitly in the sentence; niag
cuam ‘that big ol’ gibbon’ is not a shared argument. The force of the State verb,
nyoos ‘be raw’, is not to describe the state of the gibbon, but rather to describe
something about the condition in which it is to be consumed. Likewise, in the
case of example (101) above, when the final verb kub ‘be(come) hot’ is inter-
preted depictively (interpretation (b)), it has scope over the whole of the
preceding verb phrase, haus khob mis ‘drink (the) cup (of) milk’; it does not
relate simply to khob mis ‘cup (of) milk’ by itself, as the attributive reading does
(interpretation (a)).
We have already noted (§5.6.1 above) that pragmatic factors sometimes
make an ‘attributive’ interpretation more or less difficult to access in a sequence
NP VTR NP VSTATE. What factors are involved in interpreting such a sequence
as either causative or adverbial/depictive? In both of the examples below, the
final verb is the State verb siav ‘be(come) cooked’. Example (103) has a cause-
effect interpretation, and example (104), a depictive (or attributive) one. For
(104) the labelling reflects the depictive interpretation.
(103) yusA cub movO/S siav

‘One steams the rice (so that it is) cooked.’ (Cause-Effect SVC)
(104) lawvA noj nqaijO nyoos, tab.sis pebA noj nqaijO

3PL eat meat raw but 1PL eat meat
siav
be(come).cooked
‘They eat meat raw but we eat meat cooked.’ (Depictive)
[Also: ‘They eat raw meat but we eat cooked meat.’ (Attributive)]
270 Chapter 5
The different interpretations—cause-effect or depictive—are clearly related

to a number of factors; the meanings of both the first and second verbs in these
examples play a part. Firstly, the verbs must be semantically compatible with
one or the other interpretation: for a cause-effect interpretation to be possible,
V1 must express the kind of action that can cause the state/change-of-state
described by V2. In other words, you can make rice ‘cooked’ (siav) by ‘steaming
it’, but you can’t make meat ‘cooked’ (siav) by ‘eating it’. Conversely, ‘cooked’
(siav) does describe a way you can ‘eat meat’, but it does not describe a way you
can ‘steam rice’.
Secondly, as noted above, only certain types of transitive verbs can occur at
all as V1 in a Cause-Effect SVC, namely those that express an action in which
the subject, an Agent, causes a change in the object, a Patient or Theme (i.e.
Affective, Effective, and Transfer verbs; see §5.2.1.2 above). Other transitive
verbs that express actions which either do not result in a change in the object,
or else in which there is less interest in any such change, would not be inter-
preted as ‘causing’ verbs. Verbs such as hais ‘speak’ (as in hais lus Hmoob ‘speak
Hmong’) do not cause any change in the object. Verbs such as noj ‘eat’ and haus
‘drink’ do cause some change in the object, but it is not one that attracts much
attention. The real focus of interest is on the effect of these actions on the
subject. A cause-effect interpretation would thus not be available for transitive
verbs involving little impact on the object, such as these.
5.7 Transitive Verbs Used Descriptively and Adverbially
5.7.1 Transitive Verbs Used Descriptively

As discussed in Chapter 3, §3.4, examples of the Disposal SVC consist of transi-
tive verbs that share both their A and O arguments. The O appears between the
(first) two verbs, as in the following examples:

3SG throw 3SG CLF crossbow leave
‘He threw his crossbow away.’
(106) [tus tsov]A/A tom [tus noog]O/O noj

In this type of SVC, both verbs generally express some form of impingement on
the object. The action described by the first verb leads up to, and culminates in,
the action described by the second.
There is a very common sentence type that appears structurally identical

to the Disposal SVC, with the basic pattern NP A/A VTR NP O/O VTR, but that dif-
fers in a number of important ways. Firstly, V1 in these sentences is not an
Impingement verb. Secondly, the action described by the first verb is not
thought of as culminating in that described by the second. Instead, in sen-
tences of this type, the second verb describes an action which is unrealised at
the time referred to. Here are some examples:
(107) koj puas muaj paj.ntaub muag?

2SG Q have needlework sell
‘. . . do you have any needlework (to) sell?’ (WHD 18:16)
(108) [cov poj-niam] ua mov noj

CLF.PL woman make rice eat
‘The women make food (to) eat.’
(109) es koj mus nrhiav koj tus txiv yuav

CONJ 2SG go search.for 2SG CLF man obtain
‘Go (to) look for a man (to) marry.’ (DNH 450:8)
Although these sentences appear to represent some kind of serial construc-

tion involving a shared O argument, in fact the two verbs do not contribute
equally to the proposition. The primary propositional content of the sentence
as a whole is actually vested in the first verb, while the second verb serves only
as a modifier to the NP that precedes it.
The examples below show that the object NP, along with the modifying V2,
can appear sentence initially as a topic, confirming the status of the entire con-
stituent as an argument:
(110) [paj.ntaub muag] mas, lawv muaj

needlework sell TOP 3PL sell
‘As for needlework (to) sell, they have (some).’
(111) [mov noj] mas, [cov poj-niam] ua

rice eat TOP CLF.PL woman make
‘As for food to eat, the women make (it).’
This is not possible in the case of examples of the Disposal SVC:
(112) a. kuv tso nws tseg

272 Chapter 5
b. * nws tseg mas, kuv tso

3SG leave TOP 1SG release
(113) a. [tus tsov] tom [tus noog] noj

b. * [tus noog] noj mas, [tus tsov] tom

CLF bird eat TOP CLF tiger bite
5.7.2 Transitive Verbs Used Adverbially

The verb tas/tag ‘finish’ often appears as the second verb in a sequence that
seems to represent the Disposal Serial Construction:4
(114) tag.kid noj tshais tag ces Ntsias txawm

next.morning eat breakfast finish CONJ PN then
hais . . .
say
‘The next morning she finished eating breakfast, and then Njia said . . .’
(DNH 451:15)
In this example both verbs, noj ‘eat’ and tag ‘finish’, seem to govern the NP tshais
‘breakfast’ and the ‘eating’ can be thought of as culminating in the ‘finishing’.
However, there are numerous other examples of sequences with tas/tag ‘fin-
ish’, such as those illustrated below, for which such an analysis is not possible:
(115) lawv cog nplej tag lawm

3PL plant unhulled.rice finish PRF
‘They have finished planting the rice.’
(116) neeg tuag tas . . .

human die finish
‘All human beings died.’ (DNH 115:1)
These examples show that the morpheme tas/tag is not simply the second
verb in a Disposal SVC, even though, by virtue of its own meaning, its semantic
input is similar to V2 in an SVC of this type in some cases. In example (115)
4 Heimbach (1979:304) attributes the variation in tone in this morpheme to stress, tag being
the stressed form. Whatever its cause, this variation is not relevant to the present discussion.
above the NP nplej ‘unhulled rice’ is governed by the verb cog ‘plant’ but not by
tag ‘finish’; it is not that ‘they have finished the rice’, but that ‘they have finished
planting the rice’. In example (116) above, the verb preceding tas is intransi-
tive; there is simply no object to be shared. It is clear from these examples that
tas/tag is functioning adverbially, with scope over the entire predicate which
precedes it.
Another transitive verb that can function in a similar way is tso ‘release/
relinquish/send’. This verb can actually appear as a non-initial verb in a
Disposal SVC, as in the following example:
(117) wb muab txhoov ua [tej daim] tso . . .

1DU take cut make CLF.PL pieces relinquish
‘. . . we will cut it into pieces and throw (them) away.’ (DNH 117:14)
However, tso also appears very commonly in sentences in which it has an

entirely different function:
(118) tos ob peb hnub tso

wait two three day relinquish
‘. . . wait a few days and then . . .’ (Heimbach, 1979, p. 354)
(119) kuv yuav mus <ntxuav [kuv lub qhov-muag]>PURP tso

1SG IRR go wash 1SG CLF eye relinquish
kuv li los <saib>PURP
1SG as come look.at
‘First I’ll go (to) wash my eyes and then I’ll come (to) look at (you)
again.’ (DNH 401:32)
In these sentences tso ‘relinquish’, like tas/tag ‘finish’ in the sentences above,
has scope over the entire clause that precedes it. Its function is to indicate that
one action is ‘relinquished’ or completed, before the next is begun: ‘Get the
‘waiting a few days over with (and then) . . .’ / ‘I’ll finish with ‘going (to) wash my
eyes, (and then) come to see you again.’
5.8 Conclusion to Chapter Five
There are a considerable number of constructions in White Hmong that super-

ficially resemble serial verb constructions, in that their components are simply
juxtaposed, with no overt form of linkage. It is possible to isolate at least four
274 Chapter 5
distinct reasons why this is so. Firstly, White Hmong is a highly paratactic lan-
guage; the semantic relationship between propositions and events does not
always have to be expressed lexically. Secondly, White Hmong is a language
with no inflectional morphology. Even in the absence of lexical signals of, say,
subordination, a change in role from predicate to argument may be signalled
in some languages by gerundive or subjunctive morphology; in White Hmong,
no such signals are available. Thirdly, the phenomenon of semantic ‘binding’
between a main verb and its complement, which has significant syntactic con-
sequences regarding the status of the complement as a separate clause, results
in many cases of subordination having important features in common with
serialization. These features include the inability of the subordinate junct to
take independent peripheral arguments, and the lack of a complementizer,
resulting in the juxtaposition of the two juncts. Finally, White Hmong is a lan-
guage in which verbs have a great variety of functions. For example, they can
function attributively or adverbially within the clause. These particular func-
tions are not considered to involve serialization, because there is no sharing
of arguments.
Verbs in White Hmong may be juxtaposed within a single clause or over a
series of clauses. However, only those that share a common core argument, do
not involve complementation, and work together to express a single proposi-
tion, are regarded as serialized in this analysis.
CHAPTER 6
Conclusion
The term ‘serial verb construction’ has been used to refer to a diverse array
of constructions in a wide variety of different languages: from constructions
expressing a series of events to those expressing a single event; from construc-
tions involving syntactic coordination to those involving syntactic subordina-
tion (Chapter 2, §2.1). The term is used here to describe a series of two or more
verbs, none of which is syntactically embedded as an argument of another,
which work together to express a single event within a single proposition
(Chapter 2, §2.2 and §2.3).
The notion of two or more verbs working together in a particular language
to express a single event within a single proposition raises a number of impor-
tant questions. These questions relate to issues such as: an appropriate syn-
tactic analysis of the constructions concerned and the boundaries within
which they might differ syntactically (Chapter 2, §2.4); the common prop-
erties and functions of the SVC types identified in the language concerned
(Chapter 3, Introduction), and the properties and functions of each of these
types (Chapter 3, §3.1–3.4; the relationship between function and part-of-
speech classification (Chapter 4), and the similarities and differences between
the SVC types identified and other constructions to which they are superfi-
cially similar (Chapter 5). All these issues have been dealt with in depth in
various parts of this study, and the conclusions reached concerning them are
briefly summarised here.
6.1 The Notions of a Single Event and of Joint Predication
Section §2.2 of Chapter Two described research by Givón (1987, 1991) regarding
the pause probabilities within serial verb constructions. On the basis of that
evidence, Givón concludes that serial verb constructions simply represent a
different way of expressing propositions, not a totally different way of mentally
segmenting the real world, implying a totally different conceptualisation of the
notion of ‘event’. However, the factors that lead to two or more conceptually
separable events being thought of, and treated as, one event are the factors of
common and cultural association, allowing for some variation between speech
communities regarding the kinds of event types that can be serialized (§2.2.1).

276 Chapter 6
The fact that serial verb constructions consist of two or more verbs working
together to express a single event means that they can be viewed as constituting
a bridge between the syntax and the lexicon. Like lexical items, serial verb
constructions express commonly or culturally associated ideas within a single
unit. However, unlike lexical items, the component parts of serial verb con-
structions are transparent and must be processed individually, in order that
the construction as a whole be understood. In this way serial verb construc-
tions are clearly a syntactic phenomenon (§2.2.1).
The interpretation of serial verb constructions as expressing a single event
can be seen as iconically related to their syntactic structure. The two (or
more) verbs are united by the fact that they occur together in a single clause,
sharing all core and clause-layer operators, and having in common at least
one core argument and all peripheral arguments. Furthermore, the lack of
lexical linkage in the form of conjunctions or complementizers means that
the verbs appear physically close together, reinforcing their semantic one-
ness (§2.2.2).
Finally, in this analysis the criterion that the two verbs in a serial verb con-
struction work as joint predicators of a single proposition (§2.3) is taken to be
an important issue in delimiting SVCs in White Hmong. In cases in which the
second of the two verbs is part of a complement of the first, the two verbs are
not seen as acting as joint predicators. This criterion allows us to delimit a dis-
tinct group of constructions as serial verb constructions in this language, and
to thereby identify a distinct function that the strategy of serialization fulfils
(see §6.4 below).
6.2 The Syntactic Analysis of SVCs in White Hmong
The syntactic analysis of serial verb constructions in White Hmong proposed in

this study (Chapter 2, §2.4) involves the linkage of either clausal nuclei (verbs)
or cores (verbs plus their core arguments). When nuclei are linked, they share
all core arguments, and the verbs are in a cosubordinate relationship (depen-
dent with regard to nuclear layer operators). When cores are linked, the verbs
share at least one core argument, and are also in a cosubordinate relationship
(dependent with regard to core layer, modal operators). All peripheral argu-
ments and operators are shared by all juncts in all SVC types, consistent with
the fact that the juncts constitute a single clause. In this respect core layer
serial verb constructions are seen to differ from the notion of core juncture
proposed in the revised version of Role and Reference Grammar (RRG), which
portrays core juncture as linkage between separated clauses (core and periph-
Conclusion 277
ery) rather than between separate cores, provided there is some ‘intersection’
of core arguments. The RRG notion of cosubordination, however, has been
found to be particularly useful in characterising the relationship between the
predicates in most SVC types; while there is evidence to show that serial verb
constructions in White Hmong do not involve a coordinate relationship, nei-
ther do they exhibit the embedding associated with subordination.
6.3 The Common Properties of SVCs in White Hmong
One of the most basic problems in the study of serial verb constructions is
to determine which of the numerous sentence types containing concatenated
verbs might warrant the title ‘serial verb construction’ in a particular language.
Are there any particular syntactic and semantic properties shared by some of
these sentence types that set them clearly apart from others that have a similar
‘arrangement’ of components?
On the basis of the evidence from White Hmong, particularly that described
in Chapter 3, the following set of features is proposed in the Introduction to
that chapter, as the defining syntactic and semantic features of a serial verb
construction in this language:
i) A series of two or more concatenated verbs (with or without intervening

core arguments);
ii) Each verb can occur on it own as the only predicate in the clause;
iii) No overt form of linkage;
iv) No pause between the juncts;
v) At least one shared core argument;
vi) Shared core operators (i.e. deontic modality);
vii) Shared clausal operators (e.g. tense, epistemic modality, illocutionary
force);
viii) A single set of peripheral adjuncts; the verbs belong to a single clause;
ix) No junct is embedded as an argument of another;
x) The verbs are joint predicators within a single proposition, and express a
single event.
In addition, the available evidence seems to point to the conclusion that the
verbs in a serial verb construction in White Hmong share a single syntactic
subject.
This study has shown how this set of features serves to define a particular
group of concatenated verb sentences in White Hmong that can justifiably be
278 Chapter 6
set apart from other sentences involving concatenated verbs in this language.
That justification relates largely to their common function.
6.4 The Functions and Semantics of SVCs
As noted in the Introduction to Chapter 3, all four constructions identified

as major SVC types in this study of White Hmong do seem to share a com-
mon element in terms of their function. This common element is the type of
elaboration of detail that these constructions express within the bounds of a
single clause, to portray a single event. In the case of Cotemporal SVCs, this
elaboration of detail relates to action (including motion), whether transitive
or intransitive, in which the focus is entirely on the actor and the action. In
all other types—Cause-Effect SVCs, Attainment SVCs with transitive verbs,
and Disposal SVCs—the construction serves to provide detail concerning both
the action and its outcome. All of these three types facilitate the expression of
an effective, intentional, transitive action and its outcome in relation to the
O argument. The Attainment type can also occur with intransitive verbs, in
which case the elaboration relates to the intransitive process and its outcome
for the S argument. In all cases, the structure of multiple verbs within a single
clause provides precisely the kind of rich and intricate portrayal of events that
Westermann (1930, p. 126) described as occurring in the Ewe:
. . . the Ewe people describe every detail of action or happening from

beginning to end, and each detail has to be expressed by a special verb:
they dissect every happening and present it in its several parts, . . .
In addition to identifying this shared function of all of the main types of verb
serialization in White Hmong, Chapter 3 also reveals their key differences. In
Section §3.1, the Cotemporal SVC type is described. In this type, the actor—
the S/A argument—is always shared. In Cotemporal Motion SVCs (§3.1.1), the
verbs convey different facets of the motion event (e.g. Manner of Locomotion
or Transport, Path, Source, Goal and Deixis). Amongst Cotemporal SVCs that
include action verbs are those that describe Cotemporal Action and Motion
(§3.1.2.1), Cotemporal Action and Stance (§3.1.2.2), and Cotemporal Action
(§3.1.2.3). This final type is the least common, and the least productive; it
occurs only when there is clearly a common or cultural association that allows
the two actions to be construed as contributing to a single event. Action verbs
appearing in a Cotemporal SVC are always atelic, as are the majority of Motion
Verbs; notable exceptions are the telic Source and Goal verbs, which always
Conclusion 279
appear clause finally in these constructions. While Cotemporal SVCs can con-
tain both transitive and intransitive verbs, the kind of transitive verbs that
occur are those that are low in effectiveness in relation to the O argument, or at
least give little attention to any outcome for the O argument. In a Cotemporal
SVC, the elaboration of detail thus relates entirely to what the actor (the S/A
argument) does and how s/he does it.
In all of the other three types of serialization identified in Chapter 3, the
pragmatic function of elaboration of detail is associated with focus on both
an action or process and its outcome. In most cases, this involves the effective,
intentional action of an A argument and the outcome of that action in relation
to the O argument.
The first of these types discussed is the Cause-Effect SVC type (§3.2). The
shared argument is in O function in relation to the first verb, and (nearly
always) in S function in relation to the second. In these SVCs, the outcome of
the action described by the first verb is a change in the O argument, a change
that is described by the second verb. V1 is always an archetypal transitive
verb—an Affective, Effective, or Transfer verb—the A argument of which is an
Agent whose action directly impinges on the O argument, which is a Patient
or a Theme. V2, on the other hand, is almost always intransitive—a (change
0f) state or activity verb—the single argument of which is either a Patient or
a Theme. In the few examples found of a structurally transitive verb occurring
as V2, the A argument of this verb is still a Theme. It is thus not hard to see
why it is the O argument of V1—likewise a Theme—rather than the A argu-
ment—an Agent—that is interpreted as the shared argument in this construc-
tion, even though evidence points to the A argument being the subject of the
entire clause (§3.2.4). The Cause-Effect SVC in White Hmong involves a transi-
tive action and its outcome (§3.2.2) and thus portrays an extremely direct form
of causation in comparison with other constructions expressing causation in
this language. This is related to the syntax of the construction, which conveys
both cause and effect within a single clause, rather than separating these ele-
ments with a complementizer and/or a clause boundary.
In the third type of SVC discussed in Chapter 3, the Attainment SVC (§3.3),
the second verb may be understood to express the attainment of a goal, which
is described by the first. When the event involves an intransitive process that
has an intrinsic goal, the elaboration of detail allows attention to both this telic
process and its outcome for the S argument. When the event involves a transi-
tive action, it facilitates focus on the action of the A argument and its outcome
for the O argument, both of which are shared. In this case, the construction
expresses effective, intentional action that results in the A argument attaining
some kind of goal (either intrinsic or extrinsic) in relation to the O argument.
280 Chapter 6
The second verb in this kind of SVC is always an achievement verb. The first
verb may be an activity verb that has an extrinsic goal, or an accomplishment
verb, which has an intrinsic goal. Finally, it may be an activity verb without any
kind of goal, in which case the Attainment SVC is interpreted as indicating that
the performance of the action itself is the goal attained, and the implication
is that the subject has the ability to perform the action described by V1. The
fact that the Attainment SVC in White Hmong involves juncture at the nuclear
level is obviously related to the fact that it performs a fundamentally aspectual
function: the expression of the attainment of a goal. However, this merges into
a modal function when no particular goal is involved. This may have been the
pathway to the grammaticalization of the morpheme tau in its common post-
verbal role expressing the deontic modality meaning ‘can’.
Disposal SVCs are the fourth and final type of SVC described in Chapter 3
(§3.4). This type always involves two or more transitive verbs and the sharing
of two core arguments: A=A and O=O. Here the elaboration of detail facilitates
focus once again on both the effective, intentional action of the A argument
and the outcome of that action in relation to the O argument: in this case the
literal or metaphorical ‘disposal’ of the O argument. It is generally the case
that all verbs in these sentences express a high degree of impingement on the
object, with the second and subsequent verbs expressing the way in which the
object is destroyed, consumed, or relinquished. However, in Disposal SVCs in
which V1 is the verb muab ‘take’ (§3.4.6), it is not always the case that the object
is actually highly affected. These sentences seem to have a distinct discourse
function; they bring the object into a more topical position in the clause, and
simply express something about what happens to it, or how it is dealt with.
It must be noted that the elaboration of detail facilitated by these four major
types of serial verb constructions in White Hmong is often supported and
enhanced by complementary devices such as reduplication and repetition,
and by adverbial and elaborate expressions. The vivid portrayal of an event
can also be made even more intricate by the process of linking or interweaving
serial verb constructions together. These devices allow the speaker to create an
intricate verbal panorama, as the scene shifts through the various components
of a process or action, or from a process or action to its outcome.
Several of the ways in which verb serialization in White Hmong provides
elaboration to the clause can be seen to be related to ways in which verbs in
this language lack lexical complexity in certain domains. For example, verbs
expressing locomotion and transport never incorporate notions of deixis,
impingement verbs do not necessarily entail an effect on the O argument,
accomplishment verbs do not in themselves incorporate the notion of attain-
ment of goal, and Transfer verbs are structurally monotransitive rather than
Conclusion 281
ditransitive. The strategy of verb serialization does not raise the level of lexical
complexity, but it does serve to increase the semantic complexity of the clause
as whole.
One additional way in which serial verb constructions increase the com-
plexity of the clause as a whole is the valency increasing function (Chapter 4).
In this case, one verb serves to introduce an unshared argument into the clause
that is not a part of the argument structure of the other verb, thereby increas-
ing the clausal valency. The kinds of arguments introduced in this way in White
Hmong include arguments expressing Location of action, Goal of motion and
transfer, Topic of speech or thought, Terminus, Extent, Comitative, and so on.
The kind of SVC most commonly used to perform this valency increasing func-
tion in White Hmong is the Cotemporal SVC, although the Attainment SVC
and the Disposal SVC are also used in more limited ways.
6.5 The Relationship Between Function and Part of Speech
The function of introducing arguments such as Locatives, Goals, Comitatives,

and the like into the clause is one that is clearly associated with adpositions
rather than with verbs in many languages. This fact has led several scholars
dealing with verb serializing languages to propose that, when a morpheme
functions like an adposition, it should be analysed as an adposition in that
particular environment. It may be homophonous with a morpheme used else-
where as a verb, but this is simply attributed to a synchronic or historical deri-
vation relationship, and is not seen by these scholars as having any bearing on
correct part-of-speech classification.
However, it is argued in Chapter 4 (§4.1) that, while function is intricately
bound to grammatical behaviour, it does not constitute an appropriate crite-
rion for determining word class. The function of a particular morpheme may
well actually result in some changes in its grammatical behaviour, and possibly
even in its part-of-speech classification. However, it is these formal changes that
are the real indices of any change, say, from verb to adposition. Furthermore,
while performing as a valency increasing morpheme may, indeed, represent a
radical change in function from that normally expected of a verb in a language
that does not use the strategy of serialization, such a function is well within
the range of what would be expected of a verb in a language that does—one in
which two or more verbs can work together to express a single event.
Syntactic tests (§4.2) show that, in the majority of cases, the so-called
‘coverbs’ in White Hmong exhibit the grammatical behaviour that would be
expected of them as verbs in the serial verb constructions of the relevant type.
282 Chapter 6
This is shown to sometimes be the case even when these morphemes are used
in highly metaphoric, non-literal senses. However, in any case where such a
morpheme is used to express Location, Extent, Goal, etc. of the clause as a
whole, and does not have a predication relationship with the subject, there
seems to be a concomitant change in grammatical properties. Note that predi-
cation of the subject is a feature of all three types of SVC from which the sen-
tences in question are derived: Cotemporal, Attainment, and Disposal. If an
erstwhile verb does not predicate the subject, and if it has no other shared
argument with the other predicate in such a clause, its only function in the
clause is to introduce its own object argument. In this case its grammatical
properties tend to indicate that it has undergone true part-of-speech change.
6.6 Superficially Similar Constructions
In Chapter Five a wide variety of constructions in White Hmong are e xamined—

constructions that resemble serial verb constructions in that they contain verbs
that are simply juxtaposed with no overt lexical linkage in the form of conjunc-
tions or complementizers. However, these constructions are not regarded as
serial verb constructions according to the criteria proposed in this analysis.
One type of Elaborate Expression in White Hmong has the same structure
as one kind of SVC (§5.1). That is, Elaborate Expressions of the form V NP V
NP could simply be thought of as rather poetical or euphonious examples
of the Cotemporal SVC type; they particularly resemble SVCs that express
Cotemporal Actions. Nevertheless, many other kinds of Elaborate Expressions
involve structures of quite a different kind, such as those that do not contain
predicates at all. For this reason the phenomenon needs to be described quite
separately from that of serialization in this language.
White Hmong is a highly paratactic language, which relies to a considerable
extent on pragmatic factors for interpretation at the discourse level. Separate
clauses in various syntactic and semantic relationships, expressing distinct
events, can be simply juxtaposed, and the interpretation of the semantic rela-
tionship between them assigned entirely to factors such as context and real
world knowledge (§5.2).
It is not only juncts in a paratactic relationship but also those in a hypo-
tactic relationship that can be simply juxtaposed in White Hmong. Even in
the absence of a complementizer, no morphological signals of subordination
(such as gerundive or subjunctive morphology) occur to indicate the nature of
the relationship in this language. However, syntactic tests can be used to reveal
Conclusion 283
differences between serialization, on the one hand, and subordination without

a complementizer, on the other.
In spite of the fundamental differences between serialization and subor-
dination, some examples of subordination that involve simple juxtaposition
of the main and subordinate juncts also exhibit other points of similarity to
serial verb constructions. A high degree of ‘binding’ between a superordinate
and subordinate junct (§5.3–§5.5) has significant implications regarding the
status of the subordinate junct as a separate clause. For example, a highly
‘bound’ complement may not have a truth-value independent of that of the
main verb and/or may not be able to exhibit certain peripheral arguments
independently. Factors such as these obviously contribute to the difficulty in
distinguishing these constructions from serial verb constructions. However, it
is argued here that similarities such as these are the result of ‘convergent evolu-
tion’ rather than ‘common ancestry’.
Finally, White Hmong is a language that utilises verbs in a wide variety of
roles, including attributive and adverbial roles (§5.6 and §5.7). Although adver-
bial constructions, in particular, bear a considerable degree of resemblance
to certain types of serial construction, with the two verbs occurring within a
single clause, they are not regarded as examples of serialization in this analysis
because there is no shared argument. Verbs functioning in an adverbial role
predicate the entire proposition, rather than any particular argument within
that proposition.
Because serial verb constructions are often employed to perform func-
tions not normally associated with verbs in some languages, such as a valency
increasing or adverbial function, it is sometimes assumed that any verb in a
serializing language that performs a function not normally associated with
verbs in other languages must be part of a serial construction. However, as in
Chapter 4, it is also shown in Chapter 5 that this is not necessarily the case.
For example, verbs in White Hmong that function adverbially (§5.6.2 and
§5.7.2), those that express modal meanings (§5.4), and those that introduce
instrumental arguments in the Purpose Construction (§5.5.1.3), are not actu-
ally serial verb constructions in White Hmong. Sentences containing such
verbs often exhibit significant similarities to serial verb constructions, but
nevertheless are shown to represent distinct constructions. The phenom-
enon of serialization does, indeed, allow verbs to perform functions that
they would not be able to perform in a language that does not utilise the
strategy of employing more than one verb to express a single proposition.
However, this should not be taken to mean that other constructions in the
language cannot also allow verbs to be used in a variety of ways. It is not
284 Chapter 6
surprising that these other constructions do bear a considerable resem-

blance to serial verb constructions; the constructions may certainly be simi-
lar in many respects in both form and function, but are not identical.
6.7 Directions for Further Research
A particularly interesting issue that has emerged in this study is related to both
the differences and similarities between verb serialization and other types of
juxtaposition of verbs involving both paratactic and hypotactic relationships
between the juncts. Some scholars who have drawn a distinction between seri-
alization and other forms of juxtaposition have tended to draw a fairly clear
cut one, such as two actions/events vs. one. However, much of the evidence
discussed in Chapter Five of this study has shown that such a clear cut dis-
tinction cannot always be drawn; there are good reasons, in other words, why
scholars have had considerable difficulty in enunciating a clear set of criteria
for verb serialization in the language(s) they investigate that divides this phe-
nomenon neatly from other ‘superficially similar constructions’.
The evidence available seems to point to the conclusion that the human
mind does not perceive all situations in discrete terms, as involving one event
or two, which can then be equally discretely expressed as one proposition or
two, in one clause or two. Instead, there are many types of situation, for exam-
ple certain situations involving manipulation and causation, modality, pur-
pose and intention, multiple related actions by a single subject, etc., which we
see as neither wholly one event, nor wholly two, but as something in between.
This indeterminacy in perception is often reflected by an indeterminacy of
form in the linguistic expression. As pointed out by Givón (1980, 1985, 2001b),
constructions involving relationships between predicates may have more or
fewer features of separate clauses, rather than being absolute realisations of
either one clause or two:
Along the functional continuum, the propositional-semantic features of

event integration shade gradually into the more discourse-pragmatic fea-
tures of cross-event coherence. Along the parallel syntactic continuum,
the grammatical bonds of inter-clausal dependence become gradually
looser. (Givón, 2001a, p. 328; emphasis in the original)
It has been shown in this study (Chapter 2, §2.4) that the RRG tripartite analy-
sis of juncture types requires some refinement, to allow for at least two distinct
kinds of core juncture: core juncture expressing one event within a single
Conclusion 285
clause, and core juncture expressing two events over two clauses, allowing
independent peripheral arguments for each junct. Both types involve the inter-
section of cores, but they are at opposite ends of the spectrum with respect to
the status of the juncts as separate events.
It may well be the case that what is really required is not simply one further
distinction within the core juncture type such as this, but rather a full incor-
poration of a scalar notion of clause-hood into the theory of juncture types.
Such an incorporation would work towards providing a syntactic account of
the varying degrees of separation/integration between clauses, depending on
factors such as their relative degree of binding.
Deeper investigation would be extremely valuable into the syntactic con-
sequences of indeterminacy regarding clause-hood in many linguistic forms
in White Hmong, and a careful comparison of these various forms with the
syntax of serial verb constructions, which have been characterised as describ-
ing a single event and involving a single proposition within a single clause.
In a similar vein, a fuller investigation of the precise syntactic relationship
between the verbs in all the various types of serial verb constructions and their
corresponding degree of semantic ‘oneness’ would also doubtless reveal more
of this non-discrete nature of the clause. This study is no more than a first
attempt at understanding just one part of this fascinating topic in this delight-
ful language.
Sources for White Hmong Examples1
Johnson, C. (1981). Hmong folk tales. St Paul: Linguistics Department, Macalester

College. (HFT)
———. (Ed.) (1985). Dab neeg Hmoob: Myths, legends and folk tales from the Hmong of
Laos. St Paul: Linguistics Department, Macalester College. (DNH)
Lis, N.P. (n.d.). Vim leejtwg: Lhub, nco, tusiab, and chimsiab [Because of Whom? Love,
memories, offence, and anger]. Sydney: Published by the author (4 Janali Ave,
Bonnyrigg, NSW 2177 Australia). (VLT)
Lis, Y. (n.d.). Notes on White Hmong grammar [in White Hmong]. Unpublished paper,
Sydney. (WHG)
Overseas Missionary Fellowship. (1976). Kawm ntawv Hmoob: Phau 2 [Learning Hmong:
Book 2]. Thailand. (KNH)
Southeast Asian Summer Studies Institute. (1985). Introduction to White Hmong, Class
Materials. University of Michigan. (IWH)
Strecker, D. & Vang, L. (1986). White Hmong dialogues (Southeast Asian Refugee
Studies Occasional Papers 3). Minneapolis: Center for Urban and Regional Affairs,
University of Minnesota. (WHD)
Tsawb, Y.T. & Strecker, D. (1986). Excerpt from Piav Tus Txheej Txheem Kab Tshoob Kev
Kos [Outline of Marriage Rites] (Translated and annotated by David Strecker). In
B. Johns & D. Strecker (Eds.), The Hmong World. (THW)
Vang, L.P. (1985). Tus tsov thiab tus qav [The tiger and the frog]. Unpublished folk tale,
Southeast Asian Summer Studies Institute. University of Michigan. (LPV)
1 Capitals in parentheses following a reference show the abbreviation used to cite that refer-
ence in the text.
Bibliography
Aikhenvald, A.Y. (2000). Classifiers: A Typology of Noun Categorization Devices. Oxford:

Oxford University Press.
——— (2006a). Serial verb constructions in Tariana. In A.Y. Aikhenvald & R.M.W. Dixon
(Eds.), Serial verb constructions: A cross-linguistic typology (Explorations in Linguistic
Typology Vol. 2) (pp. 178–201). Oxford: Oxford University Press.
——— (2006b). Serial verbs constructions in a typological perspective. In
A.Y. Aikhenvald & R.M.W. Dixon (Eds.), Serial verb constructions: A cross-linguistic
typology (Explorations in Linguistic Typology, Vol. 2) (pp. 1–68). Oxford: Oxford
University Press.
Aikhenvald, A.Y., & Dixon, R.M.W. (2006). Serial verb constructions: A cross-linguistic
typology (Explorations in Linguistic Typology Vol. 2). Oxford: Oxford University Press.
Andersson, S.-G. (1972). Aktionalität im Deutschen: Eine Untersuchung unter Vergleich
mit dem Russischen Aspektsystem. Uppsala: Acta Universitatis Upsaliensis.
Andruski, J.E., & Ratliff, M. (2000). Phonation types in production of phonological
tone: The case of Green Mong. Journal of the International Phonetic Association,
30(1–2), 37–61. doi: 10.1017/S0025100300006654
Ansre, G. (1966). The verbid—A caveat to ‘serial verbs’. Journal of West African
Languages, 3(1), 29–32.
Armstrong, D. (1981). The Ancient Greek aorist as the aspect of countable action. In
P.J. Tedeschi & A. Zaenen (Eds.), Tense and aspect (pp. 1–12). New York: Academic
Press.
Awobuluyi, Ọ. (1973). The modifying serial construction: A critique. Studies in African
Linguistics, 4(1), 87–111.
Bamgboṣe, A. (1973). The modifying serial construction: A reply. Studies in African
Linguistics, 4(2), 207–217.
——— (1974). On serial verbs and verbal status. Journal of West African Languages,
9(1), 17–48.
Benedict, P.K. (1972). Sino-Tibetan: A conspectus (J.A. Matisoff Ed.). Contributing
Editor: James A. Matisoff. Cambridge: Cambridge University Press.
——— (1975). Austro-Thai: Language and culture (with a glossary of roots). New Haven,
Connecticut: Human Relations Area Files Press.
Bertrais-Charrier, Y. (1964). Dictionnaire Hmong (Mèo-Blanc-Français). Vientiane:
Mission Catholique.
Bisang, W. (1992). Das Verb im Chinesischen, Hmong, Vietnamesischen, Thai
und Khmer. Vergleichende Grammatik im Rahmen der Verbserialisierung, der
Grammatikalisierung und der Attraktorpositionen. Tübingen, Germany: Gunter Narr.
——— (1993). Classifiers, quantifiers and class nouns in Hmong. Studies in Language,
17(1), 1–51.
Bibliography 289
——— (2009). Serial verb constructions. Language and Linguistics Compass, 3(3),
792–814.
Bohnemeyer, J., Enfield, N.J., Essegbey, J., Ibarretxe-Antuñano, I., Kita, S., Lüpke, F.,
& Ameka, F.K. (2007). Principles of event segmentation in language: The case of
motion events. Language, 83(3), 495–532.
Bohnemeyer, J., Enfield, N.J., Essegbey, J., & Kita, S. (2011). The macro-event prop-
erty: The segmentation of causal chains. In J. Bohnemeyer & E. Pederson (Eds.),
Event representation in language and cognition (pp. 43–67). New York: Cambridge
University Press.
Bohnemeyer, J., & Van Valin, R.D., Jr. (2009). The macro-event property and the layered
structure of the clause. Paper presented at the 2009 International Conference on
Role and Reference Grammar, August 7–9, University of California, Berkeley.
Bradley, D. (2007). East and Southeast Asia. In C. Moseley (Ed.), Encyclopedia of the
world’s endangered languages (pp. 349–424). London Routledge.
Bruce, L. (1988). Serialisation: From syntax to lexicon. Studies in Language, 12(1), 19–49.
Bybee, J. (1985). Morphology: A study of the relation between meaning and form.
Amsterdam: John Benjamins Publishing.
——— (2003). Mechanisms of change in grammaticalization: The role of frequency.
In R. Janda & B.D. Joseph (Eds.), Handbook of Historical Linguistics (pp. 603–623).
Oxford: Blackwell.
Chao, Y.-R. (1968). A grammar of spoken Chinese. Berkeley: University of California
Press.
Christaller, J.G. (1875). A grammar of the Asante and Fante language called Tshi [Chwee,
Twi]. Basel: Printed for the Basel Evangelical Missionary Society.
——— (1933). Dictionary of the Asante and Fante language called Tshi (Twi). Basel:
Basel Evangelical Missionary Society.
Clark, M. (1977). Ditransitive goal verbs in Vietnamese. Mon-Khmer Studies, 6, 1–38.
——— (1978). Coverbs and case in Vietnamese (Pacific Linguistics Series B. 48). Canberra:
Research School of Pacific Studies, The Australian National University.
——— (1979a). Coverbs: Evidence for the derivation of prepositions from verbs—new
evidence from Hmong. University of Hawaii Working Papers in Linguistics, 11(2), 1–12.
——— (1979b). Synchronically derived prepositions in diachronic perspective: Some evi-
dence from Hmong. Paper presented at the 12th International Conference on Sino-
Tibetan Languages and Linguistics, 19–21 October, École Normale Supérieure, Paris.
——— (1980a). Derivation between goal and source verbs in White Hmong. University
of Hawaii Working Papers in Linguistics, 12(2), 51–59.
——— (1980b). Source Phrases in White Hmong (Laos). University of Hawaii Working
Papers in Linguistics, 12(2), 1–59.
——— (1982). Some auxiliary verbs in Hmong. In B.T. Downing & D.P. Olney (Eds.),
The Hmong in the West (pp. 125–141). Minneapolis/St. Paul: Southeast Asian Refugee
Project, Center for Urban and Regional Affairs, University of Minnesota.
290 Bibliography
——— (1989). Hmong and areal South-east Asia. In D. Bradley (Ed.), Papers in South-
east Asian linguistics No. 11: South-east Asian syntax (pp. 175–230). Canberra: Pacific
Linguistics.
——— (1992). Serialization in Mainland Southeast Asia. Paper presented at the Third
International Symposium on Language and Linguistics, Bangkok.
Claudi, U., & Heine, B. (1985). From metaphor to grammar: Some examples from Ewe.
Afrikanistishe Arbeitspapiere, 1, 17–54.
Comrie, B. (1976). Aspect: An introduction to the study of verbal aspect and related
problems. Cambridge: Cambridge University Press.
Crowley, T. (1987). Serial verbs in Paamese. Studies in Language, 11(1), 35–84.
Culas, C. (2009). The ethnonyms of the Hmong in Vietnam: Short history (1856–1924)
and practical epistemology. In C. Culas & F. Robinne (Eds.), Inter-ethnic dynamics
in Asia: Considering the other through ethnonyms, territories and rituals (pp. 13–42).
London: Routledge.
Dahl, Ö. (1981). On the Definition of the Telic-Atelic (Bounded–Nonbounded)
Distinction in Tense and Aspect. In P.J. Tedeschi & A. Zaenen (Eds.), Syntax and
semantics (Vol. 14, pp. 79–90). New York: Academic Press.
Dixon, R.M.W. (2006). Complement clauses and complementation strategies in typo-
logical perspective. In R.M.W. Dixon & A.Y. Aikhenvald (Eds.), Complementation:
a cross-linguistic typology (pp. 1–48). Oxford: Oxford University Press.
Downer, G.B. (1967). Tone-change and tone-shift in White Miao. Bulletin of the School of
Oriental and African Studies, 30(3), 589–599.
Dowty, D.R. (1979). Word meaning and Montague grammar: The semantics of verbs and
times in generative semantics and in Montague’s PTQ (Vol. 7). Dordrecht: D. Reidel.
Durie, M. (1988). Verb Serialization and “Verbal-Prepositions” in Oceanic Languages.
Oceanic Linguistics, 27(1/2), 1–23.
——— (1997). Grammatical structures in verb serialization. In A. Alsina, J. Bresnan &
P. Sells (Eds.), Complex predicates (pp. 289–354). Stanford, CA: CSLI Publications.
Enfield, N.J. (2003). Linguistic epidemiology: Semantics and grammar of language con-
tact in mainland Southeast Asia. London: Routledge.
——— (2007). A grammar of Lao. Berlin: Mouton de Gruyter.
——— (2009). Review of the book Serial verb constructions: A cross-linguistic typology
ed. by Alexandra Y. Aikhenvald and R.M.W. Dixon. Language, 85(2), 445–451.
——— (2015). The Utility of meaning: What words mean and why. Oxford: Oxford
University Press.
Enwall, J. (1994). A myth become reality: History and development of the Miao written
language (Vol. 1). Stockholm: Institute of Oriental Languages, Stockholm University.
Esposito, C.M. (2012). An acoustic and electroglottographic study of White Hmong
tone and phonation. Journal of Phonetics, 40(3), 466–476.
Fei, X.T. (1999). The pattern of diversity in unity of the Chinese nation (in Chinese).
Beijing: Central University for Nationalities Press.
Bibliography 291
Foley, W.A., & Olson, M. (1985). Clausehood and verb serialization. In J. Nichols &
A. Woodbury (Eds.), Grammar inside and outside the clause (pp. 17–60). Cambridge:
Cambridge University Press.
Foley, W.A., & Van Valin, R.D., Jr. (1984). Functional syntax and universal grammar.
Cambridge: Cambridge University Press.
Fuller, J.W. (1985). Topic and comment in Hmong (PhD Dissertation). University of
Minnesota.
Fulop, S.A., & Golston, C. (2008). Breathy and whispery voice in White Hmong.
Proceedings of Meetings on Acoustics, 4, 1–10. DOI: 10.1121/1.3033931
Geddes, W.R. (1976). Migrants of the mountains: The cultural ecology of the Blue Miao
(Hmong Njua) of Thailand. Oxford: Clarendon Press.
Givón, T. (1973). The time-axis phenomenon. Language, 49(4), 890–925.
——— (1975). Serial verbs and syntactic change: Niger-Congo. In C.N. Li (Ed.), Word
order and word order change (pp. 47–112). Austin: University of Texas Press.
——— (1980). The binding hierarchy and the typology of complements. Studies in
Language, 4(3), 333–377.
——— (1984). Syntax: A functional-typological introduction (Vol. 1). Amsterdam: John
Benjamins.
——— (1985). Iconicity, isomorphism, and non-arbitrary coding in syntax. In
J. Haiman (Ed.), Iconicity in syntax (pp. 187–219). Amsterdam: John Benjamins.
——— (1987). Serial verbs and the mental reality of ‘event’: Final project report pre-
sented to the Research Programs Division, National Endowment for the Humanities.
——— (1991). Serial verbs and the mental reality of ‘event’: Grammatical vs. cognitive
packaging. In E.C. Traugott & B. Heine (Eds.), Approaches to grammaticalization
(Vol. 1, pp. 81–127). Amsterdam: John Benjamins.
——— (2001a). Syntax: An introduction (Revised ed. Vol. 2). Amsterdam: John Benjamins.
——— (2001b). Syntax: An introduction (Revised ed. Vol. 1). Amsterdam: John
Benjamins.
Greenberg, J.H. (1953). Historical linguistics and unwritten languages. In A.L. Kroeber
(Ed.), Anthropology today (pp. 265–286). Chicago: University of Chicago Press.
Haas, M.R. (1964). Thai-English student’s dictionary. Stanford, California: Stanford
University Press.
Haiman, J. (1980). The iconicity of grammar: Isomorphism and motivation. Language,
56(3), 515–540.
——— (Ed.). (1985). Iconicity in syntax: Proceedings of a Symposium on Iconicity in
Syntax, Stanford, June 24–6, 1983. Amsterdam: John Benjamins Publishing.
Hansell, M. (1987). Serial verbs and complement constructions in Mandarin: A clause
linkage analysis. Davis Working Papers in Linguistics, 2, 38–54.
——— (1993). Serial verbs and complement constructions in Mandarin: A clause
linkage analysis. In R.D. Van Valin, Jr (Ed.), Advances in role and reference grammar
(pp. 197–233). Amsterdam: John Benjamins.
292 Bibliography
Haspelmath, M. (2011). Pre-established categories don’t exist: Consequences for lan-

guage description and typology. Linguistic Typology, 11(1), 119–132.
Haudricourt, A.-G. (1951). Introduction à la phonologie historique des langues Miao-
Yao. Bulletin de l’Ecole française d’Extrême-Orient, 44(2), 555–576.
——— (1966). The limits and connections of Austroasiatic in the Northeast. In N.H. Zide
(Ed.), Studies in comparative Austronesian linguistics (pp. 44–56). The Hague: Mouton.
——— (1971). Les langues Karen et les langues Miao-Yao. Asie du Sud-Est et Monde
Insulindien, 2(4), 25–52.
Heimbach, E.E. (1979). White Hmong—English dictionary (Revised ed.). Ithaca, NY:
Cornell University.
Hopper, P.J., & Thompson, S.A. (1980). Transitivity in grammar and discourse. Language,
56(2), 251–299.
Hopper, P.J., & Traugott, E.C. (2003). Grammaticalization (2nd ed.). Cambridge:
Cambridge University Press.
Huffman, M.K. (1987). Measures of phonation type in Hmong. The Journal of the
Acoustical Society of America, 81(2), 495–504. doi: 10.1121/1.394915
Ikegami, Y. (1985). ‘Activity’-‘accomplishment’-‘achievement’: A language that can’t
say, ‘I burned it, but it didn’t burn’ and one that can. In A. Makkai & A.K. Melby
(Eds.), Linguistics and philosophy: Essays in honor of Rulon S. Wells (pp. 265–304).
Amsterdam: John Benjamins.
Jaisser, A. (1984). Complementation in Hmong (Unpublished MA thesis). San Diego State
University.
——— (1987). Hmong classifiers: A problem set. Linguistics of the Tibeto-Burman Area,
10(2), 169–176.
——— (1989). Evaluation in a Hmong narrative. Paper presented at the 22nd
International Conference on Sino-Tibetan Languages and Linguistics, University of
Hawai’i at Manoa.
Jansen, B., Koopman, H., & Muysken, P.C. (1978). Serial verbs in the creole languages.
Amsterdam Creole Studies(2), 125–159.
Jarkey, N. (1987). An investigation of two alveolar stop consonants in White Hmong.
Linguistics of the Tibeto-Burman Area, 10(2), 57–70.
——— (1991). Serial verb constructions in White Hmong: A functional approach. PhD
thesis, University of Sydney.
——— (2006). Complement clause types and complementation strategy in White
Hmong. In R.M.W. Dixon & A.Y. Aikhenvald (Eds.), Complementation: a cross-
linguistic typology (pp. 115–136). Oxford: Oxford University Press.
——— (2010). Cotemporal serial verb constructions in White Hmong. In M. Amberber,
B. Baker & M. Harvey (Eds.), Complex predicates: Cross-linguistic perspectives on
event structure (pp. 110–134). Cambridge: Cambridge University Press.
Bibliography 293
Johns, B., & Strecker, D. (1982). Aesthetic language in White Hmong. In B.T. Downing &
D.P. Olney (Eds.), The Hmong in the West: Observations and reports (pp. 160–169).
Minneapolis: Center for Urban and Regional Affairs, University of Minnesota.
Johnson, C. (1981). Hmong folk tales. St Paul: Linguistics Department, Macalester
College.
——— (Ed.). (1985). Dab neeg Hmoob: Myths, legends and folk tales from the Hmong of
Laos. St Paul: Linguistics Department, Macalester College.
Karttunen, L. (1971). Implicative verbs. Language, 47(2), 340–358.
Keenan, E.L., & Li, C.N. (1975). Towards a Universal Definition of ‘Subject’. In C.N. Li
(Ed.), Subject and Topic (pp. 303–333). New York: Academic Press.
Kosaka, R. (2002). On the affiliation of Miao-Yao and Kadai: Can we posit a Miao-Dai
family? Mon-Khmer Studies, 32, 71–100.
Langacker, R.W. (1983). Foundations of cognitive grammar. Bloomington: Indiana
Linguistics Club.
Lee, G.Y., & Tapp, N. (2010). Culture and Customs of the Hmong. Santa Barbara,
California: ABC-CLIO, LLC.
Lemoine, J. (2005). What is the actual number of the (H)mong in the world? Hmong
Studies Journal, 6, 1–8.
Lewis, M.P., Simons, G.F., & Fennig, C.D. (Eds.). (2013). Ethnologue: Languages of the
world (Seventeenth ed.). Dallas, Texas: SIL International. Online version: http://
www.ethnologue.com.
Li, C.N. (n.d.). The apectual system of Hmong. Unpublished paper, University of
California, Santa Barbara.
Li, C.N., & Thompson, S.A. (1973). Serial verb constructions in Mandarin Chinese:
Subordination or coordination? In C. Corum, T.C. Smith-Stark & A. Weiser (Eds.),
You take the high node and I’ll take the low node: Papers from the Comparative Syntax
Festival (pp. 96–103). Chicago: Chicago Linguistic Society.
——— (1974a). Co-Verbs in Mandarin Chinese: Verbs or Prepositions? Journal of
Chinese Linguistics, 2(3), 257–278.
——— (1974b). An explanation of word order change SVO→SOV. Foundations of
Language, 12(2), 201–214.
——— (1981). Mandarin Chinese: A functional reference grammar. Berkeley: University
of California Press.
Li, C.N., Thompson, S.A., & Thompson, R.M. (1982). The discourse motivation for the
perfect aspect: The Mandarin particle le. In P.J. Hopper (Ed.), Tense-aspect: Between
semantics and pragmatics (pp. 19–44). Amsterdam: John Benjamins.
Lord, C. (1973). Serial verbs in transition. Studies in African Linguistics, 4(3), 269–296.
——— (1974). Causative constructions in Yoruba. Studies in African Linguistics
(Supplement 5), 195–204.
294 Bibliography
Lyman, T.A. (1974). Dictionary of Mong Njua: A Miao (Meo) language of Southeast Asia.
The Hague: Mouton.
Lyons, J. (1977). Semantics (vols I & II). Cambridge: Cambridge University Press.
Matisoff, J.A. (1969). Verb concatenation in Lahu: The syntax and semantics of ‘simple’
juxtaposition. Acta Linguistica Hafniensia, 12(1), 69–120.
——— (1973). The grammar of Lahu. Berkeley: University of California Press.
Matthews, S. (2006). On serial verb constructions in Cantonese. In A. Aikhenvald &
R.M.W. Dixon (Eds.), Serial verb constructions: A cross-linguistic typology
(Explorations in Linguistic Typology, Vol. 2) (pp. 69–87). Oxford: Oxford University
Press.
Mortensen, D.R. (2003). Hmong elaborate expressions are coordinate compounds.
University of California at Berkeley. Unpublished. Retrieved from http://www.pitt
.edu/~drm31/ (Accessed 13 April 2013)
Mottin, J. (1978). Eléments de grammaire hmong blanc. Bangkok: Don Bosco Press.
Mourelatos, A.P.D. (1981). Events, processes, and states. In P.J. Tedeschi & A. Zaenen
(Eds.), Tense and aspect (14 ed., pp. 191–212). New York: Academic Press.
Nylander, D.K. (1981). Serial and pseudoserial verb constructions in Krio. Montreal
Working Papers in Linguistics, 17, 101–118.
Olson, M. (1981). Barai clause juncture: Toward a functional theory of interclausal rela-
tions. Unpublished PhD thesis, Australian National University.
Owensby, L. (1986). Verb serialization in Hmong. In G.L. Hendricks, B.T. Downing &
A.S. Deinard (Eds.), The Hmong in Transition (pp. 237–243). New York and
Minneapolis: Centre for Migration Studies and Southeast Asian Refugee Studies
Project, University of Minnesota.
Paul, W. (2008). The serial verb construction in Chinese: A tenacious myth and a
Gordian knot. The Linguistic Review, 25(3–4), 367–411.
Pawley, A. (1973). Some problems in Proto-Oceanic grammar. Oceanic Linguistics,
12(1/2), 103–188.
——— (1987). Encoding events in Kalam and English: Different logics for reporting
experience. In R.S. Tomlin (Ed.), Coherence and grounding in discourse (pp. 329–
360). Amsterdam: John Benjamins.
——— (1993). A language which defies description by ordinary means. In W. Foley
(Ed.), The role of theory in language description (pp. 87–129). Berlin: Mouton de
Gruyter.
——— (2009). On the origins of serial verb constructions in Kalam. In T. Givón &
M. Shibatani (Eds.), Syntactic complexity: Diachrony, acquisition, neuro-cognition,
evolution (pp. 119–144). Amsterdam: John Benjamins.
Pfeifer, M.E., & Lee, S. (2004). Hmong population, demographic, socioeconomic, and
educational trends in the 2000 Census. In B. Thao, L. Schein & M. Niedzweick (Eds.),
Bibliography 295
Hmong 2000 census publication: Data and analysis (pp. 3–11). Washington, DC:
Hmong National Development, Inc., and the Hmong Cultural and Resource Center.
Piau, J.A. (1985). The verbal syntax of Kuman. Unpublished MA thesis, Australian
National University.
Ratliff, M. (1986a). An analysis of some tonally differentiated doublets in White Hmong
(Miao). Linguistics of the Tibeto-Burman Area, 9(2), 1–33.
——— (1986b). Two-Word Expressives in White Hmong. In G.L. Hendricks, B.T.
Downing & A.S. Deinard (Eds.), The Hmong in Transition (pp. 219–236). New York
and Minneapolis: Center for Migration Studies Press and Southeast Asian Refugee
Studies Project, University of Minnesota.
——— (1987). Tone sandhi compounding in White Hmong. Linguistics of the Tibeto-
Burman Area (Hmong-Mien Languages), 10(2), 71–105.
——— (1990). The influence of geographical change on grammar: The case of Hmong
spatial deictics. Paper presented at the 23rd International Conference on Sino-
Tibetan Languages and Linguistics, University of Texas, Arlington.
——— (1991). Cov, the underspecified noun, and syntactic flexibility in Hmong.
Journal of the American Oriental Society, 111(94), 694–703.
——— (1992). Meaningful tone: A study of tonal morphology in compounds, form
classes, and expressive phrases in White Hmong. DeKalb, Illinois: Northern Illinois
University, Center for Southeast Asian Studies.
——— (1998). Ho Ne (She) is Hmongic: one final argument. Linguistics of the Tibeto-
Burman Area, 12(2), 97–109.
——— (2010). Hmong-Mien language history. Canberra: Pacific Linguistics.
Riddle, E.M. (1989). Serial verbs and propositions in White Hmong. Linguistics of the
Tibeto-Burman Area, 12(2), 1–13.
Sakuragi, T., & Fuller, J.W. (2013). Shape and function in Hmong classifier choices.
Journal of psycholinguistic research, 42(4), 349–361.
Schachter, P. (1974). A non-transformational account of serial verbs. Studies in African
Linguistics, Supplement 5, 253–270.
Sebba, M. (1987). The syntax of serial verbs: An investigation into serialisation in Sranan
and other languages. Amsterdam: John Benjamins.
Shafer, R. (1964). Miao-Yao. Monumenta Serica, 23, 398–411.
Shibatani, M. (1976). The grammar of causative constructions: A conspectus. In
M. Shibatani (Ed.), The grammar of causative constructions (pp. 5–41). New York:
Academic Press.
Shibatani, M., & Pardeshi, P. (2002). The causative continuum. In M. Shibatani (Ed.),
The grammar of causation and interpersonal manipulation (pp. 85–126). Amsterdam:
John Benjamins.
Singh, M. (1991). The perfective paradox: Or how to eat your cake and have it too.
Berkeley Linguistic Society, 17, 469–478.
296 Bibliography
Smalley, W.A. (1976). The problems of consonants and tone: Hmong (Meo, Miao). In
W.A. Smalley (Ed.), Phonemes and orthography: language planning in ten minority
languages in Thailand. Pacific Linguistics Series C (pp. 85–123). Canberra: Linguistic
Circle of Canberra.
Smalley, W.A., Vang, C.K., & Yang, G.Y. (1990). Mother of writing: the origin and develop-
ment of a Hmong messianic script. Chicago: University of Chicago Press.
Smith, C.S. (1997). The parameter of aspect (Vol. 43). Dordrecht: Kluwer Academic
Publishers.
Solnit, D.B. (2006). Verb serialization in Eastern Kayah Li. In A.Y. Aikhenvald & R.M.W.
Dixon (Eds.), Serial verb constructions: A crosslinguistic typology (pp. 144–159).
Oxford: Oxford University Press.
Sposato, A. (2012). Relative clauses in Xong (Miao-Yao). Journal of the Southeast Asian
Linguistics Society, 5, 49–66.
Stahlke, H. (1970). Serial Verbs. Studies in African Linguistics, 1(1), 60–99.
Stewart, J.M. (1963). Some restrictions on objects in Twi. Journal of African languages,
2(2), 145–149.
Tai, J.H.-Y. (1984). Verbs and times in Chinese: Vendler’s four categories. Papers from the
Parasession on Lexical Semantics (pp. 289–296). Chicago: Chicago Linguistic Society.
Talmy, L. (1975). Semantics and syntax of motion. In J.P. Kimball (Ed.), Syntax and
semantics (Vol. 4). New York: Academic Press.
——— (1991). Path to realization: A typology of event conflation. Berkeley Linguistics
Society, 17, 480–519.
Tapp, N. (2002). Cultural Accommodations in Southwest China: The “Han Miao” and
Problems in the Ethnography of the Hmong. Asian Folklore Studies, 61(1), 77–104.
doi: 10.2307/1178678
Thepkanjana, K. (1986). Serial verb constructions in Thai. Unpublished PhD disserta-
tion, University of Michigan.
Thompson, L.C. (1965). A Vietnamese Grammar. Seattle: University of Washington Press.
Thompson, S.A. (1973). Resultative verb compounds in Mandarin Chinese: A case for
lexical rules. Language, 49(2), 361–379.
Tsujimura, N. (2003). Event cancellation and telicity. Japanese/Korean Linguistics, 12,
388–399.
Van Valin, R.D., Jr. (1986). Pragmatics, island phenomena, and linguistic compe-
tence. Chicago Linguistic Society (Papers from the Parasession on Pragmatics and
Grammatical Theory), 22(2), 223–233.
——— (1987). Recent developments in Role and Reference Grammar: The layered
structure of the clause and juncture. Davis Working Papers in Linguistics, 2, 1–5.
——— (2005). Exploring the syntax-semantics interface. Cambridge: Cambridge
University Press.
Bibliography 297
——— (2007). Recent developments in the Role and Reference Grammar theory of
clause linkage. Language and Linguistics, 8(1), 71–93.
Van Valin, R.D., Jr., & LaPolla, R.J. (1997). Syntax: Structure, meaning, and function.
Cambridge: Cambridge University Press.
Vang, C.K., Yang, G.Y., & Smalley, W.A. (1990). Life of Shong Lue Yang: Hmong ‘Mother of
Writing’ (Keeb kwm Soob Lwj Yaj: Hmoob ‘Niam Ntawv’). Minneapolis, MN: Center for
Urban and Regional Affairs, University of Minnesota. CURA 90–1.
Vendler, Z. (1957). Verbs and times. The philosophical review, 66(2), 143–160.
Wechsler, S. (2003). Serial verbs and serial motion. In D. Beermann & L. Hellan (Eds.),
Proceedings of the Workshop on Multi-Verb Constructions, Trondheim Summer School.
Trondheim: Norwegian University of Science and Technology.
Wen, B., Li, H., Gao, S., Mao, X., Gao, Y., Li, F., . . . Jin, L. (2005). Genetic structure of
Hmong-Mien speaking populations in East Asia as revealed by mtDNA lineages.
Molecular Biology and Evolution, 22(3), 725–734. doi: 10.1093/molbev/msi055
Westermann, D. (1907). Grammatik der Ewe-Sprache. Berlin: Diedrich Reimer.
——— (1930). A study of the Ewe language. London: Oxford University Press.
Wheatley, J. (1984). The role of verb serialization in word order change. Paper pre-
sented at the 10th Annual Meeting of the Berkeley Linguistics Society, University of
California at Berkeley.
——— (1990). Burmese. In B. Comrie (Ed.), The world’s major languages (pp. 834–854).
New York: Oxford University Press.
Yang, D. (1980). Dictionnaire français-hmong blanc. Paris: Comité National dE̓ ntraide &
Jacques Lemoine.
Yang, K.-L. (1998). Problems in the interpretation of Hmong surnames. Paper presented
at the International Conference on the Hmong/Miao in Asia, Aix-en-Provence.
11–13 September.
Zacks, J.M., & Tversky, B. (2001). Event structure in perception and conception.
Psychological Bulletin, 127(1), 3–21.
Index of Authors
Aikhenvald, A.Y. 1, 33, 71, 81, 82, 84, 118, 127, Golston, C. 16
128, 188, 193 Greenberg, J.H. 9
Andersson, S.-G. 150
Andruski, J.E. 16 Haiman, J. 80
Ansre, G. 73, 189, 198 Hansell, M. 95, 140, 160, 161, 162, 237
Armstrong, D. 153n Haspelmath, M. 85, 196
Awobuluyi, Ọ. 75 Haudricourt, A.-G. 9
Heimbach, E.E. 12, 32–33, 47, 59, 65, 68, 177,
Bamgboṣe, A. 75, 124, 140, 192, 193, 259 208, 215, 248, 267, 272, 273
Benedict, P.K. 9 Heine, B. 194
Bertrais-Charrier, Y. 40, 54, 59, 65 Hopper, P.J. 130, 196
Bisang, W. 1, 26, 35, 76, 81, 82, 83, 185–86 Huffman, M.K. 15, 16
Bohnemeyer, J. 78, 80, 98, 125
Bradley, D. 10 Ikegami, Y. 154n
Bruce, L. 78, 79, 96, 104, 118, 128
Jaisser, A. 34, 59, 65, 69, 176, 242, 247,
Chao, Y.-R. 161 248–49, 253
Christaller, J.G. 70, 71, 72, 73, 141, 171 Jansen, B. 127
Clark, M. 5, 40, 43, 54, 59, 65, 76, 139, 198, Johns, B. 233, 235, 236
199, 207, 208, 237, 238–39 Johnson, C. 4
on ‘synchronically derived prepositions’
189, 190, 194, 195, 196, 197, 198, 199, 200, Karttunen, L. 136
201, 207, 208, 212, 213, 230 Keenan, E.L. 143
Crowley, T. 71, 98, 125, 127, 140, 141, 142 Kita, S. 98
Culas, C. 8 Koopman, H. 127
Kosaka, R. 9
Dahl, Ö. 150, 152n, 153, 154
Dixon, R.M.W. 1, 83 LaPolla, R.J. 2, 85, 96, 105, 106
Downer, G.B. 17, 21 Lee, G.Y. 8, 10
Dowty, D.R. 105 Lee, S. 10
Durie, M. 1, 76, 79, 82, 96, 118, 128, 142, 189, Lemoine, J. 9, 10
195, 198 Lewis, M.P. 7, 8, 10, 11
Li, C.N. 5, 59, 64, 65, 67, 68, 143, 201
Enfield, N.J. 56, 64, 66, 67, 79, 98, 147, 150, and Thompson, S.A. 73, 74, 75, 76, 139,
166, 167, 168, 169, 173, 193 177, 178, 189, 190, 194, 199, 205, 237, 259,
Enwall, J. 11 261
Esposito, C.M. 15, 16 Lord, C. 75, 76, 171, 188, 202
Essegbey, J. 98 Lyman, T.A. 11, 16
Lyons, J. 33
Fei, X.T. 10
Fuller, J.W. 34, 40, 59, 65, 110, 121, 138, 143, Matisoff, J.A. 73, 233
144, 145, 199 Matthews, S. 82–83
Fulop, S.A. 16 Mortensen, D.R. 234, 236
Mottin, J. 32, 33, 40, 52, 59, 65, 199, 200
Geddes, W.R. 10 Muysken, P.C. 127
Givón, T. 1, 3, 65, 66, 76, 77, 78, 190, 196, 245,
246, 251, 252, 275, 284 Nylander, D.K. 139
Index of Authors 299
Olson, M. 1, 85, 86, 93, 100, 126, 128, 193 Tapp, N. 7, 8, 9, 10

Owensby, L. 259 Thepkanjana, K. 71, 76, 83, 98, 125, 127, 128,
129, 191, 192–93, 249, 254
Pardeshi, P. 135 Thompson, L.C. 71
Paul, W. 75, 76 Thompson, R.M. 68
Pawley, A. 76, 77, 78, 186, 189 Thompson, S.A. 68, 130, 160, 161
Pfeifer, M.E. 10 and Li, C.N. 73, 74, 75, 76, 139, 177, 178,
Piau, J.A. 95 189, 190, 194, 199, 205, 237, 259, 261
Traugott, E.C. 196
Ratliff, M. 9, 10, 15, 16, 17, 18, 21, 23, 24, 25, 27, Tsujimura, N. 154n
28, 30–32, 36, 40
Riddle, E.M. 16, 76, 102, 103, 163, 259, 260–61 Vang, C.K. 23
Van Valin, R.D. Jr. 2, 85, 86, 87, 90, 91, 92, 93,
Sakuragi, T. 34 96, 98, 105, 106, 130, 166, 173
Schachter, P. 71, 75, 125, 127, 128, 140–41 and Foley, W.A. 85, 86, 93, 100, 101, 105
Sebba, M. 100, 127, 128, 139, 140, 171, 181, 190, and LaPolla, R.J. 2, 85, 96, 105, 106
194, 202, 237, 259 Vendler, Z. 105, 128, 149
Shibatani, M. 135
Singh, M. 154n Wechsler, S. 154n
Smalley, W.A. 11, 15, 16, 22, 23, 208, 209 Westermann, D. 70, 73, 105, 171, 278
Smith, C.S. 105, 106, 154n
Strecker, D. 233 Yang, D. 65
Yang, G.Y. 23
Tai, J.H.-Y. 154n, 160, 237 Yang, K.-L. 8
Talmy, L. 108, 154n
Index of Languages and Language Families
A-Hmao (West Hmongic, Hmong-Mien/ Kalam (Papuan) 77

Miao-Yao) 9 Khmer (Mon-Khmer, Austroasiatic) 43, 168,
Akan (Kwa) 70, 72, 125 185, 186, 189
Akuapem dialect of Akan (Kwa) 141 Kmhmu Cwang (Mon-Khmer, Austroasiatic)
Austroasiatic languages 9 168
Austronesian languages 9 Korean (Koreanic) 154n
Austro-Thai languages 9 Kuman (Papuan) 95, 96
Kwa languages 75
Barai (Papuan) 100
Belhare (Tibeto-Burman, Sino-Tibetan) Lahu (Tibeto-Burman, Sino-Tibetan) 73,
91n 189, 233
Blue Mong 8 Lao (Tai-Kadai) 10, 56, 66, 67, 68, 125, 154n,
see also Mong Njua 168, 169, 173
Bunu (West Hmongic, Hmong-Mien/ Lisu (Tibeto-Burman, Sino-Tibetan) 189
Miao-Yao) 9, 10
Mainland Southeast Asian languages 1, 76,
Cantonese (Sinitic, Sino-Tibetan) 82–83 82, 83
Chinese (Sinitic, Sino-Tibetan) 5, 26, 75, Mandarin 43, 68, 71, 73, 74, 75, 76, 95, 189,
154n, 160, 189, 199 190, 261
bǎ construction 177, 178, 183
Eastern Kayah Li (Tibeto-Burman, coverbs 194, 205
Sino-Tibetan) 71, 82, 128 resultative verb compounds 160–161, 183
English (Germanic, Indo-European) 73, 76, Miao (Hmongic) languages 9
77, 90, 91, 136, 149, 152n, 180, 266 Miao-Yao / Hmong-Mien languages 8, 9, 10
Ewe (Kwa) 70, 72, 99, 105, 171, 198, 278 Mong Leng 8
see also Mong Njua
Gitua (Oceanic, Austronesian) 141–142n Mong Njua (West Hmongic, Hmong-Mien/
Green Mong 8 Miao-Yao) 8, 11, 15
see also Mong Njua
Nepali (Indo-Aryan, Indo-European) 91n
Hindi (Indo-Aryan, Indo-European) 154n Niger-Congo languages 189
Hmong Dô (West Hmongic, Hmong-Mien/
Miao-Yao) 7 Oceanic languages 141n
Hmongic (Miao) languages 9
Hmong-Mien / Miao-Yao languages 8, 9, 10 Paamese (Oceanic, Austronesian) 71, 98,
Hmong Sua (West Hmongic, Hmong-Mien/ 125, 141, 142
Miao-Yao 7 Pa-Hng (Hmongic, Hmong-Mien/
Hmu (Hmongic, Hmong-Mien/ Miao-Yao) 10
Miao-Yao) 10
Ho Ne (Hmongic, Hmong-Mien/ Qo-Xiong (Hmongic, Hmong-Mien/
Miao-Yao) 10 Miao-Yao) 10
Japanese (Japonic) 154n Sinitic languages 9

Jiongnai (Hmongic, Hmong-Mien/ Southeast Asian languages 5, 70, 73, 189, 233
Miao-Yao) 10 Sranan (Carribean Creole) 100, 171, 172, 181
Index of Languages and Language Families 301
Tai-Kadai languages 9 Ulithian (Oceanic, Austronesian) 142n11

Tairora (Papuan) 77
Tamil (Dravidian) 154n West African Languages 75, 198
Tariana (Arawak) 71 West Hmongic languages 9
Thai (Tai-Kadai) 5, 43, 68, 71, 83, 98, 125, 126,
154n, 233, 247, 249 Yao (Mienic) languages 9
resultative constructions 127, 128, 129, 130 Yoruba (Benue-Congo) 98, 124, 171, 192
Tibeto-Burman languages 9, 189
Tok Pisin (Neo-Melanesian Creole) 77
Twi dialect of Akan (Kwa) 70, 72, 73, 171, 198
Index of Subjects
accomplishments 29, 105, 115, 149, 150, 153, change of location 129, 130, 132, 151, 158, 182,
154n, 155, 157, 158, 162, 184, 186, 280 212, 227
effective 153–156, 160, 182 of object 217, 221
projected 150n Change of Location verbs 129, 130, 132, 158,
see also activities, goal-oriented 182, 227
achievements 29, 128, 159, 160, 162, 166, 182, Change of Stance verbs 116, 253, 256, 259
212, 223, 259, 280 change of state (see also states) 46, 128, 129,
activities 182, 259 132, 133, 151, 158, 182, 212, 227
goal-oriented 149–152, 159, 160 classifiers 28, 31, 32, 33–5, 36–40, 180
see also goal, extrinsic collective 36, 37
Affective verbs 131, 132, 135, 182, 270, 279 mensural 33, 35
aktionsart see aspect, situation sortal 33, 34
allophones class nouns 26–28
of consonant phonemes 11, 12, 22 see also nominal prefixes
of vowel phonemes 13 clausal juncture 88, 89–92
ambitransitive verbs 111, 210 clause types 43–56
aspect 65, 87, 91, 192, 198, 223, 224 containing locative arguments 50–53
imperfective 67, 68 ditransitive 53–56
situation 103, 105 existential 43–45
viewpoint 106 intransitive 46–47
Attainment Serial Verb Constructions 116, nominal subject complements 45–46
147–170, 175, 182, 185, 225, 258 possessive 49–50
nature of goal in 159–160 reciprocal 47–48
type of verbs in 148–159 stative intransitive 46–47
comparison with Mandarin Complements transitive 48–53
of Result 160–161 Cognition verbs 49, 224–225
autonyms complementation 2, 3, 69, 83, 274
for Hmong sub-groups 7, 8n4 resembling Cause-Effect SVCs 241
for other Hmongic groups 10n9, see also complements
10nn11–13 complement clause serialization 128
complementizer kom 137, 242, 243, 244, 253
binding 3, 246, 274, 283, 285 Complement of Result Construction in
emotional attachment of subject to Mandarin 160–61
outcome 3, 252, 265 complements
scale 245, 252–53, 265 activity type 83
and binding 3, 245, 252, 266, 274
causation of Causative verbs 128, 246, 247
direct 131, 133, 135, 169, 182, 279 of implicative verbs 136
unintentional 134, 247 of Manipulative verbs 242
causatives 135, 242, 247 of Modal verbs 251, 253
Causative verbs 128, 135, 246, 247 of non-implicative verbs 242, 243, 245
Cause-Effect Serial Verb Constructions of Perception verbs 248, 249
124–147, 176, 182, 245–47, 266 of Permissive verbs 248, 250
type of causation in 134–139 of Purpose Constructions 260, 262
type of verbs in 127–134 of Speech Act verbs 220
subject in 139–147 as sub-propositional 82, 83, 84
without a complementizer 82, 83, 242, features, binary in RRG 92

243, 244, 245, 246, 248, 249, 250, 253, 274 future 59–61, 69
compound words 26, 27, 28
conceptual events 77, 78–79 Goal
core of Action 50
coordination 92, 99–100 arguments 56, 130, 179, 197, 209, 211, 264
cosubordination 91, 97, 99, 101, 123, 124, of Motion 41, 42, 43, 107, 108, 110, 111, 112,
139, 174, 175, 241 113, 114, 181, 183, 209, 220, 222
core juncture 88, 89, 90, 91, 92, 96, 97, 98, of Speech 220
99, 118, 120, 136–137, 173, 240, 284–85 temporal 210
Cotemporal Serial Verb Constructions 103, of Transfer 50, 52, 55, 217, 219, 220, 281
104, 106–24, 181, 203, 205, 213, 221, 234, verbs 50, 52, 110, 111, 197, 207, 218, 278
254, 256, 257–58, 278–79 facilitating interpretation of Source 51
involving action 85, 106, 115–118, 181, 236, goal
278, 282 intrinsic 114, 150, 182, 279, 280
involving motion 106–115, 152, 183, 184, attainment of 114, 147, 151–152, 160, 162,
254, 278 165, 182, 212, 222, 225, 258, 259, 280
coverbs 74, 188–96, 281 extrinsic 150, 153, 157, 175, 182, 280
Goal verbs 50, 52–53, 110, 111, 197, 207, 218,
Deictic Motion verbs 52, 206, 209, 213, 221, 278
222, 253, 254, 255, 257–59
deixis 41, 111, 112, 135, 181, 186, 213, 221, 222, illocutionary force 81, 87, 91, 92, 93, 199, 239,
254, 278, 280 277
deontic modality 62, 63, 81, 91, 93, 138, 139, imperative 56–57
166, 183, 235, 277, 280 intention 59, 60–61, 99, 134, 150, 265, 284
Disposal Serial Verb Constructions 170–181, lack of 150, 244, 247
183, 185, 217, 221, 231, 270, 271, 272, 273, interrogative 57–58, 64, 168
278, 280, 281 ISO-639 codes 7n1, 9nn6–7, 10n9, 10nn11–13
type of verbs in 171–3
shared object in 173–4 jokes and riddles in White Hmong 240, 267
juncture 86, 87, 88, 93, 98, 100, 101, 118, 119,
elaborate expressions 3, 117, 186, 233, 234, 160–161, 173, 174, 175, 239
235, 236–37, 241, 280, 282 in Attainment Serial Verb
elaboration of detail in serial verb Constructions 160, 164, 280
constructions 2, 3, 104, 278, 279, 280 in Cause-Effect Serial Verb
epistemic modality 60, 81, 87, 92, 93, 239, Constructions 136–137
277 clausal 88, 89–92
events core 88, 90, 100, 119, 138, 143, 173, 174, 239
conceptual 77, 78–79 in Cotemporal Serial Verb Constructions
distinct 78, 80, 96, 103, 114, 117, 120, 118, 119
162–163, 182, 237, 285 in Disposal Serial Verb Constructions
single 1, 2, 76, 78, 80, 81, 104, 116–17, 135, 174
181, 186, 187, 275, 276, 278 in juxtaposed clauses 234, 237, 239
event types 3, 78, 79, 275 in Mandarin Complement of Result
typical / recognisable 78, 79, 118 Constructions 160
exonyms for Hmong sub-groups 8 nuclear 136, 164, 170, 185
extent 108, 112, 214, 215, 223, 225, 231, 232, RRG theory of 86, 87, 93, 98, 284, 285
281, 282 in serial verb constructions 93, 100, 101,
Extent, vertical 212 175
304 Index of Subjects
lexical complexity 135, 186, 218, 280, 281 RRG theory of 86, 90
likelihood 3, 58n, 59, 60, 62, 252 subordinate 74, 75, 91–92, 122, 138, 261,
Location 41, 42–43, 50, 51, 52, 53, 76, 203, 262–63, 274, 277, 282–83
204, 206, 208, 231, 232, 281, 282 types 90, 101
Location verbs 50–51 nominal prefixes 23, 26, 27, 28, 31, 70
locative arguments 50, 51, 55, 119, 203, 264 see also class nouns
inner 50, 51, 52, 189, 203, 214, 264 Noun Phrase structure 31–33
outer 53, 120, 189, 203, 206, 208, 219, 232, nuclear juncture 87, 88, 90, 91, 93, 94, 118,
240, 264 119, 160, 164–65, 173, 174, 183, 190n, 280
Locative Phrases 20, 42, 50, 52, 53, 54, 109, coordination 95, 101
110, 111, 203, 204, 206, 208, 209, 264n cosubordination 95, 101, 160
Locomotion verbs 108, 135, 181, 206, 207, numerals 31, 32, 36
208, 209, 220, 280
distinguished from Transport verbs 206 operators 81, 87, 91, 92, 120, 138, 139, 162, 163,
164, 165, 166, 234, 235, 239
Macro-Event Property (MEP) 81, 98–99, 101 clause layer 91, 92, 100, 186, 199, 237, 239,
Manner of Motion 108, 109, 111, 112, 114, 183, 240, 276
278 core layer (deontic modal) 91, 97, 138,
measure words 33, 35 139, 175
MEP see Macro-Event Property orthography 2, 7, 11–15, 16, 23n
Modal verbs 30, 65, 167, 249, 250, 251, representation of tones in 14
252–53, 265
modality paired words see elaborate expressions
deontic 62, 63, 81, 91, 93, 138, 139, 166, 183, partitives 35
235, 277, 280 Path verbs 50, 109, 113, 114
epistemic 60, 81, 87, 92, 93, 239, 277 Perception verbs 248–50
mood 56, 57, 70, 87 phonology 11
morphotonemes 23, 26, 30 plurals, marked by collective classifier 37,
47, 48
negative 58–59, 64, 65, 67, 168 polysyllabic morphemes 26
irrealis negative 57, 58 possessor 31
negative imperative 57 prepositional-verbs 189, 195
scope of 234, 235, 240, 241, 260 see also coverbs
nexus 90, 92, 93, 96, 97, 98, 99, 100, 118, 120, Purpose Constructions 253–59, 261, 262,
164, 165, 166, 174, 175 263, 264, 265, 283
in Attainment Serial Verb and binding scale 265
Constructions 164 comparison with Mandarin 261
in Cause-Effect Serial Verb differences from serial verb
Constructions 137–139 constructions 261, 262, 263
coordinate 72, 73, 75, 90–92, 96, 100, 120, implication of result in 256
122, 138, 165, 166, 175, 261, 277 with muab ‘take’ 259
cosubordinate 90, 91, 92, 96, 97, 98, 99, similarities to serial verb
103, 120, 122, 138, 165, 175, 276, 277 constructions 264
in Cotemporal Serial Verb
Constructions 118 quantifiers 31, 32, 35
in Disposal Serial Verb Constructions non-numeric 32
174
embedded 175 Recipient 50, 54, 179, 180, 186, 220
non-embedded 92, 122, 138 reciprocal clauses 47–48
reduplication 23, 26, 28–30, 46, 47, 67, 68, status 12, 58, 62, 76, 87, 91, 128, 194, 210, 263,
152, 158, 192, 198, 280 266, 271, 274, 283, 285
riddles and jokes in White Hmong 240, 267 subject
Role and Reference Grammar (RRG) 2, 85, properties 142, 143, 144, 145, 146
86, 87, 88, 92, 93, 94, 162–163, 276 semantic 127, 140–41
revised theory of 96 single grammatical in SVCs 139, 147
Roman Popular Alphabet (RPA) 7, 11 syllables 14, 22, 23
RPA see Roman Popular Alphabet polysyllabic morphemes 26
serial verb types in White Hmong see telic / telicity 105, 111, 114, 115, 116, 150, 153,
Attainment Serial Verb Constructions, 154, 157, 254, 255, 256
Cause-Effect Serial Verb Constructions, tense 61, 65, 68, 70, 81, 87, 91, 93, 99, 106, 192,
Cotemporal Serial Verb Constructions, 277
Disposal Serial Verb Constructions tone 8, 11, 14, 15, 16–17, 20, 21, 22, 23–25, 30,
shared argument 81, 103, 124, 136–137, 140, 31, 40, 42, 227
141, 142, 173, 174, 176, 199, 239, 247, 249, auditory qualities of 11, 15, 16
279 co-occurrence with initials 16
in all Serial Verb Construction types 103 in loanwords from Chinese 16
in Cause-Effect Serial Verb morphological functions 23
Constructions 124, 125, 126, 136–137, in names for Hmong sub-groups 8
140, 141, 142, 144, 169, 246, 247, 279 orthographic representation of 14
not co-referential 136–137, 173, 174, 239 patterns in two-word expressives
in Disposal Serial Verb 24–25
Constructions 174, 176, 180 phonation types of 16
in juxtaposed clauses 239 phonemic 14, 15
in linked serial verb constructions 184, phonetic 15
185 and Spatial Deictics 20, 42
related to single event 81 unmarked 16
and valency-increasing serialization 199, tone change 16, 20
210, 231, 232, 282 morphotonemic alternation 16, 20, 21,
Source 41, 42–43, 50, 51, 52, 53–55, 107, 108, 40, 42
109, 111, 113, 130, 209, 214, 230 sandhi 16, 17–18, 20, 31
of motion 42, 43, 50, 51, 52, 107, 108, 109, Transfer verbs 50, 52, 54, 55, 108, 133, 217,
111, 112, 113, 119, 130, 206n11, 209 218, 219, 220, 221, 222, 279, 280, 281
of transfer 41, 50, 52, 53–55, 230 distinguished from Transport verbs
Source verbs 50, 51–52, 109, 110, 113, 114, 214, 108
230, 254 inherent Goal type 55, 217, 218
Spatial Deictics 20, 40–43, 52, 109, 110, 199, inherent Source type 52, 54
200, 209 physical vs. metaphorical 108
Speech verbs 55, 220, 224–225 Transport verbs 108, 111, 135, 181, 183, 206,
Stance verbs 50, 227, 228 207, 208, 209, 220, 278, 280
states (see also change of state) 46, 47, 50, distinguished from Locomotion
105, 128, 129, 133, 150n, 203, 205, 268, verbs 206n12
270, 279 distinguished from Transfer verbs 108
State verbs 129, 133, 178, 204, 236, 266, 267, truth-value 3, 57, 83, 126, 256, 257, 260, 283
268, 269 two-word expressives 24–25
ambiguous 267
used adverbially 267 Valency-increasing serial verb constructions
used attributively 266 188, 189, 191, 193, 195, 196, 197, 199, 201,
used depictively 268, 269 202–3, 205, 207, 209, 211, 213
306 Index of Subjects
verbal-prepositions 189 vowels

see also coverbs diphthongs 13, 14, 30
voice quality 14, 15 monophthongs 13, 14
breathy 15, 16, 17
creaky 15, 16 words 23, 26–30

Serial Verbs in White Hmong (PDFDrive)

Uploaded by

Copyright:

Available Formats

Serial Verbs in White Hmong (PDFDrive)

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

Serial Verbs in White Hmong (PDFDrive)

Uploaded by

Copyright:

Available Formats

Serial Verbs in White Hmong

Brill’s Studies in Language,

Alexandra Y. Aikhenvald (Cairns Institute, James Cook University)

The titles published in this series are listed at brill.com/bslc

Library of Congress Cataloging-in-Publication Data

Jarkey, Nerida, author.

Copyright 2015 by Koninklijke Brill nv, Leiden, The Netherlands.

This book is printed on acid-free paper.

1 Introduction to White Hmong Language 7

1.7.8 Deontic Modality with Postverbal tau 63

2 Introduction to Serial Verb Constructions 70

3 Productive Serial Verb Constructions in White Hmong 102

3.4 Disposal Serial Verb Constructions 170

4 Valency-Increasing SVCs and Grammaticalization 188

5 Superficially Similar Constructions 233

5.5.3 Syntactic Similarities Between SVCs and Purpose

Sources for White Hmong Examples 287

1 The consonant phonemes and their allophones I 12

About This Study

This study is a substantially revised version of work that originally began

© koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292390_002

constructions in the language, including juxtaposed clauses, Purpose Con-

About the Data Used for the Study

language, the vernacular spoken by our (generally) unschooled narra-

The problem of Hmong language consultants, particularly (though not only)

. . . the phenomena of NP zero anaphora and serialization of unmarked

She continues in a footnote:

Tony Diller (personal communication) tells of recording natural Thai

The problem of interpretation is particularly serious for the analytic (iso-

About the Glossing and Translation of Examples

Introduction to White Hmong Language

© koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292390_003

9 North Hmongic; classified as ‘Xiangxi Miao’ in Chinese terminology; includes ‘Western

Vietnam—and perhaps around 320,000 outside Asia—the majority in the

1.2 Phonology and Orthography

1.2.1 The Consonant Phonemes and their Allophones

1.2.1.1 Notes on the Realisation of Consonants

table 1 The consonant phonemes and their allophones I

Labial Apico-Dental Apico- Lamino-

Stops & Unaspirated /p/ /pl/ /t/ /tx/ /d/ /ts/

table 2 The consonant phonemes and their allophones II

Apico Post- Dorso Velar Uvular Glottal

Stops & Unaspirated /r/ /c/ /k/ /q/ [q] /ʔ/

1.2.2 The Vowel Phonemes and their Allophones

1.2.2.1 Notes on the Realisation of Vowels

/i/ /w/ /u/

/ai/ /aw/ /au/

1.2.3.1 Orthographic Representation of Tones

1.2.3.2 The Auditory Description of Tones in Citation Form

table 3 Tone symbols, pitch, voice quality and length

Symbol Pitch/Contour Brief Description

b 55 high level, modal, short

1.2.3.3 Notes on the Realisation of Tones

1.2.3.4 The Unmarked Tone

1.2.3.5 Tone Change

table 4 Tone sandhi changes

Tone of Citation Form Tone Following ‘b’ or ‘j’

high falling ‘j’

As shown in Table 4, these changes result in the neutralisation of three tones

© koninklijke brill nv, leiden, ��5 | doi ��.��63/9789004292390_002

© koninklijke brill nv, leiden, ��5 | doi ��.��63/9789004292390_003

9 North Hmongic; classified as ‘Xiangxi Miao’ in Chinese terminology; includes ‘Western