Complex Predicates

Complex predicates are multi-predicational, but monoclausal structures.

They have proven problematic for linguistic theory, particularly for pro-
posed distinctions between the lexicon, morphology, and syntax. This
volume focuses on the mapping from morphosyntactic structures to event
structure, and in particular, the constraints on possible mappings. The vol-
ume showcases the ‘coverb construction’, a complex predicate construction
which, though widespread, has received little attention in the literature.
The coverb construction contrasts with more familiar serial verb construc-
tions. The coverb construction generally maps only to event structures
such as those of monomorphemic verbs, whereas serial verb constructions
map to a range of event structures differing from those of monomorphemic
verbs. The volume coverage is truly cross-linguistic, including languages
from Australia, Papua New Guinea, Southeast Asia, the Indian subcontin-
ent, East Africa and North America. The volume establishes a new arena
of research in event structure, syntax, and cross-linguistic typology.

MENGISTU AMBER BER is a senior lecturer in Linguistics at the University

of New South Wales (Sydney, Australia). He is the co-editor (with
P. Collins) of Language Universals and Variation (), and (with
H. de Hoop) of Competition and Variation in Natural Languages: The Case
for Case (), and the editor of The Language of Memory in a Cross-
Linguistic Perspective ().

BRETT BAKER is a senior lecturer in Linguistics at the University of New

England (Armidale, Australia). He is the author of Word Structure in
Ngalakgan () and the co-editor (with Ilana Mushin) of Discourse and
Grammar in Australian Languages ().

MARK HARVEY is a senior lecturer in Linguistics at the University of

Newcastle (Australia). His previous publications include Proto Mirndi: A
Discontinous Language Family in Northern Australia (), A Grammar
of Gaagudju, a Language of the Alligator Rivers Region, Northern Territory
(), and A Grammar of Limilngan: A Language of the Mary River
Region, Northern Territory ().
Complex Predicates
Cross-linguistic Perspectives on
Event Structure

Edited by

MENGISTU AMBERBER , BRETT BAKER ,

AND MARK HARVEY
CAMBR IDGE UNIVERSITY PR ESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,
São Paulo, Delhi, Dubai, Tokyo
Cambridge University Press
The Edinburgh Building, Cambridge CB RU, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/

© Cambridge University Press 

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.

First published 

Printed in the United Kingdom at the University Press, Cambridge

A catalogue record for this publication is available from the British Library

Library of Congress Cataloguing in Publication data

Complex predicates : cross-linguistic perspectives on event structure /[edited by] Mengistu
Amberber, Brett Baker, Mark Harvey.
p. cm.
ISBN ---- (hardback)
. Grammar, Comparative and general–Verb phrase. . Grammar, Comparative and
general–Syntax. I. Amberber, Mengistu, – II. Baker, Brett J. (Brett Joseph),
– III. Harvey, Mark, – IV. Title.
P.C 
′.–dc 

ISBN ---- Hardback

Cambridge University Press has no responsibility for the persistence or

accuracy of URLs for external or third-party internet websites referred to
in this publication, and does not guarantee that any content on such
websites is, or will remain, accurate or appropriate.
 Contents

List of figures and tables page vi

List of contributors vii
Preface viii

 Introduction: complex predicates 

MENGISTU AMBERBER , BRETT BAKER , AND MARK HARVEY

 Complex predicate formation 

BRETT BAKER AND MARK HARVEY

 The light verb jungle: still hacking away 

MIRIAM BUTT

 Events and serial verb constructions 

WILLIAM A . FOLEY

 Cotemporal serial verb constructions in White Hmong 

NERIDA JARKEY

 Activity incorporates in some Athabaskan languages 

KEREN RICE

 Warlpiri verbs of change and causation: the thematic core 

MARY LAUGHREN

 Complex predicates in Wambaya: detaching predicate

composition from syntactic structure 
RACHEL NORDLINGER

 Compound verbs and ideophones in Wolaitta revisited 

AZEB AMHA

 The structure of the light verb construction in Amharic 

MENGISTU AMBERBER

Index 

v
 List of figures and tables

Figure . Types of LCS-syntax relations page 

Figure . Gentner and Boroditsky’s Division of Dominance 
Figure . Morphosyntactic structure of the Warlpiri inflected
verbal constituent 
Figure . Event structure of stative predicate 
Figure . Event structure of ‘change’ predicate 
Figure . Internally controlled cause change event 
Figure . Externally caused change event 
Figure . Copula verb and nominal predicate 

Table . Ngan’gityemerri finite verb inventory 

Table . Slave classificatory verbs 
Table . Coindexation of arguments in SVC types 
Table . Verb types and ordering in cotemporal motion SVCs 
Table . Stance and spatial configuration verbs 
Table . CHANGE of state and CAUSE change of state
dependent verbs 
Table . Converb markers in Wolaitta 
Table . V verbs in Wolaitta in Adams 
Table . Asymmetrical V verbs in Wolaitta 
Table . Symmetrical set in Wolaitta 
Table . Transitivity of the light verb construction 
Table . Coverbs derived from activity intransitives 

vi
 Contributors

MENGISTU AMBERBER School of Languages and Linguistics, The University

of New South Wales, Sydney, . Australia
AZEB AMHA Leiden University Centre for Linguistics, Leiden University,
van Wijkplaats , P.O. Box ,  RA Leiden. The Netherlands
BRETT BAKER School of Behavioural, Cognitive, and Social Sciences,
University of New England, Armidale, NSW . Australia
MIRIAM BUTT Fachgruppe Sprachwissenschaft, Universität Konstanz, Fach
D, D- Konstanz, Germany
WILLIAM FOLEY Department of Linguistics, School of Letters, Arts and
Media, University of Sydney, NSW, . Australia
MARK HARVEY School of Humanities and Social Science, The University of
Newcastle, Callaghan , NSW. Australia
NERIDA JARKEY Department of Japanese, School of Languages and Cultures,
University of Sydney, NSW, . Australia
MARY LAUGHREN Linguistics Program, School of English, Media Studies
and Art History, The University of Queensland, . Australia
RACHEL NORDLINGER School of Languages and Linguistics, University of
Melbourne, VIC . Australia
KEREN RICE Department of Linguistics, University of Toronto,  St.
George Street, Toronto, Ont. MS H. Canada

vii
 Preface

Most of the contributions to this volume are based on papers presented at the
Complex Predication and the Coverb Construction session held in Brisbane
(Queensland, Australia),  July , as part of the Australian Linguistic
Society annual conference.
We wish to thank all the contributors for their support and patience. We
would also like to thank the anonymous reviewers who assisted in review-
ing the individual contributions. Many thanks also to the three anonymous
reviewers commissioned by Cambridge University Press for their crit-
ical feedback on the prospectus for the volume. Special thanks are due to
Helen Barton, Commissioning Editor (Linguistics and Anthropology) at
Cambridge University Press, for guidance and encouragement.
We gratefully acknowledge the financial support from the Australian
Research Council, Discovery Grant (DP).
M. Amberber, B. Baker, M. Harvey
March, 

viii
1 Introduction

MENGISTU AMBERBER , BRETT BAKER, AND

MARK HARVEY

What is a complex predicate? There is currently no widely accepted answer

to this question, no agreed set of criteria which allow an analyst to classify
Construction A as a ‘complex predicate’, and Construction B as ‘not a com-
plex predicate’. This volume does not pretend to offer the final definitive
answer to this basic question, but it does aim to further delimit the range of
possible answers.
The volume does this in two ways. First, it provides detailed data on con-
structions usually classified as ‘complex predicates’ in a range of languages
from Australia, East Africa, Papua, South and Southeast Asia, and North
America. In particular, it provides detailed data on a hitherto little described
construction – the coverb construction.
Coverb constructions are common among Australian, East African,
Iranian, and Oceanic languages. The construction involves two constitu-
ents: a coverb and a verb. Coverbs must be analysed as a distinct part-
of-speech class (Amberber, Baker, and Harvey ). They share some
characteristics with verbs – they are inherently predicational and they are
not derived from any other part of speech. However, they differ from verbs
in being inherently non-finite.
The volume also aims to delimit the range of possible answers by pro-
viding a detailed examination of the mapping between complex predicates
of various types and event structure, in the sense of Rappaport Hovav and
Levin (). This is a central focus for all of the papers in the volume. This
mapping has not previously been as prominent a focus of research.
The seminal work in the field of complex predicates is Alsina, Bresnan,
and Sells () – a collection of papers which covers a range of complex
predicate constructions including resultatives, V-V compounds, preverb +
V combinations, Noun + light verb constructions, noun incorporation, and
the way-construction in English, among others. At the outset, the complex
predicate is defined as ‘predicates which are multi-headed; they are com-
posed of more than one grammatical element (either morphemes or words),
each of which contributes part of the information ordinarily associated with
a head’ (Alsina et al. : ).

1
2 Introduction

Since Alsina et al. (), works such as Ackerman and Webelhuth (),
Andrews and Manning (), Hale and Keyser (), and Hinrichs et al.
() have contributed greatly to theoretical research in the field. There
have also been a number of important works on individual languages or
language groups: Bowern (), Butt (), McGregor (), Schultze-
Berndt (), and Wilson (). This volume brings together and further
develops empirical and theoretical analyses and questions from this body of
literature. The volume also proposes new lines of analysis.
Any analysis of ‘complex predicate’ naturally turns on the analysis of the
term ‘predicate’. This term has a pedigree going back to ancient and modern
logic, but in linguistics it refers to the ‘part of a clause or sentence tradi-
tionally seen as representing what is said of, or predicated of, the subject’
(Matthews : ). The subject is usually regarded as more definite and
determined than the predicate ( Jespersen : ). In formal semantics,
the standard Fregean view is that a predicate is an ‘unsaturated expression’
and must combine with an entity in order to form a proposition (Bowers
).
A predicate is most typically realised by a verb phrase and ‘combines with
the subject NP to make up the complete sentence’ (Trask : ). In John
bought a book, the predicate is the whole VP ‘bought a book’ and not only
the verb ‘bought’. This does not mean that the predicate ‘bought a book’
is more complex than the predicate ‘went’ in John went. A predicate is said
to be complex with reference to the head of the predicate. The head of the
predicate is normally an X category, whereas a complex predicate is multi-
headed. Thus, the term ‘complex predicate’ properly construed is shorthand
for complex predicate head.
Given this, it is necessary to consider what counts as the head of a predi-
cate. We may consider the English constructions (a) I will walk and (b)
I walk. Is it the case that the predicate in (a) but not in (b) is multi-headed?
This type of periphrastic construction involving an auxiliary and main verb
has been analysed as a complex predicate in the literature (Müller ).
However, it stands to reason that if ‘will walk’ is treated as a kind of complex
predicate, then ‘walk-ed’ should also be analysed in the same way. The fact
that tense is marked by an inflectional affix in ‘walk (PAST)’ but periphrasti-
cally in ‘walk (FUTURE)’ does not have any deep syntactic consequences. For
the purpose of delimiting the object of inquiry, we consider only elements of
the multi-headed predicate that make a significant lexical–semantic contri-
bution including, in particular, information that is relevant to determining
the argument structure of a clause.
It is important to note here that at a sufficiently abstract level of analysis,
every predicate can be treated as complex. In the Generative Semantics of
the seventies virtually every lexical verb, including monomorphemic ones,
was analysed as a complex form at Deep Structure (Lakoff ). Famously,
for example, the English verb ‘kill’ was analysed as ‘cause to become dead’.
 Amberber, Baker, and Harvey 3

Even though the identification of semantics with Deep Structure was

gradually abandoned as a viable research programme in mainstream genera-
tive grammar (Fodor ), some key elements of the programme, including
its decompositional approach, have been incorporated into many contem-
porary studies, including the influential works of Baker (), Jackendoff
(), Pustejovsky (), Hale and Keyser (, ), Rappaport Hovav
and Levin (), the neo-Lakovian approach of Pesetsky (), and a
number of neo-constructionist approaches such as Marantz (), Borer
(a,b), and Ramchand (), just to name a few. It can also be seen in
the widespread practice of decomposing verbs into syntactically transparent
but abstract entities such as voice (Kratzer ), and (little) v in most stud-
ies within the Minimalist Program of Chomsky ().
Restricting ourselves to multi-headed forms where the heads make an
intuitively non-trivial lexical–semantic contribution, we find that the term
‘complex predicate’ is commonly used to refer to a variety of constructions
including: periphrastic causatives (Mary made him go to the shop), verb par-
ticle combinations (the child picked it up), resultatives (they hammered the
iron flat), consider + predicate combinations (she considers him (to be) intel-
ligent), and restructuring constructions (typical of infinitival constructions
in Romance).
The extent to which any or all of these constructions can be analysed as
involving complex heads is not always clear, and the formal analyses of the
constructions are still controversial. For example, Wurmbrand () has
specifically argued that German Clause Union/restructuring constructions
do not involve complex V-V heads (as assumed by many scholars), but rather
are derived through regular VP complementation.
The enduring theoretical interest in complex predicates is undoubtedly
due to the fact that in some aspects they pattern with prototypical words,
whereas in other aspects they pattern with prototypical phrases. Complex
predicates exhibit word-like properties in terms of argument structure com-
position and in sometimes having lexicalised meanings. They exhibit phrase-
like properties in allowing certain syntactic operations, such as movement,
to manipulate their internal structure.
This presents a major challenge to grammatical theory, particularly if the
lexicon is formally segregated from syntax, as enshrined by various princi-
ples such as the Lexical Integrity Hypothesis (LIH). The LIH assumes that
‘no syntactic rule can refer to elements of morphological structure’ (Lapointe
), and that ‘words are built out of different structural elements and by
different principles of composition than syntactic phrases’ (Bresnan and
Mchombo : ).
The LIH has been explicitly rejected in a number of frameworks (Marantz
, Borer a,b). There is also a range of different interpretations of the
LIH which attempt to curtail its domain of application in various ways – for
example by prohibiting the direct syntactic manipulation of word-internal
4 Introduction

(X) categories, but nevertheless allowing the syntax to have some access
to sub-lexical features. Two leading morphologists have aptly summarised
the current state of play in this domain as follows: ‘we have available to us
not only multiple theories of syntax to consider, but also multiple theories of
word formation. It is impossible to reassess the LIH without considering a
multitude of possibilities’ (Lieber and Scalise ).
Whatever the ultimate fate of the LIH, the contributions to this volume
support the view that its fundamental insight will need to be incorporated
into linguistic theory. This insight is that there is no single mapping pro-
cedure that will proceed from any system of structural analysis to reliably
produce the full set of predicate meanings that are associated with the total
range of monoclausal structures found in human language. Whatever sys-
tem of structural analysis is adopted, it appears that it will be necessary to
posit at least two mapping procedures in order to account for the full set of
predicate meanings that can be associated with monoclausal structures. One
mapping procedure derives ‘word-like’ meanings, and the other procedure
derives ‘phrase-like’ meanings.

The contributions
Brett Baker and Mark Harvey argue in Chapter  that complex predicates
fall into two main classes in terms of their event structure configuration. The
key claim is that some complex predicates (for example, coverb construc-
tions in many Australian languages) involve the merging of argument struc-
ture at the level of a Jackendovian-type Lexical Conceptual Structure (LCS),
( Jackendoff , ). Such predicates are shown to have the semantic and
morphosyntactic range of monomorphemic verbs in more familiar languages
such as English. On the other hand, other complex predicates (for example,
serial verb constructions (SVCs)) involve the coindexation of argument
structure at LCS. Coindexation allows for a wider range of event structures,
including many structures which cannot be expressed by monomorphemic
verbs.
By appealing to a richly articulated level of conceptual structure and the
independently motivated tools of ‘merger’ and ‘coindexation’, Baker and
Harvey provide a detailed analysis of coverb constructions, tying together
a number of descriptive generalisations (for example, verb serialisation may
enable the ‘raising’ of non-subcategorised arguments, something which co-
verb constructions never do) within a single overarching framework.
While the insights behind the notions of ‘merger’ and ‘coindexation’ have
reflexes in other frameworks (for example, the processes of Event Fusion and
Argument Fusion in Lexical-Functional Grammar (LFG) parallel ‘merger’
and ‘coindexation’, as pointed out in Nordlinger (this volume)), Baker and
Harvey show how a variety of empirical consequences fall out from the
application of these two simple operations.
 Amberber, Baker, and Harvey 5

In their analysis of coverb constructions, Baker and Harvey take mono-

clausality as one of the defining properties of complex predicates. That is,
while complex predicates have two or more heads, these heads function as
a single predicate in a monoclausal configuration. This property of complex
predication plays a central role in the analysis of light verbs and complex
predication by Miriam Butt in Chapter  (this volume).
While the empirical focus of Butt’s study is Urdu, the main purpose of
her analysis is to provide a novel and cross-linguistically valid understand-
ing of light verbs with particular reference to the relationship between light
verbs and complex predicates. Butt identifies a number of salient properties
of light verbs: (a) they are form identical with a main verb, (b) they have
a marked morphosyntactic behaviour distinguishable from both auxiliaries
and main verbs, and (c) they serve to modify the event encoded by the main
predicator in a way that is different from other types of verbs (including aux-
iliaries, modals or main verbs).
These and other distributional factors strongly suggest that light verbs
have a semi-lexical status (neither fully lexical nor fully functional) and that
they should be treated as a separate syntactic class. Of course, this conclu-
sion, if true, has non-trivial consequences not only to the traditional typol-
ogy of part-of-speech categories, but also to all current models of grammar
where the light verb carries out a central theoretical function (as in the
Minimalist Program and Distributed Morphology, among others).
Although many studies – both within the formal-generative and the
typological-functionalist traditions – assume that monoclausality is a cru-
cial property of complex predicates, it does not mean that the notion itself
is always conceptually clear. The use of monoclausality and other notions
such as ‘event’ (single vs. multiple eventhood) in the analysis of complex
predicates in general and serial verb constructions in particular is sharply
critiqued by William Foley in Chapter . Foley starts his discussion with
Aikhenvald’s () definition of SVCs, according to which the sequence of
verbs in an SVC occurs within a single clause, and the verbs are interpreted
as expressing a ‘single event’.
On the basis of data from Yimas and other Papuan languages, Foley ques-
tions the extent to which any expression in an SVC constitutes a single event.
For example, what does it mean to say that an expression such as ak-mpi-
wul ‘push down (into the water)’ in Yimas encodes a single event? In other
words, how can we distinguish a single event from multiple events on neces-
sary and sufficient grounds? Foley argues that the number of ‘events’ cannot
be determined in relation to a criterion of ‘simplicity’, however this criter-
ion is to be constructed. The event encoded by the aforementioned verb in
Yimas is patently not simple (typically requiring multiple agents, a complex
path, and figure-ground configuration).
Foley invites us to reconceptualise the notion of eventhood (and indeed
other related terms) within a wider, and arguably more cognitively salient,
6 Introduction

notion of linguistic organisation. For this purpose, Foley appeals to the

notion of Division of Dominance as developed by Gentner and Boroditsky
() in the domain of early word learning. The idea is that there are two
types of principles – cognitive dominance and linguistic dominance – that
guide children in their acquisition of lexical meaning. By virtue of a person’s
perceptual engagement with the world, linguistic units can be used to label
certain items in a rather straightforward manner (cognitive dominance).
This is how the meaning of concrete nouns such as the word dog is acquired.
On the other hand, the meaning of a word such as the English although does
not fall out from a person’s perceptual engagement with the world, thus must
be learned on a language-particular basis (linguistic dominance).
If this is on the right track, the prediction is that there is a fundamen-
tal asymmetry between the distribution of nouns and verbs: all languages
will have a linguistic unit (root) for concrete entities – ‘dog’, for example –
whereas a lot more cross-linguistic variation would be expected with respect
to verbal concepts such as ‘kill’. Foley then goes on to show how this asym-
metry plays out in the formation of SVCs, demonstrating that the term SVC
actually refers to a heterogeneous class of predicates both within and across
languages.
The issue of eventhood and SVCs is also examined in Chapter  by Nerida
Jarkey, who examines the SVCs of White Hmong (spoken in Mainland
Southeast Asia). Four distinct types of SVCs – referred to as Cotemporal,
Disposal, Pivotal, and Attainment – are identified on the basis of two formal
criteria: the coindexation relations between the predicates and the relative
order of predicates and arguments.
The focus of Jarkey’s chapter is the Cotemporal SVC that is character-
ised by the coindexation of the subject argument of all verbs in the serial
complex. Thus, in the White Hmong equivalent of ‘The Hmong crossed the
Mekong River, escaping Laos and going to Thailand’, the subjects of the
three motion verbs cross, escape and go are coindexed. Jarkey shows that the
main function of the Cotemporal SVCs is to focus attention on and highlight
what the subject argument does and how s/he does it, rather than simply
encoding the event as motion, state or action. As such, their function is simi-
lar to adverbial expressions in portraying a particular action in a vivid and
dynamic manner, as Jarkey demonstrates.
The fundamental question of the nature of the potential mappings
between structure and events is brought into particular focus in Chapter 
by Keren Rice. Working on the Athapaskan languages, which are famous
for their complex morphological structure, Rice examines a construction
which she terms the activity incorporate construction, which has the mean-
ing ‘do X while Y-ing’. This construction differs formally from those exam-
ined elsewhere in this volume, in that it is a noun incorporation structure,
with X being a verb and Y being an incorporated noun. The events speci-
fied by X and Y occur simultaneously and have the same subject. Thus, for
 Amberber, Baker, and Harvey 7

example, in the Koyukon language the equivalent of ‘they are going along
shouting’ involves the incorporation of a noun ‘shout’ into a verbal complex
based on the ‘go’ verb root.
Rice argues that the activity incorporate construction exhibits properties
that suggest that it is a kind of complex predicate. These properties include:
(a) argument structure (the two predicational elements have the same sub-
ject within a monoclausal structure); (b) aspect (there is a single marking
for aspect); and (c) phonological status (the verb word behaves as a single
phonological unit). Further, she notes that in terms of its predicate struc-
ture, this noun incorporation structure classes with SVCs.
By contrast with SVCs and noun incorporation, the coverb construction
is relatively understudied. This neglect is most obvious in formal theoretical
linguistics, notwithstanding some earlier seminal studies on this phenom-
enon (for example, Nash , Hale  and Simpson , among others).
Therefore, a closer investigation of the coverb construction deploying for-
mal linguistic methodologies is important both for understanding the coverb
construction in its own right, and also for providing evidence that bears on
some current issues actively pursued in formal syntactic theories. This is
what Mary Laughren does in Chapter  on the Warlpiri verbal complex.
Her empirically rich and formally fine-grained analysis provides a bet-
ter understanding of the Warlpiri verbal complex, particularly those involv-
ing preverbs (= coverb). This has implications for some broader theoretical
questions about the organisation of grammar. In terms of the current debate
between so-called ‘lexicalist’ and ‘syntactocentric’ approaches to word for-
mation, Laughren defends the view that the Warlpiri data is best accounted
for if the grammar has a level of syntax (S-Syntax) which is formally distinct
from lexical syntax (L-syntax) in the sense of Hale and Keyser (, )
and much related work.
While Laughren’s discussion assumes familiarity with recent studies of
event structure within formal generative theories (Travis , Folli, Harley
and Karimi , among others), the descriptive insights are easily accessi-
ble to anyone familiar with the broad terrain of work on predicate argument
structure in the past twenty-five years (see Levin and Rappaport Hovav
() for an excellent review).
Laughren argues that the verbal constituent in Warlpiri has a complex
structure. She proposes an inner ‘thematic core’ which must contain a verb
and may additionally contain a preverbal element drawn from a particular
set, and an outer periphery consisting of ‘outer’ preverbal constituents that
modify the thematic core in largely productive ways. The mapping between
morphological form and the inner thematic core of a preverb construction
is not necessarily one-to-one. Thus, very closely related predicates may be
expressed by different structures, such as wanti- and para-karri: while both
roughly have the same meaning – ‘fall’ – the former is a simplex V, whereas
the latter is a PV-V complex.
8 Introduction

Throughout her discussion, Laughren shows that the morphosyntactic

and semantic properties of complex predicates in Warlpiri exhibit impor-
tant similarities to complex predicates in other languages such as Persian.
Although there are a number of unresolved issues (for example, the role of
some verbalisers in the derivation of complex predicates), Laughren’s study
provides a model for the kind of work that needs to be carried out if our
understanding of this fascinating construction is to be advanced.
The formation of complex predicates in another Australian language,
Wambaya, is the focus of Chapter  by Rachel Nordlinger. The two
Wambaya constructions investigated in detail are: (a) the associated motion
construction, which is analogous to coverb constructions in other languages;
and (b) a serial verb construction (involving the combination of two lexical
verbs). In the associated motion construction an obligatory lexical verb is
combined with a ‘directional marker’ (‘go/away’ or ‘come/towards’). This is
an auxiliary, which is inflected for person agreement and tense/aspect.
The semantic contribution of the directional auxiliary is not fixed, but
rather depends on the nature of the main verb. When the main verb encodes
a motion event, the directional auxiliary specifies the direction of motion.
Motion verbs in Wambaya have no specification for direction of motion.
On the other hand, when the main verb encodes a non-motion event, the
direction affix marks a sequential motion event, resulting in two sub-events.
Likewise, in the motion serial verb construction which involves the verb
yarru ‘go’, the contribution of the motion verb yarru depends on the nature
of the verb it combines with. Thus, when it combines with a motion verb it
specifies concurrent motion (‘go sneaking’ vs. ‘sneaking’), and with a non-
motion verb it indicates sequential motion (‘go and (then) swim’).
Hence, while the associated motion construction and the yarru seriali-
sation are formally distinct, they are semantically identical with respect to
the constraints on their contribution to Predicate Composition in complex
predicates. Nordlinger argues that this co-occurrence of formal distinc-
tion with semantic identity is evidence that the theoretical representation
of syntax should be distinct from the theoretical representation of Predicate
Composition, as in theories such as Lexical-Functional Grammar.
As previously stated, the coverb construction is not limited to Australian
languages. This construction is also found in a number of Ethiopian lan-
guages belonging to the Cushitic, Omotic, and Semitic families, where it is
known in the literature as a compound verb or composite verb. The compound
verb construction in these languages involves two predicative elements,
a closed class of inflecting verbs and an open class of coverbs (also known
as ‘converbs’). The most productive inflecting verbs that occur in the com-
pound verb construction are roughly equivalent to the English verbs say and
do/make – often contrasting in transitivity. This construction is the focus of
the last two contributions of the volume. Chapter  by Azeb Amha focuses
 Amberber, Baker, and Harvey 9

on the Omotic language Wolaitta and Chapter  by Mengistu Amberber

investigates the Ethio-Semitic language Amharic.
Amha shows that the compound verb construction in Wolaitta that
involves two predicative entities (V + V) allows a wider range of V verbs
(about ten) compared to other languages of the area. While bearing verbal
inflection is not a defining formal criterion of the V verbs, tense–aspect
and mood are typically marked on the V. Importantly, Amha argues that,
contrary to previous studies on Wolaitta, the V verbs should not be iden-
tified simply as an auxiliary verb. She provides two main reasons for this.
First, treating the V verbs as ‘auxiliary’ in the compound construction but
as ‘lexical’ elsewhere is unmotivated, as the verbs have exactly the same form
and distribution. Second, the V verbs do not simply mark tense, aspect, and
modality, but have argument structure and make a clear lexical semantic
contribution to the complex predicate.
Amberber makes the same point for Amharic – the V component is not
merely a marker of tense–aspect–mood features in the compound verb con-
struction. There is considerable overlap between the compound verb con-
struction of Wolaitta and what Amberber refers to as the light verb construction
in Amharic. Nevertheless, there are also some important differences.
In Amharic, as in other Semitic languages, all inherently predicational
word forms, whether finite or non-finite, are derived from a consonantal
root. This consonantal root cannot itself appear as a word. There is an exten-
sive range of derivational structures, known as binyan. The coverb is formed
by derivation of the root into one of these binyan. In most languages, the
coverb binyan is closed, and new coverbs cannot be freely created. However,
in Amharic, and to some extent in Tigrinya and Qafar, the coverb binyan is
open. Virtually any inherently predicational root in the language can derive
a coverb binyan.
With regard to the compound verb construction in Amharic, Amberber
argues that the light verbs have the same function as valency-encoding deri-
vational affixes (causative, inchoative, passive-reflexive). He argues that both
the light verb construction (coverb binyan + light verb) and directly inflected
verbal binyan can be analysed in terms of a single morphosyntactic struc-
ture, which he presents within the framework of Distributed Morphology
(Halle and Marantz () and subsequent work).

The contributions to this volume demonstrate that empirical and theoretical

research proceed most profitably in tandem. Ultimately, it is possible that
the various formal mechanisms employed in the contributions may turn out
to be notational variants. Nevertheless, the contributions show that the exer-
cise is important in itself, as each theory forces us to unearth and zero in on
a set of facts that might otherwise be left in the background in competing
frameworks.
10 Introduction

The contributions obviously do not cover all issues pertinent to the

understanding of complex predicates and coverb constructions. However,
they present not only a range of empirically rich analyses of data from dif-
ferent languages, but also an interesting array of theoretical perspectives on
complex predicates with important implications for current debates on the
syntax–semantics interface.

References
Ackerman, F. and Webelhuth, G. . A Theory of Predicates. Stanford, CA: CSLI
Publications.
Aikhenvald, A. . Serial Verb Constructions in Typological Perspective. In
Aikhenvald, A. and Dixon, R. (eds) Serial Verb Constructions: A Crosslinguistic
Typology, –. Oxford: Oxford University Press.
Alsina, A., Bresnan, J., and Sells, P. (eds), . Complex Predicates. Stanford, CA:
CSLI Publications.
Amberber, M., Baker, B., and Harvey, M. . Complex Predication and the
Coverb Construction. In Siegel, J., Lynch, J., and Eades, D. (eds) Language
Description, History and Development. Linguistic Indulgence in Memory of
Terry Crowley, –. Amsterdam: John Benjamins.
Andrews, A. and Manning, C. . Complex Predicates and Information Spreading
in LFG. Stanford, CA: CSLI Publications.
Baker, M. . Incorporation: A Theory of Grammatical Function Changing.
Chicago: University of Chicago Press.
Borer, H. a. In Name Only: Structuring Sense, vol I. Oxford: Oxford University
Press.
b. The Normal Course of Events, vol II. Oxford: Oxford University Press.
Bowern, C. . Bardi Verb Morphology in Historical Perspective. PhD disserta-
tion, Cambridge, MA: Harvard University.
Bowers, J. . Predication. In Baltin, M. and Collins, C. (eds) The Handbook of
Contemporary Syntactic Theory, –. Blackwell: Oxford.
Bresnan, J. W. and Mchombo, S. A. . The Lexical Integrity Principle: Evidence
From Bantu. Natural Language and Linguistic Theory : –.
Butt, M. . The Structure of Complex Predicates. Stanford, CA: CSLI.
Chomsky, N. . The Minimalist Program. Cambridge, MA: MIT Press.
Fodor, J. . Three Reasons for not Deriving ‘Kill’ from ‘Cause to Die’. Linguistic
Inquiry :  –.
Folli, R., Harley, H., and Karimi, S. . Determinants of Event Structure in
Persian Complex Predicates. Lingua , : –.
Gentner, D. and Boroditsky, L. . Individuation, Relativity and Early Word
Learning. In Bowerman, M. and Levinson, S. (eds) Language Acquisition and
Conceptual Development, –. Cambridge University Press.
Hale, K. L. . Warlpiri and the Grammar of Non-configurational Languages.
Natural Language & Linguistic Theory .: –.
Hale, K. L. and Keyser, S. J. . On Argument Structure and the Lexical
Expression of Syntactic Relations. In Hale, K. and Keyser, S. J. (eds) The View
 Amberber, Baker, and Harvey 11

From Building : Essays in Linguistics in Honor of Sylvain Bromberger, –.

Cambridge, MA: MIT Press.
. Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT
Press.
Halle, M. and Marantz, A. . Distributed Morphology and the Pieces of
Inflection. In Hale, K. and Keyser, S. J. (eds) The View from Building : Essays
in Linguistics in Honor of Sylvain Bromberger, –. Cambridge: MIT Press.
Hinrichs, E., Kathol, A., and Nakazawa, T. (eds.) . Complex Predicates in
Nonderivational Syntax. San Diego: Academic Press.
Jackendoff, R. . Semantic Structure. Cambridge, MA: MIT Press.
. Foundations of Language. Oxford: Oxford University Press.
Jespersen, O.  []. The Philosophy of Grammar. New York: Norton.
Kratzer, A. . Severing the External Argument From its Verb. In Rooryck, J. and
Zaring, L. (eds) Phrase Structure and the Lexicon, –. Dordrecht: Kluwer.
Lakoff, G. . Irregularity in Syntax. New York: Holt, Rinehart, and Winston.
Lapointe, S. . A Theory of Grammatical Agreement. PhD dissertation,
Amherst: University of Massachusetts.
Levin, B. and Rappaport Hovav, M. . Argument Realization. Cambridge:
Cambridge University Press.
Lieber, R. and Scalise, S. . The Lexical Integrity Hypothesis in a New
Theoretical Universe. Lingue e Linguaggio , –.
Marantz, A. . No Escape From Syntax: Don’t Try Morphological
Analysis in the Privacy of Your Own Lexicon. Penn Working Papers in
Linguistics :: Proceedings of the st Annual Penn Linguistics Colloquium,
Dimitriadis, A., Siegel, L., Surek-Clark, C., and Williams, A. (eds), –.
Philadelphia: University of Pennsylvania.
Matthews, P. H. . Oxford Concise Dictionary of Linguistics. Oxford: Oxford
University Press.
McGregor, W. . Verb Classification in Australian Languages. Berlin: Mouton de
Gruyter.
Müller, S. . Complex Predicates. In Brown, K. (ed.) Encyclopaedia of Language
and Linguistics, nd edn., –. Elsevier: Oxford.
Nash, D. . Warlpiri preverbs and verb roots. In Swartz, S. (ed.) Papers in
Warlpiri Grammar: In Memory of Lothar Jagst. Work Papers of SIL-AAB,
Series A vol , –. Berrimah, NT: SIL-AAB.
Pesetsky, D. . Zero Syntax: Experiencers and Cascades. Cambridge, MA: MIT
Press.
Pustejovsky, J. . The Syntax of Event Structure. Cognition : –.
Ramchand, G, . Verb Meaning and the Lexicon. Cambridge: Cambridge
University Press.
Rappaport Hovav, M. and Levin, B. . Building Verb Meanings. In Butt, M. and
Geuder, W. (eds) The Projection of Arguments, –. Stanford, CA: CSLI.
Schultze-Berndt, E. . Simple and Complex Verbs in Jaminjung: A Study of
Event Categorization in an Australian Language. PhD dissertation, Catholic
University Nijmegen, The Netherlands.
Simpson, J. . Warlpiri Morpho-Syntax. A Lexicalist Approach. Dordrecht/
Boston/London: Kluwer Academic Publishers.
12 Introduction

Trask, L. . A Dictionary of Grammatical Terms. London: Routledge.

Travis, L. . Agents and Causes in Malagasy and Tagalog. In Erteschik-Shir, N.
and Rapoport, T. (eds) The Syntax of Aspect: Deriving Thematic and Aspectual
Interpretation, –. New York: Oxford University Press.
Wilson, S. . Coverbs and Complex Predicates in Wagiman. Stanford, CA: CSLI
Publications.
Wurmbrand, S. . How Complex are Complex Predicates ? Syntax : –.
2 Complex predicate formation*

BRETT BAKER AND MARK HARVEY

. Introduction
The term ‘complex predicate’ has a wide usage, including, for example,
serial verb constructions, light verb constructions, and particle + verb con-
structions, among others. An examination of the data provided by analysts
in their discussions of complex predicates shows that monoclausality is the
critical factor in determining whether a construction involves a complex
predicate or not. Complex predicates are monoclausal structures involving
two or more predicating morphemes. Butt (this volume) is explicit on this
point.
[T]he term complex predicate is used to designate a construction that
involves two or more predicational elements (such as nouns, verbs, and
adjectives) which predicate as a single element, i.e. their arguments map
onto a monoclausal syntactic structure.
We show that monoclausality as a criterion does not determine a unitary
set of predicate structures. Rather, we show that there are two quite distinct
ways of combining predicate information within monoclausal structures. We
call one method ‘merger’ because the predicate information from the con-
tributing constituents merges where they have common conceptual struc-
ture. This method produces predicate structures whose range classes with
the range of predicate structures found in monomorphemic predicates. We
propose that there are constraints on the conceptual structure of monomor-
phemic predicates which also apply to merger constructions. We discuss the
constraints on monomorphemic predicates in Section ..
We call the other method ‘coindexation’ because relations among the
contributing predicates are constrained only by a requirement that some of
their arguments must be coindexed. This method produces multi-predicate

* We would sincerely like to acknowledge the important contribution of the following schol-
ars to the ideas developed in this paper, even though they would not necessarily agree with
our arguments: Mengistu Amberber, John Beavers, Bill Foley, Cliff Goddard, Andrew
Koontz-Garboden, Beth Levin, Rachel Nordlinger, Nick Reid, Eva Schultze-Berndt,
and audiences at Manchester, USC, Nijmegen, ALS  in Brisbane, and LSA  in
Anaheim.

13
14 Baker and Harvey

structures whose range classes with multi-clausal structures. The range may
overlap to some degree with that of monomorphemic and merger predicates,
but always extends to conceptual structures which cannot be expressed by
monomorphemic and merger predicates. Given that coindexation structures
are multi-predicational, they are not subject to the constraints on the con-
ceptual structures of monomorphemic and merger predicates.
The distinction we draw between two classes of conceptual structure
associated with monoclausal structures is independent of morphological or
syntactic oppositions. As we will see, there are some common associations
between each of the classes of conceptual structure and particular morpho-
syntactic structures. However, there are no bi-directional one-to-one corres-
pondences between a particular conceptual structure class and a particular
morphosyntactic structure.
We may illustrate the opposition in conceptual structure classes by
comparing two apparently similar morphosyntactic constructions, whose
predicate interpretations are very different. The first construction is a co-
indexation construction, in this case a serial verb construction. These are
well known in the literature on complex predicates. The following example
from Barai (Papua New Guinea) is typical.
() fu burede ije sime abe ufu
he bread DEF knife take cut
‘He cut the bread with the knife.’ (Barai – Foley and Olson : )
In Barai, abe ‘take’ is a ‘light’ verb which adds the function ‘with x [instru-
ment]’ to the clause. This is a very common pattern with serial verb con-
structions (Aikhenvald ).
We illustrate merger constructions with a less well-known class of com-
plex predicates: the class of ‘coverb’ constructions. Consider the following
example from the Australian language Marra (Heath ).
() rang=ng-anyi Ø-manuga
hit=SG.S/SG.O-TAKE.PC MA-rock
‘I hit a rock.’ (Not: ‘I hit it with a rock.’)
The coverb construction involves two different kinds of verbal words – a
coverb, in this case rang ‘hit’, and an inflected finite verb, here nganyi which
means ‘I was taking it’ when used as an independent verb. Coverbs consti-
tute a basic word class. They are non-derived, inherently predicational, and
inherently non-finite. That is, they do not inflect for tense, mood or agree-
ment, but may inflect for aspect.
In the coverb construction, the coverb generally conveys the main lexical
meaning in the complex predicate. The finite verb usually, but not neces-
sarily, functions as a light verb. It conveys tense, aspect, mood, and agree-
ment, and some very general predicate information (‘generic’ in the terms
 Complex predicate formation 15

of Schultze-Berndt : ). As we will see (section .), it also determines

argument structure.
Both () and () involve two inherently predicational constituents, and in
both the ‘take’ verb is a light verb. Yet their interpretations are very different.
In (), the ‘take’ verb adds an instrumental argument. By contrast, in (), it does
not add an instrumental argument. Rather, it provides the necessary tense,
aspect, and agreement information that the coverb cannot itself provide.

. Merger constructions

Among languages which have complex predicate structures, there are some
languages which have only merger structures. This includes many languages
of northern Australia. We illustrate the range of conceptual structures char-
acteristic of merger constructions with data from Marra. In Marra, the
class of verbs is a small closed class with  members (Heath : –).
The coverb class is an open class. The great majority of ‘verbal’ predicates
are expressed through the coverb construction. Nearly all the languages of
north-central and north-western Australia have this same pattern of organ-
isation for verbal predicate meanings.
In all languages with the coverb construction, a sizeable number of coverb
lexemes can co-occur with two or more different finite verbs. In the great
majority of cases, the alternation between finite verbs has a semantically pre-
dictable effect on the meaning of the complex predicate. In every language
with the coverb construction, the construction codes alternations in transi-
tivity. Monovalent coverbs, such as dirra ‘be tied up’ and birli ‘go in’ may be
combined with both transitive and intransitive finite verbs.
() dirra=nga-jurliyi
be.tied.up=SG.S-BE.PR
‘I am tied up.’
() birli=gu-lini
go.in=SG.S-GO.PC
‘He went in.’
() dirra=nan-bili-ju
be.tied.up=SG.O-PL.S-DO.FUT
‘They are going to tie me up.’

Coverb constructions are found in many languages of northern Australia, including Warlpiri
(where they are commonly called ‘preverb’ constructions), as well as many languages of the
Ethio-Semitic family (e.g. Amharic: Amberber: this volume), Kurdish, Persian (Megerdoomian
, Folli et al. ), and many languages of Papua New Guinea and South America.

It is unclear whether there are languages which permit only coindexation. The well-known
serialising languages of West Africa such as Twi may be of this type, but we lack sufficient
information to be certain.
16 Baker and Harvey

() birli=nga-Ø-ganji
go.in=SG.S-SG.O-TAKE.PP
‘I put it in(side).’
As illustrated, the finite verb determines the argument structure of the
overall merger construction.
Significantly, these alternations apply productively to loaned coverbs in all
the Australian languages we have examined. In Marra, for example, loaned
verbs take a range of light verbs. In general, the light verb appears to be
selected on an analogical basis: a loaned coverb takes the same light verb as
native coverbs with related meanings. Hence, we assume that the verb bendi-
jimap ‘bandage’ (tr.), borrowed from Kriol (an English-lexifier creole spoken
across northern Australia) takes the light verb jujunyi ‘do’ () because co-
verbs with a similar meaning, such as dirra ‘be tied up’ do also ().
() bendijimap=nan-bili-ju
bandage=SG.O-PL.S-DO.FUT
‘They are going to bandage me up.’
We should not be particularly surprised that a light verb such as jujunyi
‘do’, with  of the total of  coverbs listed in the dictionary (Heath
), can derive new loans. What is most striking about Marra is that
even light verbs with a small class of coverbs can derive new loans. For
instance, the light verb janyi ‘tell’ takes just  attested coverbs, yet it
too can derive new loans (), again apparently by analogy with coverbs of
related meaning ().
() ringimap =nan-bili-yi
ring=SG.O-PL.S-TELL.PP
‘They rang me.’
() gaw=nan-bili-yi
call=SG.O-PL.S-TELL.PP
‘They called me.’
Apart from transitivity alternations, manner specifications are another
common function of coverbs, as illustrated in ()–() (Heath ).


Except where otherwise indicated, Marra examples are taken from the first author’s field-
notes, . Marra finite verbs have complex, largely irregular inflectional paradigms for
tense/aspect/mood (see Heath ). Finite verbs are cited in their citation form in Heath
(), which is the Past Continuous Durative form, for most verbs. We use the stand-
ard orthography now in use for teaching and documentation purposes by Marra speakers,
hence the representational differences from Heath ().

A reviewer asks whether () and () include the meaning of ‘bite’ as well as the meaning
of the coverb. We assume that, in keeping with the behaviour of this construction in Marra
 Complex predicate formation 17

() jag=nga-Ø-bayngarli
chew.briefly=SG.S-SG.O-BITE.PC
‘I chewed it briefly.’
() buny=nga-Ø-bayngarli
suck=SG.S-SG.O-BITE.PC
‘I was sucking on something.’
() gil=nga-lini
crawl=SG.S-GO.PC
‘I was crawling along.’
() jarlarla=nga-lini
walk.around=SG.S-GO.PC
‘I went for a walk.’
In some languages, such as Marra, coverbs never appear independently.
They are always in a dependency relationship with some finite verb. Indeed,
speakers do not recognise some coverbs independently of the inflected finite
verb. If the linguist pronounces these coverbs independently, speakers will
insist that the finite verb must be pronounced also. In other cases, speakers
will recognise the coverb and even give a sense for it.
In other languages, such as Jaminjung (Schultze-Berndt ) and
Wagiman (Wilson ), coverbs have relatively greater independence. They
can appear immediately after the finite verb (approximately  percent of
clauses). They appear as independent words, and occur in imperatives and
in non-initial clauses without an accompanying finite verb. However, in no
Australian languages do coverbs as a class have the full range of possibilities
for inflection and independence that finite verbs have. Clearly, coverbs form
a separate part-of-speech class in these languages.

. Coindexation constructions

One of the best-known classes of coindexation constructions is serial verb
constructions. Serial verb constructions are found as an areal feature of many
West African languages such as Twi (Lord ), East Asian languages such
as Cantonese (Matthews and Yip ), Oceanic languages such as Ambae

and other languages, they do not, but we have not explicitly tested these examples with
speakers. The translations are those given by Heath in the dictionary section of his gram-
mar of Marra (). In general, as we discuss in section ., the meanings of coverb com-
plexes merge the meanings of the constituent predicates, such that more specific meanings
(such as ‘chew briefly’) tend to override more general meanings (such as ‘bite’); moreover,
many finite verbs can be regarded as realising a semantically ‘bleached’ meaning in coverb
constructions, compared to their meaning as independent verbs. We assume this is the case
with these examples.
18 Baker and Harvey

(Crowley, ), Papuan languages (Foley ), and also many Caribbean
creoles (Sebba ).
As with the term ‘complex predicate’, the term ‘serial verb’ has been
applied to a wide variety of constructions with many kinds of semantic
structures involved. Here we focus on some representative serial verb struc-
tures, such as benefactive marking with ‘give’, comitative marking and object
marking with ‘take’, and complementisers with ‘say’. None of these can be
conveyed by merger constructions.
The following examples are from Twi (Lord : ; citing Christaller
:  and : , respectively). In () and () we see the ‘give’ verb
being used to introduce a non-subcategorised dative argument into a mono-
clausal structure.
() ageŋkwã no wú mãã y e- ŋ
Saviour the die GIVE us
‘The Saviour died for us.’
() daŋ mu hɔ n-sõ m-mã wɔŋ
house in there NEG-be.large NEG-GIVE them
‘The house is not large enough for all of them.’
In () and (), we find the ‘take’ verb de being used to introduce a non-
subcategorised argument – in this case an instrumental – into a monoclausal
structure.
() o-de n’ensa be-ñkum o-didi
he-TAKE his-hand left he-eat
‘He eats with his left hand.’
() o-de adarre o-tya duabasa
he-TAKE hook he-cut branch
‘He cut off a branch with a hook.’
This function – the introduction of non-subcategorised arguments into mono-
clausal structures – is a prominent feature of serial verb constructions in West
African languages and Caribbean creoles (Lord ), as well as Southeast Asian
languages (e.g. Cantonese: Matthews and Yip ). This function is not uni-
versally a prominent characteristic of serial verb constuctions – it is not typical
of serial constructions in Oceanic languages (Crowley ). Merger construc-
tions, as exemplified by the coverb construction, never have this function.
Other characteristic functions of serial verb constructions are the intro-
duction of information on direction () and manner (). That is, functions
which are again realised by adjuncts in other languages.

Crowley (: –) provides evidence that the Paamese constructions in ()–() are
monoclausal serial verb constructions and not multi-clausal conjoined constructions.
 Complex predicate formation 19

() ni-suvulu ni-hiitaa netano

SG:DIST.FUT-climb.down SG:DIST.FUT-descend down
‘I will climb down.’ (Paamese – Crowley : )
() inau na-muasi-ko Ø-gaiho
SG SG:REAL-hit-SG SG:REAL-hard
‘I hit you hard.’ (Paamese – Crowley : )
Aspect distinctions are often realised by means of serial verbs:
() teeviti Ø-mule Ø-metau
David SG:REAL-stay SG:REAL-afraid
‘David is (habitually) afraid.’ (Paamese – Crowley : )
Serial verb constructions also commonly convey all kinds of resultative
and causative meanings:
() inau nuas vuas he:mat
inau ni-uasi vuasi hee-mate
SG SG:DIST.FUT-hit pig SG:DIST.FUT-die
‘I will hit the pig to death.’ (Paamese – Crowley : )
() ne-sakini-e ko-musau
SG:REAL-cause-SG SG:REAL-sing
‘I made you sing.’ (Paamese – Crowley : )

. Constraints on monomorphemic predicates

We have proposed that the differences between merger constructions and
coindexation constructions follow from differences in the way that the con-
stituents contribute to the overall complex predicate. Specifically, we pro-
pose that merger constructions class fundamentally with monomorphemic
predicates. We group merger constructions and monomorphemic predicates
together in a class of ‘simplex event’ constructions.
In this section, we argue that there are constraints on simplex events,
following similar proposals in the tradition of Dowty (); e.g. Van Valin
and LaPolla (), Rappaport Hovav and Levin (), and Levin and
Rappaport Hovav (). Our analysis relies on Jackendoff’s (, ,
) theory of Lexical Conceptual Structures (hereafter LCS). LCSs
are formal decompositions of the meaning of event lexemes, construc-
tions, and clauses. Jackendoff recognises that conceptual structure can-
not be approached from a single perspective. Levin and Rappaport Hovav
(: –) discuss three of the principal approaches – which they call
‘localist’, ‘causal’ and ‘aspectual’. The localist approach claims explicitly that
20 Baker and Harvey

all verbs can be represented in terms of predicates of location or motion

(Levin and Rappaport Hovav : ). Early work representing this pos-
ition includes Gruber () and Jackendoff (). In later work Jackendoff
(: ) proposes that LCSs have distinct ‘tiers’, which represent dis-
tinct bases for the analysis of conceptual structure. He proposes an ‘action
tier’ for Actor–Patient relations – the causal approach – and a ‘thematic
tier’ for motion and location – the localist approach. Jackendoff () does
not propose a representation for the aspectual approach (though Jackendoff
(, ) contain developments of such an approach), but as we will see
(section .), there is evidence that there must be a level of representation
for aspect in lexical conceptual structure.
In this chapter, we are principally concerned with representations on the
thematic tier. This is because relations of motion and location are the most
frequent criteria in distinguishing between the two classes of complex predi-
cates. This in turn is presumably because relations of motion and location
are more commonly constant across the various appearances of a particular
lexeme than are aspectual or causal relations (Gruber ). A classic exam-
ple is the verb ‘hit’. In all uses, there is motion from the location of one entity
x towards the location of another entity y. Neither aspectual nor causal rela-
tions are constant across all uses of ‘hit’. We do not, however, intend to deny
the importance of aspectual and causal relations, and we discuss them at
various points.
The thematic tier consists of ‘conceptual functions’, predicates with
very general meanings and their arguments. The arguments of the concep-
tual functions correspond to the arguments (overt or implicit) of a clause.
Conceptual functions can themselves be the arguments of other conceptual
functions. The major conceptual functions relevant to event structure are
BE, BECOME, CAUSE and MOVE. These functions are similar in rele-
vant respects to the sets of basic functions found in other work in predicate
decomposition, such as Dowty (), Rappaport Hovav and Levin (),
Van Valin (), and others.
We propose that there are two major constraints on simplex event
structures, whether realised as monomorphemic predicates or as merger
constructions.
() The major Predicate functions – CAUSE, BECOME, MOVE,
BE – may appear only once in the LCS of the overall complex
predicate.


Jackendoff ( : –) distinguishes two motion functions – a MOVE function and a
GO function. The difference between the two is that the GO function has a Path expres-
sion, whereas the MOVE function does not. We discuss this issue and others more fully in
section ..

See especially Levin and Rappaport Hovav () for an overview of work in this tradition.
 Complex predicate formation 21

() The major Predicate functions must appear in the following sequen-
tial order:

CAUSE > {MOVE

BECOME > BE

We propose that stative predicates are characterised by a BE predi-

cate, whereas activity predicates are characterised by a MOVE predicate
( Jackendoff ). We propose that the distinctions between the various
types of activities are coded by subscripting the MOVE predicate with a
manner specification. Within the class of events involving activity predi-
cates, there is a basic opposition between those involving a Path expres-
sion (motion predicates), and those not involving a Path expression. Other
categories of events, such as achievements and accomplishments may be
derived by the application of additional predicates such as BECOME and
CAUSE.
Given the constraints in () and (), we can derive the following simplex
event structures:
() Intransitive state predicates, e.g. ‘be wet’

 BE ( 
 Thing  ,  place  )
  
 State

() Intransitive activity predicates, e.g. ‘tremble’

 
 Event MOVE (  Thing  )


() Intransitive motion predicates, e.g. ‘walk’

 Event MOVE (  Thing  ,[ Path ])

   


Levin and Rappaport Hovav (: ) propose a similar set of ‘event structure templates’
for simple predicates, following on from the agenda-setting work of Vendler () and
Dowty (), though they distinguish just five basic types (state, activity, achievement,
and two kinds of accomplishment, corresponding to our (), (), (), and (), though
we do not distinguish between their ‘internally caused’ and ‘externally caused’ accom-
plishments). They do not distinguish the various transitivity sub-types of activities and
accomplishments, nor between motion and non-motion activities without further augmen-
tation of the template –‘event structure templates may be freely augmented up to other
possible templates in the basic inventory of event structure templates’ – but we feel the two
accounts are broadly comparable.
22 Baker and Harvey

() Intransitive accomplishments, e.g. ‘sink’ and achievements, e.g.

‘shatter’

( )
 
 Event BECOME  Thing  ,  State BE (  Thing  ,[ Place ]) 




() Causatives of intransitive activity predicates, e.g. ‘shake something’


(  
 Event CAUSE  Thing  ,  Event MOVE (  Thing  ) 



)
() Transitive motion predicates, e.g. ‘walk a dog’

( 
 Event CAUSE  Thing  ,  Event MOVE (  Thing  , [ Path ]) 

)
() Transitive accomplishment predicates, e.g. ‘build a house’


( 
(
 Event CAUSE  Thing  ,  Event BECOME  Thing  ,  State BE (  Thing  , [ Place ])
    ) )
There is one other licit simplex event LCS, but its structure differs from
those in ()–(). Unlike all of these constructions, it involves simultan-
eous functions – MOVE and BECOME. This special merger construction is
examined in section ..
Allowing for this special construction, we propose that there are no
licit LCSs for simplex event structures, beyond those listed in ()–().
The constraints predict that the following kinds of structures should not
be licit simplex event structures, either because they involve an illicit
ordering of predicates, or because they involve duplication of predicate
functions.
() Inceptives of activity, e.g. ‘start to laugh’

( 
 Event BECOME  Thing  ,  Event MOVE (  Thing  ) 



)


This entails that the object position of transitive activity verbs like ‘hit’ does not find a
straightforward reflection in the structure, a problem also noted by Levin and Rappaport
Hovav ().

We have addressed only the major predicate functions here. We omit from consideration
minor types such as ‘EXT’ (extend), ‘CONF’ (configure) etc., as well as adjuncts, and add-
itional possibilities for the ‘Path’ argument discussed in Jackendoff ().
 Complex predicate formation 23

() Inceptives of motion, e.g. ‘start to go’


(  
 Event BECOME  Thing  ,  Event GO (  Thing  , [ Path ]) 



)
() Associated state with state, e.g. ‘be sitting drunk’
 
 State  State BE (  Thing  , [ Place ])  State BE (  Thing  , [ Place ]) 
 

() Associated state with activity, e.g. ‘be sitting laughing’

  BE (   
 Thing  , [ Place ])  Event MOVE (  Thing  ) 
  
 State  State

() Associated motion with state, e.g. ‘go along drunk’


( 
 Event GO  Thing  ,  State BE (  Thing  ,[ Place ]) 



)
() Associated motion with activity, e.g. ‘go along laughing’

(  
 Event GO  Thing  ,  Event MOVE (  Thing  ) 



)
() Associated motion with causatives and inceptives of a transitive verb,
e.g. ‘go along causing to/starting to eat’

   [ Thing ], 
    
 Event MOVE [ Thing ] ,  Event CAUSE  
 Event BECOME ([ Thing ] ,  State BE ([ Thing ],[ Place ]) )
     
  


( 
(
 Event MOVE  Thing  ,  Event BECOME  Thing  ,  State BE (  Thing  ,  Place  )
    ))
Because these are not licit simplex event structures, we predict that
in no language should we find such events expressed as monomorphe-
mic predicates. We also predict that merger constructions should not be
able to produce such event structures. Coindexation constructions, on
the other hand, may be able to code all or any of ()–(), depending
upon the operation of other constraints. We show that this is the case in
section .
24 Baker and Harvey

. The formal structure of merger

In the preceding sections . and ., we showed that the semantic range
covered by coverb constructions is a subset of that covered by serial verb
constructions. In this section, we show how this difference in ranges follows
from the differing ways that predicate information is combined in the two
constructions.
Following Wilson (), we propose that merger takes place at the level
of conceptual structure. The LCSs of two predicates merge to produce a
single LCS. The general principle is that predicate information from the
constituents is maintained in the merged predicate. Wilson presents the fol-
lowing description of conceptual merger.
What actually happens is that the coverb’s LCS fuses into the verb’s LCS
wherever it happens to fit. Where it happens to fit will depend upon the
particular LCSs involved. For instance, the light verb -ge- ‘putLT’ con-
tains a State as part of its LCS. The coverb guk ‘sleep’ is a State. So
when the LCSs are fused together, the LCS of guk is fused with the State
entity in the LCS of ge-. But the LCS of bort ‘die’ consists of an Event
which is expanded as a BECOME function. So when it is fused with the
LCS of -ge-, it is fused with the BECOME Event there, and not with the
State. In short, I propose that the LCS of the coverb is fused with some
part of the LCS of the inflecting verb with which it can unify. (Wilson
 : –)
Using the examples from Marra preceding, repeated here, we present the
basic structure of conceptual merger.
() birli=nga-Ø-ganji
go.in=SG.S-SG.O-TAKE.PP
‘I put it in(side).’
() birli=gu-lini
go.in=SG.S-GO.PC
‘He went in.’
The LCS of the Marra coverb birli ‘go in’, following Jackendoff (: )
can be characterised as in ().
() birli ‘go in’: [Event MOVE ([Thing x ], [Path IN])]
The LCS of the verb lini is given as ().
() lini ‘go’: [Event MOVE ([Thing x ], [Path ])]
The LCS of the light verb ganji is given as ().
() ganji ‘take’: [Event CAUSE ([Thing y ], [EventMOVE ([Thing x ], [Path ])])]
 Complex predicate formation 25

When coverbs and finite verbs merge, the LCS of the finite verb takes
the LCS of the coverb as an argument, at the highest shared predicate
function. Since the LCS of birli contains a MOVE function, it can merge
successfully with a light verb containing a MOVE function. In the case of
ganji, this MOVE function is in turn an argument of a CAUSE function,
and hence the only argument of MOVE becomes an internal argument of
CAUSE. This gives us the syntactic result that birli+ganji is a transitive
verb where the only argument of birli surfaces as the object of the CAUSE
predicate.
() birli+ganji ‘put in’: [EventCAUSE ([Thing y ], [MOVE ([Thingx ], [PathIN])])]

In the process of merger, the LCS of birli merges with the LCS of ganji
at the point where the two LCSs share a predicate in common. As already
explained, this is at MOVE. When birli combines with lini ‘go’, the result is
straightforward. We simply get the LCS that birli has anyway, with the only
argument of MOVE surfacing as the subject of the coverb construction.
() birli+lini ‘go in’: [Event MOVE ([Thing x ], [PathIN])]

Having a predicate function in common is not, however, sufficient to

ensure a successful merger. We may consider the following Wagiman
examples.
() bak ‘break’: [Event BECOME ([State BE ([Thing ], [PlaceAT (broken)])])]
() yu ‘be’: [State BE ([Thing ], [Place ])]

() *bak yu ‘be broken’: [State BE ([Thing ], [PlaceAT (broken)])]

The bak ‘break’ and yu ‘be’ predicate have a common predicate func-
tion BE. However, as Wilson (: , ) points out, this merger fails
because the lexical structure of the coverb is not preserved in the merger
construction. Wilson uses Lexical-Functional Grammar to encode this as a
constraint on merger. We express the constraint as follows.
() The LCSs of the finite and non-finite constituents must be merged
at the level of the highest major predicate function in the LCS of the
non-finite predicate(s).
The constraint expresses the intuition that the central motivation for
merger constructions is to enable the non-finite constituent to head a clause.
Consequently, its conceptual structure is maintained in the merged LCS.
In addition to this constraint, the predicates resulting from merger,
being simplex event predicates, are subject to the constraints in () and
(), already discussed. The effects of these constraints may be illustrated

Except where otherwise indicated, Wagiman examples are taken from Harvey’s
fieldnotes.
26 Baker and Harvey

by comparing the acceptable combination of the motional activity coverb

birli ‘go in’ with the finite verb lini ‘go’, against the unacceptable combin-
ation of another non-motional activity coverb wir ‘whistle’ with the same
finite verb.
() *wir=gu-lini
whistle=SG.S-GO.PC
‘He went along whistling.’
The obvious interpretation of this combination is the associated motion
meaning ‘go along whistling’, which is perfectly semantically and pragmat-
ically plausible. However, it is not possible to compose this meaning within
a merger predicate.
() wir ‘whistle’: [Event MOVE<WHISTLE> ([Thing ])]
() lini ‘go’: [Event MOVE ([Thing ], [Path ])]
() *wir+lini: [Event MOVE< WHISTLE> ([Thing ], [Path ])]
‘move along a path by means of whistling’
As shown in (), the result of merging ‘whistle’ and ‘go’ is the seman-
tically incoherent ‘move along a path by means of whistling’. In order to
express the associated motion meaning, two simultaneous MOVE predicates
are required – one describing the whistling and the other describing motion
along a path. However, the constraint in () prohibits multiple appearances
of the MOVE predicate. Consequently, non-motion activity coverbs cannot
generally combine with a motion verb in Marra.
The ungrammaticality of () does not, however, follow from any
inherent prohibition on the merger of non-motion functions with motion
functions. Rather it follows from the real-world implausibility of the
resultant combination. We may consider the following examples from
Wagiman.
() warratj-ja ga-ba-yu yurrup-pa
dance-ASP PRES-PL.S-be stand-ASP
‘They are dancing, standing upright.’
() jahan-gu warratj-ja g-i-ya
what-DAT dance-ASP PRES-SGS-go
‘Why are you dancing along/going dancing along?’ (Wilson : )

It should be noted that this combination is unacceptable only in the interpretation ‘go
whistling’. In other north Australian languages, the ‘go’ verb has an additional light verb
entry ‘DO/BE for a long time’ (e.g. Jaminjung; Schutlze-Berndt : ). In these lan-
guages the combination of activity coverb lacking a Path expression with the ‘go’ verb is
perfectly acceptable in the meaning ‘DO activity for a long time’.
 Complex predicate formation 27

() warratj-ja Ø-di-nginy

dance-ASP SGS-come-P.PFV
‘She came dancing.’ (Wilson : )
The coverb warratj ‘dance (of women)’ may plausibly describe both
motional and non-motional situations. When describing a non-motional
situation – dancing on the spot – it takes the ‘be’ verb. When describing a
motional situation, it takes either the ‘come’ or the ‘go’ verb. With changes
in real-world circumstances other mergers might become plausible. For
example, in the weightless environment of space a sentence like ‘She sneezed
herself into the next compartment of the space capsule’ might be perfectly
acceptable. In this case the movement of sneezing would be sufficient to
engender a path.
Our analysis of the class of activities differs somewhat from that of
Jackendoff. Jackendoff (: –, –) proposes two activity func-
tions – a MOVE function and a GO function. The difference between the
two is that the GO function has a Path expression, whereas the MOVE func-
tion does not. This opposition is different in kind from the other oppositions
between predicate functions. Predicate functions are not otherwise distin-
guished by whether or not they require or prohibit a particular expression.
We suggest that this is not a well-motivated basis for opposition.
Further, the analysis that manner of motion [MOVE] is necessarily con-
ceptually distinct from motion along a path [GO] is problematic. Some
predicates, at least, seem inherently to involve both. The paradigm example
is ‘walk’. This is a manner of motion, but it must necessarily be realised
along a path. While it is possible to run, hop, skip, jump, and dance on the
spot, it is not possible to walk on the spot.
Proposing a distinction between MOVE and GO functions also requires
additional formal theoretical structure. In order to describe situations where
a manner of motion predicate encodes a path, Jackendoff requires adjunction
rules which add a GO function (: ).
() GO-adjunct rule (version )
[VPVh…PP] may correspond to

 GO α , 
 ([ ] [ Path ]) 

(
 AFF [ ] i ,
α
) 

 
  WITH/BY MOVE ([α ])   
 h 


The ‘be’ verb has two meanings in Wagiman. One is to signal stativity, the other to signal
atelicity (section .). In (), it signals atelicity.
28 Baker and Harvey

The rule, as presented by Jackendoff, requires an overt expression of the

endpoint of the motion – the PP in Jackendoff’s formalisation. It is unclear
how Jackendoff would account for examples such as (), where there is no
overt endpoint expression. We may also note that this adjunction rule has no
overt linguistic realisation in many cases ( Jackendoff : ).
Given these issues, we depart from Jackendoff’s analysis and propose
that there is only a MOVE function, common to all activity predicates. This
MOVE function may license an optional Path expression. Whether an indi-
vidual predicate does license a Path expression is subject to real-world con-
siderations and to language-specific variation.
In those cases where a Path expression is not licensed, a non-finite
constituent encoding a MOVE predicate must select a verb other than ‘go’ or
‘come’ in order to construct a successful merger. In Marra, a coverb such as
wir ‘whistle’ selects the mindini ‘do/say (thus)’ verb (see section .).
() wir=nga-mindini
whistle=SG.S-DO.PC
‘I was whistling.’
We propose that this is possible because the verb ‘do’ is the generic activ-
ity verb, being simply a MOVE predicate, without further specification as to
the manner of movement.
() mindini ‘do’: [Event MOVE ([Thing ])]
() wir+mindini ‘whistle’: [Event MOVE<WHISTLE> ([Thing ])]
In merger constructions, the finite predicate is commonly less extensively
specified than the non-finite predicate. Activity coverbs normally have either
a manner subscript, e.g. <WHISTLE>, or a specification in the Path expres-
sion, e.g. IN. The ‘go’ verb, on the other hand, has neither. However, as
Wilson (: ) points out, it is not necessary that finite predicates should
bear less specification than non-finite predicates in some quantifiable way.
The following examples from Wagiman illustrate this.
() durdut-ta ba-di-nya
run-ASP PLS-come-PAST
‘They came running.’

() durdut bula-ndi

run:PFV leave-PAST (Wilson : –)
‘She ran away from him/She ran away and left him.’
The LCSs of these predicates are set out in ()–():
() di ‘come’: [Event MOVE ([Thing ], [Path TOWARD ([Place HERE])])]
 Complex predicate formation 29

() bula ‘leave’: [Event MOVE ([Thing ], [Path FROM ([Thing])])]

() durdut ‘run’: [Event MOVE<RUN> ([Thing ], [Path])]
The merged LCSs are set out in () and ():
() durdut di ‘come running’: [Event MOVE<RUN> ([Thing ], [Path TOWARD ([Place HERE])])]

() durdut bula ‘run away from’: [Event MOVE<RUN> ([Thing ], [Path FROM ([Thing])])]

The verbs di ‘come’ and bula ‘leave’ do not have light verb entries in
Wagiman. Further, it is not evident how their degree of specification might
be quantified as against the coverb durdut ‘run’. Nonetheless, neither of these
factors prevents them from entering into merger constructions.
In keeping with the general principle of preserving predicate informa-
tion, merged structures may involve compound functions, where these are
plausible, as in (), where the DOWN function from the coverb lek ‘move
down’ is compounded with the TOWARD function from the finite verb di
‘come’.
() lek ‘move down’: [Event MOVE ([Thing ], [Path DOWN])]

() lek di ‘come down’: [Event MOVE ([Thing ], [Path DOWN TOWARD ([Place HERE])])]

. Merger and motional inchoatives

The merger constructions discussed in section . all involve sequential rela-
tions among the major predicate functions. The constraint in () gener-
ally prohibits simultaneous relations among the major predicate functions.
However, there is one circumstance where the constraint does not prohibit
simultaneous relations. This is when motion to an endpoint induces a change
of state in an entity contemporaneous with that endpoint. In this case, mer-
ger can support simultaneous MOVE and BECOME predicates.
We provide examples of this type of merger construction from Wagiman
(Wilson, : , , ):
() bak Ø-linyi-ng lari
break SG-fall-PP arm
‘He fell and broke his arm.’
[lit. ‘His arm broke in falling.’]
() menuny burbur bak ga-ba-du-n
maybe wing break NP-PLS-cut-PR
‘Maybe they break its wings by cutting them.’

Wilson translates this sentence as ‘They broke its wing by spearing’, but ‘spear’ is a separ-
ate verb re, while the verb used here, du-, centrally refers to cutting, not spearing.
30 Baker and Harvey

The coverb bak ‘break’ is an inchoative coverb (Wilson : –),

and has the following LCS:
() bak ‘break’: [Event BECOME ([State BE ([Thing ], [PlaceAT (broken)])])]
The LCSs of the ‘fall’ and ‘cut’ verbs are as follows:
() linyi ‘fall’: [Event MOVE ([Thing ], [Path DOWN])]

() du ‘cut’: [Event CAUSE ([Thing ], [Event MOVE ([Thing ], [Path TO ([Place IN [Thing]])])])]

The LCS of the coverb may be merged with the LCSs of these verbs to
produce the following:
() bak linyi ‘break by falling’: [Event MOVE ([Thing ]i, [Path DOWN])]
[Event BECOME ([State BE ([Thing ] i, [PlaceAT (broken)])])]

() bak du ‘break by cutting’:

[Event CAUSE ([Thing ], [Event MOVE ([Thing ], [Path TO ([Place IN [Thing] i ])])])]
[Event BECOME ([State BE ([Thing] i, [Place AT (broken)])])]

The BECOME function is monovalent, and its argument must be co-

indexed with either the argument of the MOVE function or the argument
of the TO function. If the TO function has no argument then there is no
choice. If the TO function has an argument, then considerations of simul-
taneity determine coindexation. The MOVE and BECOME functions are
simultaneous at the endpoint of the motion, rather than during its trajec-
tory. Consequently, if there are distinct trajectory and endpoint arguments,
the argument of the BECOME function is coindexed with the endpoint
argument.
We propose that simultaneous MOVE and BECOME functions may only
appear in a merger construction when they are independently supplied by
the contributing constituents. If they are not independently supplied then
the merger construction will not be interpretable with an inchoative mean-
ing. In the available materials, there are no examples of a stative coverb mer-
ging with a motion verb and the resulting merger construction having an
inchoative interpretation. We predict that such combinations cannot be so
interpreted.
() mele-ma Ø-linyi-ng
black-ASP SGS-fall-PP
*‘S/he became black by falling.’
This interpretation is not impossible for real-world reasons. A situation
where a white-skinned person fell into a large puddle of black mud could
plausibly be described this way.
 Complex predicate formation 31

In addition to the straightforward examples of motion inchoatives, such

as () and (), there are other less immediately evident examples. The fol-
lowing examples are from Ngaliwurru (closely dialectal with Jaminjung –
Schultze-Berndt p.c.).
() darnku gani-yu
full SGS.SGO-do.PP
‘He has become full.’ (‘He is full/He has had a feed.’)
() darnku gani-minda-ny
full SGS.SGO-eat-PP
‘He has eaten (food) to becoming full.’ (‘He has eaten till full.’)

() darnku gani-ngarna-ny

full SGS.SGO-give-PP
‘She gave him (food) and he become full.’ (‘She fed him./She gave him a feed.’)

The coverb darnku is an inchoative coverb ‘become full’, and not a sta-
tive coverb ‘be full’. This is shown by the fact that it occurs in intransitive
constructions with the ‘do, say, become’ verb and not the ‘be’ verb. The mer-
ger constructions in () and () are motional inchoatives. Their LCSs are
set out following:
() [CAUSE ([x], [MOVE ([ food ] [TO (stomach of [x])])])]
[BECOME ([BE ([x], [AT (full)])])]
() [CAUSE ([ y], [MOVE ([ food ] [TO (stomach of [x])])])]
[BECOME ([BE ([x], [AT (full)])])]

. ‘Be’ and ‘do’ verbs

Any analysis of complex predicate constructions necessarily involves
some consideration of how these are to be distinguished from auxil-
iary constructions. Auxiliary verbs are traditionally analysed as non-
predicational. Unlike light verbs, they lack an argument structure.
Consequently, monoclausal constructions involving an auxiliary verb and
a main verb are not complex predicate constructions. Thus the English
perfect HAVE + V-en and progressive BE + V-ing constructions are
not complex predicate constructions, as ‘have’ and ‘be’ have no effect on
argument structure in these constructions. Rather, they provide aspec-
tual information.
In English, there is a very clear distinction between auxiliaries, which
have no argument structure, and light verbs, which do have an argument
structure. The two classes may co-occur – ‘She should have given you an
32 Baker and Harvey

answer’. Butt (this volume) shows that the two classes are similarly distinct
in Urdu and may co-occur. She argues that a distinction must be drawn
cross-linguistically between ‘auxiliary’ and ‘light verb’.
However, in many languages with the coverb construction, it is not imme-
diately evident whether this distinction is applicable. The issue as to whether
the distinction is applicable or not arises in relation to verbs translated with
the classic auxiliary meanings ‘be’ and ‘do’. Verbs translated with these
meanings have a wide range of uses, which vary somewhat from language
to language, and determining their language specific meaning is complex.
A full analysis of the semantics of ‘be’ and ‘do’ verbs is beyond the scope of
this chapter.
The only detailed discussion of these verbs in a language with the coverb
construction is Schultze-Berndt (: –, –, –, –) on
Jaminjung. We consider the Jaminjung data here, as it is illustrative of the
general patterns in languages with the coverb construction.
Schultze-Berndt suggests that there is a distinction between auxiliary
function and light verb function, at least for the ‘be’ verb. We may con-
sider the following example, where the coverb takes the continuous suffix
-mayan.
() en janyungbari burlug-mayan ga-yu gugu
and another drink-CONT SGS-BE.PRES water
‘And the other one is drinking water.’ (Schultze-Berndt : )
In this case the Jaminjung construction burlug-mayan ga-yu and its
English translation is drink-ing are morphologically isomorphic. Both involve
a derived progressive/continuous non-finite form burlug-mayan ~ drink-ing
and a finite ‘be’ verb.
The two constructions are not only morphologically isomorphic, they also
appear to be predicationally isomorphic. The Jaminjung ‘be’ verb appears
to lack an argument structure. The coverb burlug ‘drink’ is apparently biva-
lent. Bivalent coverbs cannot otherwise combine with monovalent verbs
(). The ‘be’ verb in () signals atelic aspect (Schultze-Berndt : ).
Schultze-Berndt therefore proposes that the ‘be’ verb is an auxiliary in this
construction.
However, she states that the analysis of continuous coverbs and this aux-
iliary interpretation requires further research (: ). She presents two
reasons why further research is required. First, with non-continuous cov-
erbs, the ‘be’ verb behaves straightforwardly as a light verb:
() bayirr ga-yu
be.supported SGS-BE.PRES
‘It is supported.’ (Schultze-Berndt : )
Second, continuous coverbs do not require the ‘be’ verb in Jaminjung:
 Complex predicate formation 33

() jarr-mayan=biya gan-arra-m=ngarndi ba-ngawu

put.down.one-CONT=NOW SGS.SGO-put- PRES =SFOC IMP-see
‘She keeps putting them down one at a time, look!’ (Schultze-Berndt : )

Polyvalent coverbs such as jarr-mayan ‘put.down.one-CONT’ may combine

with polyvalent verbs such arra ‘put’. Given that the apparent mismatch in
() is not categorically required, we suggest that it is not a sufficient basis
for distinguishing two morphosyntactic functions of the ‘be’ verb.
Further, we propose that there are in fact no mismatches in Jaminjung
where a bivalent coverb combines with a monovalent verb. The examples of
this mismatch listed in Schultze-Berndt () all involve the coverbs bur-
lug ‘drink’ or thawaya ‘eat’. Examination of the combinatorial possibilities
of burlug show that it is not best translated with a bivalent ‘drink’ meaning.
Rather its distribution exactly parallels that of darnku ‘become full of food’
(–). As such it is better translated with a monovalent meaning some-
thing like ‘become full of water’. Similarly, the coverb thawaya codes the
intransitive ‘eat’ meaning. The transitive ‘eat’ meaning is coded by the finite
verb minda ‘consume’.
Nonetheless, it must be recognised that the semantic contribution of
‘be’ in () is distinct from its contribution in (). This must be for-
mally modelled with two distinct lexical entries – one for stative ‘be’ and
the other for atelic ‘be’. As discussed in section ., Jackendoff does not
provide a formal representation for aspectual information in lexical con-
ceptual structure. In his model, it would presumably be an independent
tier. It would be on this tier that the aspectual information from ‘be’ in
() would be integrated into the LCS of the merger construction. In the
absence of any formalisation of this tier, we do not examine this issue
further.
The semantics of the ‘do’ verb are particularly complex (Schultze-Berndt
: –). In Jaminjung, it conveys the meanings ‘become’ and ‘say’
as well as ‘do’. This set of meanings is attested in a heterogeneous range
of languages, particularly the ‘do/say’ combination (Schultze-Berndt
: –). This argues that it is not a chance collocation, but follows
from general principles. Schultze-Berndt (: ) proposes that this verb
has the following meaning.
() x internally causes, and gives immediate evidence of, an event E
The Event in this representation could be an utterance, a cognate object
noun (‘speech/word’), or a coverb. She proposes, further, that ‘do’ is the
default light verb in Jaminjung (: ). If there is a positive motivation
for another light verb, then ‘do’ is not used.
The LCS formalisms adopted in this chapter do not allow for default
defined meanings such as that in (). Rather, meanings must be
positively determined. Given that the core meaning of this verb is an
34 Baker and Harvey

unspecified activity, we propose that its representation is as a simple

MOVE predicate.
() do: [Event MOVE ([Thing ])]
In many languages, including Jaminjung, the same form has another lexi-
cal entry with an inchoative meaning.
() do: [Event BECOME ([State BE ([Thing ], [Place])])]
This additional lexical entry is licensed by universal metaphor, where
inchoative relations are analogically assimilated to motional relations
( Jackendoff : ).
The analysis of the ‘be’ and ‘do’ verbs presented here is undoubtedly pre-
liminary in nature, and we agree with Schultze-Berndt that much further
research is required. However, we may note that there is no clear evidence
for an opposition between an auxiliary category and a light verb category in
Jaminjung. This appears to be true in many languages with the coverb con-
struction. In the absence of clear evidence for this opposition, we analyse
coverb constructions as involving light verbs only.

. Monoclausality: the dissociation between simplex

event structure and clause structure
At this point, it is useful to reconsider the notion ‘complex predicate’,
with which we started this chapter, now that we have introduced the dif-
ferent types of constructions equally labelled ‘complex predicates’ in the
literature. Jackendoff’s () model allows us to represent the relation
between Lexical Conceptual Structure and syntax in a more sophis-
ticated way. It appears that the parameters ‘constituting a licit simplex
event structure’ and ‘being a clausal predicate’ must be allowed to vary
independently. This produces at least four kinds of LCS:Syntax relations,
shown in Figure ..
Type (a) represents a simple, monomorphemic predicate heading a sin-
gle clause. Type (b) represents the merger construction, where two pred-
icates jointly contribute to an event which corresponds in its semantic
range to the events realised by monomorphemic predicates. Type (c) is a
standard construction where multiple predicates are realised by multiple
clauses. Type (d) is a multi-predicational, but monoclausal construc-
tion, as in the classic serial verb construction. In this case, we regard
the conceptual structure as one in which there are multiple events. This
is contrary to the commonly held view, as summarised by Aikhenvald
(: ).


We note that further research may establish that there are some languages where the verb
in a coverb construction is clearly an auxiliary.
 Complex predicate formation 35

Type (a): simple, monomorphemic predicate – monoclausal structure

[ Event ] Lexical Conceptual Structure

[ ]Word Syntax

Type (b): classic coverb construction – monoclausal structure

[ Event ] Lexical Conceptual Structure

[[ ] Wdi [ ] Wdj ] Clause Syntax

Type (c): multiple predicates – multiple clauses

[ Eventi ] [ Eventj ]... [ Eventn ] Lexical Conceptual Structure

[[ ] Wdi] Clausei [[ ] Wdi ] Clausej ... [[ ]Wdi ] Clausen Syntax

Type (d): multi-predicational, but monoclausal construction – classic SVC

[ Eventi ] [ Eventj ] [ Eventk ]... [ Eventn ] Lexical Conceptual Structure

[[ ] Wdi [ ] Wdj [ ] Wdk... [ ] Wdn ] Clause Syntax

Figure . Types of LCS-syntax relations

A serial verb construction (SVC) is a sequence of verbs which act together

as a single predicate, without any overt marker of coordination, subordin-
ation, or syntactic dependency of any other sort. Serial verb constructions
describe what is conceptualised as a single event.
In terms of permissible event types and sequences, classic serial verb con-
structions class with multi-clausal structures. They differ from multi-clausal
structures in that serial verb constructions are always constrained in terms
of some argument coindexation requirement.

. The range of oppositions between merger

and coindexation constructions
We have proposed that coindexation constructions are multi-predicate con-
structions and consequently not subject to the constraints which affect mer-
ger constructions. This leads to a number of differences both obvious and
less immediately evident. Firstly, coindexation constructions may permit

It is, however, clearly necessary to recognise at least three different types of coindexa-
tion relationships between predicates below the level of completely independent sentences.
Theories of syntax need to accommodate nuclear junctures, core junctures, and clause
chaining, in addition to completely independent sentences (Foley and Olson ). Foley
(this volume) discusses these issues.
36 Baker and Harvey

multiple instances of one of the major predicate functions. As such, coin-

dexation constructions may code causatives of transitives.
() nayphon hây tháhan khâ nákrian
general give soldier kill student
‘The general made the soldiers kill the students.’ (Thai – Pongsak, p.c.)
By contrast, in languages with the merger construction, complex predi-
cates cannot form the causatives of transitives. The only way of forming the
causatives of transitives is with a biclausal construction, as in ().
() nan-gu-yi nani na-boj judum=nga-mi nana rayi
sgO-SGS-tell.PC MA.OBL MA.OBL-boss shoot=SGS-DO.FPUN MA.REL bird
‘The boss told me to shoot those birds.’

In coindexation constructions, there is theoretically no upper limit on the

number of verbs in a single clause, subject to pragmatic and semantic well-
formedness. While not common, examples with three or more verbs may be
found, as in ():
() o da mɔng la saao de bing bare ko ma
.s PAST stir FACT food take put leave give me
‘S/he made food and left it there for me.’ (Dagaare – Bodomo, : )
In merger constructions, the maximum attestation is two non-finite predi-
cates. This is illustrated in the following example from Jaminjung (Schultze-
Berndt, : ):
() munuwi-ni gabarl yurl gani-mangu\ wirib
bee-ERG go.close chase SG:SG-HIT.PST dog
‘The bees came up close chasing him, the dog.’
[lit. ‘The bees chased him up close, the dog.’]
In all examples of this type – two coverbs and a single verb – one coverb
modifies the other. Many languages, such as Marra, do not allow more than
one coverb per clause.
Coindexation constructions permit the major predicate functions to
appear in sequential orders other than that specified in ().
() dbëhna-noh-me-r
sick-die-REM.PAST-SGM
‘He was sick and died.’ (Alamblak – Bruce : )
In (), the BE (sick) predicate precedes the BECOME (BE (AT
dead)) predicate. Coindexation constructions also permit simultaneous
 Complex predicate formation 37

temporal relations between all the major predicate functions. The

following examples are from Ngan’gityemerri, with () involving two
simultaneous MOVE predicates, and () involving simultaneous MOVE
and BE predicates.
() nga-ganim-fifi
SGS-go.PR-smoke
‘I’m going along smoking.’ (Ngan’gityemerri – Reid p.c.)
() ngi-rim-fifi ngi-rribem-fifi ngi-bem-fifi
SGS-sit.PR-smoke SGS-stand.PR-smoke SGS-lie.PR-smoke
‘I am sitting smoking.’ ‘I am standing smoking.’ ‘I am lying smoking.’
(Reid : )

For an example of associated motion with a transitive predicate, we may

consider the following from Gurr-goni (Green, : ).
() njirr-rre+rrmi-rri njiwurr-ma-nay gut-djardi wana
MIN:A.AUG.O-pound+REDUP-PRE AUG.S-go.along-PRE IV-rain big
‘We went along being pelted by heavy rain.’
[lit. ‘Big rain pounded us while we went.’]

This example is also of interest because the first person entity would in
most theories bear different thematic relations to the two predicates – the
theme/patient of ‘pound’ and the agent of ‘go’. In some theories, it might be
classified as the theme of ‘go’, thereby giving it the same thematic relation
to both predicates. However, there are examples of serial verb constructions
where a single entity bears distinct roles under any theory of thematic
relations.
() wǒ qiú tā dàibiǎo wǒ
SG beg SG represent SG
‘I begged him/her to represent me.’ (Mandarin Chinese – Li and
Thompson, : )

() wón mu otí yó

they drank wine drunk
‘They drank wine until drunk.’ (Yoruba – Bamgbose, : )
In merger constructions, an entity can bear only one thematic role within
the overall complex predicate.


We omit from consideration here reflexive and reciprocal constructions.
38 Baker and Harvey

Both merger and coindexation constructions allow the formation of

‘light verbs’. However, the process of light verb formation is quite dis-
tinct in each construction type. As we have seen, one common pattern
in serial verb constructions is that the ‘take’ verb adds an instrumental
argument to the clause:
() fu burede ije sime abe ufu
he bread DEF knife take cut
‘He cut the bread with the knife.’ (Barai – Foley and Olson, : )

The LCS of the ‘take’ verb is shown in ().

() [Event CAUSE ([Thing y ], [EventMOVE ([Thing x ], [Path TO ([Thing y ])])]

An implicature of this LCS is that entity y has possession of entity x.
It is a development of this implicature which survives in the serial verb
construction ( y has the knife and y cuts the bread J y cuts the bread
with the knife). This is an example of what is commonly analysed as
‘grammaticalisation’.
By contrast, in merger constructions light verbs are created by deleting
inner conceptual structure from the LCS of finite verbs.
() ‘takeLIGHT’: [Event CAUSE ([Thing y ], [EventMOVE ([Thing x ], [Path ])])]
This allows the merger to add additional material into the open inner
positions of the argument structure. The coverb rang ‘hit’ has the following
LCS.
() [Event CAUSE ([Thing y ], [EventMOVE ([Thing x ], [Path TO ([Thing z ])])]
This coverb may successfully unite with the light verb to form a standard
transitive predicate, as in ().
() rang=ng-anyi Ø-manuga
hit=SG.S/SG.O-TAKE.PC MA-rock
‘I hit a rock.’ (Not: ‘I hit it with a rock.’) (Marra – Heath )
In order to express an instrumental use, Marra requires an oblique noun
class prefix, one of whose functions is to mark instrumental case, as in ().
() rang=nga-nyi na-manuga
hit=SG.S/SG.O-TAKE.PC MA.OBL-rock
‘I hit it with a rock.’
Instrumentals cannot be expressed with verbs in Marra, nor in any other
coverb language that we are aware of.
 Complex predicate formation 39

The same kind of opposition is found with the ‘give’ verb. In serialising
languages commonly the ‘give’ verb adds a Benefactive argument to the
clause (Twi – Lord : ):
() a. ageŋkwã no wú mãã y e- ŋ
Saviour the die give us
‘The Saviour died for us.’
We propose that, in this case, what survives in the grammaticalised ver-
sion of ‘give’ is the thematic relation of Benefactive. In Jackendoff’s ()
model, thematic roles depend upon the ‘action tier’.
By contrast, the presence of the ‘give’ verb in a merger construction does
not encode non-subcategorised Benefactive type relations. We may consider
the previously discussed Ngaliwurru example (), repeated here as () for
convenience.
() darnku gani-ngarna-ny
full SGS.SGO-give-PP
‘She gave him (food) and he become full.’ (‘She fed him./She gave
him a feed.’) (Not: ‘She became full for/on him.’)

As indicated, this does not describe an action performed affecting some-

one else. As elsewhere, this does not follow from real-world considerations.
A malefactive interpretation – ‘s/he became full negatively affecting him’ –
at least is plausible.

. Correlations between conceptual oppositions and

morphosyntactic oppositions
In our presentation of materials thus far, the semantic opposition between
merger and coindexation has generally matched the categorial opposition
between coverb constructions and serial verb constructions. Coverb con-
structions involve the merger of conceptual structures, whereas serial verb
constructions involve coindexation.
While coverb constructions most commonly map to merger structures,
they do not necessarily do so. Thus, Ngan’gityemerri is a language which on
initial inspection appears to class structurally with the other Australian lan-
guages so far discussed – Jaminjung, Marra, Wagiman. It has a large open
class of coverbs and a small closed class of finite verbs. Most verbal clauses
involve a coverb construction. However, there are a number of differences
between Ngan’gityemerri on the one hand, and other Australian languages
on the other hand.
40 Baker and Harvey

First, as we have seen, coverb constructions in Ngan’gityemerri can

convey meanings, such as associated motion () and stance (), which cov-
erb constructions cannot generally in Australian languages. Examples ()
and () are repeated here as () and () for convenience.

() nga-ganim-fifi
SGS-go.PR-smoke
‘I’m going along smoking.’ (Ngan’gityemerri – Reid p.c.)
() ngi-rim-fifi ngi-rribem-fifi ngi-bem-fifi
SGS-sit.PR-smoke SGS-stand.PR-smoke SGS-lie.PR-smoke
‘I am sitting smoking.’ ‘I am standing smoking.’ ‘I am lying smoking.’
(Reid, : )

Second, the coverb construction in Ngan’gityemerri is syntactically dif-

ferent from that in most other Australian languages. As previously discussed
(section .), the configurationality of coverb constructions varies somewhat
among Australian languages. In some languages, such as Jaminjung and
Wagiman, coverbs have a greater degree of independence. However, even in
these languages, in at least  percent of occurrences a coverb immediately
precedes a finite verb. In other languages such as Marra, the coverb con-
struction is configurational and coverbs necessarily immediately precede a
finite verb. In Ngan’gityemerri, as illustrated in () and (), the coverb
necessarily immediately follows the verb.
Third, the inventory of finite verbs in Ngan’gityemerri is very different
from that in most other Australian languages. The inventories of Jaminjung
(Schultze-Berndt : –) and Wagiman (Wilson : ) are set out
in () and ().

() Jaminjung
Intransitive: be, be sick, burn, come, do/say, go, fall ( verbs)
Transitive: approach, be angry with, bite, bring, chop, cook, deceive,
eat, excrete, fear, follow, get, give, have, hear, hit, leave, make,
put, remove, see, ‘sing’ someone, spear, step on, swear at, take,
throw ( verbs)
() Wagiman
Intransitive: be ( distinct verb roots – historically ‘lie’ and ‘sit’), be
afraid, be egocentric/narcissistic, become, burn, come, cry, do/
say, dream, fall, go, stand, stay ( verbs)
Transitive: beget/name, bite, bring, cause, chase, cook, cut, deceive,
eat, follow, fuck, get, give, have/keep, hear, hit, leave, look for,
loose, make, put, see, sew, ‘sing’ someone, spear, step on, take,
tell off, throw ( verbs)
 Complex predicate formation 41

These inventories are very similar. There is a comparatively small num-

ber of intransitive verbs. The stance verbs are notable by their absence.
There are a large number of transitives. These Australian inventories may
be compared with the inventory of light verbs appearing in Persian coverb
constructions (Folli et al. : ).
() Persian (about  directly inflecting verbs –  appear commonly
as light verbs)
Intransitive: be, become, come, fall, go ( verbs)
Transitive: arrange, bring, carry, catch/take, collide, do/make,
entrust, give, have, hit, pass/cross, pull, scatter, show, throw, tie,
wash ( verbs)
This Persian inventory is similar to the two Australian inventories. The
Ngan’gityemerri inventory is set out in Table ..
There are two obvious differences between the Ngan’gityemerri inventory
and the usual inventory, as illustrated in (–). First, four of the seven
members of the Ngan’gityemerri intransitive inventory are stance verbs – lie,
perch, sit, stand. Second, Ngan’gityemerri has a set of formally distinctive
reflexive (detransitive) directly inflecting verbs.
() nge-riny-Ø-syirr nge-meny-syirr
SGA-HANDS-SGO-scratch sgS-HANDS.REFL-scratch
‘I scratched her.’ ‘I scratched myself.’
(Reid : )
It may be observed that these detransitive verbs bear no formal relation-
ship to their corresponding transitives. Specifically, detransitive verbs of this
kind are not part of the inventory of directly inflecting and/or light verbs in
most languages with coverb constructions.
In this respect, the inventory of Ngan’gityemerri is more similar to the
inventories of classificatory verbs in Athabaskan languages, where pairs
of controlled and non-controlled verbs show no formal relationship to one
another, as shown in Table ..
It may be noted that some Athabaskan languages have associated motion
constructions (Rice: this volume), of a similar kind to those we find in
Ngan’gityemerri, as in (–).
The differences between Ngan’gityemerri and other languages with the
usual coverb patterns are summarised in ().
() a. The coverb obligatorily follows the verb, as opposed to normally
or obligatorily preceding it.


The ‘stand’ verb in Wagiman is extremely rare. It occurs only with a couple of coverbs
and is not obligatory with these. It does not occur independently, unlike all the other finite
verbs.
42 Baker and Harvey

Table . Ngan’gityemerri finite verb inventory (Reid : )

Intransitive

arrive involving arrival/emergence

go carried out in motion
lie carried out in a lying posture
perch carried out up off the ground
sit carried out in a sitting posture
stand carried out in a standing posture
travel carried out in motion (goal-oriented motion)

Transitive
do/say speech and unspecified doing (do things, say things)
poke using long, thin things in point contact (stab, prod)
see performed with the eyes (look at, watch, keep an eye on)
slash using hinged trajectory and edge-on contact (sweep, slice)
take taking/bringing things
bash using vertical trajectory and lumpy contact (thump, crash)
feet holding things down with the feet (tread on, kick, walk on)
hands holding things within the grasp of the hands (grab, hold, grip)
heat applying heat (burn, melt, warm, light)
mouth holding things within the mouth (chew, suck, some speech verbs)
move moving things to a different place (shift, throw, push)
pull pulling things (pull, tow, lever up)
snatch acquiring things (get, pick up)
suck ingesting things (eat, drink)

Reflexive
bash.refl reflexive activity using vertical trajectory and lumpy contact
do/say.refl reflexive speech (talk to yourself, mutter under your breath)
feet.refl reflexive activity holding things down with feet
hands.refl reflexive activity holding things within the grasp of the hands
heat.refl reflexive activity by applying heat
mouth.refl reflexive activity holding things within the mouth
move.refl.dyn reflexive activity by moving things to a different place-dynamic
move.ref.stat reflexive activity by moving things to a different place-stative
see.refl reflexive activity performed with the eyes
 Complex predicate formation 43

Table . Slave classificatory verbs (Rice: –)

action by poking (with stick, hand) controlled Ø-kwi, Ø-ge

non-controlled Ø-ka
action by hand controlled h-nih
non-controlled Ø-tsi
action by foot controlled Ø-ʔéh
non-controlled Ø-táh

b. The stance verbs constitute half of the inventory of intransi-

tive finite verbs, as opposed to being absolutely or effectively
absent.
c. The inventory of transitives is matched by an inventory of for-
mally unrelated detransitives, as opposed to detransitivisation
being marked by standard morphological or syntactic structures.
d. Meanings such as ‘associated motion’ and ‘associated stance’
may be conveyed by the coverb construction in Ngan’gityemerri.
These meanings cannot be conveyed by coverb constructions in
most languages.
The features listed in () are also characteristic of the western and
northern neighbours of Ngan’gityemerri – Marranj, Marramaninjsji,
Marringarr, Marrithiyel, Matige, Murriny-Patha. This suggests that
the features in () are not a chance collocation, but follow from signifi-
cant structural differences between the coverb construction in these lan-
guages, and the coverb constructions found in the great majority of other
languages.
The patterns found in Ngan’gityemerri and its neighbours are not the
only types of departures from the usual patterning of the coverb construc-
tion. In three Australian languages, Kamu, Malak-Malak, and Matngele,
coverbs can combine with other coverbs in serial constructions to form com-
plex predicates through coindexation, as in ().
() dal-ngak-ma=gu-yang
poke-eat-IMPF=SGS-GO:PR
‘(The bird) is pecking (at the food).’
[lit. ‘(The bird) is poking, eating (the food).’] (Kamu – Harvey fieldnotes)
In summary, it is clear that there is no necessary connection between
the existence of coverbs as a part-of-speech class, their combination with
finite verbs to form coverb constructions, and the event structures which are
44 Baker and Harvey

possible within a single clause. In many languages with coverbs, only sim-
plex event structures are possible within a single clause. In a small group of
Australian languages from the north-western Northern Territory, however,
more complex event structures can be encoded with coverb constructions.
The situation with serial verb constructions is unclear. In terms of the
available evidence, it appears that a subset of the serial verb constructions
in many languages may map to coindexation conceptual structures. It is
conceivable that in some languages with serial verb constructions, only sim-
plex event structures may be possible within a single clause, suggesting that
these languages may only allow merger of LCS, and not coindexation, as
with many languages with coverb constructions. It is also conceivable that
languages with serial verbs may allow merger and coindexation as mapping
mechanisms at different levels of syntactico-semantic structure (as in the
‘nuclear’ vs. ‘core’ juncture model of Van Valin and LaPolla , for exam-
ple). These are questions for future research. There are also other morpho-
syntactic structures such as Germanic particle + verb constructions, whose
status with respect to the merger vs. coindexation distinction is likewise an
area for further research.

. Conclusion
We have shown that complex predicates may be divided into two classes in
terms of their correspondence with conceptual structure. One class of com-
plex predicates – the merger construction – groups with monomorphemic
predicates in that merger constructions only allow predicates which cor-
respond to the classic Vendlerian verb classes. Various authors (e.g. Dowty
, Rappaport Hovav and Levin ) have proposed that the Vendlerian
classes represent a constraint on what kinds of events may be lexicalised
as monomorphemic verbs in languages. Merger constructions therefore
obey this constraint, and for this reason we have characterised them as
realising ‘simplex’ event structures. The other class – coindexation con-
structions – groups with multi-clause sequences, in allowing more com-
plex event structures, which cannot in many cases be expressed by simple
monomorphemic predicates in other languages. It is not clear whether
there are limits on the complexity of event structures expressible by coin-
dexation constructions.
We have also shown that this distinction in conceptual structures is not
predictable from the morphosyntactic structure of complex predicates,
although there is a clear bias in most languages with coverb construc-
tions towards merger of conceptual structures, rather than coindexation.
The deeper question, which we have not addressed here, is why it is that
a complex predicate construction such as the coverb constructions of
Marra, Wagiman, and Jaminjungan should be so constrained. We have
 Complex predicate formation 45

similarly not attempted to explain which particular finite verbs become

selected for use in coverb constructions. Again, there are relatively clear
preferences for verbs with ‘generic’ meanings such as ‘take’, ‘get’, and
‘hit’ rather than more specific meanings, though there are exceptions to
this general pattern in most languages with multiple coverbs, such as
Marra janyi ‘tell’, or Bardi -ar- ‘spear lice’ (Bowern ) (see Amberber,
Baker and Harvey ). These are topics for future research.

References
Aikhenvald, A. . Serial Verb Constructions in Typological Perspective. In
Serial Verb Constructions: A Cross-Linguistic Typology, Alexandra Aikhenvald
and R. M. W. Dixon (eds), –. Oxford: Oxford University Press.
Amberber, M. . Verb Classes and Transitivity in Amharic [Lincom Studies in
Afro-Asiatic Linguistics ]. Muenchen: Lincom Europa.
Amberber, M., Baker, B., and Harvey, M. . Complex Predication and the Coverb
Construction. In Language Description, History and Development: Linguistic
Indulgence in Memory of Terry Crowley, Jeff Siegel, John Lynch, and Diana
Eades (eds), –. Amsterdam: John Benjamins.
Bamgbose, A. . On Serial Verbs and Verbal Status. Journal of West African
Linguistics :–.
Bodomo, A. B. . Paths and Pathfinders: Exploring the Syntax and Semantics of
Complex Verbal Predicates in Dagaare and Other Languages. PhD dissertation,
Trondheim: The Norwegian University of Science and Technology.
Bowern, C. . Bardi Verb Morphology in Historical Perspective. PhD disserta-
tion, Harvard.
Bruce, L. . The Alamblak Language of Papua New Guinea (East Sepik).
Canberra: Pacific Linguistics Series C: .
Butt, M., this volume. The Light Verb Jungle: Still Hacking Away.
Crowley, T. . Serial Verbs in Paamese. Studies in Language :– .
. Serial Verbs in Oceanic: A Descriptive Typology. Oxford: Oxford University
Press.
Dowty, D. R. . Word Meaning and Montague Grammar: The Semantics of Verbs
and Times in Generative Semantics and Montague’s PTQ. Dordrecht: Reidel.
Foley, W. A. . The Papuan Languages of New Guinea. Cambridge: Cambridge
University Press.
Foley, W. A. and Olson, M. . Clausehood and Verb Serialisation. In Grammar
Inside and Outside the Clause, Johanna Nichols and Anthony C. Woodbury
(eds), –. Cambridge: Cambridge University Press.
Foley, W. A. and Van Valin, R. D. . Functional Syntax and Universal Grammar.
Cambridge: Cambridge University Press.
Folli, R., Harley, H., and Karimi, S. . Determinants of Event Type in Persian
Complex Predicates. Lingua : –.
Green, R. . A Grammar of Gurr-goni: A Language of North Central Arnhem
Land. PhD dissertation, Canberra: ANU.
46 Baker and Harvey

Gruber, J. . Studies in Lexical Relations. PhD dissertation, MIT. [Published

as part of Lexical Structures in Syntax and Semantics. Amsterdam: North-
Holland, .]
Hale, K. and Keyser, S.J. . On Argument Structure and the Lexical Expression
of Syntactic Relations. In The View from Building : Essays in Linguistics in
Honor of Sylvain Bromberger, Kenneth Hale and Samuel Jay Keyser (eds),
–. Cambridge, MA: MIT Press.
. Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT
Press.
Heath, J. . Basic Materials in Mara: Grammar, Texts and Dictionary.
Canberra: Pacific Linguistics : Series C: .
Jackendoff, R. . Semantics and Cognition. Cambridge, MA: MIT Press.
. Semantic structures. Cambridge, MA: MIT Press.
. Parts and Boundaries. Cognition :  –.
. The Proper Treatment of Measuring Out, Telicity, and Possibly Even
Quantification in English. Natural Language and Linguistic Theory ,
: –.
. The Architecture of the Language Faculty. Cambridge, MA: MIT Press.
. Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford:
Oxford University Press.
Levin, B. and Rappaport Hovav, M. . Argument Realization.
Cambridge: Cambridge University Press.
Li, C. and Thompson, S. . Mandarin Chinese: A Functional Reference Grammar.
Berkeley: University of California Press.
Lord, C. . Historical Change in Serial Verb Constructions [Typological Studies
in Language, vol. ]. Amsterdam/Philadelphia: John Benjamins.
Matthews, S. and Yip, V. . Cantonese: A Comprehensive Grammar. London/
New York: Routledge.
Megerdoomian, K. . Event Structure and Complex Predicates in Persian.
Canadian Journal of Linguistics  (/): –. Special Issue on the Syntax
of Iranian Languages.
Ramchand, G. and Svenonius, P. . The Lexical Syntax and Lexical Semantics
of the Verb-Particle Construction. In WCCFL  Proceedings, L. Mikkelsen
and C. Potts (eds), –. Somerville, MA: Cascadilla Press.
Rappaport Hovav, M. and Levin, B. . Building Verb Meanings. In The projec-
tion of arguments, Miriam Butt and Wilhelm Geuder (eds), –. Stanford,
CA: CSLI.
Reid, N. . Complex Verb Collocations in Ngan’gityemerri: A Non-derivational
Strategy for Encoding Valency Alternations. In Changing Valency: Case
Studies in Transitivity, R. M. W. Dixon and Alexandra Aikhenvald (eds),
–. Cambridge: Cambridge University Press.
Rice, K. . A Grammar of Slave. Berlin and New York: Mouton de Gruyter.
This volume. Activity Incorporates in Some Athabaskan Languages.
Schultze-Berndt, E. . Simple and Complex Verbs in Jaminjung. Nijmegen:
Catholic University.
Sebba, M. . The Syntax of Serial Verbs: An Investigation into Serials in Sranan
and Other Languages. Amsterdam: John Benjamins.
 Complex predicate formation 47

Van Valin, R. D. . A Synopsis of Role and Reference Grammar. In Advances

in Role and Reference Grammar, Robert D. Van Valin, Jr. (ed.), –.
Amsterdam: John Benjamins.
Van Valin, R. D. and LaPolla, R.J. . Syntax: Structure, Meaning and Function.
Cambridge: Cambridge University Press.
Vendler, Z. . Linguistics in Philosophy. Ithaca, NY: Cornell University Press.
Wilson, S. . Coverbs and Complex Predicates in Wagiman. Stanford, CA:
CSLI.
3 The light verb jungle: still hacking away

MIRIAM BUTT

. Introduction to the jungle

This is a revised and updated version of Butt (), which noted that the
study of light verbs and complex predicates is fraught with dangers and mis-
understandings that go beyond the merely terminological. This chapter thus
attempts to provide some clarity by addressing how light verbs and com-
plex predicates can be identified cross-linguistically, what the relationship
between the two is and whether light verbs must always be associated with
uniform syntactic and semantic properties. Based primarily on both dia-
chronic and synchronic evidence from the South Asian language Urdu, but
also by taking cross-linguistic patterns into account, this chapter attempts
to pull together the relevant available knowledge in order to arrive at a more
definitive understanding of light verbs.
Jespersen (, Volume VI: ) is generally credited with first coining the
term light verb, which he applied to English V+NP constructions as in ().

() have a rest, a read, a cry, a think

take a sneak, a drive, a walk, a plunge
give a sigh, a shout, a shiver, a pull, a ring

The intuition behind the term ‘light’ is that although these constructions
respect the standard verb complement schema in English, the verbs take,
give, etc., cannot be said to be predicating fully. That is, one does not
actually physically ‘take’ a ‘plunge’ but rather one ‘plunges’. The verbs
therefore seem to be more of a verbal licenser for nouns. However, the
verbs are clearly not entirely devoid of semantic predicative content
either: there is a clear difference between take a bath and give a bath. The
verbs thus seem to neither retain their full semantic predicational con-
tent, nor are they semantically completely empty. Rather, they appear to
be semantically light in some manner that is difficult to identify. From a


Thanks go to the editors for bearing with me, to an anonymous reviewer for very thought-
ful comments and finally, to Veronika Walther for converting this paper, which was origin-
ally in LaTeX, to Word.

48
 The light verb jungle: still hacking away 49

diachronic perspective, the intuition has been that the light form of these
verbs developed from the main verb and that the light form lost some of
the semantic content as part of historical change (but see section .).
However, what it is precisely that the light verb contributes to the joint
predication and therefore exactly which parts of the predication are sup-
posed to have been lost as part of historical change is difficult to charac-
terise. Furthermore, there is no documented evidence of such a historical
development (cf. Bowern ).
Since Jespersen’s original coinage, the term light verb has been adopted
for analyses in a number of languages. Some (fairly) recent examples are
Grimshaw and Mester’s () analysis of Japanese suru ‘do’ (N+V con-
structions), Rosen’s () analysis of Romance periphrastic causatives with
‘make’ (V+V), Mohanan’s () analysis of Hindi N+V complex predicates,
and my own analysis of Urdu V+V complex predicates (Butt ). In these
papers, the term complex predicate is used to designate a construction
that involves two or more predicational elements (such as nouns, verbs, and
adjectives) which predicate as a single unit, i.e. their arguments map onto a
monoclausal syntactic structure.
Complex predicates can also be found in other languages and have
been written about by other authors than the ones cited above. However,
the literature discussing these constructions involves a dizzying diversity
of analyses and terminology. In descriptive grammars, the term compound
verb tends to be favoured, but is generally inappropriate as the two predi-
cational elements do not form lexical compounds by anybody’s definition
of compound. Another term is composite predicate, which seems to be a
reasonable alternative. However, complex/composite predicates are also
sometimes referred to as serial verb constructions (SVC), the prototypical
instantiation of which differs considerably in terms of syntax and seman-
tics from that of the typical complex predicate, although the precise line
of demarcation is difficult to draw (cf. Butt , Choi , Aikhenvald
 ; see section .). Complex predicates are also often lumped together
with control constructions (e.g. Huang () for Chinese ba and de), which
are instances of one clause embedded in another, hence biclausal and hence
actually very much unlike complex predicates, for which syntactic mono-
clausality is a hallmark (section ..). At the other extreme, complex pred-
icates are often classified as a form of auxiliary construction with the light
verb identified as a functional item along the lines of tense and aspect aux-
iliaries (e.g. Hacker  and Hook , , , for Hindi) or there is no
distinction drawn between auxiliary constructions and complex predicates
(e.g. Abeillé, Godard, and Sag () analyse both tense auxiliaries and
causative faire constructions in French as complex predicates; Wurmbrand
() sees both auxiliary constructions and complex predicates as a form
of restructuring; generally Government and Binding (GB) and minimalist
(MP) approaches draw no distinction between auxiliaries and light verbs,
50 Butt

treating both as an instance of raising). As argued in section ., this

appears to be a fundamental misanalysis.
Sorting through the various analyses, languages, and terms that have been
proposed is not trivial and requires a great deal of careful and detailed syn-
tactic work. This chapter aims at making a first contribution to the overall
(probably book-length) task of identifying core characteristics of light verbs
and providing solid syntactic and semantic analyses. In what follows, I thus
attempt to draw a very sharp distinction between auxiliaries and light verbs,
providing cross-linguistically relevant diagnostics along the way. The chap-
ter first presents typical characteristics of light verbs (section .) and then
establishes that light verbs are part of a syntactically monoclausal predica-
tion within a complex predicate (section ..). Section . argues further
that light verbs constitute a separate syntactic class and section . takes a
look at some available diachronic evidence before proposing an analysis in
section . which ties light verbs very closely to their main verb counterparts
and which sees them as elements which serve to modulate the main predica-
tion in a subtle manner.

. Typical characteristics of light verbs

My understanding of complex predicates and light verbs is necessarily col-
oured by the types of constructions found in South Asian languages. While
I have worked mainly on Urdu, these constructions can be found in most
of the South Asian languages (cf. Masica () on South Asia as a lan-
guage area). Furthermore, the same types have also been identified in many
other languages and language families, such as Romance, Bantu, Japanese,
Korean, and Persian.

.. Light verbs in connection with complex predicates

Under my understanding (and as in e.g. Alsina, Bresnan, and Sells ,
Alsina , Mohanan ), the term complex predicate refers to any con-
struction in which two or more predicational elements each contribute to
a joint predication. Note that this is distinct from noun incorporation, in
which an object (or other argument or adjunct) is drawn into the verbal
predication to become part of that predication, but does not add anything
else to the predication. That is, it moves in with the predicate, but it does
not contribute anything other than itself to the joint household, so to speak.
An example of noun incorporation in Hindi/Urdu is given in (), in which

Note that Wurmbrand actually confuses the issue further by drawing parallels between
German coherent verbs and Romance-type complex predicates. However, the two phenom-
ena are syntactically and semantically quite different. In particular, German coherent verbs
do not involve a unified predication (Predicate Composition), rather there are two separate
domains of predication (see section .).
 The light verb jungle: still hacking away 51

in one reading the object ‘horse’ has incorporated into the verb, thus modi-
fying it and giving rise to the sense of a general activity of horse-selling (cf.
Mohanan ).
() aaniil ghoṛe bec-taa haaii
Anil.M.NOM horse.M.PL sell-IMPF.M.SG be.PRES..SG
‘Anil sells horses./Anil does horse-selling.’ (Urdu)
In complex predicates, on the other hand, both parts of the predication
contribute something. Complex predicates can encompass either morpho-
logical or syntactic elements. A typical example of morphological com-
plex predication is that of morphological causativisation, as in (b), where
the causative morpheme -vaa is at the very least contributing the causer
‘Nadya’ and the other arguments are coming from the main verb ‘cut’.
() a. yassiin=ne paodaa kaaṭ-aa
Yassin/M.SG =ERG plant.M.NOM cut-PERF.M.SG
‘Yassin cut the plant.’ (Urdu)
b. naadyaa=ne yassiin=se paodaa kaṭ-vaa-yaa
NadyaF.SG =ERG Yassin.M.SG =INST plant.M.NOM CUT-Caus-PERF.M.SG
‘Nadya had the plant cut by Yassin.’ (Urdu)
In a comparison between Bantu and Romance, Alsina () and
Alsina and Joshi () have shown that regardless of whether the
complex predication is morphological or syntactic, the composition of
arguments of both the predicational elements works along the same prin-
ciples. Thus, in the permissive in (), it is the lexical item de ‘give’ that
is the element which at the very least contributes the extra argument (in
this case the permitter) to the joint predication, but the way in which
the joint argument structure is arrived at can be modelled in exactly the
same way as for the causative (see also Butt and King , Butt, King,
and Ramchand ).
() naadyaa=ne yassiin=ko paodaa kaṭ-ne dii-yaa
Nadya.F.SG=ERG Yassin=INST plant.M.NOM cut-INF give-PERF.M.SG
‘Nadya let Yassin cut the plant.’ (Urdu)
The fact that () is an example of a complex predicate, namely a syntac-
tically monoclausal predication consisting of two more predicational heads
(see section ..) is established in Butt (). Other examples of typical
complex predication are in () and (), namely Noun-Verb and Verb-Verb
complex predicates, respectively. Beyond these, Urdu (and other languages)
also contains Adj-V complex predicates such as ‘clean-do’ (not illustrated
here; e.g. Mohanan ).
52 Butt

() a. naadyaa=ne kahaanii yaad k-ii

Nadya.F.SG =ERG story.F.SG memory.F do-PERF.F.SG
‘Nadya remembered the story.’ (Urdu)
b. naadyaa=ko kahaanii yaad aa-yii
Nadya.F.SG=DAT story.F.SG memory.F come-PERF.F.SG
‘Nadya remembered the story (the memory of the story came to
Nadya).’ (Urdu)
() a. naadyaa=ne xat likh lii-yaa
Nadya.F.SG=ERG letter.M.NOM write take-PERF.M.SG
‘Nadya wrote a letter (completely).’ (Urdu)
b. naadyaa=ne makaan banaa dii-yaa
Nadya.F.SG=ERG house.M.NOM make give-PERF.M.SG
‘Nadya built a house (completely, for somebody else).’ (Urdu)
The light verb in the examples above is always the inflected one, but this
is not necessarily the case cross-linguistically. In my view, the ability to
carry tense/aspect information or be inflected is not a typical characteristic
of light verbs.
In the N-V complex predicates, the light verb acts as a verbaliser. That
is, it is a very productive device for drawing predicates into the language
and incorporating loan words into the verbal system (e.g. ‘phone-do’ for tele-
phone). This is particularly crucial for a language such as Urdu, which only
has a basic verb inventory of about  items. The light verb in this case
is reminiscent of the role that verbalising derivational morphology plays in
other languages (e.g. English -ify), though in Urdu, as in other languages
with complex predicates, the light verbs are used to make a distinction
between agentive and non-agentive actions, ‘do’ vs. ‘come’ in (). Also note
that the argument ‘story’ is contributed to the joint predication by the noun,
not by the light verb.
In (), the light verb combines with something that is already a verb
(historically a gerund; see Butt and Lahiri ) and generally affects the
Aktionsart of the joint predication. In () the light verb renders the event
bounded, but other subtle modifications such as benefactive readings, force-
fulness, suddenness, or inception are also possible (Hook ). In this case
it is difficult to see how the light verb contributes arguments to the joint
predication, but it does (see section ..).
The examples above show that light verbs are always part of a complex
predicate. This complex predicate may range over different types and there-
fore exhibit differing syntactic and semantic properties. As a consequence,
light verbs also do not all necessarily exhibit exactly the same syntactic prop-
erties – just as all auxiliaries or all lexical verbs are not alike, but can be
divided into subclasses based on their differing syntactic behaviour, so do
 The light verb jungle: still hacking away 53

light verbs constitute a cohesive class on the one hand, but fall into differing
subclasses on the other hand.
Note that the different types of light verbs found in Urdu can inter-
act with one another, allowing for complex complex predications as in (),
in which a N-V complex predicate is causativised and this combination
becomes part of a V-V complex predicate, which is further combined with
a permissive (see Butt, King, and Ramchand () for a full analysis of
this example).
() taaraa=ne amu=ko (bacce=se) haathii
Tara.F.SG=ERG Amu.F.SG=DAT child.M.OBL=INST elephant.M.SG.NOM
pinc kar-vaa le-ne dii-yaa
pinch do-CAUS take-INF.OBL give-PERF.M.SG
‘Tara let Amu have the elephant pinched (by the child).’ (Urdu)
A detailed investigation of which kinds of complex predicates can inter-
act with which other kinds of complex predicates remains to be undertaken,
both in Urdu and in a cross-linguistic context. Butt and Ramchand ()
point out for Urdu that only certain combinations are licit.

.. Form identity to a full verb

A central characteristic of light verbs is that they are always form identi-
cal to a main verb of the language (Butt and Lahiri ). This has already
been illustrated by the examples above. Even though the light verbs clearly
do not have the same predicational content as their full/main verb counter-
parts, they are always exactly form identical to a full verb and inflect exactly
like that full verb. This characteristic sets light verbs apart from auxiliaries
in terms of historical change, as auxiliaries may be form identical to a full
verb at the initial stages of reanalysis from verb to auxiliary, but then quickly
tend to develop away from the original form of the full verb. Examples are
the English preterite -d, which has been related to the verb do, or the Urdu
future -g- in (), which until just a few hundred years ago used to be the
independent lexical item ‘go’ (Butt and Lahiri ; see Bybee, Perkins, and
Pagliuca  for further examples).
() naadyaa=ko kahaanii yaad aa-ye-g-ii
Nadya.F.SG=DAT story.F.SG memory.F come-.SG-FUT-F.SG
‘Nadya will remember the story (the memory of the story will come
to Nadya).’ (Urdu)
Butt and Lahiri () therefore claim that light verbs are not part of the
grammaticalisation cline that is often posited (Hopper and Traugott ;
see Bowern () for a discussion of the state of the art with respect to
this line of inquiry), but that instead light verb and full verb usages must
54 Butt

be drawn from the same underlying lexical entry, whose lexical information
plays out in different ways depending on its syntactic environment. Indeed,
as discussed in Butt and Lahiri () at some length and as summarised in
section ., no evidence for continual reanalysis of a full verb form to a light
verb and thence to an auxiliary can be identified.

.. Joint predication and monoclausality

As discussed in section .., light verbs are always part of a joint predication
within a complex predicate. Indeed, it is this central characteristic that has
rendered complex predication and the representation of light verbs a tough
nut for syntactic theories. This is because a very fundamental assumption
underlying all syntactic theories has been that the main verb is the predi-
cational lynch-pin of the clause and that all other elements in the clause are
either arguments or modifying elements of some sort. However, there was
no sense that two or more predicational elements could come together to
form a joint predication, with a jointly determined argument structure.
A very simple solution is to assume that light verbs are predicationally
empty, i.e., their function is simply to license the predication of a non-verbal
element. For example, this is essentially the solution pursued by Grimshaw
and Mester () for Japanese or Cattell () for English. However, light
verbs do contribute to a joint predication in a systematic manner (this is
true for both Japanese suru (Butt ) and English light verbs (Brinton and
Akimoto ()). Approaches which seek to capture this systematic contri-
bution of the light verb to the joint predication posit some kind of Argument
Merger. Rosen (), for example, differentiates between light (empty), par-
tial, and complete merger for restructuring verbs and causatives in Romance.
Alsina (), Mohanan (), and Butt () respectively propose the
notions of Predicate Composition, Argument Merger, and Argument Fusion in
order to account for Romance and Hindi/Urdu.
Another possible idea within generative syntax is that light verbs are
actually instantiations of v (Adger : ). The idea of v goes back to
Chomsky (), who introduced it for auxiliaries and modals. As used in
current analyses within the Minimalist Program, v is a curious category: it
could be interpreted as either a functional or a lexical category, or a mixture
of both. Given the mixed nature of light verbs (some semantic information,
but predicationally dependent), v would actually seem to be quite a good
candidate for a light verb analysis (see Butt and Ramchand () for an
articulation of this idea with respect to Urdu; analyses are also being worked
out for Persian N-V complex predicates).

For an early analysis of complex predication in Romance as Clause Union see Aissen and
Perlmutter ().

The relevant papers are currently under review.
 The light verb jungle: still hacking away 55

Whichever analytical framework is chosen, the central characteristics of

complex predication in connection with light verbs has to be modelled.
For one, the form identity of light verbs to full verbs must be accounted
for (this is taken up in section .). For another, the jointly determined,
complex argument structure that represents a primary predication cor-
responding to a syntactically monoclausal structure must be represented.
In order to illustrate precisely what I mean by the latter, I provide a con-
crete analysis in terms of Lexical-Functional Grammar (LFG; Dalrymple
).
As shown in Butt (), the permissive complex predicate in () contrasts
with the superficially similar Urdu biclausal tell-construction as in () in
terms of agreement, anaphora, and control.
() naadyaa=ne yassiin=ko [paodaa kaṭ-ne=ko] kaah-aa
Nadya.F.SG= ERG Yassin=INST plant.M.NOM cut-INF=ACC say-PERF.M.SG
‘Nadya told Yassin to cut the plant.’ (Urdu)
In both the permissive and the tell-construction, there are two predica-
tional elements. However, in the permissive, the two combine to form one
syntactically monoclausal predication, as shown in (). This contrasts with
the syntactically biclausal construction in (), in which evidence from agree-
ment, anaphora, and control point towards the existence of an embedded
subject and an embedded object. Thus, in () two predicational elements,
‘let/give’ and ‘cut’ combine to form a single predicational unit, a complex
predicate. In (), on the other hand, the two verbs predicate separately,
each linking to a separate syntactic domain, with a subject in each of these
domains.
() Monoclausal permissive

GIVE/LET < agent goal CUT < agent theme >>

PRED ‘let-cut < ___ , ___ , ____ > ’

PRED ‘Nadya’
SUBJ
CASE ERG

PRED ‘Yassin’
OBJgo
CASE DAT

PRED ‘plant’
OBJ
CASE NOM

TENSE PAST
56 Butt

() Biclausal tell-construction

TELL < agent goal theme > CUT < agent theme >

SUBJ PRED ‘Nadya’

OBJgo PRED ‘Yassin’

PRED ‘say < ___ , ___ , ___ >’

PRED ‘cut < ___ , ___ >’

XCOMP SUBJ

OBJ PRED ‘plant’

Exactly how the correspondence between the a(rgument) structures and

the syntactic representation happens is determined by a theory of Argument
Merger in combination with a version of LFG’s linking theory. The details of
the analysis are not relevant here (see Bresnan () and Dalrymple ()
for a general description and Butt () and Butt, King, and Ramchand
() for a discussion of the phenomena at hand), the point of presenting the
analyses is merely to drive home the idea that complex predication involves
two or more predicational elements which predicate jointly by mapping their
combined predicational content to a single monoclausal syntactic domain.
It is not always easy to see that both parts of the complex predication are
contributing independently to the argument structure. For example, the
light verbs in examples such as in () never add to the overall valency of
the predication. However, that the light verb does make a contribution to
the overall joint predication in terms of argument structure in these cases
as well becomes evident with examples as in () and (). For one, agentive
light verbs like ‘take’ cannot ever be combined with unaccusative verbs such
as ‘go’, as shown in (). For another, the light verbs determine the case of
the subject. This is illustrated by the contrast between (a) and (b), where
the agentive light verb ‘take’ triggers the ergative case on the subject, but a
non-agentive verb does not.
() a. naadyaa gir ga-yii
Nadya.F.SG.NOM fall go-PERF.F.SG
‘Nadya fell (down).’ (Urdu)
b. *naadyaa=ne gir lii-yaa
Nadya.F.SG=ERG fall take-PERF.M.SG
‘Nadya fell (completely).’ (Urdu)
 The light verb jungle: still hacking away 57

() a. naadyaa=ne ro lii-yaa

Nadya.F.SG=ERG cry take-PERF.M.SG
‘Nadya cried (has finished and did it on purpose).’ (Urdu)
b. naadyaa ro paṛ-ii
Nadya.F.SG.NOM cry fall-PERF.F.SG
‘Nadya fell to crying (involuntarily).’ (Urdu)
A careful analysis of the complex predicates in () and () in terms of
data from agreement, anaphora, and control (Butt ) confirms that these
constructions are indeed monoclausal. Thus, this type of complex predicate
also passes the test of a complex joint predication which corresponds to a
monoclausal syntactic domain. Such tests for monoclausality are discussed
in the next section.

.. Establishing monoclausality

Complex predicates differ syntactically from control or raising constructions
as in () and (), which encompass two syntactically separate domains
of predication, but where some arguments (her/she) are shared across the
domains.
() I ordered her to play soccer.
() She seems to play soccer.
One crucial step in the identification of complex predication therefore is
the establishment of syntactic monoclausality. Whether a given structure
is monoclausal or not can only be determined on the basis of language-
dependent tests. That is to say, tests for monoclausality may vary across
languages, depending on the internal structure and organisation of the lan-
guage in question.
Some of the earliest work on complex predication stems from analyses of
Romance languages within Relational Grammar (RG). Aissen and Perlmutter
() show that Clause Union (i.e. complex predication) in Spanish and
Italian can be identified by phenomena such as clitic climbing: clitics ‘climb’
to the higher verb in complex predicates, but not in biclausal constructions,
as shown in () and () for French. Other tests include passivisation and
reflexivisation (see Rosen () for further discussion and tests, primarily
for French and Italian).


Note that within the class of monoclausal complex predicates, a further distinction can be
made, namely between primary and secondary predication. The complex predicates exam-
ined in this chapter are all examples of primary predication. Standard examples of second-
ary monoclausal predication are furnished by resultatives in languages such as English,
German, and Dutch: Paul painted the door green. Secondary predications such as resulta-
tives do not involve light verbs as far as I can tell.
58 Butt

() a. Jean a fait partir Marie.

Jean has made go Marie
‘Jean made Marie go.’ (French – Rosen : )
b. Jean l’a fait partir.
Jean her has made go
‘Jean made her go.’ (French – Rosen : )
() a. Marie a entendu Pierre réciter les poèmes.
Marie has listened Pierre recite the poems
‘Marie heard Pierre recite the poems.’ (French – Rosen : )
b. *Marie les a entendu Pierre réciter.
Marie them has listened Pierre recite
‘Marie heard Pierre recite them.’ (French – Rosen : )
Korean is an SOV language that does not have clitic climbing (let alone pro-
nominal clitics). It does, however, contain a number of constructions that look
like V-V complex predicates. How can one establish that these constructions
are indeed complex predicates? Choi () examines V-V constructions of
the kind shown in () in terms of how they behave with respect to negative
polarity items (NPI), negation, and the (non-)separability of the two verbs.
() Chelswu-Ka namwunip-ul ssel-E chiw-ess-ta
Chelswu-NOM leaves-ACC sweep-E clean-PAST-DECL
‘Chelswu has swept up the leaves.’ (Korean)
In Korean, the NPI items anwu-to ‘nobody’ and an ‘not’ together mean
‘nobody’. These items must co-occur in the same clause. If they do not, the NPI
meaning is not licensed and the sentence is ungrammatical, as shown in ().
() *Chelswu-nun [anwu-to pam-ul ilk-ess-ta]-ko
Chelswu-Top nobody chestnut-ACC eat-PAST-DECL-COMP
an malha-ess-ta.
NEG say-PAST-DECL
‘Chelswu did not say that nobody ate the chestnut.’ (Korean)
In contrast, when the NPI items are distributed across the kind of V-V con-
struction illustrated in (), the NPI reading is well-formed, indicating that
the construction must be monoclausal and therefore a complex predicate.
() anwu-to pam-ul an mek-E chiw-ess-ta.
nobody chestnut-ACC NEG eat-E clean-PAST-DECL
‘Nobody (children) has eaten up the chestnut.’ (Korean)
 The light verb jungle: still hacking away 59

The second (inflected) verb in the construction is the one that can be
identified as a light verb, given its ‘diminished’ predicational impact. Again,
this light verb is always form identical with a main verb in the language and
has been glossed with the meaning of the main verb.
The NPI test also works for Urdu (and Turkish). In addition, as already
mentioned, Butt () shows that object agreement, anaphora, and control
are good tests for monoclausality in Urdu. In sum, there are several possi-
ble tests for monoclausality, but these tests must be applied on a language-
internal basis. That is, a test like clitic climbing will not apply to languages
without pronominal clitics, the Urdu object agreement test will not work for
a language which does not allow object agreement, and the Korean NPI test
will not work for a language that does not allow the separation of NPI items.
The identification of complex predicates and light verbs therefore presup-
poses a very careful scrutiny of the syntax of a given language.

. Light verbs as a separate syntactic class

Light verbs are parts of complex predicates. While this fact has been recog-
nised, many syntactic approaches chose to classify light verbs either as a type
of auxiliary or as a main verb that works similarly to raising or control verbs
such as order, tell, or seem. This section argues that light verbs should be rec-
ognised as a separate syntactic class. Or rather, that the syntactic properties
of light verbs distinguish them from the syntactic distribution of auxiliaries
as well as main verbs. Just as with monoclausality, language-particular syn-
tactic factors must be taken into account in establishing this. The evidence
presented in this section goes through some data from Northern Australian
languages, Mandarin Chinese, and Urdu.

.. Northern Australian

Some Australian languages exhibit complex predicates which consist of a
coverb or preverb and an inflecting verb. The coverb and inflecting verb
can be shown to be part of a monoclausal complex predication (see Wilson
() for Wagiman, Bowern () for Bardi, Schultze-Berndt () for
Jaminjung). In the Northern Australian languages, the coverbs are drawn
from an open class, do not inflect and in general seem to share character-
istics with both adverbials and verbs. The inflecting verbs, on the other
hand, are drawn from a closed class that is generally quite small (Bowern
() lists a core class of about ten verbs). The inflecting verbs can be
used to predicate as a main verb, but when they are used in combina-
tion with a coverb, their predicational content is light. An example from
Jaminjung is shown in (), where the main predication is furnished by the
coverb ‘race’.
60 Butt

() burdurdubba=biya ga-ngga ngayin thanthu

race=now SG-go.PRES animal(ABS) DEM (ABS)
‘It is racing off now that animal.’ (Jaminjung – Schultze-Berndt,
)
Besides their unique syntactic distribution, these inflecting verbs also
play a unique semantic role within the language. As the examples in ()
and () show, the inflecting (light) verbs are able to influence the Aktionsart
of the joint predication: while the coverb stays constant in each of these
examples, the choice of the light verb modulates the event predication in a
subtle way.
() a. walthub ga-jga-ny
inside SG-go-PAST
‘go in’ (Jaminjung – Schultze-Berndt )
b. walthub ga-rdba-ny
inside SG-fall-PAST
‘get in’ (Jaminjung – Schultze-Berndt )
() a. bul ga-ruma-ny
appear SG-come-PAST
‘appear’ (Jaminjung – Schultze-Berndt )
b. bul gani-ma
appear SG-hit.PAST
‘appear (suddenly)’ (Jaminjung – Schultze-Berndt )
Bowern () provides a host of arguments that identify inflected verbs of
this type in Bardi as light verbs of the type found in Urdu (Butt and Geuder
). Schultze-Berndt () further offers an analysis of these light verbs
as classifiers of events. In (), for example, the coverb denotes the manner,
while the light verb supplies the event predication. In (), the coverb sup-
plies a path and the light verb supplies information about the type of motion
on that path. The coverb in () denotes a result and the light verb supplies
the cause.
() buru ga-ruma-ny
back SG-come-PAST
‘s/he came back’ (Jaminjung – Schultze-Berndt )
() ning burr-wa-na
break.off PL:SG-bite-IMPF
‘They were biting something off.’ (Jaminjung – Schultze-Berndt
)
 The light verb jungle: still hacking away 61

The available evidence from Northern Australian thus points to a distinct

syntactic class of light verbs which serve to modulate or modify the joint
event predication. Another such example comes from Mandarin Chinese, as
is shown in the next section.

.. Mandarin Chinese

The discussion in this section is based on materials and insights taken from
Scott () and on further joint work (Butt and Scott ), which takes a
close look at Chinese directionals. As illustrated in (), the Mandarin direc-
tionals are drawn from a closed set of verbs of direction. Typical examples of
usage, which are immediately reminiscent of the Australian examples in the
previous section, are shown in ().
() Directional etymons
shang ascend/up hui return/back
xia descend/down qi rise
jin enter/in kai open/apart
chut exit/out lai come/hither
guo cross/over qu go/thither
dao reach/to

() Examples of usage

pao jin run enter ‘to enter running’

na chu take exit ‘to take/extract’
fang xia put descend ‘to put down’
pa shang climb ascend ‘to climb up’
tong guo traverse cross ‘to go through/cross’
zhuan hui turn return ‘to turn back/return’
lai dao come reach ‘to arrive/come to’
zhan qi stand rise ‘to stand up’
zuo kai walk open ‘to walk away’

A cursory examination of the directional construction already reveals

some of the hallmarks of light verbs identified previously: the second verb
(the directional) is always form identical to a full verb in the language, but
is not predicating as a full verb. This is illustrated with respect to ‘descend’
in ().
62 Butt

() qing ni liu xia nide ming-pian

please you leave descend your name card
‘Please leave your name card.’ (Mandarin Chinese)
Teasing apart the relevant data for Mandarin Chinese is quite tricky.
However, a careful analysis shows that these directionals can have as many as
four usages and that one of them aligns with light verbs cross-linguistically.
Butt and Scott () concentrate on guo ‘cross’ as a case study.
This verb can be used in four different ways. Example () illustrates the
main verb use, () a verb second (V) directional use, and () an aspectual
use (experiential perfect).
() guo malu
cross road
‘to cross the road’ (Mandarin Chinese)
() pao guo qu
run cross go
‘to run across’ (Mandarin Chinese)
() wo shuai duan guo tui
I fall sever cross leg
‘I broke my leg once (but it has healed since).’ (Mandarin Chinese)
In addition to these three usages, a fourth one can be distinguished. On
the surface, this light verb usage appears to be very similar to the directional
V or the aspectual use in () and (). However, there are clearly identifi-
able syntactic and phonological differences.
() wo chi guo le
I eat cross CURR.REL.SIT.
‘I have eaten.’ (Mandarin Chinese)
When guo is used as a main verb, it receives tone and is not subject
to selectional restrictions. When it is used as a lexical V directional,
as in (), tone is preferred but not obligatory, and the directional may
take either a locative or a theme argument. The light use in (), on the
other hand, disprefers tone, is not separable from the main verb (‘eat’),
and cannot take a locative argument, only a theme. Finally, as a pure
aspect marker in (), guo receives no tone and licenses no independent
arguments.
There are thus four distinct identifiable uses of guo and we again have a
case of a syntactically distinct category which goes hand-in-hand with a dis-
tinct semantics.
 The light verb jungle: still hacking away 63

.. Urdu
The same conclusion, namely that light verbs must be acknowledged as a
distinct syntactic category, can be reached in Urdu on the basis of yet a dif-
ferent set of criteria. Urdu is an SOV language with fairly free word order
among constituents, but a rigid order within the verbal complex. As shown
in (), the light verb fits into a distinct slot in the verbal complex and no
other order is licit.
() Main Verb (Light Verb) (Passive Auxiliary) (Progressive Auxiliary)
(Tense Auxiliary)
Like the other main members of the verbal complex, namely the pas-
sive, progressive, and ‘be’ auxiliaries, the light verbs are always optional and
always independent syntactic elements. Unlike the auxiliaries, light verbs can
be reduplicated. The contrast is illustrated by () vs. () (cf. Fitzpatrick-
Cole  for Bengali).
() a. vo so jaa-tii th-ii
PRON..SG.NOM sleep go-IMPF.F.SG be.PAST-SG.F
‘She used to go to sleep.’ (Urdu)
b. vo so jaa-tii vaatii th-ii
PRON..SG.NOM sleep go-IMPF.F.SG go.REDUP be.PAST-SG.F
‘She used to keep going to sleep (at inopportune moments).’ (Urdu)

() a. vo so rah-ii th-ii

PRON..SG.NOM sleep PROG-F.SG be.PAST-SG.F
‘She was sleeping.’ (Urdu)
b. *vo so rah-ii vaahii th-ii
PRON..SG.NOM sleep PROG-F.SG PROG.REDUP be.PAST-SG.F
‘She was sleeping.’ (Urdu)
Another difference which sets light verbs apart from auxiliaries as
well as main verbs is that light verbs exhibit subtle lexical semantic dif-
ferences in terms of combinatorial possibilities with main verbs. These
differences are not necessarily predictable (unlike, for example, the
restriction that progressives tend to be incompatible with stative verbs).
An example from Urdu is provided in (), an example from Mandarin
Chinese in ().
() a. naadyaa makaan banaa pa ṛ -ii
Nadya.F.NOM house.M.NOM make fall-PERF.F.SG
‘Nadya fell to building a house.’ (Urdu)
64 Butt

b. ??naadyaa makaan banaa u ṭ h-ii

Nadya.F.NOM house.M.NOM make rise-PERF.F.SG
‘Nadya rose to building a house.’ (Urdu)
() a. guan diao/*shang shouyinji
shut fall/ascend radio
‘Switch off the radio.’ (Mandarin Chinese)
b. guan shang/? diao men
shut ascend/fall door
‘Close the door.’ (Mandarin Chinese)
In conclusion, in Urdu we again have a set of light verbs which can be
identified by a number of distinct distributional properties (phonological,
syntactic, semantic). Light verbs can thus clearly be established as a dis-
tinct syntactic category in a number of languages. As was the case with the
tests for monoclausality, the relevant tests differed from language to lan-
guage, but a close look at the language internal structure brought out very
precise criteria for differentiating light verbs from main verbs or aspectual
auxiliaries.
Despite the distributional differences that set light verbs apart from main
verbs and auxiliaries, they are always form identical to a main verb in the
language. This fact still needs to be accounted for. Before attempting to do
so in section ., the next section presents pertinent diachronic data.

. Light verbs and change

The previous sections have established that light verbs contribute to a
semantically complex but syntactically monoclausal predication and that
they form a syntactically distinct class. This section takes a look at some of
the available diachronic evidence and at what it suggests about the relation-
ship between main verbs, light verbs, and auxiliaries.
A quick look through the literature on syntactic change with respect to
light verbs reveals a dearth of relevant discussions, with the recent excep-
tion of Bowern (), who provides a fairly comprehensive survey. Most
approaches have focused on auxiliaries and/or modals (e.g. Lightfoot ,
Plank , Warner , Denison , Roberts , Roberts and Roussou
, ). Harris and Campbell (), for example, formulate the princi-
ple in (), which at first sight would appear to apply to light verbs. However,
a closer inspection of the material cited in support of the principle shows
that the data set only pertains to auxiliary formation.
() The Heir-Apparent Principle (Harris and Campbell : )
When the two clauses are made one by diachronic processes, the main
verb governs the syntax of the reflex clause.
 The light verb jungle: still hacking away 65

One approach which takes light verbs into account explicitly is

Grammaticalisation Theory. As shown in (), Hopper and Traugott
(: ) include vector or light verbs as an optional stage on the grammat-
icalisation cline.
() full verb > (vector verb) > auxiliary > clitic > affix
The inclusion of light verbs is due to a study on Hindi and Marathi by
Hook (, ) who analyses the light verbs as a stage in aspectogenesis
which will ultimately give rise to a type of aspectual auxiliary (see also Hook
() for a broader examination of the diachrony of light verbs in the South
Asian context). However, there are several problems with this assump-
tion. For one, rather than constituting a subclass of the existing auxiliaries,
the light verbs interact with all of the other auxiliaries of the verbal para-
digm (Butt and Geuder ). For another, the semantic contribution goes
beyond that of the purely functional tense/aspect kind. While light verbs
generally do signal some kind of boundedness or telicity or causation (cross-
linguistically), they also go beyond that and signal volitionality, benefaction,
forcefulness, surprise, etc. The degree to which they signal this differs from
language to language, but this component always seems to be present (again,
see Butt and Geuder () for more discussion).
Furthermore, the notion of aspectogenesis runs counter to an observable
diachronic fact, which is that light verbs always remain form identical to a
main verb in the language. This is very much unlike what is found with aux-
iliaries (and modals to some extent), which start out as a version of a main
verb (e.g. the English going future) but then quickly develop away from the
main verb in form, function, and meaning (e.g. English past tense -d from
do). This contrast is discussed in some detail in Butt and Lahiri () with
respect to data from Urdu and Bengali ‘be’ and ‘go’ and is confirmed by the
discussions in Brinton and Akimoto (), which examine the origin and
use of English N-V complex predicates such as take a bath.
The conclusion Traugott () reaches on the basis of the available
evidence is that English N-V complex predicates have been native to the
language at least since Old English (the furthest one can look back) and
that though the middle ages saw a marked increase in their use, no other
concomittant signs of grammaticalisation can be identified: the light verbs
do not diminish in form (e.g. become auxiliaries or affixes) and they do
not lead to the development of functional categories (Traugott  : ).
Indeed, the overall system appears to be quite stable given that the number
and type of light verbs involved remains relatively constant and given that
their ranking in terms of frequency of use remains stable with respect to
each other (e.g. give consistently comes third in terms of frequency of use
since early modern English). As is cross-linguistically typical, the light

The term vector verb is due to Pray () and has been applied to describe light verbs in
South Asian languages.
66 Butt

verbs are shown to contribute aspectual nuances as well as other types of

semantic information. For example, the use of give an answer as opposed to
the simple verb answer appears to signal that the action was done deliber-
ately (Traugott  : ).

.. Tracing light verbs through the ages

In order to drive the point home that light verbs do not appear to be subject
to historical change in the same way that auxiliaries are, this section traces
light verbs through some of the available historical evidence for Indo-Aryan.
This language family has a historical record of about  years. The discus-
sion on this section is based on Butt and Lahiri (), who investigate Urdu
and Bengali V-V complex predicates and contrast the available diachronic
data with that of auxiliaries based on ‘be’ and ‘go’ in the modern languages.
There is no precise dating for Indo-Aryan. However, the oldest attested
form of the language is thought to go back to  BCE. Vedic is generally
dated until about  BCE. Epic and Classical Sanskrit fall into the time from
 BCE to  CE. Together with Vedic, these are referred to as Old Indo-
Aryan. Middle Indo-Aryan includes Pali (mainly preserved in the form
of Buddhist texts), several Prakrit languages (which include non-standard
dialects of Sanskrit), Apabhramśa, and inscriptions of the Emperor Aśoka
(– BCE). The Middle Indo-Aryan period stretches from about 
BCE to  CE. The languages of the period from then on are commonly
referred to as New Indo-Aryan. As of  CE, distinct ancestors of the mod-
ern languages such as Old Hindi, Old Bengali, or Old Marathi are readily
identifiable.
It is generally agreed (e.g. Hook , Tikkanen , Hendriksen ,
Chatterji ) that the ancestral construction of the modern V-V com-
plex predicate is the Sanskrit ‘gerund’ or ‘absolutive’ in -tvā(ya), or -ya/
yā. These suffixes served as derivational morphemes which resulted in an
indeclinable participle (e.g. Whitney :–). In the more modern lit-
erature, this participle has also sometimes been referred to as a conjunctive
participle (CP).
The use of the -tvā participles was manifold and varied. Tikkanen ()
uses the constructed example in () to illustrate the various possible trans-
lations found in the literature with respect to the -tvā participle. () shows
an actual example from Vedic (an older stage of the language). One of the
uses Tikkanen lists is comparable with the modern complex predicate in
that the literal meaning of ‘go’ is not expressed.
() a. indram ārabhya cara
Indra-ACC grasp-GD go-IMP.SG
‘Having taken hold of Indra, move!’
‘Keep yourself to Indra!’ (Sanskrit – Tikkanen : )
 The light verb jungle: still hacking away 67

b. ime ta indra te vayam

PRON.DUAL DEM.PRON..SG Indra.VOC.SG PRON..SG.GEN PRON..PL.NOM
puruṣṭuta ye tvārabhya carāmasi
much-praise.PP.VOC.SG REL.PRON PRON..SG.ACC-grasp.GD go.PRES..PL
‘We here are yours, O ever-praised Indra, who wander about having
taken hold of you/who constantly keep ourselves to you.’ (Vedic –
Ṛgveda I..; Tikkanen : )
In contrast to Vedic and Sanskrit, which provide hints of a light verb use, but
no conclusive evidence, complex predication is clearly identifiable in Middle
Indo-Aryan (Hendriksen , Hook , , ). In particular, the Pali
examples in () both involve the verb ‘give’ as a finite verb which combines
with the participle of ‘make’. For both the sentences in () it would be strange
to assume that the meaning should be rendered as: ‘having led her to the her-
mitage, having made a fire, he gave (it) (to her)’. Rather, the complex predicate
benefactive reading given in the glosses is more appropriate.
() a. assamapadaṃ ānetvā aggiṃ katvā adāsi
hermitage.ACC lead.GD fire.ACC.SG make.GD give.IMPF..SG
‘… brought her to his hermitage and made a fire for her’
[‘having brought (her) to the hermitage, made a fire (for her)’]
(Pali, Jataaka Tales, Sri Lanka – Hendriksen : )
b. daruni āharitvā aggiṃ katvā dassati
sticks bring.GD fire.ACC.SG make.GD give.FUT..SG
‘Bringing wood he’ll make a fire (Benefactive use).’
(Pali – Trenckner : , cited by Hook : )
Turning to more modern times, examples of complex predicates can
be found in Old Bengali in the Caryapad (– CE), which consists of
forty-six complete songs and one incomplete song of six lines by twenty-four
different poets. Here the finite verbs ‘take’ (a) and ‘give’ (b) cannot be
interpreted in their main verb sense, but must rather be analysed as light
verbs which signal completion, much as is done in the modern language.
() a. cauṣathi koṭ ha gu ṇ -iā lehu
sixty-four rooms count-GD take
‘count sixty-four rooms (for yourself)’ (Old Bengali, Caryapad  –
Mojunder : )
b. bājule dila moha-kakhu bha ṇ -iā
Bajula.OBL give.PAST..SG rooms of illusion count-GD
‘Bajula counted the rooms of illusion (for his disciple).’
(Old Bengali, Caryapad  – Mojunder : )
68 Butt

Examples from Old Hindi are illustrated by () and (). McGregor
(: –) explicitly notes that the V-V constructions in (), which
are found in Braj Bhaaṣaa prose from around  CE, were used much as in
modern Hindi.
() a. … cori letu hai
steal.GD take.IMPF be.PRES..SG
‘… (he) steals’ (Old Hindi)
b. kāḍhi lei
pull out.GD take.PERF
‘(He) pulled out (with effort).’ (Old Hindi)
c. samudrahiṃ naaa ṣ i jaata haiṃ
ocean.OBJ cross.GD go.IMPF be.PRES./.PL
‘(They/We) cross oceans (completely).’ (Old Hindi)
() ḍ h ũ ḍẽ diye suhag kõ
seek give husband DAT/ACC
‘seeking a husband’ (Old Urdu/Punjabi, Baba Farid (–),
Verse )
Light verb constructions can thus be identified clearly and continually
over thousands of years. As was observed for N-V complex predicates
in English (Traugott ), the syntactic construction itself is relatively
stable. While the overt form of the gerundive morphology has changed,
the syntactic co-occurrence of a main predicate and an inflected light
verb remains constant, as does the choice of light verbs involved (e.g. ‘go’,
‘give’, ‘take’). Just as in English, the light verb is always form identical to
a main verb in the language. Light verbs thus appear to be historically
stable, very much unlike what has been documented for auxiliaries.
The available evidence from Indo-Aryan thus points to the idea that light
verbs do not enter the grammaticalisation cline, i.e. they are not main verbs
which have been reanalysed as light verbs and which are now prone to fur-
ther reanalysis. In her cross-linguistic survey of the diachrony of complex
predicates, Bowern (: ) also concludes that so far there has been no
evidence that documents the grammaticalisation (or reanalysis) of an auxil-
iary from a light verb.
Section . therefore explores an alternative notion which posits that
light verbs are intimately connected to their main verb counterpart in
the lexicon. They are so intimately connected that one can assume just
one lexical entry which can give rise to both light and main verb mean-
ings. In terms of the grammaticalisation cline, this idea plays out as
shown in ().
 The light verb jungle: still hacking away 69

()
main verb > auxiliary > clitic > affix(es)

light verb

Note that I assume one underlying lexical entry that can give rise to
distinct syntactic elements. This is similar to the idea that deverbal nomi-
nalisation operates on the same lexical entry which gives rise to inflected
main verbs, except that derivational morphology tends to be involved with
nominalisation (but cf. zero nominalisation in English). In the case of
light verbs, no derivational morphology is involved, rather the difference
in syntax and semantic interpretation arises out of the complex predicate
construction.
Whether a given verb predicates as a light or as a main verb is determined
by the syntactic environment (section .). In addition, as discussed in Butt
and Lahiri (), we assume that auxiliaries are derived from the main
verb, not the light verb. That is, we assume that light verbs are inert for the
purposes of historical change. This idea accounts for the fact that a light
verb always corresponds to a form identical main verb in the language and
that light verb constructions do not give rise to auxiliaries and modals.

.. The connection to preverbs/particles

Before proceeding on to the final section discussing the lexical representation
and semantics of light verbs, this section explores the relationship between
light verbs and preverbs/particles. This issue is often raised in connection
with South Asian light verbs, as the semantics that are described are remi-
niscent of the semantics associated with Germanic preverbs/particles like
German auf in aufessen ‘eat up’ or English out as in throw out.
Old Indo-Aryan employed a set of preverbs which in combination with
the main verb gave rise to a complex range of meanings. These meanings are
similar in nature to the contribution of the light verbs in V-V complex predi-
cates and to that found with Germanic verb particles. () provides a fairly
complete list of Sanskrit preverbs (see e.g. Whitney (: §) for a com-
plete list with their basic meanings), () provides some examples of usage.
() a. pariṇīya
around.lead.GD
‘having led around’ (Sanskrit)

Light verbs may, however, be prone to lexicalisation or idiomaticisation along with the main
predicate. This issue is also addressed in Brinton and Akimoto () where it is noted that
some N-V complex predicates in English have been reanalysed as idioms.
70 Butt

b. vi-kṛ
apart-do
‘scatter’ (Sanskrit)
The modern Indo-Aryan languages, among them Urdu/Hindi and
Bengali have lost these preverbs completely. Hook (, , ) and
Hook and Pardeshi () furthermore document an increase in light verb
use in South Asian languages since the Middle Ages. These two observa-
tions taken together raise the immediate question of whether these develop-
ments are related: could the more frequent use of V-V complex predicates in
modern Indo-Aryan be tied to the loss of preverbs?
() Sanskrit preverb Rough meaning

ati across, beyond, past, over, to excess

adhi above, over, on, on to
anu after, along, toward
antar between, among, within
apa away, forth, off
api unto, close upon, on
abhi to, unto, against (often with implied violence)
ava down, off
ā to, unto, at
ud up, up forth, out
upa to, unto, toward
ni down, in, into
nis out, forth
parā to a distance, away, forth
pari round about, around
pra forward, onward, forth, fore
prati in reversed direction, back to, back against,
against, in return
vi apart, asunder, away, out
sam along, with, together

Deo () provides a partial answer to this question by tracing the devel-
opment of preverbs in Indo-Aryan. She shows that in Vedic (the oldest form
of the language), the preverbs are associated with canonical directional or
adpositional meanings. However, for some preverbs, the meanings are less
 The light verb jungle: still hacking away 71

transparent (non-compositional) and the use of these preverbs is associated

with semantic notions of forcefulness, completion, inception, etc. This is
immediately reminiscent of the semantics associated with the modern light
verbs.
In Sanskrit, the preverbs can be divided into two categories: those that
have a literal prepositional (directional) semantics and those that have a
non-transparent semantics. Interestingly enough, the former are all multi-
syllabic, while the latter are monosyllabic. This is consonant with general
trends observed in grammaticalisation: forms which are less substantial are
more prone to grammaticalisation.
In Middle Indo-Aryan (Prakrit), the preverbs are reanalysed as either ver-
bal prefixes or part of a monomorphemic root. There is a marked decline
of preverbs which have a strictly directional or prepositional semantics. In
the modern languages, the only surviving preverbs are those that have been
reanalysed as a part of the verbal root. For the native speaker, these are not
identifiable as separate preverbs: they simply appear to be a part of the root.
Deo’s () study thus makes a plausible case for the idea that the use of
light verbs increased as preverbs fell out of the language. An explicit con-
nection between the semantics of light verbs and the semantics of preverbs/
particles is made by Ramchand ().

.. Summary
In sum, though there seems to be a correlation at least in Indo-Aryan
between the demise of preverbs and preponderance of light verbs, the light
verb construction, i.e. the possibility of forming complex predicates with
light verbs, seems to be stable over the ages. Cross-linguistically, there is no
attested evidence so far by which light verbs have evolved into auxiliaries.
Indeed, unlike auxiliaries, light verbs are always form identical to the main
verb they are related to. The next section attempts to posit an explanation
for why this is so.

. The lexical semantics of light verbs

This chapter so far has surveyed what I see as central properties which must
be accounted for in any analysis of light verbs. For one, a light verb is always
form identical with a main verb in the language. This form identical light


Note that the expression of causation also seems to be remarkably stable. The Urdu causa-
tive morpheme has undergone some phonological changes since Vedic, but other than its
surface appearance, it has remained remarkably the same since Vedic (similar patterns with
similar verb classes, etc.). In Latin, causation was expressed periphrastically and this is still
the case in its Romance descendants. It would therefore also be interesting to conduct an
in-depth cross-linguistic study of the diachrony of causative complex predicates, be they
morphological or syntactic (periphrastic).
72 Butt

verb enters into a joint complex predication with an element that furnishes
the main predicative content. The complex predication is syntactically mono-
clausal and the contribution of the light verb is not necessarily transparent.
Light verbs are unlike main verbs in that they are dependent on another
predicative element. That is, they seem to modulate or structure a given
event predication, but not supply their own event. This modulation can be in
terms of providing more information about the typical parts of an event: who
did the causation, what the result was, whether the event was bounded or
whether it was benefactive, sudden, agentive/volitional, accidental, etc.
A light verb does not, however, situate the main event predication with
respect to temporal or aspectual information. That is, it does not have the
functionality of a tense or aspect auxiliary, which situates a given event with
respect to speech and reference time. Light verbs also need to be differenti-
ated from passive auxiliaries. In general, light verbs add information to an
event predication. This stands in contrast to phenomena such as passives,
reflexive or middles, which operate on an existing argument structure in
such a way as to provide a different perspective on the participants of the
events, i.e. express voice.
Detailed argumentation on how complex predication differs from aspec-
tual auxiliaries, modality, reflexivisation, or passivisation goes beyond the
scope of this chapter. However, I believe that a careful look at the phenomena
in languages will always show that light verbs have a very different syntac-
tic distribution, semantic impact, and diachrony than aspectual and passive
auxiliaries, modals, and reflexives.
With respect to the diachrony, the observation is that light verbs are
always form identical to a corresponding main verb in the language and
that there is no attested instance of a light verb grammaticalising or being
reanalysed further (though lexicalisation or idiomaticisation may occur). As
argued in Butt and Lahiri (), this indicates that light verbs and their
corresponding main verbs stand in a very tight relationship towards one
another. Recall from the Introduction to this volume that one common way
to view light verbs is that they are semantically bleached versions of main
verbs. This implies a historical relationship in which one is derived from
the other, or, at the very least, a synchronic derivative relationship. In con-
trast, what Butt and Lahiri () suggest is that the lexical specification of
a handful of verbs (somewhere between five and twenty) cross-linguistically
allows for a use as either a main verb or a light verb. Some common examples
crosslinguistically are the verbs for ‘come’, ‘go’, ‘take’, ‘give’, ‘hit’, ‘throw’,
‘give’, ‘rise’, ‘fall’, and ‘do/make’. One can think of this set of verbs as passe-
partout: their lexical semantic specifications are so general that they can be
used in a multitude of contexts, that is, they ‘fit’ many constellations.
Concretely, Butt and Lahiri () posit one underlying underspecified
lexical entry that gives rise to both main and light verb usages. Exactly how
the lexical semantic information in this underspecified entry should be
 The light verb jungle: still hacking away 73

coded is a tricky question, just as most research into the appropriate lexical
semantic representations remains tricky. Light verbs do not predicate their
own event; rather they hook onto another event predication and deploy their
lexical content dependently.
Much of the work around lexical semantics involves lexical semantic
decomposition (e.g. Jackendoff ) and the postulation of sub-events (e.g.,
Hale and Keyser , Levin and Rappaport , Ramchand ). With
respect to light verbs, given that they clearly do not instantiate a full event
predication of their own and given that they often predicate about the caus-
ation or result (boundedness) of an event, one intuitive avenue of analysis
is that light verbs correspond to, or predicate parts, of an event, i.e. sub-
events (e.g. Ritter and Rosen , Butt , Butt and Ramchand ).
However, as Tantos () points out with respect to English light have as
in John had his students walk out of class, analyses which work with event
semantics in combination with lexical decomposition are problematic since
languages do not seem to compute cleanly with respect to events and sub-
events. Tantos instead proposes to use Segmented Discourse Representation
Theory’s (SDRT) notion of labelling certain segments of discourse and of
clauses and then computing with these labels. He works this out concretely
with respect to English light have, which has been analysed as a light verb
by Ritter and Rosen (). Since English light have can give rise to both
experiencer ( John had his dog die on him) and agentive ( John had his dog eat
the mouse) readings, Tantos posits an underspecified lexical entry for have.
This underspecified entry is then specified and disambiguated through an
interaction with pertinent information within the clause and within the
immediate discourse context .
It would lead too far afield to provide details of Tantos’ analysis here,
or to attempt to apply his ideas to the representation of light verbs cross-
linguistically. In terms of this chapter, the following will have to suffice.
Along with Butt and Lahiri (), I assume one underspecified underlying
entry for light verbs and their main verb counterparts. The content of this
entry will be not an argument structure, as is generally assumed (i.e. as in
the representations in section ..), but a loose collection of information
along the lines of Dowty’s () Proto-Role entailments. Sample entries are
provided in () for ‘give’, ‘do’, and ‘fall’.
() give Verb-stem agentive, some entity (concrete or abstract)
is to be transferred to a recipient/goal
do Verb-stem agentive activity, could involve some entity
(concrete or abstract)
fall Verb-stem non-agentive
In addition to the type of information in (), the verb entries are asso-
ciated with world knowledge. That is, what a falling event usually entails,
74 Butt

namely that it is involitional, that it is sudden, that it is downward. Or

what a giving event usually entails: that it is usually for the benefit of
somebody (but not necessarily) and that it generally is a considered action
(weighing the pros and cons). These further pieces of information may
influence the argument structure in terms of what kinds of arguments are
realised in the syntax. However, they are also likely to provide that extra
bit of semantic predication which is the hallmark of light verbs as in (),
namely the information whether a given action is sudden, benefactive, or
the responsibility of the actor (cf. Butt and Geuder () for a case study
with respect to ‘give’). And since it is information coming out of our world
knowledge, it is also defeasible, i.e. not every predication with the light
verb ‘give’, for example, will necessarily always have the same range of
semantic connotations.
When the verb enters the syntax as a main verb, it predicates as a full event
with a full range of argument participants. These are determined by the col-
lection of information associated with the verb stem, as in () (see Butt and
Tantos () for such a model involving Petri Nets). When the verb enters
the syntax as a light verb, i.e. is slotted into the distributional space for light
verbs in a language, then its lexical semantic content must combine with a
full event predication. That is, argument merger as outlined in section ..
must take place, as well as a modulation of the main event semantics by the
information coming from the light verb. Thus, depending on the syntactic
role of the verb, the information contained in it is deployed differently.
Not all verbs have such underspecified and flexible entries. Rather, as
already mentioned, we assume that such verbal passe-partout are confined
to a handful of verbs (somewhere between five and twenty). Interestingly, we
have been able to show for German that such flexible verbs, which can give
rise to both main and light verb uses, are also distinguished by a neural brain
response that sets them apart from verbs which have no light verb counter-
part (Briem et al. ). This is the case even when they are presented to
subjects in isolation, i.e. kommen ‘come’ vs. rennen ‘run’.

. Conclusion
This chapter has surveyed a number of differing complex predicates and
light verbs across languages. Complex predicates were defined as containing
two or more predicational elements which jointly predicate within a mono-
clausal structure. The evidence for monoclausality was seen to be language
dependent. Similarly, the chapter argued that light verbs must be acknowl-
edged as a separate syntactic category cross-linguistically, but that the pre-
cise syntax of light verbs differs across languages. The category light verb
must be established according to language-internal tests.
The function of light verbs is to modulate the event predication of a main
predicator in the clause. Different light verbs will do so in different ways and
 The light verb jungle: still hacking away 75

some of the semantic contributions are quite subtle. This is in part because
of the flexible interpretation of the underlying lexical semantics. The verbs
which allow light verb readings have lexical semantic specifications that are
of a very general nature. This allows them to appear in a wide variety of syn-
tactic contexts. The idea that light verbs and their corresponding main verbs
are derived from one and the same underlying representation accounts for
the fact that light verbs are always form identical to a main verb counterpart
in the language and that they are stable with respect to historical change.

References
Abeillé, Anne, Danièle Godard, and Ivan A. Sag. . Two Kinds of Composition
in French Complex Predicates. In Erhard Hinrichs, Andreas Kathol, and
Tsuneko Nakazawa (eds) Complex Predicates in Nonderivational Syntax, –.
San Diego: Academic Press.
Adger, D. . Core Syntax: A Minimalist Approach. Oxford: Oxford University
Press.
Aikhenvald, Alexandra. . Serial Verb Constructions in Typological Perspective.
In Serial Verb Constructions: A Cross-linguistic Typology, Alexandra Aikhenvald
and R.M.W. Dixon (eds), –. Oxford: Oxford University Press.
Aissen, Judith and David Perlmutter. . Clause Reduction in Spanish. In Studies
in Relational Grammar , David Perlmutter (ed.), –. Chicago: The
University of Chicago Press.
Alsina, Alex. . The Role of Argument Structure in Grammar. Stanford, CA: CSLI
Publications.
Alsina, Alex, Joan Bresnan, and Peter Sells (eds). . Complex Predicates. Stanford,
California: CSLI Publications.
Alsina, Alex and Smita Joshi. . Parameters in Causative Constructions. In
Papers from the th Regional Meeting of the Chicago Linguistic Society, –.
Bowern, Claire. . Bardi Verb Morphology in Historical Perspective. Doctoral
dissertation, Harvard University.
. The Diachrony of Complex Predicates. Diachronica , : –.
Bresnan, Joan. . Lexical-Functional Syntax. Oxford: Blackwell.
Briem, Daniela, Britta Balliel, Brigitte Rockstroh, Miriam Butt, Sabine Schulte im
Walde, and Ramin Assadollahi. . Distinct Processing of Function Verb
Categories in the Human Brain. Brain Research :–.
Brinton, Laurel J. and Minoji Akimoto (eds). . Collocational and Idiomatic
Aspects of Composite Predicates in the History of English. Amsterdam: John
Benjamins.
Butt, Miriam. . The Structure of Complex Predicates in Urdu. Stanford,
California: CSLI Publications.
. The Light Verb Jungle. In Harvard Working Papers in Linguistics, eds
G. Aygen, C. Bowern, and C. Quinn, –. Volume , Papers from the GSAS/
Dudley House Workshop on Light Verbs.
Butt, Miriam and Wilhelm Geuder. . On the (Semi)Lexical Status of Light
Verbs. In Semilexical Categories: On the Content of Function Words and the
76 Butt

Function of Content Words, Norbert Corver and Henk van Riemsdijk (eds),
–. Berlin: Mouton de Gruyter.
Butt, Miriam, Tracy H. King, and Gillian Ramchand. . Complex
Predication: Who Made the Child Pinch the Elephant? In Reality Exploration
and Discovery: Pattern Interaction in Language and Life, Linda Uyechi and
Lian Hee Wee (eds). Stanford, California: CSLI Publications.
Butt, Miriam and Tracy H. King. . Restriction for Morphological
Valency Alternations: The Urdu Causative. In Intelligent Linguistic
Architectures: Variations on Themes by Ronald M. Kaplan, Miriam Butt, Mary
Dalrymple, and Tracy Holloway King (eds), –. CSLI Publications.
Butt, Miriam and Aditi Lahiri. . Historical Stability vs Historical Change.
Unpublished ms. http://ling.uni-konstanz.de/pages/home/butt/.
Butt, Miriam and Gillian Ramchand. . Complex Aspectual Structure in Hindi/
Urdu. In The Syntax of Aspect, Nomi Ertishik-Shir and Tova Rappaport (eds),
–. Oxford: Oxford University Press.
Butt, Miriam and Biljana Scott. . Chinese Directionals. Talk given as part of the
Workshop Complex Predicates, Particles and Subevents, Konstanz, September,
http://ling.uni-konstanz.de/pages/home/butt/.
Butt, Miriam and Alexandros Tantos. . Verbal Semantics via Petri Nets. In
On-Line Proceedings of the LFG Conference, Miriam Butt and Tracy H. King
(eds) CSLI Publications, University of Canterbury, Christchurch, http://
cslipublications.stanford. edu/LFG/.
Bybee, Joan, Revere Perkins, and William Pagliuca. . The Evolution of
Grammar: Tense, Aspect and Modality in the Languages of the World.
Chicago: The University of Chicago Press.
Cattell, R. . Composite Predicates in English. Sydney: Academic Press Australia.
Chao, Y.-R. . A Grammar of Spoken Chinese. Berkeley, California: The
University of California Press.
Chatterji, S. K. . The Origin and Development of the Bengali Literature, vol. II.
Calcutta: Do Mehra, Rupa of Co. ( edition).
Choi, Seongsook. . Multiple Verb Constructions in Korean. Doctoral disserta-
tion, University of Sussex.
Chomsky, Noam. . Syntactic Structures. The Hague: Mouton de Gruyter.
Dalrymple, Mary. . Lexical Functional Grammar. New York: Academic Press.
Davidson, Donald. . The Logical Form of Action Sentences. In The Logic
of Decision and Action, N. Resher (ed.), –. Pittsburgh: University of
Pittsburgh Press. Reprinted in Davidson, Donald. . Essays on Actions and
Events, –. Clarendon Press, Oxford.
Denison, David. . English Historical Syntax: Verbal Constructions.
London: Longman.
Deo, Ashwini. . A Diachronic Perspective on Complex Predicates in Indo-
Aryan. Talk given as part of the Workshop Complex Predicates, Particles and
Subevents, Konstanz, September.
Dowty, David. . Thematic Proto-roles and Argument Selection. Language ,
: –.
Fitzpatrick-Cole, Jennifer. . The Prosodic Domain Hierarchy in Reduplication.
Doctoral dissertation, Stanford University.
 The light verb jungle: still hacking away 77

Grimshaw, Jane and Armin Mester. . Light Verbs and θ -Marking. Linguistic
Inquiry  (): –.
Hacker, Paul. . Zur Funktion einiger Hilfsverben im modernen Hindi.
Mainz: Akademie der Wissenschaften und der Literatur.
Hale, Kenneth and Jay Keyser. . Argument Structure. In The View from
Building , Kenneth Hale and Jay Keyser (eds), –. Cambridge,
MA: MIT Press.
Harris, Alice C. and Lyle Campbell. . Historical Syntax in Cross-Linguistic
Perspective. Cambridge: Cambridge University Press.
Hendriksen, Hans. . Syntax of the Infinite Verb Forms of Pali. Copenhagen: Einar
Munksgaard.
Hook, Peter Edwin. . The Compound Verb in Hindi. Center for South and
Southeast Asian Studies: The University of Michigan.
. The Emergence of Perfective Aspect in Indo-Aryan Languages. In
Approaches to Grammaticalization, Elizabeth Traugott and B. Heine (eds),
–. Amsterdam: Benjamins.
. Aspectogenesis and the Compound Verb in Indo-Aryan. In Complex
Predicates in South Asian Languages, ed. Manindra Verma, –.
Delhi: Manohar Publishers & Distributors.
. Where do Compound Verbs Come From? (And Where are They Going?).
In The Yearbook of South Asian Languages and Linguistics, Peri Bhaskararao
and K.V. Subbarao (eds), –. New Delhi: Sage Publications.
Hopper, Paul J. and Elizabeth C. Traugott. . Grammaticalization.
Cambridge: Cambridge University Press.
Huang, Charles T.-J. . Complex Predicates in Control. In Control and
Grammar, Richard Larson, Sabine Iatridou, and Utpal Lahiri (eds), –.
Dordrecht: Kluwer Academic Publishers.
Jackendoff, Ray. . Semantic Structures. Cambridge, MA: MIT Press.
Jespersen, Otto. . A Modern English Grammar on Historical Principles, Part VI,
Morphology. London: George Allen and Unwin Ltd.
Levin, Beth and Malka Rappaport Hovav. . Building Verb Meanings. In The
Projection of Arguments: Lexical and Compositional Factors, Miriam Butt and
Wilhelm Geuder (eds), –. Stanford, California: CSLI Publications.
Lightfoot, David. . Principles of Diachronic Syntax. Cambridge: Cambridge
University Press.
Masica, Colin. . Defining a Linguistic Area: South Asia. Chicago: The University
of Chicago Press.
McGregor, R.S. . The Language of Indrajit of Orchaa. Cambridge: Cambridge
University Press.
Mohanan, Tara. . Argument Structure in Hindi. Stanford, California: CSLI
Publications.
. Wordhood and Lexicality: Noun Incorporation in Hindi. Natural Language
and Linguistic Theory (): –.
Mojunder, Atindra. . Bengali Language Historical Grammar, Part II.
Calcutta: K.L. Mukhopadhay.
Plank, Frans. . The Modals Story Retold. Studies in Language  ():
–.
78 Butt

Pray, Bruce. . Topics in Hindi-Urdu Grammar. Berkeley, CA: Center for South
and Southeast Asia Studies.
Ramchand, Gillian. . Verb Meaning and the Lexicon: A First Phase Syntax.
Cambridge: Cambridge University Press.
Ritter, Elizabeth and Sara T. Rosen. . Deriving Causation. Natural Language
and Linguistic Theory : –.
Roberts, Ian. . A Formal Account of Grammaticalisation in the History of
Romance Futures. Folia Linguistica Historica : –.
Roberts, Ian and Anna Roussou. . A Formal Approach to ‘Grammaticalization’.
Linguistics (): –.
. Syntactic Change: A Minimalist Approach to Grammaticalization.
Cambridge: Cambridge University Press.
Rosen, Sara. . Argument Structure and Complex Predicates. Doctoral disserta-
tion, Brandeis University, Massachusetts.
Schultze-Berndt, Eva. . Simple and Complex Verbs in Jaminjung: A
Study of Event Categorisation in an Australian Language. MPI Series in
Psycholinguistics.
. Neither Noun nor Verb nor Particle. Talk given as part of the Workshop
Complex Predicates, Particles and Subevents, Konstanz, September.
Scott, Biljana. . Aspectogenesis and the Categorisation of Directionals in Chinese.
Doctoral dissertation, Oxford University.
Tantos, Alexandros. . Computing Events in Discourse: A Case Study Involving
Light have. Doctoral dissertation, University of Konstanz.
Tikkanen, Bertil. . The Sanskrit Gerund: A Synchronic, Diachronic and
Typological Analysis. Helsinki: Finnish Oriental Society. Studia Orientalia.
Traugott, Elizabeth Closs. . A Historical Overview of Complex Predicates. In
Collocational and Idiomatic Aspects of Composite Predicates in the History of
English, Laurel J. Brinton and Minoji Akimoto (eds), –. Amsterdam: John
Benjamins.
Trenckner, V. . Pali Miscellany. London: Williams & Norgate.
Warner, Anthony. . English Auxiliaries: Structure and History. Cambridge:
Cambridge University Press.
Whitney, William D. . Sanskrit Grammar. Cambridge, MA: Harvard University
Press. Ninth Issue of the Second Edition ().
Wilson, Stephen. . Coverbs and Complex Predicates in Wagiman. Stanford, CA:
CSLI Publications.
Wurmbrand, Susanne. . Infinitives: Restructuring and Clause Structure. Berlin:
Mouton de Gruyter.
 Events and serial verb constructions

WILLIAM A. FOLEY

. Introduction
Baker and Harvey (this volume) claim a fundamental distinction between
two types of complex predicate constructions: coverb constructions, typically
those involving a light verb, and serial verb constructions (henceforth SVCs)
in terms of their respective expression of eventhood. They argue that while
both coverb constructions and SVCs are monoclausal, they contrast in their
event structure: coverb constructions express a single simple event, albeit
one that may be semantically complex, while SVCs express multiple events.
In this latter claim they disagree with an often stated view about SVCs, as
summarised by Aikhenvald (: ):
A serial verb construction (SVC) is a sequence of verbs which act together as
a single predicate, without any overt marker of coordination, subordination
or syntactic dependency of any other sort. Serial verb constructions describe
what is conceptualised as a single event.
While this chapter will ultimately support Baker and Harvey’s claim about
SVCs in contrast to Aikhenvald’s, in our view not much progress can be
made in understanding SVCs while one proceeds in any analysis with
unexamined, vague, and undefined concepts like event, simple and multiple,
and monoclausality. We will show that SVCs are in no sense a unified phe-
nomenon, but manifest both different structural realisations and express
diverse types of event structures, some as simple as a coverb construction
and some much more complex. To describe such cross-linguistic variation,
we will need both more precise structural notions than clausehood and more
sophisticated semantic notions than a simple contrast between single and
multiple events.
But, first, to set the stage for our discussion, it is necessary to define what
an SVC might be. Although our ultimate conclusion will be that there is no
unified syntactic phenomenon that goes under this label, we need a start-
ing point to delineate the kinds of constructions we will be discussing and
which conventionally are termed SVCs. Comrie’s () summary of SVC

79
80 Foley

properties is particularly clear and succinct and seems a good place to start,
although ultimately we will be deconstructing each of these:

Comrie’s () summary of SVC properties:

. The sequence of verbs in an SVC occurs within a single clause.

. Either there is no marking of most verb categories (person–number,
tense–aspect–mood, polarity, etc.) except on the main verb, or all verbs
must have the same categories as the main verb.
. The verbs in the SVC are interpreted as expressing a single event.
Beyond this the SVC has two main functions: (a) to allow the expression
of more arguments of the predicate; and (b) to allow the creation of new
lexical items.
We can illustrate the operation of these defining properties for SVCs by
comparing the syntactic behavior of the SVC in (a) from Yimas, a Papuan
language of New Guinea, with that of the related clause chaining construc-
tion (henceforth CCC) in (b):

() a. arm-n kay i-ka-ak-mpi-wul

water-OBL canoe VIII SG VIII SG O-SG A-push-SEQ-put.in
‘I pushed the canoe down into the water.’
b. kay ak-mpi i-ka-wul arm-n
canoe VIII SG push-SEQ VIII SG O-SG A-put.in water-OBL
‘I pushed the canoe and put it in the water.’

(a) is an SVC (in fact, it is a single grammatical word, see the arguments
in Foley (:–)): the pronominal agreement affixes must precede
the sequence of verbs in the SVC and the tense–aspect–mood suffixes
(here the Ø allomorph of -r PERF) must follow. In this example, in which
the events denoted by the verb roots are in a temporal relationship of
sequence, they are linked by the suffix -mpi SEQ ; if the events denoted
were simultaneous, the verb roots would be simply juxtaposed. Other
than -mpi, nothing can separate the verb roots linked together in an SVC.
Example (b) is a CCC; as in all CCCs, the final verb in the chain carries
marking for tense–aspect–mood (again the Ø allomorph of -r PERF) and
the pronominal agreement prefixes for the core grammatical relations,
even though, as in this case, an NP agreed with is in a preceding medial
clause (kay ‘canoe’ is linked to the agreement prefix i- VIII SG O). The
verbs in medial clauses of CCCs have reduced inflectional possibilities, in
Yimas typically just -mpi SEQ. The verbs in CCCs can have intervening
material, for example the NP arm-n water-OBL in (a), but this is strictly
prohibited for SVCs (b):
 Events and serial verb constructions 81

() a. kay ak-mpi arm-n i-ka-wul

canoe VIII SG push-SEQ water-OBL VIII SG O-SG A-put.in
‘I pushed the canoe and put it in the water.’
b. *kay i-ka-ak-mpi arm-n wul
canoe VIII SG VIII SG O-SG A-push-SEQ water-OBL put.in

Verbal inflectional categories like imperative formation also show differential

behavior in SVCs and CCCs. In SVCs, there is a single prefix for impera-
tive which precedes the verbal concatenation and another which follows it
(a) (note that agreement for objects is realised by suffixes, not prefixes, in
imperative verb forms). Splitting an SVC marked for imperative by inter-
leaving material is ungrammatical (b); the SVC must remain indivisible.
Marking each verb root separately for imperative is also impossible in an
SVC; to do so results in a sequence of coordinated full clauses (c). In a
CCC only the final verb carries marking for imperative, although the scope
of the command is also over the medial clause (d).

() a. arm-n kay ŋaŋk-ak-mpi-wut-ɲa-i

water-OBL canoe VIII SG IMP DL-push-SEQ-put.in-IMP-VIII SG O

‘You two push the canoe down into the water!’

b. *kay naŋk-ak arm-n wut-ɲa-i
canoe VIII SG IMP DL-push water-OBL put.in-IMP-VIII SG O
c. arm-n kay naŋk-ak-na-i
water-OBL canoe VIII SG IMP DL-push-IMP-VIII SG O
naŋk-wut-ɲa-i
IMP DL-put.in-IMP-VIII SG O
‘You two push the canoe and then put it in the water!’
d. kay yawra-mpi arm-n
canoe VIII SG pick.up-SEQ water-OBL
naŋk-wut-ɲa-i
IMP DL-put.in-IMP-VIII SG O

‘You two pick up the canoe and put it into the water!’

. Events and the problems posed by SVCs

So far, so good; we seem to have a paradigmatic case in (a) of an SVC in the
terms of Comrie’s properties: single clausehood (indeed, single grammatical
82 Foley

word), shared verbal inflectional categories, single eventhood. But in what

sense does (a) express a single event? In the prototypical case, ak-mpi-wul-
‘push down (into the water)’ refers to one (or more commonly, multiple)
actor(s) causing a canoe to move linearly along the ground away from the
high ground of the riverbank toward the lower level of the river itself, so
that it descends down the edge of the riverbank and comes to float on the
water of the river. As we can see from this description, the action is any-
thing but simple (as are most events denoted by a verb root in a language),
so on what grounds can we call this a single event? Consider as a starting
point Gentner and Boroditsky’s () notion of the Division of Dominance.
They argue that two principles guide children in the acquisition of mean-
ing for words: cognitive dominance, in which the nature of human percep-
tual engagement with the world presents items to which words as labels are
largely unproblematically attached; and linguistic dominance, in which the
world does not offer salient perceptual bits for labelling and where the lan-
guage being acquired has a major say in the learning of meanings for words.
The acquisition of concrete nouns, e.g. dog, is driven by cognitive domin-
ance. Regardless of wider Quinean (, ) worries about the ontolo-
gies held by different speech communities behind the meaning of this word,
it is a fair bet that all human languages have a term for ‘dog’ and that the
core meaning for it is essentially the same (Berlin ). This is because the
salient perceptual features for the animal referred to by the noun stand out
clearly in human engagement with them. This is not to claim that in English
or any other language ‘dog’ is not a semantically complex notion; merely that
wider cognitive constraints play a determining role in ensuring that they will
be bundled together in exactly this lexeme. But this would not hold for the
meaning of the English word although, the acquisition of which is guided by
linguistic dominance. There is nothing straightforward in human percep-
tual engagement with the world that corresponds to this word, and English
speakers learn its meaning and usage because their language requires them
to do so. Indeed, if the language doesn’t, such as Papuan languages like
Yimas or Watam, it is not even obvious how to translate although, a complex
periphrastic circumlocution being the best that can be mustered. Figure .
(Gentner and Boroditsky : ) presents a summary of the different
word types according to the Division of Dominance parameters.
An important point to note about Figure . is the midpoint position of
the class of verbs. Whatever is denoted by verbs – actions, states, processes –
they do not have perceptual properties of separability and spatial-temporal
continuity – in a word, Gentner and Boroditsky’s concept of individuation –
that lie behind the meanings of nouns. Actions, states, processes do not exist
apart from the entities that instantiate them; they cannot be easily individu-
ated. Barking cannot be separated from the dog which is barking; nor does
barking have any spatial-temporal continuity should the dog cease to exist.
It is this lack of individuation or the inherent relationality of verbs which
 Events and serial verb constructions 83

Division of Dominance

← Cognitive dominance
linguistic dominance →

← open class
closed class →

proper concrete kinship verbs spatial determiners

names nouns terms prepositions conjunctions
&
other
relational
systems

Ida dog grandmother skate on the

spoon uncle enter over and

Figure .: Gentner and Boroditsky’s (: ) Division of Dominance

distinguishes them from nouns; in a word, their predicativity. It is not that

nouns are semantically simple and verbs complex; both are complex, but
verbs differ from nouns in that they entail arguments, participants prototyp-
ically doing or undergoing an event or being in a state. While nouns like dog
are often analysed in formal semantics as predicates with an argument struc-
ture, the arguments are the individual or set of individuals which belong to
the class defined by the noun. The situation with verbs denoting events, like
kill, is very different; the members of its argument structure are the doer
and undergoer of the event denoted by the verb, not an individual or even set
of individuals of the event type denoted by it.
The semantic complexity of verbs, then, is of a different character than
that of nouns. This is clearly reflected in the types of frameworks proposed to
analyse this complexity, specifically predicate decomposition as exemplified
by Baker and Harvey (this volume), building on work of Jackendoff (),
Hovav Rappaport and Levin (), Levin and Rappaport (), Foley and
Van Valin (), and Van Valin and La Polla (). Predicate decomposi-
tion is a schematic breaking down of the meaning of a verb denoting an event
into a set of atomic predicates which define its sub-events, such as in Baker
and Harvey (this volume ()–()). The atomic predicates which denote
the sub-events are the functions whose own argument structures ultimately
project the argument structure of the overall event denoted by the verb, e.g.
the first argument of CAUSE in Baker and Harvey’s formulation is the actor
of the overall event. See Levin and Hovav Rappaport () for a clear and
84 Foley

exemplary exposition of predicate decomposition approaches to the repre-

sentation of verbal meanings and the types of constraints these approaches
impose on the range of meanings that can be lexicalised in a monomorphe-
mic verb and the types of argument structures possible.
The point of this chapter is not to quibble with such approaches. They
are clearly very valuable and have yielded many important insights into the
organisation of the verbal lexicon in diverse languages. There is no doubt
that there are robust and widespread constraints on the possible meanings
bundled into monomorphemic verbs and their argument structures. Yet it
is also clear that verbal inventories as a result of the effect of the principle of
linguistic dominance show greater cross-linguistic variability than nominal
ones. There are languages such as Kuman or Kalam with around just a hun-
dred verb roots; there are no reported languages with only a hundred or so
nominal roots. All languages, for instance, have a root for ‘dog’, but by no
means do all have a root or arguably even a lexical entry for ‘kill’. Just how
the tangled unindividuated stream of events gets reported in inventories
or sequences of linguistic expressions, prototypically verbs, is subject to a
great deal of cross-linguistic variation, an effect of the principle of linguis-
tic dominance. Again, this is not to deny that there are also robust universal
cognitive constraints operative here, which we will see exemplified in the
following sections, but they do underdetermine the result to a much greater
extent than with the category of concrete nouns, for example. Formulas
like ()–() of Baker and Harvey do provide central insights into what
can be lexicalised as a given verb root in a language. But the modality is
vital here, can, not must, and it is the goal of this chapter to explore some
of the cross-linguistic parameters of variation in how the event structures
represented by ()–() of Baker and Harvey can be realised, demonstrat-
ing how SVCs and even CCCs can be alternatives to monomorphemic verb
roots.
To illustrate this point, consider the differing exponents of ‘kill’ across
four New Guinea languages. As it provides a common ground for compari-
son, consider Hovav Rappoport and Levin’s () semantic representation
for this verb through Predicate Composition:
() [[x ACT <MANNER> CAUSE [BECOME [ y <DEAD>]]]]
The meaning is clearly semantically complex: ‘someone/something does
some action in a particular way that causes someone/something to become
dead’, involving two actions, those denoted by ACT<MANNER>, and CAUSE,
a process, BECOME, and a state, DEAD. However, Yimas, like English,
expresses this all in a monomorphemic verb root:
() namot numpran na-mpu-tu-t
man PL pig SG SG O-PL A-kill- PERF
‘The men killed the pig.’
 Events and serial verb constructions 85

Numbami, an Austronesian language of the Morobe region of Papua New

Guinea, uses an SVC (Bradshaw ):
() kolapa i-lapa bola uni
boy  SG R-hit pig dead
‘The boy killed the pig.’
Numbami separately realises the causing event denoted by ACT<MANNER> in ()
with -lapa ‘hit’ and the resulting state with uni ‘dead’. The notion of CAUSE
between the two is an implicature of the SVC, i.e. the meaning of the con-
struction, but not realised by a lexical item. The meaning of -lapa ‘hit’ is
grammaticalised here; () denotes any act of killing, whether by clubbing,
stabbing, or shooting, and is not restricted to causing events with a specific
manner of action, so we can reasonably claim that this sequence represents
a lexicalisation for ‘kill’, albeit one in an SVC rather than a monomorphemic
verb root as in Yimas, but still a unique lexical item.
This claim, however, does not hold for Watam, another Papuan language,
very distantly related to Yimas. Any of the following six expressions will
translate the meaning of ‘kill’, depending on the manner of the action done
which causes the death:
() a. rug- minik- b. arig- minik-
hit die shoot die
c. rutki- minik- d. rutki- yak- minik-
slash die slash cut.open die
e. wak- minik- f. mo- minik-
sever die do die
In Watam, ‘kill’ is always expressed in an SVC involving overt verbs for the
manner of the causing event and for the resulting change of state, minik- ‘die’.
But there is no set lexicalised SVC to express this notion: different manners
of killing are denoted by varying choices of the verbs for the manner of the
causing event, not unlike English they speared/knifed/stabbed/hacked him to
death. These collocations cannot plausibly be claimed to be fixed lexicalisa-
tions as in the Numbami case, for they can be further expanded in yet more
specific SVCs, as in (d) or ():
() arig- turka- minik-
shoot pierce die
When the speaker wishes to omit mention of the specific manner of force
employed in the causing event, for example when the causing event does not
fit one of these, as in ‘kill with poison’, the generic verb of action mo- ‘do’
can be used. But this is just one more SVC expressing ‘kill’, not a more basic
86 Foley

or even lexicalised form for this: mo- minik- ‘do die’ is no more basic an
SVC than arig- minik- ‘shoot die’. Note that in English we have the follow-
ing alternative expressions:
() a. They speared him to death
b. They killed him with a spear
The first parallels the Watam SVC, with the main verb expressing the man-
ner of action of the causing event and the PP the resulting state (this is prob-
ably a family of expressions all related to put to death); the second lexicalises
the whole complex event of killing, causing act and resulting state, but puts
the manner of the act in the form of a PP realising the instrument. In Watam
only the first option is possible (R indicates realis):
() a. min ma ŋg(a)-argi-r minik-ri
 PL  SG FOC-shoot-R die-PAST
‘They shot him to death.’
b. *?min oɲjaŋ ma ŋga-mo-r minik-ri
 PL arrow  SG FOC-do-R die-PAST
‘They killed him with an arrow.’
(b) is ungrammatical because a more specific lexical verb root (a) is avail-
able. These data strongly argue that mo- minik- ‘do die’ is not a more basic
expression for ‘kill’, of which the other SVCs in () are specific expansions.
There is no lexical item ‘kill’ in Watam; specific SVCs are used to denote
distinct types of killing. Given the expandability exemplified by (d) and (),
there is no fixed sequence of lexemes covering the semantic range of ‘kill’
either, but rather strings productively produced by syntactic rules.
While these Watam SVCs are straightforwardly monoclausal (to be defined
more precisely below), fitting Comrie’s property , their event structure is
much more problematic. Are they denoting single events, as Comrie’s point 
asserts? The fact that the causing event can itself be complex, as in (d) and
() (and theoretically can be further extended) suggests otherwise. A binary
SVC like (a) and a ternary like (d) behave differently under negation:
() a. ba-(a)rig- minik-tap
NEG-shoot die-NEG
b. *ba-rutki- yak- minik-tap
NEG-slash cut.open die-NEG
c. ba-rutki- yak- ba-minik-tap
NEG-slash cut.open NEG-die-NEG
 Events and serial verb constructions 87

Binary SVCs for ‘kill’, with just a single causing action and the resulting
state, can be negated by circumfixal negation: the negative prefix ba- pre-
fixed to the first verb and the negative tense suffix -tap/-rap suffixed to
the second. This is definitely not possible for a ternary SVC like (d). In
these SVCs, ba- is prefixed to the first verb, but it must also be prefixed to
the last verb, this last in combination with the negative suffix -tap. This
indicates that, while both types may be monoclausal, the relations that hold
among the verbs in the two types may not be the same, and the difference,
we would argue, resides in the event relations. In the binary SVCs, we have
a straightforward expression for (), a causing event and its intended over-
all stereotypical result. But in the ternary SVC under discussion, we have
two causing actions, the second of which is a deeper, more localised, action
on the body resulting from the first action, and it is clearly the effect of
the second action which results in the death: one slashes at someone with
a machete and cuts them open with it and consequently the person dies.
Note that yak- ‘cut open’ is a transitive verb of action like rutki- ‘slash’, not
an intransitive process verb like minik- die’; i.e. it does not mean ‘become
cut open’. What we have here is a coordination of causing events, where the
second is a more localised and causal event that follows from the first. Even
putting aside the evidence from behaviour under negation, it is hard to see
what grounds there are for claiming that these coordinated events are single
events.
While Watam challenges Comrie’s property  about SVCs, their expres-
sion of single events, the final language of this section, Mangap Mbula
(Bugenhagen ), another Austronesian of the Morobe region of Papua
New Guinea fairly closely related to Numbami, problematises issues around
property  on monoclausality. In this language, ‘kill’ is necessarily expressed
in an overtly conjoined structure:
() a. ti-pun-i ma i-meete
 PL-hit-SG O  SG-die
‘They killed him.’
(a) cannot mean ‘they hit him and he died’; no overt pronoun ni ‘he.she.
it’ is permissible following ma. Contrasting with (a), there is a verb com-
pound formed from these two roots which means ‘diminish, take away’;
compare this with the Numbami example in ():
() b. i-pun-meete tomtom mbura-n
SG-hit-die person strength-PL POSS
‘He took the people’s strength away.’
In an early work, Bugenhagen () gives ‘kill’ as a meaning of this com-
pound as well, but in the final complete grammar (Bugenhagen ), he
88 Foley

specifically does not, and in fact explicitly says that a structure like (a)
with ma is the proper way to translate ‘kill’. Clauses linked by ma are the
normal general way to express cause–effect relationships between clauses in
Mangap Mbula:
() c. zin ti-pun-i ma i-ko
PL PL-hit-SG O SG-flee
‘They hit him and so he fled.’
Mangap Mbula has two conjunctions ma and mi, which could be described
as tighter versus looser respectively, as in this example:
() to-na i-ko mi i-miili ma i-mar kar
then-TOP  SG-flee  SG-return  SG-come village
‘After that, he fled back to the village.’
The translation here reflects a simultaneous temporal relationship between
the act of fleeing and that of returning, i.e. ‘flee back’. If one wishes to say
‘fled and then came back’, with a significant time delay between the two
events, the conjunction used would have to be to ‘and then’ rather than mi.
The conjunction ma has a strong implicative force as opposed to mi; one
can flee some place without having to return somewhere, but returning to a
place generally entails coming to it. Note that the relationship between -miili
‘return’ and -mar ‘come’ mediated by ma, MOTION plus PATH, is typical
of what would be expressed by SVCs in many languages; both -miili ‘return’
and -mar ‘come’ share the same endpoint place argument, the place arrived
at, but -ko ‘flee’ does not, nor does it require an endpoint place. It does,
though, require an origin place argument, a place fled from. But not all uses
of ma are implicative in this way:
() a. aŋ-kam Aibike ma am-la mokleene
SG-get PN PL EXCL-go garden
‘I took Aibike with me to the garden.’ (note the change of verb
agreement here; very atypical of SVCs, but expected in conjoined
clauses)
b. zin ti-ziburru pa yok ma ti-le-la
PL PL-catch.crayfish OBL water PL-enter-go
‘They were catching crayfish and heading inland.’
Example (a) cannot be translated as ‘I got Aibike and we went to the gar-
den’, nor (b) as ‘they were catching crayfish and they were heading inland’;
both of these would require the conjunction mi. Verbs in Mangab Mbula
typically require subject agreement as in (), but overt pronouns niamru ‘we
 Events and serial verb constructions 89

(DL) (EXCL)’, and zin ‘they’ are not permissible here. They would be accept-
able if the conjunction ma was replaced with mi. The conjunction ma also
conjoins nouns in formulaic NPs, demonstrating its doubtless status as a
conjunction: ti ma suga ‘tea and sugar’, zoŋ ma yaŋ ‘sun and rain’, namaana
ma kumbuunu ‘hands and legs’.
On the other hand, these clause sequences linked with ma do exhibit one
common property associated with SVCs: shared scope of negation:
() Keke i-pun ŋge ma i-meete som
Keke  SG-hit pig  SG-die NEG
‘Keke didn’t kill the pig.’ (note that som NEG has scope over both
conjuncts)
The negative particle som is always clause final in Mangap Mbula, as in many
Oceanic languages of New Guinea, and its scope must move leftward into
the clause conjoined with ma before it (if we wanted the scope of the nega-
tive to remain in its clause, i.e. to assert ‘Keke hit the pig but it didn’t die’,
the conjunctions mi or more contrastively tamen ‘however, in spite of that’
would be used). Though this looks like the shared negative polarity of SVCs
illustrated in the Watam examples in (), this may be deceptive. Spreading
negative scope is a widespread feature of CCCs as well. In many New
Guinea languages with CCCs, the negative occurs in the final clause, from
which its scope spreads leftwards to the preceding medial clauses (Reesink
). In some languages this is limited to clauses sharing the same subjects,
but in other languages there is no such restriction. English shows much the
same phenomenon, but in the opposite direction; conjoined clauses with the
shared tense specification and a zero anaphoric subject in the second clause,
a close analogue of a CCC, exhibit negative spreading rightwards (a), but
this is blocked if an overt subject occurs in the second clause, even if its ref-
erent is the same as the subject of the first clause:
() a. I didn’t hit the child and drive on (NEG scope over both
clauses/VPs)
b. I didn’t hit the child and I drove on (NEG scope over first clause)
Consequently, the behaviour of negative scope with ma is no evidence against
a conjoined clause analysis of (). This seems to be a type of CCC, albeit
one in which the clauses are knitted together more closely than in a simple
juxtaposition, the effect of a contrast between ma and mi.
Let us summarise the point of this discussion, a consideration of how
‘kill’, whose meaning is represented by (), is expressed in four languages
of New Guinea. It is expressed as a monomorphemic verb root in Yimas,
a lexicalised SVC in Numbami, a range of SVCs, productively produced
and expandable, in Watam, and a CCC in Mangap Mbula. In the terms
90 Foley

of Baker and Harvey’s typology of options for mapping from a seman-

tic notion of events to the structural notion of clause (their Figure . is
reproduced here as ()):
() Types of LCS syntax relations

a. [ Event ] b. [ Event ] Lexical Conceptual Structure

[ ]Word [[ ] Wdi [ ] Wdj ]Clause Syntax

c. [ Eventi ] [ Eventj ]... [Eventn ] Lexical Conceptual Structure

[ [ ] Wdi ]Clausei [[ ]Wdi ]Clausej..[[ ]Wdi ]Clausen Syntax

d. [ Eventi ][Eventj ][Eventk ]...[Eventn ] Lexical Conceptual Structure

[[ ]Wdi [ ]Wdj [ ]Wdk..[ ]Wdn] Clause Syntax

Yimas corresponds to type a; Numbami possibly to type b; Mangap Mbula

to type c; and Watam to type d. But do we really want to claim that the event
structure of ‘kill’ is as different as these four types suggest? If so, then () is
obviously not an adequate semantic representation of its meaning. One pos-
sible move in the face of this objection is to restrict () to languages of types
a and b. But then we have immediate and obvious problems of interlanguage
translation. Do we really want to claim that the meaning of ‘kill’ is radically
different in Mangap Mbula from Yimas? We think not. Whatever the seman-
tic structure of ‘kill’ is, it is the same in all four languages, and in none of
them is it a simple event. There are undoubtedly semantic constraints on what
can be realised in a monomorphemic lexical verb root, what can be an SVC
(though examples below from Watam indicate they must be pretty weak) and
what must be coordinated, but our knowledge in this area is woefully insuf-
ficient to allow us simply to read off from the formal cross-linguistic variation
in the data, semantic notions like simple or complex eventhood.
As the contrast in productivity and expandability of SVCs in Numbami
and Watam demonstrates, these should not be taken as a unified phenom-
enon, either semantically or structurally. Although both types are mono-
clausal, Watam SVCs unquestionably denote complex multiple events,
consisting of sequences of events with internal relations of cause and effect, a
possibility not countenanced by Numbami SVCs. In this way, Watam SVCs
 Events and serial verb constructions 91

approach the semantic properties of CCCs. On the other hand, Mangap

Mbula CCCs with ma are much more restricted semantically than Watam
SVCs, and, in fact, rather parallel to SVCs in its sister language Numbami.
Indeed, Mangap Mbula example () is as simple (or complex, depending on
how one looks at it) as the Numbami SVC in (). Thus, to really get a handle
on the typology summarised in (), we need to explore the limits of both
what can be compressed into a single clause – and, in turn, the hazy bound-
ary in Papuan languages between SVCs and CCCs – and what counts as an
single event, albeit complex, versus what counts as multiple events. In the
next two sections, we will look at each of these questions in greater detail.

. Clauses and compression

We have been using the terms ‘clause’, ‘monoclausality’, and ‘clause chains’
to this point without definition, and in order to progress much further, we
need to be more explicit. To replace the vague term ‘clause’, we now intro-
duce a distinction between the categories IP and S (Bresnan ). IP is an
endocentric phrasal category projected by its head, finite verb inflectional
categories like tense, mood, or illocutionary force; an IP can be taken roughly
as equivalent to a finite clause. An S is a non-projective exocentric category;
as such it lacks a head. Prototypically, an S constituent contains a predicate
and its arguments and adjuncts. Its contrast with IP means that S is non-
finite; in fact, the most straightforward way to understand it is as a non-finite
or small clause. We can reinterpret what has been analysed as a CCC as a
sequence of conjoined S constituents (i.e. non-finite clauses) functioning as
the complement of the I head that projects the IP, as in ():
(18) IP

I'

S I

S S. . . . S TNS
MOOD
IF

(XP)n V (XP)n V (XP)n V

Note that this explains the typical pattern of stripped down medial verbs ver-
sus a fully inflected final verb in CCCs: the final verb is simply the bearer of
the finite verbal inflectional features that constitute the I head projecting the
IP, as in this Watam example () and its corresponding structure in ():
() namtiŋ un nakan i mbo ŋg(a)-aok-or-o endau nik
boy pot big a OBL FOC-put.inside-R-V house inside
92 Foley

g-idbuku-r atki-ri
FOC-take.inside-R put-PAST
‘The boy put (it, a frog) inside a big pot and took it and put it inside the
house.’
(20) IP

S I

S S I

NP PP PP V

N NP P V NP P V V

N A D N

namti un nakan i mbo g(a)-aok-or-o endau nik g-idbuku-r atki-ri

child pot big a OBL FOC-put.inside-R-V house inside FOC-take.inside-R put-PAST

The contrast between IP and S requires a revision of Baker and Harvey’s

typology of semantics to structure mapping in (). Their type c needs to be
separated into two types, depending on whether the syntactic units involved
are S (CCCs) or IP (standard full clausal coordination) as in ():

(21)
(a) [Eventi ] [Eventj ]... [Eventn ] Lexical Conceptual Structure

[ [ ]Wdi ]Si [[ ]Wdi ]Sj... [[ ]Wdi ]Sn Syntax

(b) [Eventi ] [Eventj ]... [Eventn ] Lexical Conceptual Structure

[ [ ]Wdi ]IPi [[ ]Wdi ]IPj... [[ ]Wdi ]IPn Syntax

Furthermore, the evidence from Mangap Mbula in examples such as ()

suggests a new type which they had not considered: when the event is spread
over more than one S constituent within a single IP:

(22) [ Event ] Lexical Conceptual structure

[ [ [ ] Wdi ] Si [[ ]Wdi ]Sj ... [[ ]Wdi ]Sn ]IP Syntax

 Events and serial verb constructions 93

One constraint we could draw out from () is that a unitary event can
never be realised in multiple IPs, but can be spread out over multiple S
constituents. This claim tallies well with Bohnemeyer et al.’s () argu-
ment for a Macro Event Property which constrains what types of syntactic
structures can be construed as expressing a unitary event. In essence, the
Macro Event Property requires that temporal operators, of whatever type –
adjuncts, adverbial clauses or tense affixes – have scope over all compo-
nent sub-events within the overall unitary event. As tense is the prototype I
head which projects IPs, there is a close correlation between the structural
domain over which the Macro Event Property must hold and the category
IP; we leave it to future research to explore further the ramifications of this
correlation.
() is possible because the typical verb-final typology of Papuan languages
plus the normal heavy ellipsis of presupposed NP arguments results in CCCs
often looking indistinguishable from SVCs. Note that if we delete the XPs
from (), how do we know whether () or () is the correct structure:

CCC with Ø NPs

IP
(23)
S I

I
V

V V V

SVC IP
(24) S I

V
I

V V V

The surface homophony in many cases between SVCs and CCCs leads to
an overlap between these in many Papuan languages. This does not mean
we cannot distinguish them, as we shall see below, but it is often the case
that the same types of event structures can be realized alternatively as
either an SVC or a CCC. It is patently not the case that we can claim with
confidence that CCCs and SVCs always differ in event structure, multiple
versus single events respectively. Consider the following lines drawn from
a Yimas text:
94 Foley

() m-um mnta pu-n-tay-mpi-kwalca-k

NR DIST-I PL then PL O- SG A-see-SEQ-rise-IRR
awkura-mpi pu-n-api-k man-an
collect-SEQ  PL O- SG A-put.inside-IRR cult.house-OBL
‘Then he saw them and took off and collected them and put them
inside his (male) cult house.’
ma-m ma ŋarŋ mnta
other-I PL other  day removed then
pu-ŋkl-mampi-wampaki-k
 PL O- PC A-again-throw-IRR
‘Another day, they sent out some more people.’
m-um pu-n-mampi-awkura-mpi-api-k
NR DIST-I PL  PL O- SG A-again-collect-SEQ-put.inside-IRR
paympan
eagle V SG
‘Again he, the eagle, collected them and put them inside.’
Consider the two instances of ‘collect and put inside’ (marked in bold) spread
over three sentences. In the first token, this sequence of events is expressed
in a CCC. This is indicated by the marking of the tense -k IRR and the
pronominal agreement affixes on the second verb api- ‘put inside’, and the
stripped down verb awkura- ‘collect’ before it, marked merely with the suf-
fix -mpi SEQ. The second token, found in the final sentence, is an SVC,
clearly indicated by the pronominal prefixes preceding the sequence of both
verb roots, which are juxtaposed with the sequential suffix -mpi. Note that
the CCC and the SVC here look very similar: what distinguishes them is
the location of the pronominal agreement affixes (although there are other
tests which can be used to do so, some illustrated in section .). The real
question here is why use a SVC in the second token when a CCC was used
in the first. Clearly we do not want to argue that it is a semantic difference
in eventhood; it is the sequence of events of collecting and putting inside
in both cases. The crucial difference is the presence of mampi- ‘again’ in
the second token, the SVC. This is an incorporated adverbial (Foley ),
one which is treated like a bound affix. In order to modify both the events
of collecting and putting inside, as clearly intended here, the SVC is essen-
tial; if a CCC was used, mampi- ‘again’ would only modify the verb root
to which it was attached. To modify both in a CCC, it would need to be
repeated on both verb roots in each S of the CCC, but this is unnecessar-
ily prolix when the SVC choice is available, and in any case, via a Gricean
 Events and serial verb constructions 95

Maxim of Manner, such phrasing would implicate that an unusual situation

holds (Levinson ), which is clearly not meant here. In this example the
SVC is simply mandated structurally, by the grammatical resources avail-
able in the language. It says nothing about the way the events are to be
construed as single or multiple. The event structure is clearly the same in
both instances.
Generally, however, we can set up a paradigm of Yimas SVCs and CCCs
in terms of the types of meanings of event relationships that they do proto-
typically denote:
() a. narm pu-tpul-kamprak-r-akn
skin VII SG  PL S-hit-break-PERF- SG D
‘They hit and broke his skin.’
b. mparŋkat ya-n-park-mpi-kapik-mpi-wark-t
branch V PL V PL O- SG A-split-SEQ-break-SEQ-tie-PERF
‘He split the branches, broke them and tied then together.’
c. kaprak-mpi yaŋi-ɲan na-mp-ayŋ
cut.up-SEQ pot VIII SG-OBL V SG O- DL A-put inside
‘Having cut (it) up, they put (it) in a pot.’
d. tmal kray-mpi ya-kay-am-wat amtra
sun V SG dry-SEQ V PL O- PL A-eat-HAB food V PL
‘The sun having dried it, we always eat the food.’
Here is a paradigmatic opposition of syntactic compression tied to an intui-
tive notion of event compactness. (a) is an SVC with two verb roots simply
juxtaposed. The meaning is very specific: a very close cause–effect relation-
ship between the two events denoted by the verb roots. The effected result
must follow directly and immediately after the causing event: no period of
time can separate the spans of the causing event and the resulting state, and
the linking suffix -mpi SEQ is prohibited here. Now consider (b), also an
SVC, but this time the roots are linked by the affix -mpi SEQ , which in
sharp contrast to (a) is obligatory here. Unlike (a), no necessary cause–
effect relationship holds between the verb roots in this SVC; the events just
follow one another in time. Still, certain constraints must hold: all events
denoted by the verb roots in the SVC must be done by the same actor (n- 
SG A) and happen to the same object (mparŋkat ‘branches’), and any time
delay between the sequential events must be relatively fleeting. Any spatial
or temporal modifiers must hold of all events denoted by the verb roots in
the SVC. Finally, consider the CCC examples (c, d). Note that while in
(c), the two verbs share the same actor and object (although the Oblique
NP is only a constituent of the second S constituent in the chain), this is not
96 Foley

true of (d), which has distinct actors. There is no need of shared spatial
or temporal modifiers in Yimas CCCs. Further, there can be a very consid-
erable gap in time between the event expressed in one S constituent in the
CCC and that in the next, as in (d). As the representation of () entails,
the only crucial interdependency is one of tense; only the verb of the final S
constituent is inflected for tense, but this absolute tense specification holds
for all the S constituents in the chain.
While morphologically much simpler than its very distant relative Yimas,
when it come to SVCs, Watam is actually rather more elaborate. At first blush
this is somewhat surprising. Yimas, like unrelated Kalam (Pawley and Lane
), has a rather small, almost closed class of verbal roots. Given the prin-
ciple of Division of Dominance and the role of linguistic dominance in the
acquisition of verbs, widespread use of SVCs in Yimas to describe events is
hardly surprising: there is simply such a limited stock of core verb roots that
concatenation in SVCs is mandated for expressibility. But Watam is by no
means so lexically deficient in verb roots; it has many hundreds of them, some
with surprisingly specific meanings, such as mermero- ‘move in an S-shaped
movement along the ground (like a snake) or wave a firestick in the air to
draw geometrical shapes, typically loops (like American kids do with spark-
lers during th of July celebrations)’! In spite of this fact, Watam discourse
is awash with SVCs, of more types and of greater elaboration than those of
Yimas. Yimas SVCs very rarely exceed three verb roots; Watam SVCs of five
verb roots are common. Watam SVCs and CCCs are formally very similar.
Non-final verbs in SVCs and those of medial S constituents in CCCs are
only marked for status, by -r for Realis and ø/mbe for IRRealis. SVCs do
not constitute single words in Watam, unlike Yimas (there is a true class of
compound verbs in Watam distinct from SVCs, as determined by being sub-
ject to word level phonological rules). Again, with the very extensive ellipsis
of presupposed NPs, CCCs and SVCs often look formally alike. But there are
three features which do distinguish CCCs from SVCs. First, the final verbs
of medial S constituents in CCCs may be suffixed with the utterance internal
pre-pause echo vowel; such an echo vowel is never permitted inside an SVC:
() a. un an ajuna ŋg(a)-irki-r-i
pot PROX straight FOC-go.down-R-V
wakar ŋga-pka-r-a un burburu-ri
ground FOC-throw-R-V pot shatter-PAST
‘The pot went straight down, hit the ground and shattered.’
b. ma markum ŋga-rugu-r-(*u) minik-ri
 SG pig FOC-hit-R-(*V) die-PAST
‘He clubbed the pig to death.’
 Events and serial verb constructions 97

Second, verbs in CCCs can individually take the FOCus prefix ŋga-, as in
(a). In SVCs, only the first verb may be prefixed with this, as in (b)
and ():
() nomgai or ŋga-soki-r uŋgu-r apuk-ri
crocodile leg FOC-hold-R pull-R swallow-PAST
‘The crocodile grabbed (it, a chicken) by the leg, pulled (it) and
swallowed (it).’
Any SVC internal position of ŋga- FOC is ungrammatical:
() *or soki-r uŋgu-r ŋg(a)-apuk-ri
leg hold-R pull-R FOC-swallow-PST
Finally, verbs in medial S constituents are marked by a distinctive falling
pitch (S-final pitch), not as low and not over such a large pitch range as those
of final S constituents (e.g. IP-final pitch), but quite distinct nonetheless.
Verbs within SVCs have no such fall in pitch: all are spoken with roughly the
same pitch except the last, which takes on the distinctive falling pitch of the
end of an S or IP.
We noted above that Yimas SVCs were in paradigmatic opposition with
CCCs, with an SVC typically expressing a closer integration of events occur-
ring in a sequence than the corresponding CCC. The same holds true for
Watam, but, given both the fact that Watam SVCs are separate words rather
than a single polysynthetic word and that they are more extensive, a greater
delicacy of structure and its relationship to the unfolding sequence of events
can be discerned in Watam. Consider the SVC in (a) and its correspond-
ing CCC in (b):
() a. min ma mo ndo-r aŋgi-r agiga-r
PL SG OBJ see-R get-R take.up-R
rug-ur minik-ri
hit-R die-PAST
‘They saw him, got him, carried him up and beat him to death.’
b. min ma mo ndo-r-o aŋgi-r agiga-r-a
 PL  SG DAT see-R-V get-R take.up-R-V
rugu-r minik-ri
hit-R die-PAST
They saw him, got him and took him up, and beat him to death.’
The SVC is composed of five verbs, but the CCC contains only three con-
joined S constituents; note each verb of (a) cannot govern its own S in
98 Foley

the CCC paraphrase. In particular, the SVCs aŋgi- agiga- ‘get and take up’
and rug- minik- ‘beat to death’ may not be torn asunder and put into sep-
arate S constituents. In fact, to do so results in an ungrammatical string,
as in (b). What is required in Mangap Mbula to express this event, as a
CCC, is proscribed in Watam. Certain types of event sequence relations in
Watam require SVCs; others simply permit it. The constraints on what is
allowed do seem particularly weak in this language, as (a) indicates. But
where SVCs are required corresponds to prototypical uses of SVCs cross-
linguistically, for instance, causing action with immediately resulting state,
or motion plus path followed, or simultaneous events performed by the same
actor. These event types are exactly those expressed by the simplest SVC
types in Yimas, bare juxtaposition of verb roots (a). Watam differs from
Yimas in permitting much more complex SVCs of the type exemplified by
(a), within which substructures of the types corresponding to (a) can
be concatenated, a possibility Yimas simply will not countenance; a fact
related no doubt to the tightly integrated polysynthetic word structure of
Yimas SVCs. We can represent the basic structure of the SVC in (a) as
(), adopting a modified version of the Baker and Harvey typology:
(31) Semantics [Event] [Event] [Event]

[sub-event] [sub-event] [sub-event] [sub-event]

Syntax [ndo-] [ [a gi-] [agiga-] ] [ [rug-] [minik-] ]

'see' 'get' 'take.up' 'hit' 'die'

Each full event may correspond to an S constituent in a CCC paraphrase,

but the sub-events cannot. Of course, the Macro Event Property defined in
terms of the verbal inflection tense will hold for the whole five verb concate-
nation in either an SVC or a CCC because both constructions are dominated
by a single IP node.
But a question still remains: why would a speaker use (a) in prefer-
ence to (b) and vice versa? Because of the freedom and extent of Watam
SVCs this is not easy to answer. There do not seem to be obvious struc-
tural forces at work in this largely analytical language, parallel to the poly-
synthetic profile of Yimas responsible for (). For Watam the answer
seems mainly textual and discoursal. An SVC like (a) strongly highlights
the idea that the events follow each other in rapid succession; no spatial
or temporal modifiers can occur between the verbs in the SVC or modify
any of them independently. This again illustrates the pragmatic Gricean
heuristic principles: simple description, stereotypical exemplification, e.g.
rapid sequence of events. This rapid sequence is not highlighted in (b);
while the events could unfold as such, the description does not force such
 Events and serial verb constructions 99

an interpretation, indeed it invites one to consider an alternative reading

of perhaps some delay, and indeed independent spatial or temporal modi-
fication of each S constituent is perfectly possible here. But there is still
another, perhaps more important, factor. Looking at the distribution of
complex SVCs such as (a) across a range of narrative texts, it is very
striking that they typically occur in the final S constituent of a CCC. Even
if medial S constituents contain SVCs (as they commonly do, see (b)),
they are normally simpler than those found in the final S. This is because
the final S expresses the main point, the most highlighted, foregrounded
event of the whole sentence. Consequently if there is a string of interrelated
sequential events that the speaker wishes to highlight, the ‘punchline’, we
might say, of the sentence, then an SVC is the ideal way to do so. This is
not to claim that medial S constituents are backgrounded; they are not, in
Watam or in Yimas. Truly backgrounded information is best expressed in
subordinated clauses. But the binary foregrounded–backgrounded distinc-
tion is insufficient for our purposes here. Events described in the final S are
more important, more highlighted, than those expressed in the medials that
lead up to them: that is why they are expressed in a final S and not in yet
another medial one. In a sense, they are the climax of the sentence, and, if
complex, what better way to present them than in an SVC.

. Event sequences in types of SVCs

Our discussion of the event structure of (a), represented in (), was an
entrée to this section in which we will investigate the substructure of com-
plex event descriptions in SVCs in Watam. We will probe a few types of
SVCs with various grammatical tests to see how they behave. As will become
clear, not all Watam SVCs are created equal. Some are SVCs pure and sim-
ple, and no other structural realisation is possible. Others are SVCs by con-
venience, we might say; structural probing will reveal them to have much
weaker syntactic integration than the types of absolutely prescribed SVCs.
This distinction roughly coincides with one proposed by Van Staden and
Reesink (), which they termed component versus narrative serialisa-
tion. They noted that while component serialisation was present in all the
languages they surveyed in the Indonesia/New Guinea region and identified
as having SVCs, narrative serialisation was essentially restricted to Papuan
languages. This is undoubtedly due to the overlap in both form and meaning
between SVCs and CCCs in these languages.
Let us start with the Watam equivalent to the Yimas example in (a):
() yak kor uŋg-ur irik-ri
SG canoe pull-R go.down-PAST
‘I pulled the canoe into the water.’
100 Foley

The SVC behaves as a cohesive unit in non-finite nominalisations like

():
() yak [ kor uŋ irik ] -na to
SG canoe pull go.down-POSS enough
‘I am able to pull the canoe into the water.’
When we form an imperative version of this SVC, it is formed by simple jux-
taposition of the verb roots; the suffix -mbe used for irrealis verbs in medial
S constituents in CCCs is ungrammatical:
() a. kor uŋ irik
canoe pull go.down
‘Pull the canoe into the water!’
b. *kor uŋ-(m)be irik
canoe pull-IRR go.down
While polysynthetic Yimas has a number of verbal inflections to mark
aspect, more analytic Watam uses a range of verbal affixes and specialised
verbs in SVCs for this function. In particular, progressive is marked by a
circumfix ma-. . .-tak, while perfective uses mamai- ‘finish’ in an SVC:
() a. yak kor uŋg-ur ma-irik-tak
SG canoe pull-R PROG-go.down-PROG
‘I’m pulling the canoe into the water.’
b. yak kor uŋg-ur irki-r mamai-r
SG canoe pull-R go.down-R finish-PERF
‘I finished pulling the canoe into the water.’
Note that the progressive affixation occurs on the final verb of the SVC,
but its scope extends over both verbs; similarly, while the perfective verb is
the final verb in the SVC, all preceding verbs are within its scope. No alter-
native construal is possible. If, for instance, one wishes to say ‘I pulled the
canoe and it’s now going down into the water’, a complex sentence would be
required, with the first clause most likely subordinate:
() yak kor uŋ-rape, ma-irik-tak
SG canoe pull-SUBD PROG-go.down-PROG
‘Me having pulled the canoe, (it’s) going down into the water.’
The behaviour of negation is tricky in Watam SVCs, as it is in many seri-
alising languages. Negation is indicated by the circumfix ba-. . .-tap/rap.
 Events and serial verb constructions 101

This may not be spread over both verbs, but occurs on the final verb, with
the first verb bearing simply the prefix ba- or nothing:
() a. yak kor ba-uŋg-ur ba-irik-tap
SG canoe NEG-pull-R NEG-go.down-NEG
‘I didn’t pull the canoe into the water.’
b. yak kor uŋg-ur ba-irik-tap
SG canoe pull-R NEG-go.down-NEG
‘I didn’t pull the canoe into the water.’
However, this constraint does not seem to hold for all SVC types; in particu-
lar, for cause–effect pairs, split negative morphology does seen possible, as
in this textual example:
() namak yaoŋ an ma na ba-wandai-r wiski-rap
eye good PROX SG POSS NEG-wash-R clean-NEG
‘(He) didn’t wash this good eye of his clean.’
Turning to a more complex ternary SVC, consider ():
() namot i yor i aŋgi-r pika-r irki-ri
man a egg a get-R throw-R go.down-PAST
‘A man got an egg and threw it down.’
Again the SVC behaves as a cohesive unit in non-finite nominalisation:
() namot i [ yor i aŋgi pika irik- ]-na to
man a egg a get throw go.down-POSS enough
‘A man is able to get an egg and throw it down.’
In contrast to (), the elements in this SVC behave differentially under
imperative formation:

() yor i aŋgi-mbe pika-(*mbe) irik

egg a get-IRR throw-IRR go.down
‘Get an egg and throw it down!’
The first verb in the SVC aŋgi- ‘get’ must occur with the irrealis suffix for
medial verbs in CCCs -mbe, while the middle verb pika- ‘throw’ cannot. In
other words, when the SVC in () is subjected to imperative formation, the
first verb must revert to being within its own S constituent in a CCC, while
the second cannot, but must remain in an SVC with irik- ‘go down’. This
indicates that the semantic relationship between pika- ‘throw’ and irik- ‘go
down’ is closer – sub-events of manner of motion plus path in one unitary
102 Foley

event – but that aŋgi- ‘get’ is more weakly integrated, simply a preceding
event in the unfolding sequence, as in ():
(42) Semantics [Event] [Event]

[sub-event] [sub-event]

Syntax [a gi-] [ [pika-] [irik-] ]

'get' 'throw' 'go.down'

The behaviour under negation supports this analysis. The SVC can be
negated as a whole: in this case, the negative circumfix will be on the final
verb and the first verb will bear ba- or no affixation; the middle verb may not
carry any affixation at all:
() a. namot i yor i ba-(a)ŋgi-r pika-r ba-irik-tap
Man a egg a NEG-get-R throw-R NEG-go.down-NEG
‘A man didn’t get an egg and throw it down.’
b. namot i yor i aŋgi-r pika-r ba-irik-tap
man a egg a get-R throw-R NEG-go.down-NEG
‘A man didn’t get an egg and throw it down.’
If the middle verb does bear the negative prefix ba-, this forces a reading in
which the negative only has scope over the final two verbs in the SVC; the
first verb must remain positive in polarity:
() namot i yor i aŋgi-r ba-pika-r ba-irik-tap
man a egg a get-R NEG-throw-R NEG-go.down-NEG
‘A man got an egg but didn’t throw it down.’
() contradicts Comrie’s Property : here all the verbs in the SVC do not
agree in polarity (nor, as an aside, do verbs in SVCs necessarily agree in
person–number marking or aspect; a number of contrary examples have been
reported from languages in the Pacific (Crowley ); SVCs, of course, must
agree in tense–mood, as these are I head categories projecting the maximally
dominating IP node. The reason () is possible is that the negation only holds
of the second major event of the event sequence expressed by the SVC, as sche-
matised in (). Scope of negation is not structurally driven in Watam, as it is
in Yimas. It cannot be described as sentential or phrasal or word level, because
there is no obvious constituent structure of () that includes the second two,
but excludes the first. Rather it seems sensitive to event structure: its scope can
be any or all of the major events in the top tier of the event structure sequence
of (), and the relative placement of the negative affixes reflects this.
 Events and serial verb constructions 103

Now consider another ternary SVC of a slightly different type in ():

() namot tau da-gwe-r witki-r amb-ri
man sugarcane FOC-bite-R tear-R eat-PAST
‘The man bit the sugarcane off and ate it.’
Again the SVC straightforwardly occurs in a non-finite nominalisation:
() namtiŋ [ tau da-gwe witki am ]-na to
boy sugarcane FOC-bite tear eat-POSS enough
‘The boy is able to bite off the sugarcane and eat it.’
But when we apply the imperative test, we find the first two verbs grouped
together against the last:
() a. tau da-gwe witki-mbe am
sugarcane FOC-bite tear-IRR eat
‘Bite the sugarcane off and eat it.’
b. *tau da-gwe-mbe witki am
sugarcane FOC-bite-IRR tear eat
c. *tau da-gwe witki am
sugarcane FOC-bite tear eat
Again this is explained in terms of the event structure described in (): the
first two verbs describe a cause-resulting state relationship, bite and be torn
off, which, as we saw above, must be realised as an SVC in Watam, so that
a CCC option is not available here. On the other hand, the action of am-
‘eat’ just follows in an expected sequence from the previous event, a weaker
semantic relationship, and hence this verb is forced into its own S constitu-
ent in a CCC under imperative formation. The event structure of () can be
diagrammed as ():
(48) Semantics [Event] [Event]

[sub-event] [sub-event]

Syntax [ [gwe-] [witki-] ] [am-]

'bite' 'tear' 'eat'

Unfortunately, the evidence from negation is inconclusive here: either

arrangement of the negative affixation results in a reading of the entire SVC
being negated:
104 Foley

() a. namot tau ba-gwe-r witk-r ba-(a)m-tap

man sugarcane NEG-bite-R tear-R NEG-eat-NEG
‘The man didn’t bite the sugarcane off and eat it.’
b. namot tau gwe-r witki-r ba-(a)m-tap
man sugarcane bite-R tear-R NEG-eat-NEG
‘The man didn’t bite the sugarcane off and eat it.’
This is because of the basic principle of the language that negation must
be marked on the final verb with the negative circumfix (see () and ()),
often in combination with the prefix ba- on the initial verb. So if we want
to force a reading in which only the final verb of an SVC is negated, e.g. the
second full event in (), then a CCC is necessary:
() namot tau da-gwe-r witki-r-i, (ma) ba-(a)m-tap
man sugarcane FOC-bite-R tear-R-V SG NEG-eat-NEG
‘The man bit the sugarcane off, but (he) didn’t eat it.’
Unlike some other languages such as Barai (Olson ), which appar-
ently allow verbs in SVCs to have differing specifications for aspect, Watam
does not generally permit this: aspectual markers must have scope over the
whole SVC:
() a. namot tau da-gwe-r witki-r amb-ar mamai-r
man sugarcane FOC-bite-R tear-R eat-R finish-PERF
‘The man finished biting off the sugarcane and eating it.’
b. namot tau da-gwe-r witki-r amb-ar yoro-r
man sugarcane FOC-bite-R tear-R eat-R COMPL-PERF
‘The man bit off the sugarcane and ate all of it.’
yoro- is an aspectual verb which indicates that the event has gone to comple-
tion with the object being totally affected by it. Most Papuan languages have
this aspect category, which we gloss as completive. The one exception to this
generalisation about the scope of aspect concerns the progressive:
() namot tau da-gwe-r witki-r ma-(a)m-tak
man sugarcane FOC-bite-R tear-r PROG-eat-PROG
‘The man bit off the sugarcane and is eating it.’
In this case the progressive aspect only has scope over the act of eating, not
the whole SVC. If we wish to force a reading in which the man is currently
engaged in biting off sugarcane and eating it, we would simply affix the final
verb with the present tense suffix -ta, which as a tense I inflection must have
scope over the whole SVC:
 Events and serial verb constructions 105

() namot tau da-gwe-r witki-r am-ta

man sugarcane FOC-bite-R tear-R eat-PRES
‘The man is biting the sugarcane off and eating it.’
When negation is applied to an aspectual verb, the negative circumfix
attaches to it, as the final verb, and no negative affixes are permitted to occur
on any of the preceding verbs:
() namot tau da-gwe-r witki-r amb-ar ba-mamai-rap
man sugarcane FOC-bite-R tear-R eat-R NEG-finish-NEG
‘The man didn’t finish biting off the sugarcane and eating it.’
This, of course, is entirely to be expected. As () and its translation exem-
plify, it is the aspectual verb which is being negated, not the other events in
the SVC. They are asserted to have occurred: i.e. the man did bite the sugar-
cane and eat it; he just didn’t finish doing that.
For a final brief example, consider this quaternary SVC:
() ma tatariki-r nopai-r iri-r ga-ri
SG yell-R swim-R come-R go.up-PAST
‘He yelled, swam over and came ashore.’
Applying a new test, it is interesting to see where one can and cannot insert
overt subject pronouns; the presence of an overt subject pronoun establishes
the first verb as being in its own separate S constituent:
() ma tatariki-r (ma) nopai-r iri-r (*ma) ga-ri
SG yell-R SG swim-R come-R SG go.up-PAST
‘He yelled, (he) swam over and came ashore.’
This establishes that the first verb in the SVC of () is more weakly linked
to the rest, as it can occur in its own S in a CCC. Behaviour under impera-
tive formation confirms this:
() tatariki-mbe nopai iri-mbe ga
yell-IRR swim come-IRR go.up
‘Yell and swim over and come ashore!’
Interestingly, although a subject pronoun is proscribed there, under impera-
tive formation, the -mbe suffix is obligatory between iri- ‘come’ and ga- ‘go
up’. It is prohibited between nopai- ‘swim’ and iri- ‘come’. This indicates
that these last two form a required SVC of manner of motion plus path that
cannot be broken up, like previous examples () and (), while the other
two verbs are more weakly linked in a relationship of sequence, which in
an imperative construction forces them to revert to distinct S constituents
within a CCC. The event structure of () can be represented as:
106 Foley

(58) Semantics [Event] [Event] [Event]

[sub-event] [sub-event]

Syntax [tatariki-] [ [nopai-] [iri-] ] [ga-]

'yell' 'swim' 'come' 'go.up'

Given the conflicting evidence from subject pronoun insertion and imper-
ative formation, () is probably not a perfect representation of the event
sequence of (). It treats the relationship of the initial and final verbs to the
middle pair alike, but the rejection of subject pronoun insertion before the
final verb indicates that its relationship to the middle pair is somewhat dif-
ferent from that of the first verb. Indeed, the final verb bears a rather strong,
albeit non-binding, implicature to the middle pair: if one swims toward some
place on land, then it can reasonably be expected that one will come ashore.
But if one yells, there clearly can be no expectation that one will jump in the
water and start swimming toward some place.
Behaviour under negation confirms that the event relations are not entirely
equivalent:
() a. ma tatariki-r ba-nopai-r iri-r ba-ga-tap
SG yell-R NEG-swim-R come-R NEG-go.up-NEG
‘He yelled, but didn’t swim over and come ashore.’
b. ma tatariki-r nopai-r iri-r ba-ga-tap
SG yell-R swim-R come-R NEG-go.up-NEG
‘He didn’t yell, swim over, and come ashore.’
(a) is the expected example in which the scope of negation is over the
final two events of (), but not the first. Example (b), with the nega-
tive only realised on the final verb in the SVC, can only be felicitously
read as having scope over the entire SVC, all three events described
by it, as with (b). Again to force a restriction of negation to the final
event, minimally a CCC would be needed (a), but, given the denial
of the normal implicature of coming ashore, more likely a rather more
emphatic construction using two juxtaposed IPs, i.e. full clausal coordi-
nation, would be used:
() a. ma tatariki-r nopai-r iri-r-i, ba-ga-tap
SG yell-R swim-R come-R-V NEG-go.up-NEG
‘He yelled and swam over, and didn’t come ashore.’
b. ma tatariki-r nopai-r iri-r,
SG yell-R swim-R come-PERF
 Events and serial verb constructions 107

karir (ma) ba-ga-tap

but SG NEG-go.up-NEG
‘He yelled and swam over, but (he) didn’t come ashore.’

. Conclusion
Gentner and Boroditsky’s () proposal of the Division of Dominance and
the consequent important role of linguistic dominance in the acquisition of
verb meanings leads us to expect very significant cross-linguistic differences
in their semantics. While all languages plausibly have a word for ‘dog’ with
essentially equivalent meanings, this is not true for ‘kill’. As we saw in sec-
tion ., ‘kill’ is expressed in a variety of differing ways, from a monomor-
phemic verb root in Yimas to a CCC in Mangap Mbula. Both Numbami and
Watam uses SVCs to express ‘kill’, but in quite distinct ways. The Numbami
candidate is a complex, albeit fixed expression, a good candidate for a lexical
item. But the Watam SVCs are often flexible and expandable expressions;
they are a family of constructions, not lexical items. Hence, the Numbami
and Watam SVCs are, strictly speaking, not comparable. The surface simi-
larity of being SVCs hides deeper differences. SVCs are a number of differ-
ent things that simply look alike structurally, but on deeper probing turn out
to exhibit very divergent properties. None of Comrie’s properties of SVCs
turn out to hold of all SVCs investigated in this chapter save monoclausality,
but that criterion is clearly not sufficient, as many constructions are realised
in single S constituents, but are not SVCs. Are there any universal defining
properties of SVCs? Probably not, although the term may still prove useful
as a convenient descriptive label like reduplication.
And this variation holds within languages. Our closer look at Watam
SVCs in section . shows significant differences in their behaviour, so that
even to talk of a Watam SVC is somewhat of a misnomer: some are para-
phrasable in CCCs, some are not; some require the scope of negation to
include all elements, some do not; some allow subject pronoun insertion,
some do not, etc. But when we look more closely at the meanings of the
different types of SVCs, we find that those whose unity is inviolable cor-
respond to the most prototypical SVCs in other languages, and commonly
describe events that are realised as monomorphemic verbs or verb plus satel-
lite structures in Talmy’s () typology of macro-events, such as causing
action and resulting state or manner of motion plus path. This strongly sug-
gests that the principle of cognitive dominance and hence robust universal
constraints such as those discussed in Baker and Harvey (this volume) also
play an important role in the semantics of events, as we might expect, given
Gentner and Boroditsky’s () midpoint positioning of the class of verbs
in their scale of the Division of Dominance. It is exactly the perceptual/
cognitive relations which most interest humans, e.g. causation and motion,
108 Foley

which must be realised in an SVC in Watam. Language-specific facts cast

through the principle of linguistic dominance stipulate that the realisation
of events expressing cause–effect or motion–path in Watam are SVCs, and
open expandable SVCs at that, rather than monomorphemic verb roots like
kill or enter. But the fact that such event types cannot be expressed in CCCs
is due to wider, presumably cognitive, constraints on how an event can be
expressed: sub-events of a single, albeit complex, event cannot be spread over
distinct S-constituents in Watam. But they can be in Mangap Mbula. But
even there sub-events of a single, but complex, event must be expressed in
S constituents linked with the close conjunction ma, rather than the weaker
less implicative mi. In both Watam and Mangap Mbula, sub-events of com-
plex, but nonetheless coherent and overall unitary, events must be expressed
in syntactically more compact structures than those available to sequences of
distinct events. Similar constraints apply in the two languages, though the
linguistic realisations differ. So, in a nutshell, in the domain of the structure
of events, cognitive dominance proposes, but linguistic dominance disposes.

References
Aikhenvald, A. . Serial Verb Constructions in Typological Perspective. In
Aikhenvald, A. and Dixon, R. (eds) Serial Verb Constructions: A Crosslinguistic
Typology, –. Oxford: Oxford University Press.
Baker, B. and Harvey, M. This volume. Complex Predicate Formation.
Berlin, B. . Ethnobiological Classification. Princeton: Princeton University
Press.
Bohnemeyer, J., Enfield, N., Essegbey, J., Ibarretxe, I., Kita, S., Luebke, F., and
Ameka, F. . Principles of Event Encoding: The Case of Motion Events.
Language : –.
Bradshaw, J. . Subject Relationships Within Serial Verb Constructions in
Numbami and Jabêm. Oceanic Linguistics : –.
Bresnan, J. . Lexical-Functional Grammar. Oxford: Blackwell.
Bugenhagen, R. . Agentivity and Iconicity in Mangap-Mbula. In Harlow, R and
Hooper, R. (eds) VICAL : Oceanic Languages; Papers from the Fifth Inter-
national Conference on Austronesian Linguistics, –. Auckland: Linguistic
Society of New Zealand.
. A Grammar of Mangap-Mbula: An Austronesian Language of Papua New
Guinea. Pacific Linguistics C-.
Comrie, B. . Serial Verbs in Haruai (Papua New Guinea) and Their Theoretical
Implications. In Bouscaren, J., Franckel, J., and Robert, S. (eds) Langues et
langage: Problèmes et raisonnement en linguistique, mélanges offerts à Antoine
Culioli, –. Paris: University Presses of France.
Crowley, T. . Serial Verbs in Oceanic: A Descriptive Typology. Oxford University
Press.
Foley, W. . The Yimas Language of New Guinea. Stanford, CA: Stanford
University Press.
 Events and serial verb constructions 109

Foley, W. and Van Valin, R. . Functional Syntax and Universal Grammar.
Cambridge: Cambridge University Press.
Gentner, D. and Boroditsky, L. . Individuation, Relativity and Early Word
Learning. In Bowerman, M. and Levinson, S. (eds.) Language Acquisition and
Conceptual Development, –. Cambridge: Cambridge University Press.
Hovav Rappaport, M. and Levin, B. . Building Verb Meanings. In Butt, M.
and Geuder, W. (eds) The Projection of Arguments: Lexical and Compositional
Factors, –. Stanford: CSLI Publications.
Jackendoff, R. . Semantic Structures. Cambridge, MA: MIT Press.
Levin, B. and Rappaport Hovav, M. . Argument Realization. Cambridge
University Press.
Levinson, S. . Presumptive Meanings: The Theory of Generalized Conversational
Implicature. Cambridge, MA: MIT Press.
Olson, M. . Barai Clause Junctures. PhD dissertation, Australian National
University.
Pawley, A. and Lane, J. . From Event Sequence to Grammar: Serial Verb
Constructions in Kalam. In Siewierska, A. and Song, J. (eds) Case, Typology
and Grammar, –. Amsterdam: Benjamins.
Quine, W. . Word and Object. Cambridge, MA: MIT Press.
. Ontological Relativity and Other Essays. New York: Columbia University
Press.
Reesink, G. . Being Negative can be Positive. In Huttar, G. and Gregerson, K.
(eds) Pragmatics in Non-Western Perspective, –. Arlington, TX: Summer
Institute of Linguistics.
Talmy, L. . Toward a Cognitive Semantics (two volumes). Cambridge, MA: MIT
Press.
Van Staden, M. and Reesink, G. . A Functional Approach to Verb Serialization.
Paper presented at the Ninth International Conference on Austronesian
Linguistics, Canberra.
Van Valin, R. and La Polla, R. . Syntax: Meaning, Structure, Function. Cambridge:
Cambridge University Press.
5 Cotemporal serial verb constructions
White Hmong

NERIDA JARKEY

. Introduction
White Hmong is minority language spoken in Mainland Southeast Asia,
belonging to the Miao-Yao language family. It is an analytic, isolating lan-
guage, and generally monosyllabic, with some multi-syllabic words due
to compounding and borrowing. Syllable structure is predominantly CV,
with each syllable bearing one of seven phonemic tones, represented in the
Romanised orthography by a syllable-final consonant letter. The order of core
constituents is generally AVO for transitive clauses and SV for intransitive
clauses. Constituent order codes the syntactic function of core arguments.
Topical elements can be fronted and noun phrases that are recoverable from
the context are often omitted. Head modifier order is predominant, with a
few significant exceptions (for example, possessives precede nouns). Nouns
are not marked for number; they co-occur with noun classifiers that assist in
signalling definiteness and specificity.
Like many other languages of the Southeast Asian region, Hmong is
rich in complex predicates involving serial verb constructions (SVCs). As
explained by Baker and Harvey (this volume), SVCs are monoclausal but
multi-predicational. That is, they involve two or more distinct predicating
morphemes, linked together in a single clause by virtue of the fact that they
share one or more argument positions through coindexation.
SVCs are thus significantly different from merger constructions, also
introduced in Chapter  and discussed extensively in this volume. According
to Baker and Harvey’s analysis (see also Wilson ), while both of these
types of complex verb construction are monoclausal, merger constructions


I am ever grateful to those who have helped me learn Hmong, especially Cua Lee, her hus-
band Sao and brother-in-law Yeng, and her dear son William. I would also like to express
my gratitude to Bob Dixon and Sasha Aikhenvald for inviting me as a Visiting Fellow to the
Research Centre for Linguistic Typology, La Trobe University, during which time I wrote
the first draft of this chapter. Many thanks also to Mark Harvey and Mengistu Amberber
for thoughtful comments on my first draft, and to Mark and an anonymous reviewer for
further, very helpful comments.

Where S is the single argument of an intransitive verb, A is the subject of a transitive verb,
and O is the other argument of a transitive verb (see Dixon : –).

110
 Cotemporal SVCs in White Hmong 111

involve the fusion of the Lexical Conceptual Structure (LCS) of two

predicates, resulting in a single LCS, and thus a single predicate. Because
the predicates in SVCs are linked by the coindexation of arguments rather
than by fusion, they maintain their status as separate predicates with inde-
pendent LCSs.
The Hmong data discussed in this chapter address an interesting and
important issue concerning Baker and Harvey’s analysis of these ‘two quite
distinct ways of combining predicate information within monoclausal
structures’, that is, merger constructions and SVCs. This issue relates
to the clear-cut distinction Baker and Harvey draw between single events
(expressed by the merger construction and by monomorphemic predicates)
and multiple events (which, they argue, are expressed not only by multiple
predicates realised by multiple clauses, but also by the multi-predicational
but monoclausal structure of SVCs). Baker and Harvey thus explicitly reject
the view that ‘[s]erial verb constructions describe what is conceptualised as
a single event’ (Aikhenvald : ). This is a view, however, that is widely
and convincingly argued by scholars working on a broad range of typologi-
cally distinct languages in which SVCs occur (e.g. Aikhenvald : –;
Bruce : ; Durie : ; Enfield a; Jarkey : – et pas-
sim, ; Pawley ).
In this chapter, I suggest that the issue is, to some extent at least, a dif-
ference of approach, resulting in a different use of terminology, rather than
a genuine disagreement. I attempt to clarify the terminology below, and go
on to show that the Hmong data presented throughout this chapter support
the notion that SVCs do, indeed, convey what is conceptualised by speakers
as a single ‘event’, in spite of the fact that they clearly involve more than one
predicate. I focus in particular on one SVC type in Hmong – Cotemporal
SVCs – and explain how the multiple predicates in these constructions func-
tion, together with other devices, to portray events in the detailed and elabo-
rate ways that are highly valued by speakers and writers in this language.

. The notion of ‘event’

Baker and Harvey’s use of the word ‘event’ is defined by the formal con-
straints on what can be expressed by the single lexical conceptual structure
of a single predicate, either a coverb construction or a monomorphemic
predicate. Given this definition, it is, indeed, quite reasonable to argue that
SVCs involve more than one ‘event’, because they involve more than one
predicate. This sense of the word will be referred to here, as in Baker and
Harvey, as a ‘simplex event’.
Proponents of the notion that SVCs describe a single ‘event’, however, do
not use the term in this way at all. Their focus is not on the formal con-
straints on what can be expressed within a single predicate universally,
but rather has tended to be on the language-specific constraints on what is
112 Jarkey

construed, and thus presented, as a single event, in a particular utterance

type. This sense of the word is referred to here as a ‘conceptual event’.
A ‘conceptual event’ is both a cognitive and a cultural construct. It is a
cognitive construct in that it is a means by which we deal with our temporal
experiences as a metaphor of the way we deal with our physical experiences.
Just as we classify our physical world into discrete entities, so we classify our
temporal world into discrete events. At the same time, a conceptual event
is also a cultural construct in that, while our common cognitive make-up
results in significant commonalities in what can constitute an event token
across languages, cultural differences can result in variation with regard to
precisely what are considered the salient boundaries of eventhood in some
cases. (See, in particular, Section .. for discussion and examples from
Hmong.)
Bohnemeyer et al. (: ) point out that events (that is, conceptual
events) have mereological structures: both parts of events and combinations
of events can themselves be considered instances of events (see also Zachs
and Tversky ). Thus, an identical extra-linguistic stimulus might be
packaged as a single event in one syntactic construction in one context, but
packaged as more than one event in another syntactic construction in a dif-
ferent context.
These authors assume that the packaging of extra-linguistic stimuli into
discrete events relates to a speaker’s construal of their temporal proper-
ties: properties such as their beginning and end, their duration, and their
occurrence relative to other ‘events’ (p. ). They propose a universal
semantic test for identifying what is conceptualised and thus packaged as
a single event in any language: the ‘Macro Event Property’ (MEP). 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 adverbi-
als, temporal clauses, and tenses necessarily have scope over all subevents
encoded by the construction. (p. )
SVCs in Hmong all have the MEP. This results from the nature of the
syntactic relationship between the predicates and arguments in the construc-
tion, which is well described in the terms of the theory of the layered struc-
ture of the clause, developed in Olson (), and Foley and Olson (),
and further refined within the framework of Role and Reference Grammar
(RRG) by Foley and Van Valin (), Van Valin (, ), and Van
Valin and LaPolla (). According to this theory, a clause consists of three
‘layers’: the ‘nucleus’ (the predicate); the ‘core’ (the nucleus plus the core
arguments), and the ‘clause’ as a whole (the nucleus and the core, plus addi-
tional, non-core (peripheral) arguments, including beneficiaries, outer loca-
tives, and temporal expressions such as time adverbs, etc.). Each layer of the
 Cotemporal SVCs in White Hmong 113

clause can be modified by certain semantic ‘operators’: aspect, for example,

has scope only over the nucleus; deontic modals over the core; and tense,
epistemic modals, evidentials, and illocutionary force over the entire clause.
SVCs in Hmong involve the simple juxtaposition of two or more tokens
of either the nuclear or the core layer within a single clause. In addition to
sharing at least one core argument, the predicates also share all non-core
arguments, including time expressions, and all core and peripheral opera-
tors including deontic modality, tense, epistemic modality, evidentiality, and
illocutionary force. The fact that both the peripheral-layer time expressions
and the peripheral operator of tense are shared confirms that these construc-
tions possess the MEP. Furthermore, the sharing of the core layer operators
(deontic modal concepts such as the desire, obligation, intention, or ability of
the actor) indicates that the packaging involved in SVCs in Hmong is even
tighter than in the case of a construction that has the MEP alone.
With one exception involving nuclear juncture (see ..), SVC types
in Hmong all involve a special type of ‘core cosubordination’ in the terms
of RRG. The term ‘cosubordinate’ is used in this theoretical framework to
refer to a linkage type in which neither junct is embedded as an argument of
the other, but in which the juncts are, nevertheless, dependent on each other
by virtue of the fact that they share all operators at the level of juncture (Van
Valin ). The term ‘core’ linkage indicates that the point of linkage is
within the core – that is, there must be at least one shared core argument.
Cross-linguistically, core cosubordination usually involves the linkage of
two clauses, each potentially able to bring its own peripheral elements (e.g.
location or time expressions) into the construction. However, SVCs of the
types that occur in Hmong are cases of a special kind of core cosubordina-
tion, in which bare cores rather than full clauses are linked, with a single
periphery (see also Jarkey : ). Examples of juxtaposed verbs express-
ing a sequential relationship – included under the term SVCs in some analy-
ses (e.g. Sebba ) – would be instances of regular core cosubordination
rather than of this special type. They would clearly not have the MEP. This
type of juxtaposition does not occur in Hmong, and is not regarded as a
serial verb construction in this analysis.
If our definition of ‘event’ is tied to the predicate, we come up with the
notion of a ‘simplex event’ used by Baker and Harvey. This notion stands
in opposition to the ‘complex events’ that occur in all multi-predicate struc-
tures, whether they be SVCs or multiple clauses. It is reminiscent of the
Aristotelian alignment of the proposition with the verb (Givon : –,
regarding Pawley ). On the other hand, if our definition of ‘event’ relates
to the cognitive and cultural construction of the temporal world, we arrive
at the notion of a ‘conceptual event’. This notion is not constrained by the
number of predicates involved. It is, however, limited to predicates that share
temporal properties – properties such as their beginning and end, their
duration, and their occurrence relative to other ‘events’ – within a single
114 Jarkey

clause. In this sense, the multi-predicate, monoclausal structures involved in

SVCs are, indeed, closer to single predicate clauses than to multiple predi-
cates realised by multiple clauses.

. The expression of a ‘conceptual event’ with

more than one ‘simplex event’ in Hmong
This clarification of the distinction between a ‘simplex event’ and a ‘con-
ceptual event’ raises a further important question: In what circumstances is
more than one ‘simplex event’ used to express a single ‘conceptual event’?
Elsewhere ( Jarkey , ), I have shown that, in the case of White
Hmong, this can be related to a tendency for single verbs not to express high
levels of transitivity, either structurally, in the sense of the number of core
arguments introduced, or semantically, in the sense of ‘the effectiveness with
which an action takes place’ (Hopper and Thompson : ).
For example, structurally, there are no three place predicates in Hmong.
That is, there is no verb that can simultaneously introduce both a Theme
argument and a Recipient argument into the clause (akin to a structure like
he sent me a book in English) without support such as that of a serial verb
construction (Jarkey : –). The kind of SVC used for this purpose
in Hmong is described briefly in .. below, and is referred to in this ana-
lysis as a Disposal SVC (following the traditional use of the term ‘disposal’
for a somewhat similar construction in Mandarin (see Li and Thompson
: –)).
When it comes to the notion of semantic transitivity, even verbs that would
be thought of as very high on the transitivity scale in many languages do not
seem to be quite as high on the scale in Hmong. For example, the notion of
affect is rarely encoded unambiguously in verbs that express impingement
in this language. To ensure that a verb such as ‘shoot’ or ‘cut’ is interpreted
as successful, a second verb expressing the resulting change in the O argu-
ment is generally added to the clause. This SVC type is referred to here as
a Pivotal SVC (see .. below). Another strategy for indicating successful
impingement is to use a Disposal SVC (..), indicating the way in which
the O argument is destroyed, consumed, or relinquished.
Semantic transitivity is also raised by the use of SVCs in the case of telic
predicates, like ‘sew a skirt’ or ‘go to market’, in Hmong. Although predi-
cates like these imply an intrinsic terminal point or goal, the attainment of
that goal cannot be assumed in Hmong, even when the verb is used in the
past tense. A further telic achievement verb, such as ‘get’ or ‘arrive’, must
be included in an Attainment SVC to ensure this interpretation (see ..

Further evidence in support of closer alignment of SVCs with single predicate clauses than
with multi-clause structures comes from Givon’s () study of the pause probabilities
between words within single predicate clauses, between the components of SVCs, and
between the components of multiple clauses.
 Cotemporal SVCs in White Hmong 115

below). Similar observations have been made about other languages of the
Asian area (Chinese (Smith, ), Japanese (Ikegami, , )), and in
these languages, too, it is often the case that a serial verb construction or
other multi-verb construction is used to raise the transitivity sufficiently to
express successful impingement or attainment of a goal.
In this chapter, I introduce the discussion of another factor that is clearly
relevant, in the case of Hmong at least, to the expression of a single ‘con-
ceptual event’ with multiple ‘simplex events’. This factor is the preference
that speakers of this language have for the detailed and elaborate portrayal
of events.
The focus of the chapter is on an SVC type in Hmong that strongly sup-
ports this kind of portrayal: the Cotemporal SVC type (see .. below).
When telic motion verbs occur in Cotemporal SVCs, they function to
introduce noun phrases and locative phrases expressing path information –
source, route, and goal – into the clause, as well as to provide deictic infor-
mation. However, there is another major function of this common SVC type
in Hmong that has not been widely discussed before: that of teasing out the
details of the process and manner of events involving motion, stance, and
atelic action. The use of multiple predicates to achieve this function enhances
the rich and intricate depiction of events much favoured in this language.
This is explained and illustrated in section . of this chapter, after an over-
view of all SVC types in section ..

. Types of serial verb constructions in White Hmong

We can identify four distinct types of SVC in White Hmong on the basis of
two key structural criteria: first, the coindexation relations that hold between
the predicates, and second, the order of the predicates and arguments. These
four types are described in this analysis as Cotemporal, Disposal, Pivotal,
and Attainment SVCs. The analysis of these four types presented in this
section is a summary and refinement of the one presented in Jarkey ().
This summary serves to contextualise the further discussion of Cotemporal
SVCs, which is the focus of this chapter.
A word about the nature of some of the O arguments that can occur is
helpful, before we begin to examine the various types of SVCs in White
Hmong. Some motion and stance verbs in this language (e.g. mus ‘go’, pw
‘lie’) take O arguments that function as core (‘inner’) locatives, indicating the
location, route, source, or goal of the subject of the verb. These O arguments
can take the form of noun phrases that express place: proper nouns such as
Thai teb ‘Thailand’ and common nouns such as khw ‘market’. Alternatively
they can be locative phrases, composed of a ‘spatial deictic’ followed by a
noun phrase (Jarkey : – and : –; Ratliff ).
‘Spatial deictics’ are not prepositions; they do not signal the role of the
noun phrase in relation to the verb. They are a closed class of morphemes
116 Jarkey

in Hmong that indicate some salient aspect of the spatial properties of the
noun phrase they precede, in relation to the speaker or to the perspective
that the speaker chooses to adopt. Examples include ntawm ‘place nearby’,
pem ‘place up’, nram ‘place down’, tim ‘place across’, and hauv ‘place inside’.
When combined with verbs that take core locative arguments, it is thus the
meaning of the verb, not the meaning of the spatial deictic, that results in
the interpretation of the semantic role of the O argument: mus ntawm khw
(go place.nearby market) ‘go [to] market (nearby)’ not ‘go near the market’;
mus tim khw (go place.across market) ‘go [to] market (across there)’ not ‘go
across the market’.

.. Cotemporal serial verb constructions (S/A = S/A)

Both transitive and intransitive verbs can co-occur in a single Cotemporal
SVC in Hmong. This SVC type involves coindexation of the subject argu-
ment of all verbs – A if the verbs are transitive and S if the verbs are intransi-
tive. The fact that this core argument is shared, and that the O arguments
of transitive verbs readily intervene between the verbs in this construction,
indicates that the juncture is at the core level. The fact that deontic modality
operators have scope over both juncts indicates that they involve core cosub-
ordination (see examples (), (), and ()).
This structure is symbolised in () and illustrated in (). Verbs in the
serial construction are underlined, as are their nearest equivalents (not nec-
essarily verbs) in the English translations.
() A/S V (O) V (O) V (O) …
() cov hmoob hla dej na.koom dim hauv
COLLCLF Hmong cross river Mekong escape inside
nplog.teb mus thai.teb
Laos go Thailand
‘The Hmong escaped from Laos to Thailand over the Mekong River.’
Example () shows a Cotemporal SVC involving three motion verbs – hla
‘cross’, dim ‘escape’, and mus ‘go’ – each one introducing a separate argu-
ment – route, source, and goal respectively – of a single motion event – that
is, a single ‘conceptual event’ – undertaken by the subject cov hmoob ‘the
Hmong’. This sentence cannot be interpreted as a sequence of ‘conceptual
events’, nor even a sequence of ‘simplex events’. None of these facets of the
event ‘precedes’ or ‘follows’ another; instead the crossing of the river is pre-
cisely what constitutes the escape from Laos and the journey to Thailand.
The non-iconic ordering of the verbs (expressing route, then source, then
goal) lends support to this point.
In the most common type of Cotemporal SVC, all verbs are motion verbs,
as in example () (see .. below for further examples). Next in frequency
 Cotemporal SVCs in White Hmong 117

are those that involve a combination of action and motion verbs (associated
motion (..)), or action and stance verbs (associated stance (..)). The
use of a serial construction to express cotemporal actions is fairly rare and is
highly constrained, but does occur (..).

.. Disposal serial verb constructions (A = A; O = O)

In Disposal SVCs in Hmong, all verbs must be transitive and both subject
and object arguments are coindexed. The object intervenes between the
verbs, indicating that this SVC type involves core level juncture. Deontic
modals have scope over both cores, so the relationship is core cosubordinate
(see Jarkey : –).
In (), the coindexed arguments are highlighted and in () the serialised
verbs are underlined.
() A V O V V …
() tus tsov tom liab noj
CLF tiger bite monkey eat
‘The tiger gobbled up some monkeys.’
Verbs in this SVC type tend to be fairly high on the scale of seman-
tic transitivity (Hopper and Thompson ), indicating how the subject
sets out to manipulate or affect the object. Quite often, this initial verb
describes an action, such as tom ‘bite’ in (), that is intended to impinge
in some significant way on the object. Even so, as mentioned above, such
a verb is not necessarily interpreted as fully effective without the support
of the other verb(s) in the construction. The second and subsequent verbs
raise the transitivity of the clause by describing how the subject subse-
quently destroys or consumes the object (as in noj ‘eat’ in ()), or at the
very least, relinquishes it.
This sense of ‘disposal’ is exploited in Hmong as a way of introduc-
ing both a theme and a recipient argument simultaneously into a clause.
The verb xa ‘send’, for example, is supported by the verb pub ‘give’, in
example ():
() nws xa ib qho khoom pub kuv
SG send one CLF goods give SG
‘He sent some things to me.’
Sequences of three or more verbs sometimes occur in the disposal
construction:
() muab hlaws pov tseg
take burntr throw.away abandon
‘Burn [it] up completely.’
118 Jarkey

.. Pivotal serial verb constructions (O = S)

In Pivotal SVCs only two verbs occur; the first is transitive and the second,
intransitive. Coindexation involves just one argument: the object of the first
verb is, simultaneously, the subject of the second. This argument appears
between the two verbs, showing that the juncture is at the core level. Once
again, the scope of deontic modals over both juncts tells us that core cosub-
ordination is involved (Jarkey : –).
The ‘pivotal’ role of the coindexed argument is symbolised in (). An
example of the construction is shown in ().
() A V O→S V
() lawv tua liab tuag
PL shoot monkey die
‘They shot some monkeys dead.’
The first verb in a Pivotal SVC is always fairly high in semantic transitiv-
ity, and is often an impingement verb such as tua ‘shoot’ in (). Even so, as
explained in section ., without the support of the other verb in the SVC, a
verb such as this would not necessarily be understood to be fully effective. It
is the second verb that ensures this more highly transitive interpretation.
The semantic relationship between the two verbs is one of direct caus-
ation. Thus the second verb describes a change brought about in the co-
indexed noun phrase as a direct result of the action described by the first
verb. This may be a change of state (as in tuag ‘die’ in ()), a change of
location, or the inception of an intransitive action (as in ()).
() nws ntaus tus dev khiav kiag
SG hit CLF dog run completely
‘She hit the dog [and it] ran away.’

.. Attainment serial verb constructions (A = A; O = O)

Attainment SVCs are like Pivotal SVCs in that they involve only two verbs.
Structurally, these verbs are both transitive. The subject is always coindexed,
as is the object. Thus, the Attainment SVC also has something in common
with the Disposal SVC, in that it involves two coindexed arguments, both
with the same role with respect to each verb.
This SVC type differs significantly from all others, however, in that the
two verbs must appear contiguously. The shared object cannot intervene
between them, indicating that the level of juncture is nuclear. Scope tests
suggest that nuclear coordination is involved (Jarkey : –).
The structure is symbolised in () and exemplified in (), (), and ().
() A/S V V (O)
 Cotemporal SVCs in White Hmong 119

Table . Coindexation of arguments in SVC types

Cotemporal NP V (NP) V

Disposal NP V NP V
Pivotal NP V NP V
Attainment NP V V (NP)

() nws xaws tau daim tiab no

SG sew get CLF skirt this
‘She sewed this skirt.’
() lawv mus txog tom khw lawm
PL go arrive over.there market PERF
‘They have arrived at the market.’
() kuv nrhiav tau kuv nti nplhaib
SG search.for get SG CLF ring
‘I found my ring.’
The first verb in the Attainment SVC must have some kind of goal. It is
often a telic verb, which has an intrinsic goal, like xaws ‘sew’ in () and mus
‘go’ in (). It may also be an atelic verb whose object represents an extrin-
sic goal, such as nrhiav ‘search for’ in () or raws ‘chase’. The second verb
is always a telic achievement, like tau ‘get’ in () and () and txog ‘arrive’
in (). It indicates that the subject has successfully ‘attained’ the goal of
the first verb. This relationship between the two verbs can be thought of as
an aspectual one, and so the fact that the verbs appear contiguously in this
tightly integrated, nuclear structure is not at all surprising. In all cases, the
SVC serves to raise the level of semantic transitivity.
Table . shows the differences between the coindexation of arguments
in these four main SVC types in Hmong. Coindexed arguments are high-
lighted. For the purposes of this representation, each construction is shown
with just two verbs, and arguments are represented as noun phrases (NPs),
even though locative phrase core arguments can also occur.
In three of these four SVC types – Disposal, Pivotal, and Attainment –
the first verb is structurally transitive but the second serves to raise the level
of semantic transitivity, asserting the effectiveness with which this action
takes place in relation to the object. In a Disposal SVC, the second verb
describes how the subject destroys, consumes, or relinquishes the object, in
a Pivotal SVC, it portrays the way in which the subject changes object, and
in an Attainment SVC, how the subject attains a goal, whether intrinsic or
120 Jarkey

extrinsic, in relation to the object. The use of more than one ‘simplex event’
to express a single ‘conceptual event’ in these SVC types is clearly related to
this function of transitivity raising.
The factor that licences the occurrence of more than one ‘simplex event’ in
the expression of a single ‘conceptual event’ in the case of Cotemporal SVCs
is, however, somewhat different. In this case, it is related to the description
of actions that are fairly low, not high, in semantic transitivity. The multiple
verbs that appear in these constructions function not to express the effective
transfer of an action in relation to an object, but rather to portray the action
of the subject in detail, providing colour and elaboration to the description
of the event.

. Cotemporal serial verb constructions

As noted above, both transitive and intransitive verbs can co-occur in
Cotemporal SVCs, and all verbs share a common subject argument (A/S).
This SVC type expresses cotemporal motion (as described in ..), associ-
ated motion and action (..), associated stance and action (..) and, in a
very limited way, cotemporal action (..).
Within this one broad SVC type, we can see a considerable range in terms
of the verb types and combinations that can occur. What these various com-
binations have in common, however, is not only that the component ele-
ments occur cotemporally, but also that they each function to draw attention
to what the subject – the shared argument – does, and how s/he does it.
Unlike other SVC types in the language, the interest is not on what the sub-
ject achieves or attains, nor on how the subject affects or changes the object.
Instead, the focus is entirely on the subject’s action – teasing out the details
of its spatial location, path, extent, and orientation, its manner, its duration,
and its process.
Not surprisingly, then, we find that verbs in Cotemporal SVCs often them-
selves lexically encode information about location, direction, and manner and,
even if structurally transitive, are not particularly high on the scale of semantic
transitivity. Furthermore, within the bounds of their occurrence within a sin-
gle clause, the verbs are fairly loosely aligned, allowing the O argument (a noun
phrase or locative phrase) of any verb to appear directly after it, and adver-
bial expressions to be scattered liberally between the verbs. These adverbial
expressions tend to express manner and duration, in keeping with the focus of
the constructions as a whole on the detail of the process involved.

.. Cotemporal motion SVCs

Bohnemeyer et al. () build on Talmy’s theory of ‘lexicalisation pat-
terns’ and examine the characteristics of complex motion events in which a
 Cotemporal SVCs in White Hmong 121

Table . Verb types and ordering in cotemporal motion SVCs

Manner of motion Motion with respect to ground

Atelic &
Atelic Telic Telic
. Manner
. Transport of motion . Route . Source . ‘Return’ . Goal

coj ‘take (s.o) ya ‘fly’ raws ‘chase’ tawm rov mus ‘go’
along’ dhia taug ‘leave’ ‘return’ tuaj ‘come’
cab ‘lead’ ‘jump, ‘follow (s.t. dim los
thauj dance’ stationary) ‘escape’ ‘come
‘transport/ nkag lawv ‘follow sawv ‘get home’
give a ride’ ‘crawl’ s.t moving’ up/rise’
nqa ‘carry’ tig ‘turn hla ‘cross/ thim
ris ‘carry on around’ pass’ ‘retreat’
back’ khiav dhau ‘pass poob ‘fall’
kwv ‘run’ through’
‘carry on mus kev nce ‘ascend’
shoulders’ ‘walk’ nqis
‘descend’
ncig
‘encircle’

‘figure’ moves in relation to a number of referential ‘grounds’ (Talmy ,

) in a range of genetically and typologically distinct languages. Their
focus is on the kinds of ‘path’ functions (departing from a ‘source’, pass-
ing along a ‘route’, arriving at a ‘goal’) that can be syntactically combined
in a single construction that has the macro-event property. They conclude
that languages that allow maximally dense packaging, with all ‘path’ func-
tions able to be incorporated into a single MEP construction, ‘are either
satellite-framed [in] Talmy’s typology, or they have SERIAL VERB or MULTI-
VERB constructions that permit combinations of multiple location-change-
denoting verb phrases in single macro-event expressions’ (Bohnemeyer et
al. : ).
SVCs expressing cotemporal motion in Hmong are an excellent exam-
ple of such ‘maximally dense packaging’ in motion events. They allow a
wide range of motion verb types to occur together in a single clause, at least
in part as a strategy for expressing path functions. There are, however,
strict constraints on the order in which these verbs appear, as shown in
Table ..
Sentences can be constructed in which all six of these slots are filled, but
in natural text no more than three verbs generally occur together in a single
Cotemporal SVC. Examples () to () illustrate how these verbs are used
in the order indicated in Table . and how O arguments, in the form of
122 Jarkey

noun phrases and locative phrases, as well as adverbs, can intervene between
the verbs. In some examples, reduplication of one verb takes place, or more
than one SVC type or other kinds of juxtaposed verbs occur together, result-
ing in sentences with more than three predicates. In all the examples below,
only the verbs in the Cotemporal SVC are underlined. The number below
each of these verbs indicates its type, in accordance with the numbering
in Table .: [] transport, [] manner of motion, [] route, [] source, []
‘return’, and [] goal.
() nws coj me.nyuam khiav khiav tuaj
SG take.along child run run come
[] [] [] []
‘She fled, bringing her children along.’

() …nws txawm tig rov qab los rau tom teb lawm
SG then turn.around return back come to over.there field PERF

[] [] []

‘… then he turned (and) came back to the fields.’

() thov nce rov tuaj …

request ascend return come
[] [] []
‘Please come back up …’
() cov Hmoob xav hla dej Na.Koom dim hauv Nplog.teb mus
COLLCLFHmong want cross river Mekong escape inside Laos go
[] [] []
Thailand
Thia.teb
‘The Hmong wanted to escape from Laos to Thailand over the Mekong River.’

() …yuav.tau rov mus quib lub qhov.rooj dua

should return go open CLF door again
[] []
‘… should go back to open the door again.’
The last two examples illustrate the fact that deontic modals – xav ‘want’
in () and yauv tau ‘should’ in () – have scope over all verbs in a serial
verb construction of this type. If the shared subject argument were repeated
between the verbs, or if a pause intervened, the scope of the modal would
 Cotemporal SVCs in White Hmong 123

no longer extend over more than one junct. The fact that it does extend
over all juncts in these examples confirms the claim that they are in a core
cosubordinate relationship.
In all of examples () to () we can see that, regardless of which
verbs appear in a particular Cotemporal Motion SVC, the order of the
verbs is consistent with that presented in Table .. First come the atelic
verbs: transport verbs like nqa ‘carry’ and manner of motion verbs such
as ya ‘fly’. These verbs cannot by themselves take route, source, or goal
arguments. Following these are the route verbs, some of which are atelic –
such as raws ‘chase’ and taug ‘follow something stationary’ – and others –
including hla ‘cross’ and ncig ‘encircle’ – which are telic. Next come source
verbs such as tawm ‘leave’ and dim ‘escape’, which are telic, and finally the
goal verbs, mus ‘go’, tuaj ‘come’, and los ‘come home’. Like source verbs,
these are all telic.
It is possible for more than one verb from a single category to occur within
one Contemporal Motion SVC. When two route verbs occur, example ()
shows that an atelic verb such as taug ‘follow something stationary’ will pre-
cede a telic verb such as nqis ‘descend’.
() …taug taug dej nqis hav mus
follow follow water descend valley go
‘… follow the river down the valley (away).’
The examples above show that a key ordering principle in Cotemporal
Motion SVCs is that atelic verbs precede telic verbs. If the order of motion
verbs is reversed, and a telic motion verb precedes an atelic one, a purpose
rather than a cotemporal interpretation results (Jarkey : –).
While these Cotemporal Motion SVCs in Hmong clearly serve to intro-
duce path functions, they also elaborate on the motion event in other ways.
A range of adverbial expressions can occur in a Cotemporal Motion SVC.
These sometimes appear before all the verbs, but more often intervene
between them. The adverb qab ‘back’ is extremely common, very frequently
appearing with the verb rov ‘return’, as shown in example (). Manner
adverbs also appear, often in reduplicated form. Adverbial expressions are
highlighted in italics in examples () to ().
() ces lub nraus ntawd thiaj maj.mam poob rov qab los txog …
CONJ CLF drum that so slowly fall return back come arrive
‘And so then the drum fell slowly back to the earth.’

() tus kwv txawm nqes nthab tsuag tsuag los …

CLF young.brother then descend storage.platform quickly quickly come
‘The younger brother then came down very quickly (from) the storage platform …’
124 Jarkey

() …tus pog laus no taug kev mus-mus los-los ncig

CLF grandma old this follow path go-go come-come encircle
puv lub tsev.khw
fill CLF market
‘… this old woman followed the path here and there all around the market.’

In () the manner adverb maj mam ‘slowly’ appears before the entire SVC,
while in () tsuag ‘quickly’ is reduplicated between the route verb nqes ‘descend’
and the goal verb los ‘come’. In (), the derived adverbial expression mus-mus
los-los (literally ‘go-go come-come’) meaning ‘here and there’ occurs between
the two route verbs taug ‘follow a stationary object’ and ncig ‘encircle’.
Another type of adverbial expressing manner that can appear between
the elements in these SVCs is a construction called a ‘two-word expres-
sive’. This form of elaborate language utilises sound symbolism, including
iconic morphotoneme combinations, to convey expressive meanings (Ratliff
: ff.). As noted in section ., tone is symbolised by a syllable-final
consonant letter in this orthography. Notice the repetitive tones and conso-
nants in the two-word expressives, italicised below:
() …tus me.nyuam mob raws.plab uas yeej khiav khawv.khuav mus rau
CLF child sick diarrhoea REL defeat run energetically go to
tom tsev.dej
over.there bathroom
‘… a child sick with diarrhoea who cannot make it to the bathroom on time.’

[lit.‘… a child sick with diarrhoea which defeats running energetically off to
the bathroom.’]
() ces niag tsov txawm nce kig.kuag kig.kuag mus
and.then great tiger then ascend growling growling go
‘Then that ol’ tiger went up growling [and] growling.’
Cotemporal Motion SVCs involving just two verbs sometimes appear
back to back, creating a four-word sequence of the form ABAC or ABCB –
one of a number of four-word patterns involving repetition, and considered
to add elegance and balance to the language (see section ..). In example
(), khiav mus ‘run go’ is followed directly by khiav los ‘run come’, resulting
in the pattern ABAC.
() neeg khiav mus khiav los teeb rooj teeb khoom li ub li no
people run go run come set.up table set.up produce like that like this
‘People ran here and there, setting up tables [and] setting up produce all over the place.’

Notice that the ABAC pattern of the repeated Cotemporal Motion SVC
is reflected in the pattern of both of the following elements: teeb rooj teeb
khoom ‘set up tables set up produce’ and li ub li no ‘like that like this’.
 Cotemporal SVCs in White Hmong 125

The repetition of the entire Cotemporal SVC, like all the various adver-
bial, colourful, and repetitive expressions above, functions to draw attention
to the manner and process of the motion event. It is yet another strategy
used with this SVC type to flesh out the action of the subject in a vivid and
detailed way.

.. Associated motion and action SVCs (Cotemporal Motion and Action)
Cotemporal SVCs that express simultaneous motion and action in Hmong
are a variation of the Cotemporal Motion SVC type described in .. above.
In this sub-type of Cotemporal SVC, an action verb appears instead of a
manner of motion verb (in position  in Table .) to express a different kind
of information about the ‘manner’ of the motion event.
Like manner of motion verbs, action verbs in this construction are always
atelic; they express activities such as haus ‘drink, smoke’, tham ‘chat’, and qw
‘shout’. As in Cotemporal Motion SVCs, the atelic verbs always precede the
telic verbs.

() ces nplej thiab pob.kws 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.’
() hnub peb nas.tsuag fawb hnyuj.hnyo mus txog
day three rat search.about softly go arrive
‘On the third day a rat went searching softly about all the way there.’

() …nqa ib lub yeeb.thooj haus puj.pauv tuaj

carry one CLF pipe inhale puff.puff come
‘… bringing a pipe along, puffing on it’

Example () shows clearly that Associated Motion and Action SVCs are
simply a variant of the Cotemporal Motion type, the action verb haus ‘drink,
inhale’ appearing between the atelic transport verb nqa ‘carry’ and the telic
goal verb tuaj ‘come’, in just the same position as a manner of motion verb
would appear in a Cotemporal Motion SVC.
As seen in the examples above, both O arguments and adverbial elem-
ents readily intervene between the verbs in Associated Motion SVCs, just
as they do in Cotemporal Motion SVCs. The appearance of an adverb after
the action verb in an Associated Motion SVC seems, in fact, to be the rule
rather than the exception; notice the very common occurrence in these
SVCs of a two-word expressive adverb describing some aspect of the process
of the action verb such as its manner. Examples above include zom zaws ‘all
126 Jarkey

together’, hnyuj hnyo ‘softly’, and puj pauv ‘puff puff’. An adverb expressing
duration of time (such as ib hmo ‘all night long’) can also occur, clearly with
scope over all verbs in the construction:
() …ces cov Khuaj coj lawv maub ib hmo mus …
and.then COLLCLF Khua take.along PL grope one night go
‘… and then the Khua people took them along, groping (through the jungle)
all night long …’

As with Cotemporal Motion SVCs (see example () above), Associated

Motion SVCs sometimes appear back to back, creating a four-word sequence
of the form ABAC or ABCB. In () the verbs maub mus nrhiav mus ‘grope
go search go’ form an ABCB pattern:

() …ces ho yuav maub mus nrhiav mus noj dua

and.then then will grope go search go eat again
‘…and then [it] will go groping [and] searching [in order to] eat again.’

Again, the effect is similar to that of the adverbial expressions, vividly

drawing attention to the manner in which the action and motion event is
played out over time.

.. Associated posture SVCs (Cotemporal Stance and Action)

As well as co-occurring with verbs of motion, action verbs can also occur
in a Cotemporal SVC with a state predicate expressing stance. These SVCs
involve just the two elements – stance and action – the action component
appearing second. The action verbs in these SVCs are mostly atelic, but can
be telic (as in () below). They are invariably low on the scale of semantic
transitivity.

() peb mam li zaum.tsaws sib tham txog …

SG so like sit RECIP chat about
‘So we sit chatting with each other about …’

() …thaum chim siab nyuaj siab xav zaum.tsaws quaj …

when angry liver difficult liver want sit cry
‘…when (I) feel angry and upset, (I) want to sit crying …’

() tiam.sis, vim neeg ho txawj sawv.ntsug mus.kev, txawj siv

but because humans then be.able stand walk be.able use
lawv ob txhais tes los ua ub ua no …
PL two CLF hand CONJ do that do this
 Cotemporal SVCs in White Hmong 127
‘However, because humans were able to walk standing, [we] could use our hands
to do this and that …’

Like Cotemporal SVCs involving motion, those involving stance and

action are clearly instances of core cosubordination. This analysis is sup-
ported by the scope of the deontic modal operators xav ‘want’ and txawj
‘be able’ over both verbs in examples () and () respectively.
Another stative predicate, nyob ‘be located, stay’ is included in this analy-
sis along with the stance verbs that occur in this SVC type. Although it does
not actually indicate a particular posture adopted by the subject, it enters
into the same construction with an action verb following.
() no ces tus kwv txawm nyob ntawm ntug dej tos
this CONJ CLF young.brother then stay nearby edge water wait
Txiv.Nraug.Ntsuag txawm mus suav suav ntses
the.orphan.boy then go count count fish
‘And so then the younger brother stayed at the edge of the stream waiting.
The Orphan Boy went to count the fish.’

As in Cotemporal Motion SVCs (..) and Associated Motion SVCs

(..), a variety of elements can intervene between the predicates in
an Associated Posture SVC. These include the O argument of the state
predicate (indicating the location of the subject), as shown in () above
and () below, and adverbial elements such as duration expressions,
as in ().
() nws sawv.ntsug ntawm qhov.cub ua mov noj
SG stand nearby kitchen make rice eat
‘She stood in the kitchen cooking.’
() …sawv.ntsug ib hnub tsuj cov av pob.zeb tawv tawv
stand one day step.on COLLCLF earth stone hard hard
ua rau Tub.Ntsuag ko.taw mob heev
make to PN foot hurt very
‘… standing all day stepping on hard stony ground made Tu Nzua’s feet
hurt a lot.’
The use of the state verb with the action verb in this sub-type of
Cotemporal SVCs functions to increase the sense of the action unfolding


This construction needs to be distinguished from another construction involving nyob
‘be located, stay’, in which nyob and the noun phrase or locative phrase it introduces
appears after another verb or verb phrase to indicate the location of the entire event (Jarkey
: –).
128 Jarkey

over time, adding both detail and a strong sense of duration to the event
described.

.. Cotemporal actions

Cotemporal Action SVCs are the least productive of all SVC types in Hmong.
While it is conceivable that new combinations could be introduced, they
are highly constrained, expressing combinations of actions that are limited
to culturally ‘recognizable event types’ (Durie, : ). As Aikhenvald
(: –) explains:
This is somewhat similar to how the ‘name-worthiness’ of an activity
provides a reason for nominal and verbal lexical compounds: for instance,
in English, compounds like mountain-climbing or berry-picking are coined as
names of recognizable activities.
As shown above, Cotemporal Motion SVCs and Associated Motion and
Action SVCs that involve just two verbs are sometimes partially redupli-
cated, to create a sequence of the form ABAC (e.g. khiav mus khiav los (run
go run come) ‘run about everywhere’ in example ()) or ABCB (e.g. maub
mus nrhiav mus (grope go search go) ‘go groping and searching around’ in
example ()). Similar patterns expressing two actions also occur, as shown
in example (), repeated here as (), with the action rather than the motion
verbs underlined.
() neeg khiav mus khiav los teeb rooj teeb khoom li ub li no
people run go run come set.up table set.up produce like that like this
‘People ran here and there, setting up tables [and] setting up produce all over the place.’

As mentioned above, the ABAC pattern of the Cotemporal Action SVC

in this example mirrors that of the Cotemporal Motion SVC before it
and of the adverbial expression after it. This repetitive patterning works
beautifully to depict the liveliness and business of the market place in this
example.
Further examples involving partial reduplication with action verbs are
shown in () to (), all of the form ABAC.
() tua nas tua noog
shoot rodent shoot bird
‘hunt small game’
() npaj mov npaj zaub
prepare rice prepare vegetables
‘prepare food’
 Cotemporal SVCs in White Hmong 129

() ua noj ua haus

do eat do drink
‘make a living’
From one point of view, sequences such as these could be analysed as SVCs
involving cotemporal actions. From another, they can be viewed as examples
of a type of aesthetic construction that occurs commonly in Hmong and in
a number of other languages of Mainland Southeast Asia, sometimes called
‘elaborate expressions’ (e.g. Matisoff ), and referred to in Hmong as lus
ua txwm ‘paired words’ ( Johns and Strecker ). These expressions are
made up of a balanced group of four words in which the first and third elem-
ent, and the second and fourth element, are either identical to each other or
share a common meaning and often also a common sound. In spite of being
regarded as somewhat elegant and aesthetic, they are by no means the pre-
serve of poetic or stylised language, but very much a part of everyday speech
in White Hmong.
Paired word sequences are not always based on verbs as in the examples
above. Sequences based on nouns are also common (e.g. khwv iab khwv
daw (toil bitter toil salty) ‘arduous toil’; Johns and Strecker  : ).
For this reason if no other, they clearly need to be described independ-
ently of Cotemporal SVCs, even though the two constructions can be
thought of as overlapping in many instances when verbs and their objects
are involved.
An effect quite similar to a paired word expression occurs when a
Cotemporal SVC involving two actions is fully reduplicated, resulting in an
ABAB pattern, as in () below.
() ces Ntxawm dhia loo los nres hauv plawv quaj cem quaj
and PN jump INTENS CONJ stop inside middle cry scold cry
cem hais.tias …
scold COMP

‘And Nzeu leapt up and stood in the middle, crying [and] scolding [and] crying
[and] scolding, saying …’

Cotemporal action verbs also appear unreduplicated, but accompanied by

a paired word adverbial expression, as in example ().
() yuav pom tej me.nyuam quaj dhia rau ub rau no xav tau
will see COLLCLF child cry jump to there to here want get
yam ub yam no
thing there thing here
‘[You] will see children crying [and] stamping all over the place, wanting to have this
[and] that.’
130 Jarkey

Even in examples like this one, in which a Cotemporal Action SVC

involves no repetition or reduplication, and so bears no structural similar-
ity to paired words, it shares two common elements with paired words.
Firstly, the actions combined in these sequences must be strongly and
commonly associated with one another. A combination such as that in
() – quaj dhia (cry jump) ‘cry [and] jump, throw a tantrum’ – is cer-
tainly strongly and commonly associated with the behaviour of young
children. Secondly, like paired word expressions involving action verbs,
Cotemporal Action SVCs describe actions which, while not always exactly
simultaneous, are still broadly cotemporal. They cannot be interpreted as
occurring in a simple sequence. One can imagine the children either cry-
ing and stamping together or else alternating rapidly between the two. In
this example, the bedlam of the situation is graphically portrayed not only
by the Cotemporal Action SVC quaj dhia ‘cry jump’ and its accompanying
paired word adverbial rau ub rau no ‘to there to here’, but also by another
paired word adverbial in the following, juxtaposed clause, yam ub yam no
‘thing there thing here’.
All of the examples above depict the actions described in an intense and
lively way. This is exactly what we would expect of Cotemporal Action SVCs,
given the clear function of other types of Cotemporal SVCs: to draw out the
detail of the manner and duration of the process involved. Furthermore, it is
precisely the two common elements mentioned above – the frequent associa-
tion of the two actions and their cotemporal occurrence – that license multi-
ple actions to be viewed as a single ‘conceptual’ event and thus allow them to
occur together within a single clause in this language.
Both of these elements are evident in examples (), (), and () below.
() kuv mam zov ntsia koj mus
SG so.then guard watch SG go
‘So then I’ll watch over you going.’
() luag mus da.dej ua.si los yus txaj.txaj.muag …
others go bathe play CONJ INDEF.PRON be.embarrassed
‘Others go to swim (and) play but you are embarrassed …’
() nws dhia tshov qeej
SG jump blow bamboo.pipes
‘He dances (while) playing the pipes.’

In (), the actions ‘guard (and) watch’ can be easily seen to meet the cri-
teria for combination; these two actions are very commonly associated with
one another and, in a sense, the second might be thought of simply as an
elaboration of the meaning expressed by the first. The actions in (), ‘swim
(and) play’ are actions that are often strongly associated in cultures in which
 Cotemporal SVCs in White Hmong 131

water play is a common recreational activity. The inclusion of the second

verb ua si ‘play’ in this sequence clearly serves to enliven the description of
the action described by the first verb da dej ‘swim, bathe’.
The final example, () above, links two actions that might not be thought
of as frequently collocated in another context, but which, again, are very
much part of the same ‘conceptual event’ in Hmong culture. When a per-
former plays the qeej ‘bamboo pipes’, an integral part of the performance is
the way in which he sways his body and lifts his feet to the rhythm of the
music (Jarkey : –). The use of the first verb dhia ‘jump, stamp,
dance’ along with the phrase tshov qeej ‘blow the pipes’, brings vividly to
mind the details of the action involved. A pair of actions less commonly asso-
ciated in Hmong culture, such as dhia mloog nkauj ‘dance listen song’ is odd
and not likely to occur; although easily thought of as potentially occurring
cotemporally, this combination has none of the familiarity, nor the intensity,
necessary to be expressed in a Cotemporal Action SVC.
Some of the situations exemplified above may be seen as universally
stereotypical events, others less so. However, because of the strong asso-
ciation of these actions in Hmong culture, it is appropriate to conceive of
them as together portraying one ‘conceptual event’ in the Hmong language
(Aikhenvald : –; Bruce ; Diller : –, –; Durie
; Enfield a, b; Pawley ). Their semantic connection is
reinforced by their syntactic integration in a core cosubordinate, serial rela-
tionship within a single clause, as demonstrated by the scope of the modal
operator txawj ‘be able’ in ().
() nws txawj dhia tshov qeej
SG be.able dance blow bamboo.pipes
‘He is able to dance (while) playing the pipes.’

. Conclusion
We are dealing here with what is broadly one SVC type: all the examples
discussed in section . of this chapter express cotemporal components of
one ‘conceptual event’, all are alike in that the subject (A/S) is the coindexed
argument, and all involve a core cosubordinate, serial relation between the
predicates. Within this single, broad SVC type, we can still see considerable
variation with regard to the range of verb types, the combinations that can
occur, and the kinds of meanings conveyed.
Despite their diversity and their breadth of meaning, one key function is
shared by all Cotemporal SVCs in White Hmong, a function that is much
more significant than simply expressing motions, states, or actions that co-
occur in time and that are attributable to a single subject. That function is to
focus attention on the process of what the subject does and how s/he does it.
Cotemporal SVCs use many verbs that are low in semantic transitivity – verbs
132 Jarkey

that describe the detail of the action itself rather than what effect that action
has. These SVCs combine multiple ‘simplex events’ in order to tease apart
the various aspects of a single ‘conceptual event’. They appear together with
other strategies, such as the use of adverbial expressions, repetition, redupli-
cation, and paired words, all of which function to draw out the details of the
process, direction, manner, and duration of an action and to portray that
action in a lively and vivid way.

References
Aikhenvald, A.Y. . Serial Verb Constructions in Typological Perspective. In
Aikhenvald, A. Y. and Dixon, R. M. W. (eds) Serial Verb Constructions: A
Cross-linguistic Typology, –. Oxford: Oxford University Press.
Bohnemeyer, J., Enfield, N.J., Essegbey, J., Ibarretxe-Antuañano, I., Kita, S., Lüpke,
F., and Ameka, F.K. . Principles of Event Segmentation in Language: The
Case of Motion Events. Language. , : –.
Bruce, L. . Serialization: From Syntax to Lexicon. Studies in Language. ,
: –.
Diller, A. V. N. . Thai Serial Verbs: Cohesion and Culture. In Aikhenvald, A.
Y. and Dixon, R. M. W. (eds) Serial Verb Constructions: A Cross-linguistic
Typology, –. Oxford: Oxford University Press.
Dixon, R. M. W. . Ergativity. Cambridge: Cambridge University Press.
Durie, M. . Grammatical Structures in Verb Serialization. In Alsina, A.,
Bresnan, J., and Sells, P. (eds) Complex Predicates, –. Stanford,
CA: CSLI Publications.
Enfield, N. J. a. Functions of ‘give’ and ‘take’ in Lao Complex Predicates. In
Bauer, R. S. (ed.) Collected Papers on Southeast Asian and Pacific Languages,
–. Canberra: Pacific Linguistics.
b. Cultural Logic and Syntactic Productivity: Associated Posture
Constructions in Lao. In Enfield, N. J. (ed.) Ethnosyntax: Explorations in
Culture and Grammar, –. Oxford: Oxford University Press.
c. Semantics and Combinatorics of ‘sit’, ‘stand’, and ‘lie’ in Lao. In Newman,
J. (ed.) The Linguistics of Sitting, Standing, and Lying, –. Amsterdam: John
Benjamins.
Foley, W. A. and Olson, M. . Clausehood and Verb Serialisation. In Nichols,
J. and Woodbury, A. C. (eds) Grammar Inside and Outside the Clause, –.
Cambridge: Cambridge University Press.
Foley, W. A. and Van Valin, R. D. . Functional Syntax and Universal Grammar.
Cambridge: Cambridge University Press.
Givón, T. . Serial Verbs and the Mental Reality of ‘event’: Grammatical vs
Cognitive Packaging. In Heine, B. and Traugott, E. (eds) Approaches to
Grammaticalization, –. Amsterdam: John Benjamins.
Hopper, P. and Thompson S. A. . Transitivity in Grammar and Discourse.
Language. , : –.
 Cotemporal SVCs in White Hmong 133

Ikegami, Y. . Activity-Accomplishment-Achievement: A Language That Can’t

Say, ‘I burned it, but it didn’t burn’ and one that can. In Essays in Honor of
Rulon S. Wells. Trier: LAUT Series A, –.
. What Does it Mean for a Language to Have No Singular-Plural Distinction?
Noun-verb Homology and its Typological Implication. In Geiger, R. A. and
Rudzka-Ostyn, B. (eds) Conceptualizations and Mental Processing in Language,
–. Berlin and New York: Mouton de Gruyter.
Jarkey, N. . Serial Verbs in White Hmong: A Functional Approach. PhD disser-
tation, University of Sydney.
. Process and Goal in White Hmong. In Tapp, N. and Lee, G. Y. (eds)
The Hmong of Australia: Culture and Diaspora, –. Canberra: Pandanus
Books.
. Complement Clauses and Complementation Strategy in White Hmong.
In Dixon, R. M. W. and Aikhenvald, A. Y. (eds) Complementation: A Cross-
linguistic Typology, –. New York: Oxford University Press.
Johns, B. and Strecker, D. . Aesthetic language in White Hmong. In Downing,
B. T. and Olney, D. P. (eds) The Hmong in the West: Observations and Reports,
–. Minneapolis: Center for Urban and Regional Affairs, University of
Minnesota.
Li, C. N. and Thompson, S. A. . Mandarin Chinese: A Functional Reference
Grammar. Berkeley: University of California Press.
Matisoff, J. A. . The Grammar of Lahu. University of California Publications in
Linguistics . Berkeley: University of California Press.
Olson, M. . Barai Clause Junctures. PhD dissertation, Australian National
University.
Pawley, A. . Encoding Events in Kalam and English: Different Logics for
Reporting Experience. In Tomlin, R. (ed) Coherence and Grounding in
Discourse, –. Amsterdam: John Benjamins.
Ratliff, M. . The Influence of Geographical Change on Grammar: The Case of
Hmong Spatial Deictics. Paper Presented at the rd International Conference
on Sino-Tibetan Languages and Linguistics, Arlington, University of Texas.
. Meaningful Tone: A Study of Tonal Morphology in Compounds, Form
Classes, and Expressive Phrases in White Hmong (Monograph Series on
Southeast Asia, Special Report No. ). DeKalb, IL: Northern Illinois
University, Center for Southeast Asian Studies.
Sebba, M. . The Syntax of Serial Verbs: An Investigation into Serialisation in
Sranan and Other Languages (Creole Language Library ). Amsterdam: John
Benjamins.
Smith, C. S. . Event types in Mandarin. Linguistics : –.
Talmy, L. . Lexicalization patterns. In Shopen, T. (ed.) Language Typology
and Syntactic Description, vol. III: Grammatical Categories and the Lexicon,
–. Cambridge: Cambridge University Press.
. Towards a Cognitive Semantics. Cambridge, MA: MIT Press.
Van Valin, R. D. Jr. . Pragmatics, Island Phenomena, and Linguistic
Competence. Papers from the Parasession on Pragmatics and Grammatical
Theory. Chicago Linguistic Society , : –. University of Chicago.
134 Jarkey

. A Synopsis of Role and Reference Grammar. In Van Valin, R. D.

Jr. (ed.) Advances in Role and Reference Grammar, –. Amsterdam;
Philadelphia: John Benjamin.
. Exploring the Syntax-Semantics Interface. Cambridge: Cambridge
University Press.
Van Valin, R. D. Jr. and LaPolla, R. J. . Syntax: Structure, Meaning and
Function. Cambrige: Cambridge University Press.
Wilson, S. . Coverbs and Complex Predicates in Wagiman. Stanford, CA: CSLI.
Zachs, J. and Tversky, B. . Event Structure in Perception and Conception.
Psychological Bulletin , : –.
6 Activity incorporates in some
Athabaskan languages

K ER EN R IC E

In work on complex predicates, there is general agreement that the term can
be defined as below.
[T]he term complex predicate is used to designate a construction that
involves two or more predicational elements (such as nouns, verbs, and
adjectives) which predicate as a single element, i.e. their arguments map
onto a monoclausal syntactic structure. (Butt, this volume.)

I examine a type of complex predicate, the activity incorporate construc-

tion, in two Athabaskan languages of Alaska, United States, Ahtna and
Koyukon. This construction, shown in (), has the meaning ‘do X while
Y-ing’. In (), the verb stem, the final element of the verb, is italicised: it
generally translates as ‘X’ of ‘do X while Y-ing’. The activity incorporate, a
verbal prefix, is bolded: it is ‘Y’ of ‘do X while Y-ing’.

() a. Koyukon (Jetté and Jones )

sel-he-ghe-d-o-l-de ɬ
shout-PL.S-qualifier-qualifier-progressive-l voice/valence-plural go
‘They are going along shouting.’ ()
ze ɬ-, sel- ‘shouting, noise, howling’ ()
b. Ahtna (Kari )
ɬu-sel-d-a-l-de ɬ
around-shout-qualifier-progressive-l voice/valence-plural go


I follow the orthography of the sources. Note the following: gh = voiced velar fricative, kk =
voiceless aspirated uvular stop, gg = voiceless unaspirated uvular stop, nh = voiceless nasal;
’ = glottal stop, glottalisation of previous consonant; u = low back round reduced vowel. In
the examples, the incorporate is bolded and the stem in italics. I use the following abbrevia-
tions: SG = singular, PL = plural, S = subject, DO = direct object, OO = oblique object. Page
numbers from the source are included following translations. ‘G’ indicates gender, ‘O’ a dir-
ect object, and ‘P’ a postpositional object. The verbal morphophonemics is complex, and
there is often not an obvious relationship between the underlying and surface forms. Note
that there is no masculine/feminine distinction, and I follow the sources in their translations
of third person subjects.

135
136 Rice

‘They (mosquitoes) are whining about.’ ()

sel ‘shouting’ 
The activity incorporate construction shares many properties with the
serial verb construction (e.g. Aikhenvald , Baker and Harvey this vol-
ume); however, rather than two verbs it involves a verbal element, the verb
stem, and a nominal element, the incorporate.
In this chapter I examine properties of the activity incorporate construc-
tion as a whole as well as the properties of each of the major pieces within
it. I begin with a brief tutorial on the structure of the verb word in these
languages.

. Some background

The verb word in Athabaskan languages is notable for its complexity. A
schematic representation is given in (), with pieces that are of concern to
this chapter in bold.
() preverb-incorporate-quantifier-pronominal-qualifier-aspect-subject-
voice/valence-verb stem
The version of the verb word in () is simplified compared to the schema
in Kari () for Ahtna and Jetté and Jones () for Koyukon, but it is
sufficient for our purposes here; see the sources for details. Starting at the
right edge, the verb stem carries the main meaning of the verb word. Voice/
valence markers, called classifiers in the Athabaskan literature, can be lexi-
calised, occurring as part of the basic lexical entry, and they play import-
ant productive functions, indicating transitivity/causativity (I use the term
transitivity) and middle marking; the productive functions are of importance
here. First person singular and second person subjects occur before voice/
valence. These are preceded by elements that provide aspectual information;
see section ... and ... Qualifiers may be productive, indicating noun
gender classes and aspectual information such as inceptivity and conativity,
or they may be lexicalised as part of the basic lexical entry of the verb word.
Pronominals include direct objects and first person plural and third person
plural subjects. Quantifiers mark habituality and distributivity. Incorporates
include nouns that indicate subjects, direct objects, and obliques, and a set of
these is the focus of this chapter. Preverbs represent concepts such as direc-
tion, location, and relation.
In the Athabaskan literature, the lexical representation of the verb word is
called the ‘verb theme’: I use the term ‘(basic) lexical entry’. The lexical entry
includes a root and a lexicalised voice/valence marker (which may be null,
Ø), and can include other lexicalised material such as qualifiers and preverbs
where these occur in every verb word constructed on the particular lexical
entry. For instance, the lexical entry of ‘scream, yell, shriek, cry out; whistle,
 Activity incorporates in some Athabaskan languages 137

horn sounds’ in Koyukon consists simply of a null voice/valence marker (Ø)

and a root: Ø+kk’ok (Jette and Jones: ); that of ‘be alive, move, twitch,
have a spasm’ has a lexicalised qualifier (ghe) as well as the voice/valence
marker (Ø) and the root (no): ghe+Ø+no (p. ). The entry ‘heal’ (p. )
also includes a lexicalised preverb (no) as well as a voice/valence marker (D)
and the root: no#D+neek (where ‘#’ indicates a boundary type). The pieces
of the lexical entry must always occur in any verb words that are formed on
it. This can be seen with the entry, ne+(D)+t’ukk ‘singular flies’ (p. ). A
few verb words that are built on this entry are given in (), with the qualifier
ne and the verb stem, t’ukk, highlighted.
() naat’ukk ‘It arrived flying.’ ()
ts’aanaat’ukk ‘It flew out.’ (preverb: ts’aa ‘out into the open’)
nonot’ukk ‘It flew down.’ (preverb: no ‘down’)
Roots are obligatorily suffixed, forming a stem, and, unless otherwise noted,
I refer to the verb stem in this chapter rather than to the root.
There is debate in the Athabaskan literature about whether what I am call-
ing the verb word is a morphological or a syntactic construct, and I abstract
away from this controversy, noting that the verb word is clearly a unit from a
phonological perspective. It occurs on its own; other words cannot intervene
within it; most elements within the verb word do not, in general, appear as
independent words.
Overall there is productivity in verb word formation, part of what makes
a syntactic analysis of the verb word possible. Verb stems and preverbs com-
bine productively, subject to semantic restrictions (see Kari (, ) on
Ahtna and Axelrod () on Koyukon for detailed discussion). The mor-
pheme span that includes the qualifiers, aspect, subject, and voice/valence
combine with each other to make the underlying morphology quite opaque
from the surface perspective, but there is general agreement in the literature
on morpheme identification.
There are constructions within the languages that can be considered to
be complex predicates, and I present some here as an introduction to the
languages. Ahtna material is from Kari () and Koyukon material from
Jetté and Jones (). I do not fully gloss examples here, but focus on parts
that are relevant to the complex predicate. The verb stem is italicised, and
the other part of the complex predicate bolded.
In (), complex predicates formed through the addition of the causative
voice/valence marker ( ɬ ) are illustrated.
() Causatives
a. Koyukon
too de-l-tsuhtl
water qualifier-voice/valence-splash
138 Rice

‘The water is making a splashing sound.’ ()

too de- ɬ -tsuhtl
water qualifier-causative voice/valence-splash
‘S/he is causing the water to make a splashing sound.’ ()
b. Ahtna
nen’ ghi-ghi-na’
earth qualifier-aspect-move
‘The earth is shaking.’ ()
ɬts’ii ts’abaeli d-ghe- ɬ -naa
wind tree qualifier-aspect-causative voice/valence-move
‘Wind is moving the trees.’ ()
The examples in () represent another kind of complex predicate, one
formed with middle marking. This has two forms, l or d, depending on a
variety of conditions that are not relevant at this point. The examples below
illustrate passives.
() Middle: low elaboration (d, l )
a. Koyukon
ya-l-aa-tl-ghaanh
. DO-qualifier-aspect-voice/valence-kill singular object
‘He killed it/him, he beat him up.’ ()
l-aa-l-ghaanh
qualifier-aspect-l voice/valence-kill singular object
‘He got killed; he is aching all over from overexertion.’ ()
b. Ahtna
i-ghi-ghaan
. DO -aspect-make plural objects
‘S/he made them.’ ()
a-d-ghaan
aspect-d voice/valence-make plural objects
‘They were made.’ ()
There are a large number of preverbs in these languages, and they too
enter into complex predicates with the verb stem. A few examples are given
in (), from Koyukon.
() a. no-ghee-’onh ‘S/he handled, moved, carried compact O down.’ ()
no-ghee-yo ‘S/he came down (from upstairs, down off
something).’ ()
 Activity incorporates in some Athabaskan languages 139

b. ho-le-’onh ‘S/he carried compact O up a slope.’ ()

ho-l-yo ‘S/he went up (to the top of) the bank.’ ()
Two preverbs are illustrated in (), no ‘down’ (a) and ho ‘up a slope, to the
top’ (b), combined with two verb stems, ‘onh ‘handle compact object’ and
yo ‘singular go’. The preverbs require different aspect markers, gh with no
‘down’ (a) and l with ho ‘up a slope, to the top’ (b), overriding the aspect
marker (n) that would occur if no preverb or other aspect-affecting material
were present (see Axelrod  and Kari ).
The examples in () through () involve complex predicates in one way or
another. First, they are comprised of two (or more) predicational elements –
a verb (the verb stem) and a productive voice/valence marker, a verb and a
preverb. The morphosyntactic structure is monoclausal, with a single mark-
ing for arguments and aspect within the verb word.
The goal in the remainder of this chapter is to examine a different type of
complex predicate, the activity incorporate construction, illustrated in ().
In this construction, a noun representing certain types of activities is incor-
porated into a verb word where the verb stem is drawn from classes of verbs
that involve motion and body stance. The article is structured as follows.
Section . introduces the activity incorporate construction, considering
the construction as a whole, and sections .–. address restrictions of the
incorporate and restrictions on the incorporator. Section . examines some
other incorporate constructions. Section . concludes.

. The activity incorporate construction as a

complex predicate
.. Introduction
As discussed above, some Athabaskan languages have a construction that
I call the activity incorporate construction with the meaning ‘do X while
Y-ing’, where the events specified by X and Y are simultaneous and carried
out by the same subject. The construction is noted by Jetté in his monumen-
tal writings on Koyukon:
Combined-Action Verbs. These express the simultaneous performing of
two actions: that of going, rendered by the verb, and another performed
while going, and rendered by an accidental [i.e. an incorporate; KR] prefix.
(Jules Jetté; cited from Jetté and Jones (: ))
Examples in this chapter come from two languages, Ahtna and Koyukon,
with a focus on Koyukon, and unlabelled data is from Koyukon. All
material is drawn from the excellent dictionaries of these languages, Kari
() for Ahtna and Jetté and Jones () for Koyukon. These dictionar-
ies are carefully done with systematic checking for productivity; however
it is important to note that all material is from dictionaries and thus any
140 Rice

conclusions about absence of structures must be regarded as conjectural.

The work illustrates what is possible with dictionaries, and also the restric-
tions on such work.
I begin with a discussion of the properties of the activity incorporate
construction as a whole, and then examine the types of verbs that occur in
the construction, followed by discussion of the activity incorporates. I refer
to the verb component variably as verb stem or as verb, recognising the com-
plexities present in lexical entries, discussed in section .. Further, given
that there is generally other information present besides just the verb stem,
the inflection, and the activity incorporate, I refer to the verb word minus
the activity incorporate as the verb frame or the incorporating frame.

.. Morphological trappings

Before addressing the activity incorporate construction as a complex
predicate, some morphological trappings of the construction need to be
identified. First, a prefix ghe (~gh) often occurs in this construction. This
morpheme is identified by Kari (), following Jetté (), as a qualifier
that does not have an identifiable meaning associated with it. It can be seen
in the Koyukon examples in (), simply glossed ‘qualifier’. Jetté and Jones
(: ), on Koyukon, list this morpheme as part of the activity incor-
porate construction in the dictionary entry, INC#(ghe)+de+le+ ‘derived
intransitive with incorporate of oral noise’ (p. ), where ‘INC’ identifies
the incorporate, ‘(ghe)’ the qualifier under discussion, ‘de’ a qualifier, and ‘le’
the voice/valence marker (the latter two are discussed immediately below).
The contribution of the qualifier ghe, if any, is not clear, and I do not discuss
it further.
Second, in both languages the construction generally occurs with a
second qualifier, this one of the form de (~d ). It is called a gender marker
in the dictionaries; in Rice (a) I call it a noun class marker and
in this chapter I gloss it ‘qualifier’. Kari ( : ) on Ahtna and Kari
( : xciii) on Koyukon note the existence of active gender systems in
these languages, where gender prefixes indicate the subject of an intransi-
tive and the object of a transitive. The gender prefix d refers to, among
other things, words, songs, stories, news, and names (Jetté and Jones
 : ), and it occurs in many examples of the activity incorporate
construction. It too is visible in the examples in (). See the dictionaries
and Rice () for discussion.
Finally, the construction occurs with a voice/valence prefix of the form l
or le. In the examples in () and (a) above, this is glossed as l voice/valence.
This element is used productively to mark a middle of a transitivised verb.
Jetté and Jones (: ) list it as a required part of the activity incorp-
orate construction in Koyukon. See section .. for discussion of voice/
valence.
 Activity incorporates in some Athabaskan languages 141

.. Properties of complex predicates

In this section I address properties of the activity incorporate construction
that suggest that it is a complex predicate as defined above.

... Argument structure

Complex predicates exhibit the argument structure of a single clause, and,
in terms of argument structure, the activity incorporate construction has a
single marking for subject. While the verb word contains two predicational
elements, the verb stem and the activity incorporate, the subject functions
as subject of both. Thus, verb words with translations such as ‘I walked in
singing’ exist, as in (), while those such as ‘I walked in while you were sing-
ing’ are not expressed with this construction but are periphrastic in form.

... Aspect
The activity incorporate construction has a single marking for aspect, associ-
ated with properties of the verb frame rather than the incorporate. To discuss
this property of the construction, a brief introduction to the aspect system
is required. Aspect is marked in various ways throughout the verb word of
an Athabaskan language: the form of the stem is related to aspect, there are
prefixes that indicate a range of aspectual information, and prefixes such as
preverbs and quantifiers play a role in the aspect system. A verb word that
includes an activity incorporate has the aspectual marking that would be there
even if the activity incorporate were absent. The examples in () illustrate this.
In (a), the verb word is perfective, with a marker n that indicates that it is an
achievement. (b) is also perfective, also with n. The presence of n is deter-
mined by the preverb, a postposition do ‘entering an opening in, blocking the
opening of’. The major difference between (a) and (b) is that the verb word
in (b) includes an activity incorporate (k’eleek ‘singing’) that is not present
in (a). In addition, the verb stems differ, but this difference is not relevant to
the question at hand. The complex morphophonemics make it difficult to see
easily exactly what is going on: what is important is that the aspectual infor-
mation is not affected by the presence of the activity incorporate.
() a. ye-do-nee-yo
.OO-entering-aspect-singular go
‘He entered.’ ()
b. ye-do-k’eleek-he-ghe-d-aa-l-daatl
.OO-entering -singing-.PL.S-qualifier-qualifier-aspect-l
voice/valence-plural go
‘They came in singing.’ ()
The examples in () parallel those in (). The major difference between
(a) and (b) is that (b) has an incorporate, dlukk ‘laugh’. Abstracting away
142 Rice

from the morphology that accompanies the activity incorporate construction

(section .. – the qualifier ghe, the qualifier d, voice/valence l ), these are
identical, with the same stem and aspect, visible as le before the stem in (a),
but obscured by the morphophonemics in (b).
() a. le-do ‘He is sitting, he is staying.’ ()
b. dlukk-ghedaadle-do ‘He is sitting down laughing.’ ()

... Phonological status as a single word

A third indication of the status of the activity incorporate construction as
a complex predicate is the phonological status of the verb word as a single
word. While there are proposals in the Athabaskan literature that the level
of the verb as a word is not appropriate from a morphosyntactic perspec-
tive (see, for instance, Speas (), Jelinek and Willie (), Rice (a),
Rice and Saxon () for various versions of a syntactic analysis for differ-
ent languages of the family), there is agreement that from the phonological
perspective the verb word is a single unit. Proposals for richer internal struc-
ture for the verb word treat it as a single clause, with more than one verbal
element (e.g. verb stem, voice/valence) but not more than a single sentence.
Sentential material cannot intervene between pieces of the verb word. For
example, conjunctions are found in both Ahtna and Koyukon, but they do
not occur word-internally. Two Koyukon conjunctions are illustrated in ();
while these carry the meaning that two events are simultaneous, as in the
activity incorporate construction, they never appear within the verb word,
but rather conjoin independent clauses.
() a. ts’en’
‘and, as, and so, simultaneously, having, being, because of’
‘the conjunction should be used to connect verbs expressing
co-existent actions’ (Jules Jetté, cited in Jetté and Jones (: ))
b. dehuyeekk’e
‘simultaneously, at the same time as, just as’ ()
The above properties suggest the complex predicate status of the activity
incorporate construction. Both the verb stem and the incorporate are predi-
cational, but the construction in which they appear is monoclausal.

.. Summary: complex predicate properties

In summary, the activity incorporate construction has properties of a com-
plex predicate by Butt’s definition (this volume). First, it involves two major

It is interesting that neither dictionary has a category of complementiser, although comple-
mentisers have been identified in other languages of the family.
 Activity incorporates in some Athabaskan languages 143

predicational elements, a verb and a noun. Second, it is monoclausal, with

a single subject and single marking of situation aspect, determined by the
verb frame. Both pieces of the construction carry restrictions in terms of
the semantic classes of items allowed. The activity incorporate construction
is asymmetric, with the stem restricted to verbs of motion and body stance
(and locution; see section .), and the incorporate restricted to verbs of oral
activity; see the following sections. In section ., I examine the verb stem,
or the incorporating frame.

. Properties of the incorporating frame

The incorporating frame consists of a verb stem and other morphological
material that makes up the verb word, as discussed in section .. In this
section I examine characteristics of the verb stem.

.. Verb stem classes

Two major classes of verbs serve as incorporators, motion and body stance
verbs. These are light verbs, providing information about motion and body
position, with further semantic content supplied by other material in the verb
word. A number of examples of the activity incorporate construction are given
below, organised by incorporating frame. The incorporated noun is on the
first line, with a verb word that includes the incorporate on the second line.
Examples involving verbs of motion from Koyukon are given in () and from
Ahtna in (). Morpheme-by-morpheme translations are relatively rough.
() Verbs of motion as incorporator: Koyukon
a. k’eleek ‘sing, song’ ()
k’eleek-ghe-d-o-l-kkaa ɬ
song-qualifier-qualifier-progressive-l voice/valence-paddle
‘He is paddling along singing.’ ()
b. he-k’eleek-ghe-daa-ghe-ge-kkaa ɬ
upstream-song-qualifier-qualifier-progressive-.SG.S.l
voice/valence-paddle
‘I sing while canoeing upstream.’ ()
c. sel ‘shouting, noise, howling’ ()
sel-ghe-z-o-l-de ɬ
shouting-qualifier-.PL.S-progressive-l voice/valence-plural go
‘We are going along shouting.’ ()
d. yoyeesk ‘whistling, whistle (sound)’ ()
yoyeesk-ghe-d-o-l-ho ɬ
whistle-qualifier-qualifier-progressive-l voice/valence-singular go
144 Rice

‘He is walking along whistling.’ ()

e. kkenaa ‘word, talking, speech’ ()
kkenaa-ghe-d-o-l-ho ɬ
talk-qualifier-qualifier-progressive-l voice/valence-singular go
‘He is walking and talking.’ ()
() Verbs of motion as incorporator: Ahtna (Kari )
a. c’udyiis ‘whistle’ ()
na-c’udyiis-de-l-yaa
back-whistle-qualifier-l voice/valence-singular go
‘He returned whistling.’ ()
b. ko ɬ dogh ‘he is making a potlatch speech’ ()
ɬ u-hdogh-d-a-l-yaa ɬ
around-potlatch speech-qualifier-progressive-l voice
valence-singular go
‘He is walking around while making a potlatch speech.’ ()
c. i’dliis, c’eliis ‘song’ ()
i’dliis-h-d-a-l-de ɬ
song-.PL.S-qualifier-progressive-l voice/valence-plural go
‘They are going along singing.’ ()
d. duut ‘chattering sound, call of sandhill crane’ ()
ɬ u-duut-d-a-l-de ɬ
around-chatter-qualifier-progressive-l voice/valence-plural go
‘They (flock of geese, cranes) are going about honking, chatter-
ing.’ ()
e. neke-duut-d-a-l-de ɬ
circle-chatter-qualifier-progressive-l voice/valence-plural go
‘They (cranes) are flying in a circle calling.’ ()
f. sel, se ɬ ‘shout, shouting’ ()
ɬ u-sel-d-a-l-de ɬ
around-shout-qualifier-progressive-l voice/valence-plural go
‘They (mosquitoes) are whining about.’ ()
g. dzuuggi ti-se ɬ -d-ghe-l-ggaac
princess out-shout-qualifier-aspect-l voice/valence-singular go
‘The princess ran out shouting.’ ()
Verbs of body stance are possible incorporators, as in () and ().
 Activity incorporates in some Athabaskan languages 145

() Verbs of body stance: Koyukon

a. dlukk ‘laughter, smile’ ()
dlukk-ghe-daa-dle-do
laugh-qualifier-qualifier/aspect-l voice/valence-singular sit
‘He is sitting down laughing.’ ()
b. tsegh, tseh ‘crying, weeping, tears, grief’ ()
tsegh-ghe-daa-dle-do
crying-qualifier-qualifier/aspect-l voice/valence-singular sit
‘She is sitting down crying; he sits weeping.’ (, )
c. k’onh ‘eating’ ()
k’onh-ghe-daa-dle-do
eating-qualifier-qualifier/aspect-l voice/valence-singular sit
‘She is staying and eating.’ ()
() Verbs of body stance: Ahtna
tsagh, tsax ‘crying’ ()
y-a-tsagh-da-gha-l-da’
OO-for-cry-qualifier-aspect-l voice/valence-sit, perfective
‘He stayed mourning for him.’ ()
Verbs of motion and body stance, light verbs, are the core classes that serve
as incorporating frame in the activity incorporate construction. In addition,
verbs of locution serve as incorporator; see section . on this and other cases
that resemble the activity incorporate construction but do not share the full
range of properties of this construction. The restrictions on the verb are
similar to those found in many languages with serial verbs; see for instance,
Aikhenvald () and section . for discussion.

.. Transitivity
The verb stem in the activity incorporate construction is lexically intransi-
tive. This is perhaps surprising with motion verbs, as transitive verbs exist
in this category, the so-called classificatory verbs found across the family,
or verbs of handling. Like going verbs, handling verbs are light in terms
of meaning, specifying the type of object under consideration, with infor-
mation about the manner of handling coming from other pieces of the verb
word. No examples appear in the dictionaries of handling verbs with an
activity incorporate. Thus, verb words such as ‘s/he laughingly handed me
things’ or ‘s/he cried carrying the baby around’ are not found in the activity
incorporate construction in the dictionaries. In addition, transitivised verbs
built on lexically intransitive verbs of motion and body stance do not occur
146 Rice

in the construction. For instance, intransitive verbs ‘sit’ and ‘go’ can be tran-
sitivised through the use of the transitiviser voice/valence marker ɬ (with a
predictable allomorph tl ). In the Koyukon examples in () and (), the (a)
form is intransitive, with inflection for aspect, and the (b) form includes the
transitiviser. Note that the transitiviser has allomorphs tl (b) and ɬ (b).
() a. le-do ‘He is sitting, staying.’ ()
b. yee-tl-do ‘He is causing her to sit, stay, he is keeping him.’
() a. nee-yo ‘He arrived, came.’ ()
b. yeenee- ɬ -yo ‘He arrived walking him, made him walk.’ ()
I have found no examples of such a verb with an activity incorporate in the
dictionaries. See section .. for discussion.

.. Properties of the subject

Based on the examples in the dictionaries, the subject of the activity incor-
porate construction is human or a dog.

.. Summary: the incorporating frame

Incorporating verbs are light verbs drawn from intransitive verbs of motion
and body stance and, based on the corpus, have higher animate, normally
human, subjects.

. Properties of incorporates

Just as the incorporating verb is restricted in the activity incorporate con-
struction, so too is the incorporate. Incorporates are drawn from the class
of oral activities, and include words with translations like ‘sing’, ‘laugh’,
‘cry’, ‘talk’, ‘whistle’, ‘call’, ‘chatter’, ‘bark’, ‘shriek’, and ‘eat’. In this section
I examine properties of the incorporate. I first show that the activity incor-
porate is a noun related by a type of conversion process to an activity verb.
I then show that activity nouns are not formed from verbs of other aspec-
tual classes. Following this I discuss restrictions on activity nouns: they are
conversions from intransitive or detransitivised verbs of oral activity with
agentive subjects.

.. Lexical category of the incorporate

Activity incorporates are lexically nouns, as discussed in Rice (). There
are a number of sources of evidence for this analysis. First, many activity
incorporates are used as nouns independently, as in the examples in ()
from Koyukon.
 Activity incorporates in some Athabaskan languages 147

() a. k’onh aade dzaan k’e-ghee-yo

eating without day unspecified DO.aspect-singular go
‘He went all day without eating.’ ()
b. dlukk kk’e naa-l-’onh
laugh like qualifier-aspect-compact object is in position
‘He is smiling, he has a smile on his face; he looks like he is going
to laugh.’ (, )
c. k’eleek yen’ehedeeleeyh
song . PL.S sing
‘They are singing a song.’ ()
d. tseh ts’e dohudaanee’onh
crying to  went verbally
‘She turned to crying.’ [lit.‘she went [verbally] to grief’] ()
Second, activity incorporates appear as subjects and objects in incorpor-
ating constructions. In (), tseh ‘crying’ is an incorporated subject. The
verb stem, -taanh, is a classificatory stem, involving handling of an animate
object, and the voice/valence marker, ɬ, marks the transitivity.
() a. haa-tseh-ye-d-ee- ɬ -taanh
away-crying-.DO-qualifier-aspect-ɬ voice/valence-handle animate
‘She went away crying’ [lit. crying took her away] (–)
The examples in () have the transitive nen ‘move sg. compact O quickly,
throw compact O’ (p. ); the stem form is perfective. ‘O’ is a device used
to indicate transitivity, and the incorporate is a direct object.

() a. ye-ts’e ts’aa-kkenaa-daa-n-ee- ɬ -nenh

OO-to out into open-word-qualifier-qualifier-aspect- ɬ voice/
valence-move singular compact O
‘She spoke sharply to him, jumped at him.’ [lit. ‘threw words at
him’] ()
b. ho-dlukk-de-ghee- ɬ -nenh
up and out-laugh-qualifier-aspect- ɬ voice/valence-move singular
compact O
‘He laughed out, chuckled.’ [lit. ‘threw a laugh out’] ()
One further example is given in (). This verb word is built on a lexically
transitive verb O+G+Ø+ggotl ‘rip, tear, jerk, pull, yank O abruptly’ (p. )
(‘G’ indicates that the verb allows marking for gender).
148 Rice

() no-yooyeesk-do-de-ggotl
‘He whistled loudly.’ ()
This example has d voice/valence, or middle marking. It is not clear why
some of the others have d while others do not, but this takes us beyond the
scope of this chapter.
Continuing with the properties of activity incorporates as nouns, the
nouns can be possessed, with a suffix -e’. Possession is indicated by a hyphen
before the stem.
() k’eleek, -k’eleeg-e’ ‘song, incantation, chant’ ()
-zel-e’ ‘shouting, noise, howling’ ()
The properties discussed above indicate that the activity incorporate is
a noun: it is related to independent nouns, it serves as a subject and object
incorporate, it has a possessed form. In the next section, we will see that
while the activity incorporates relate to verb forms, there is not a systematic
morphological relationship between them.

.. Relationship of activity incorporates to verb stems

Activity incorporates usually are related to verb stems. However, there is no
straightforward relationship between the form of the activity incorporate
and that of the verb stem.
A number of verb stems are formed on a single root, determined by
aspectual properties. See Kari () and Axelrod () and the dictionaries
for details. In comparing the form of activity incorporates with the form of
roots and stems, no systematic formal relationship between them is found.
Some activity incorporates are identical to the root and some to the verb
stem that is called the durative imperfective, while some are different from
either. The list in () gives Koyukon roots, durative imperfective forms,
and activity incorporates. Some nouns have prefixes in addition to the stem;
these are separated from the stem by a hyphen to facilitate comparisons.
() root durative imperfective noun
dlukk dlukk dlukk ‘laugh, smile widely’ /
‘smile, laughter’ (–)
yeesk yeesk yo-yeesk ‘whistle’ ()
tlaa tlaa ye-tlaa ‘bark’ / ‘barking’ ()


Ahtna is similar to Koyukon. For instance, the incorporate tsagh ‘cry’ is identical to the
durative imperfective, as is the stem of the incorporate c’udyiis ‘whistle’. However, there
are differences with ‘laugh’ (root, durative imperfective dlok’; incorporate dlo’, dlok), ‘sing’
(root, durative imperfective lii; incorporate c’eliis, i’dliis), and ‘talk’ (durative imperfective
yaes ~ naes; incorporate hnae, hne).
 Activity incorporates in some Athabaskan languages 149

tsaah tsaah/tseh tseh ‘cry, weep, mourn, lament’ /

‘crying, weeping, tears, grief’ (–)
hon honh k’-onh ‘eat’ / ‘eating’ ()
lee lee k’e-leek ‘sing’ / ‘song, incantation,
chant’ (–)
yo – kkenaa ‘talk’ / ‘word’ ()
zee ɬ – ze ɬ ‘shout, yell, holler, call out,
yelp, go shouting’ / ‘shouting, noise,
howling’ (–)
Activity incorporates often share the root or the durative imperfective stem
form with the related verb, as seen in (). However, this is not always the
case, as is evident in the last three lines. There is not a clear unique syn-
chronic morphological relationship between the verb stem and the form of
the activity incorporate, and, while it is noted in the Athabaskan literature
that a single root can often form both a verb stem and a noun stem, I know of
no attempts to identify a systematic relationship between them. Historically,
activity incorporates derive from verb stems. For instance, Jetté and Jones
(: ) note of the incorporate kkenaa ‘word, talking, speech’ that it is ‘a
unique noun’ that derives from the verb theme he+ne+Ø+(y)o° ‘sg. talks’,
with the ne prefix appearing as a stem-initial consonant. They note of k’onh
‘eating’ that it derives from the verb k’e+onh ‘eat’, again indicating that the
noun form derives from the verb.
Nominalisation of this sort, a type of conversion, is not synchronically
productive. A first indication of this comes from the lack of a systematic
relationship in form between the verb and noun stems. Further indications
that this conversion is not productive, but rather that the forms are lexically
listed, are examined in following sections.

.. Aspectual restrictions on activity incorporates

Activity incorporate nouns are related to verbs drawn from what Kari calls
the operative verb theme category and, occasionally, the onomatopoetic
verb theme category. Axelrod (: ) defines operative verb themes as
referring to activities that take place over a period of time or to activities
done as a means of employment, giving ‘to pick berries’, ‘to pluck feath-
ers’, ‘to cry’, and ‘to rain’ as examples; she defines the onomatopoetic verb
theme category as ‘the production of a sound or noise generated by an
object, instrument, or animal’ (p. ). Verb theme categories are seman-
tic groupings of verbs with shared formal and semantic characteristics. A
particular verb theme, or lexical entry, is assigned to a verb theme category
based on its aspectual patterns and derivational potential, among other
characteristics; see Kari () and Axelrod () for detailed discussion.
150 Rice

In Rice (a), I argue that verb theme categories can be thought of as

what is more commonly known in the linguistic literature as situation
aspect classes. While there is controversy in the literature about whether
the verb theme categories are based in situation aspect (see, for instance,
Smith (), Rice (a), and Wilhelm () for different perspectives),
there is agreement that there are verb classes, each with overall coherent
morphological and semantic characteristics that relate to aspect, although
with some fuzziness. In the following discussion, I assume that the verb
theme categories represent situation aspect classes and examine the activ-
ity (operative), accomplishment (conversive), and achievement (motion)
classes, showing that while nouns related to activities can often be incor-
porated into the construction under discussion, nouns related to accom-
plishments and achievements are not incorporable in this construction. I
set aside the onomatopoetic verbs for lack of full information, although
they represent activities based on various criteria. More specifically, I
examine cases where there is a noun/verb pair of the sort examined so far,
where the verb is drawn from the different situation aspect classes. While
activities and accomplishments often have pairs, only activities occur in
this construction.

... The incorporation of activities

As discussed above, activity incorporates are nouns, often with a related
verb form. The Koyukon verb stems in () above are of the activity class,
in the sense defined above. While many activities are like those in (), with
incorporated forms, there are also activities that do not appear to have an
incorporated form. I return to a discussion of what type of activities can be
incorporated in section ...

... The absence of accomplishments as activity incorporates

Just as many activity situation aspect verbs have related noun forms, so do
many accomplishments. However, unlike activities, accomplishment-related
nouns are absent as incorporates in the construction under discussion. The
Koyukon verbs in () are accomplishments (Kari’s conversive verb theme
category). The first line is a verb entry, with its meaning, and the second
line the related noun. The final morpheme in the verb entry is the root.
Preceding it is a lexicalised voice/valence marker, and, in (d), the pos-
sibility of a gender marker (recall that ‘G’ indicates that the verb can take a
gender marker). I choose these items for two reasons. First, there is a related
noun stem. Second, they have meanings that one could imagine would be
possible in the construction under discussion, namely an event that occurs
simultaneously with the event or state specified by the main verb. On the
third line, we see that either there is no incorporated form listed (a, b, c)
or the form of the incorporate (d).
 Activity incorporates in some Athabaskan languages 151

() Accomplishments in Koyukon

a. verb: Ø+tson’ ‘to defecate, have bowel movement’ ()
noun: tson’, -tsone’ ‘excrement, faeces, dung’ ()
no listed incorporate
b. verb: D+doots ‘defecate, have small hard faeces’ ()
noun: doots ‘dung, small dry faeces, droppings of rabbit, moose,
porcupine’ ()
no listed incorporate
c. verb: Ø+lets ‘to urinate, pee’ ()
noun: ɬ ets, -ledze’ ‘urine’ ()
no listed incorporate
d. verb: G+D+ten ‘freeze, become chilled, numb with cold’ ()
noun: ten ‘ice’ (found in compounds) ()
incorporate: tenh’
While the accomplishment-related nouns do not appear in the activity
incorporate construction, a similar meaning, involving simultaneous activi-
ties, is achieved in other ways. Consider the following examples, the first
built on the lexical entry Ø+tson’ in (a) and the second involving an incor-
porate related to ‘freeze’ in (d).
() hukkaadaanee-tson’ ‘He defecated all over the place, floor.’ ()
hu -kkaa ‘all over the place’
The stem in () is tson’, with any sense of movement conveyed through a
preverb, the postposition kkaa ‘all over’, with an object hu, indicating what is
translated as ‘place’.
() hebe-no-tenh’-ee-de-daatl
.PL.OO-iterative-ice-aspect-voice/valence-plural go
‘They returned thoroughly chilled.’ [(lit. ‘ice went back with them’]
()
In (), the verb stem is daatl ‘plural go’. The incorporate tenh’ is the sub-
ject, and hebe is a third person plural oblique object that translates as the
subject in English.
In both dictionaries, nouns are found that relate to accomplishments, and
they may appear as incorporates. However, they represent entities rather
than events, and do not occur in the activity incorporate construction. It is
interesting to note that the dictionaries treat nouns related to accomplish-
ments differently than activity incorporates in how they are listed; while
the reasons underlying this choice are not presented in the dictionaries, the
152 Rice

different patternings likely form part of the justification for different types
of listings.

... The absence of achievements

Just as accomplishments fail to occur in the activity incorporate construc-
tion, so too do achievements. Unlike accomplishments, achievements do not,
in general, have nominal or incorporated forms.

... Aspectual restrictions: summary

The nouns that appear in the activity incorporate construction relate to
activity verbs (hence the name of the construction); while nouns related to
accomplishments exist, they are not found in this construction even if there
is a plausible meaning.

.. Transitivity restrictions

In this section I examine constraints on transitivity of activity incorporates.
Both intransitive and transitive activity verbs exist, but only intransitive
activity verbs have corresponding activity incorporates. In the presentation
of data, I follow the format in the dictionaries, with boundary types (#, +)
included. The final morpheme is the root, preceded by a lexicalised voice/
valence marker. Material to the left of that is part of the basic lexical entry of
the verb. Recall that the abbreviation ‘P’ indicates an object of a postposition
and ‘O’ indicates transitivity, or the presence of a direct object.
Intransitive activity verbs with activity incorporate counterparts are listed
in ().
() Intransitive activities (Koyukon)
a. Ø+tsaah ‘cry, weep, mourn, lament’ ()
b. Ø+zee ɬ ‘shout, yell, holler, call out, yelp,
go shouting’ ()
c. yoo#de+ ɬ +yeesk ‘whistle’ ()
d. (ne#)k’e+Ø+dlukk ‘laugh, smile widely’ ()
(note k’e lexicalised indefinite object)
e. P+e#de+Ø+lee ‘sing P’ ()
Some of these items require comment. The prefix k’e, in ‘laugh, smile widely’,
indicates an indefinite or unspecified argument, in this case a direct object
(see Jetté and Jones () and Thompson () for discussion). This prefix
is lexicalised with certain verbs, as in ‘laugh, smile widely’. While with most
verbs, this prefix provides one option for expressing a direct object, with
this verb it is always present; no other object is possible. In ‘sing’, there is an
oblique object, symbolised by ‘P’. This is often the prefix k’e, making a more
literal translation ‘sing something’.
 Activity incorporates in some Athabaskan languages 153

Transitive activities also occur in these languages, as in the Koyukon

examples in (). Related nouns are included when one is listed in the
dictionary.
() transitive activities (Koyukon):
a. O+D+noon ‘drink’ ()
b. O+Ø+zes ‘sip, drink (hot liquid, alcoholic beverage)’
()
c. O+G+Ø+kk’us ‘chew, gnaw O (hard, resistant substance
such as dry fish or gum), chew gummy
O’ ()
d. O+G+Ø+ghos ‘chew, gnaw (hard) O’ ()
e. O+G+ ɬ +ts’oots’ ‘suck on O (food, candy, thumb)’ ()
f. O+G+Ø+tl’ootl ‘suck, lick O with sucking, smooching
sound, draw on, siphon, vacuum O (with
pipe, straw, bone)’ ()
g. O+ ɬ +ts’eesh ‘scrape and stretch O (wet skin)’ ()
h. O+de+ne++zenh ‘slice, score, cut O (especially fish) with
knife’ ()
i. O+G+Ø+ghon ‘make pl. O, kill, beat up pl. O’ ()
j. O+oo+ne+Ø+(h)aa ‘pick O (berries, spruce gum, pitch)’ ()
k. O+G++k’uts ‘pluck O (feathers, hair, grass)’ ()
l. O+oo+G+Ø+kkat ‘buy O, pay for O, obtain O by trade’ ()
m. O+G+ ɬ +tson ‘sniff, smell O’ ()
tsaanh, -tsaanh ‘scent, odor, aroma’ ()
n. O+ ɬ +ggoo ɬ k ‘scale O (fish)’ ()
ggo ɬ k ‘fish scales’
o. O+de+ ɬ +lon ‘tan O (skin), work O (wet hide) to soften
it’ (–)
k’elaah ‘skin tanning’
p. O+G+ ɬ +zook ‘take, handle pl. O, marry O (pl. women),
comb (hair)’ ()
tleezook, -tlezooge’ ‘headdress made of a band with upstand-
ing feathers, wooden comb’ (tlee ‘head’ +
zook ‘comb’, –) ()
q. O+G+Ø+luh ‘scrape, rake O, smooth O by scraping’
()
bek’eelugh-e ‘bone or iron scraper for smaller skins (that
with which we scrape something)’ ()
154 Rice

These transitive activities sometimes have related stem nouns ( m–p), but
they do not have related incorporate forms listed in the dictionary. Some
of the nouns are entities (m, n, p), rather than events, while the activity
incorporates represent events; see below on (). The different patterning
of nouns related to transitive and intransitive activities suggests that incor-
porates in the activity incorporate construction are further delimited in that
they are based on intransitive activities.
An apparent exception exists to this generalisation about transitivity, given
in (). (a) is a verb form, ‘eat’, (b) the related activity noun form, ‘eat-
ing’, and (c) shows this noun as an incorporate in the activity incorporate
construction.
() a. verb form: O+G+Ø+(h)on ‘eat O, go while eating O’ ()
form with another object
k’egheehon’ ‘He ate something; he ate a meal.’
()
b. noun form, incorporate: k’onh ‘eating’ ()
c. k’onh-ghe-d-aa-dle-do
eating-qualifier-qualifier-aspect-l voice/valence-verb stem
‘She is staying and eating.’ ()
derived from k’e+onh ‘eat’ (indefinite object + eat)
The verb on which k’onh ‘eating’ is formed is transitive, as indicated by ‘O’
in (a). There is discussion in the dictionary about the use of the indefinite
object k’e, introduced briefly above. Jetté makes the following remark, cited
in Jetté and Jones (: –):
It represents an object, or subject which it is not desired to mention, or
which would be needless to express and is quite extensively used … Its
adjunction, as object, to a transitive verb, forms the corresponding intransi-
tive, or ‘inclusive’, as k’esonh I eat (intr.) from esonh I eat (trans).
Jetté thus treats this morpheme as a detransitiviser; see also Thompson
(). There is a difference between the verb ‘laugh’ in (d) that always
occurs with this morpheme and those referred to in the quote in that in
examples such as () the verb can occur without k’e and with a specific dir-
ect object. The noun k’onh ‘eating’ is lexicalised, indicating an activity, and
as such can appear in the activity incorporate construction. In the activity
verbs in (), only one, (o), has a similar related nominal form, with k’e
included.
The presence of k’onh ‘eating’ suggests that transitive activity verbs with
unspecified objects – representing generalised activities – may be, poten-
tially, nominalisable. However, as discussed earlier, conversion-type nom-
 Activity incorporates in some Athabaskan languages 155

inalisation is not productive. Rather productive nominalisation involves a

suffix, -e, as in in ().
() a. O+G+Ø+luh ‘scrape, rake O, smooth O by scraping’ ()
bek’eelugh-e ‘bone or iron scraper for smaller skins’
(that with which we scrape something) ()
b. leggunh ‘it (food) dried up, he is skinny, emaciated’ ()
leggun-e ‘dried fish’ (that which is dried) ()
Productive nominalisation is not based on the kind of conversion of a stem
found with the activity incorporates discussed in ... If nominals similar
to k’onh were productively formed, one might imagine that they could be
used in the activity incorporate construction, subject to further constraints
discussed in section ...
To summarise, noun counterparts of intransitive activity verbs may
be incorporable; counterparts of transitive activity verbs are not incor-
porable. The one noun related to a transitive activity that is incorporable
in the activity incorporate construction contains the unspecified object
k’e as part of the lexicalised noun, a morpheme that functions as a
detransitiviser.

.. Subject restrictions

So far we have seen that the activity incorporate construction occurs with
nouns related to activity predicates that are intransitive. An additional con-
straint concerns the thematic role of the subject of the activity incorporate in
this construction.
Verb forms of incorporable activities have agentive subjects. Activity
verbs with non-agentive subjects do not generally have incorporated forms.
() shows Koyukon activity verbs with non-agentive subjects. These do not
have an incorporated form in the dictionaries, although with (b) a mean-
ing is imaginable (e.g. go around being itchy).
() Activity verbs with non-agentive subjects
a. G+ ɬ +get ‘rot, become rotten, have sores’ ()
no incorporated form given
b. G+Ø+gus ‘be itchy, tingly, prickly, ticklish’ ()
ghee-gus ‘he was itchy, ticklish’
no incorporated form given

In Ahtna productive nominalisations are formed with a suffix –(y)i: uk’e’sc’aayan-i ‘table’
(that on which we eat, ); kasighiɬ’aa-yi ‘seal’ (that which has its head sticking up, ); uyii
c’udelyiis-i ‘whistle’ (that in which there is a whistling noise, ).
156 Rice

Some activity verbs with non-agentive subjects have noun counterparts,

but they do not occur in the activity incorporate construction. An example
is given in (). (a) is an activity verb; (b) is its incorporated form. Note
that the noun konh ‘rain’ is a subject, in a different relationship with the verb
than the activity incorporates.
() a. ɬ +kon ‘rain’ 
eɬ-konh ‘It is raining’; gheeɬ-kon’, gheeɬ-konh ‘It was raining’ ()
b. he-konh-ne-ghee-detl ‘big raindrops, scattered rain fell’ ()
no-konh-te-hoyh ‘It is raining hard, coming down in sheets’
()
To summarise, nominal counterparts of intransitive activities with agent-
ive subjects are potentially incorporable in the activity incorporate construc-
tion, while those with non-agentive subjects are not.

.. Further restrictions on activities: oral activities

The incorporates in the construction under study have the following charac-
teristics: they are lexicalised nouns that relate to intransitive or detransitiv-
ised activity verbs with agentive subjects. In this section I further examine
intransitive activities with agentive subjects in order to highlight an add-
itional restriction, one that is likely obvious from the examples above: not
only are activity incorporates related to intransitive activity verbs, but they
are a restricted set semantically. The incorporates relate to oral activities –
for instance, ‘sing’, ‘laugh’, ‘shout’, ‘cry’, ‘whistle’, ‘potlatch speech’, ‘chatter’,
‘bark’, ‘eating’. Other intransitive activities do not appear to have incorpor-
ates that appear in this construction, as illustrated below.
Some additional intransitive activity verbs with nominal counterparts
are given in (). These nouns are not attested in the activity incorporate
construction.
() a. verb Ø+zaah ‘spit, drool’ ()
noun saakk ‘saliva, spittle’
b. verb k’e+le+dzeets ‘dance fast native dance, to dance (any
non-native dance)’ ()
noun k’edzes ‘a dance, dancing’ ()
c. verb so#de+le+tseeyh ‘be happy, gregarious, to look good’
()
noun sots’eeyh ‘joy, happiness, gladness’
While the nouns in () are related to intransitive activity verbs, they are
not shown in the activity incorporate construction in the dictionaries. These
nouns differ from activity incorporates in one of two ways: the noun is an
 Activity incorporates in some Athabaskan languages 157

entity rather than an event (a), or the noun does not involve an oral activity
(b, c).
One might ask if it is possible to convey the general meaning of the activity
incorporate construction – simultaneous activities carried out by the same
subject – in other ways. There is a construction where the activity occurs as
the stem of the verb word, there is not an incorporate, nor is there a verb of
motion, but instead motion is indicated through the use of preverbs. Some
examples are given in ().
() a. ye-do-k’e-e-l-dzeets ‘She came in dancing.’ ()
b. kk’o-k’e-e-deeyh ‘he is going about while eating’ ()
c. k’e+le+lun’ ‘go acting crazy, drunk, erratic’ ()
ye-do-k’e-e-l-lun’ ‘he came in acting crazy, hyperactive,
drunk’ ()
tlee-k’e-e-l-lun’ ‘he went out acting crazy’ ()
(a) is built on k’e+le+dzeets ‘dance’. The preverb, the postposition do ‘enter-
ing’, conveys the motion. In (b), the stem is -deeyh, a form of the verb stem
‘eat’ (Jetté and Jones (: ) note the irregularity of this verb form), with
motion conveyed by the preverb kk’o, called perambulative in the literature,
indicating motion around and about, in various directions, in no fixed direc-
tion (p. ). (c) is similar, with motion marked by a preverb, the postpos-
ition do ‘entering’ in one example and the preverb tlee ‘out’ in the other.
In addition, a similar meaning can be expressed by a construction with an
incorporate as subject, together with a classificatory verb stem. The example
in () involves the stem -taayh, the progressive form of ‘handle animate
object’ with the subject sots’eeyh ‘happiness’, an incorporate.
() kk’o-sots’eeyh-ye-e- ɬ -taayh
perambulative-happiness-.DO-progressive-ɬ voice/valence-handle
animate
‘He is going around very happy.’ [lit. ‘happiness carries him around’]
()
Thus, while the use of the activity incorporate construction is limited, other
possibilities exist for expressing that two activities are carried out simultane-
ously by the same subject.
Returning to verbs of oral activity, given the absence of a productive rela-
tionship between the verb and noun form, it is not surprising to find that
verbs in the oral activity class do not always have incorporate counterparts.
The Koyukon examples in () are intransitive oral activity verbs without an
activity noun counterpart in the dictionary.


The Ahtna cognate ts’uux can occur incorporated; see the example in (b).
158 Rice

() a. Ø+ts’ookk ‘scream, squeal, screech, shriek’ ()

b. hu+Ø+ts’eet ‘tell a lie, make a false statement (either deliber-
ately or by mistake)’ ()
Further, given the existence of the noun k’onh ‘eating’, one might expect
similarly formed nouns for some of the verbs in (); for instance, one might
expect that ‘drinking’ could be formed in a parallel way, but no such noun
is given in the dictionaries. In addition, there are verbs of oral activity with
related nouns where those nouns do not occur in the activity incorpor-
ate construction. For instance, in Ahtna a noun kos ‘cough’ exists (p. ),
related to the verb d+l+kos ‘cough’ (p. ), but no incorporated form is
listed, and no examples of this noun in the activity incorporate construction
are shown. A different case involves the Koyukon noun saak ‘saliva, spittle’,
related to Ø+zaah ‘spit, drool’ and the root zaakk ‘spit’ (p. ). This noun
is not shown in the dictionary as occurring in the activity incorporate con-
struction, although it has an incorporated form and can appear as an object.
However, while the noun is a stem noun of the type found as activity incor-
porates, as the activity incorporates are, its semantics is different from that
of the activity incorporate: as noted above, it represents an entity, the object
of the spitting, rather than the act of spitting. Given this semantics, it is to
be expected that it cannot occur in the activity incorporate construction.
To summarise, the activity incorporate construction involves a noun of
oral activity and a verb stem of motion or body stance. The events speci-
fied by the verb and the activity noun are carried out simultaneously. The
activity incorporates are related to activity verbs by a type of non-productive
conversion, and they are lexically listed.

.. Grammatical role of incorporate

In this section I discuss the grammatical role of the incorporate in the activ-
ity incorporate construction by examining voice/valence marking in the
activity incorporate construction. The voice/valence element that is found
in the construction, l, is used productively to express a middle of a transi-
tivised form, as introduced in section . (see Arce-Arenales, et al. (),
Thompson (), and Rice (b) on middles). (a) is an intransitive
verb, (b) a transitive formed on the intransitive, with the transitive marker
ɬ , and (c) a middle form, a reflexive, built on (b), with the surface voice/
valence marker l. The voice/valence markers are bolded in these forms.
() a. intransitive
nee-yo ‘He arrived, came.’ ()
b. transitive
yee-nee- ɬ -yo ‘He arrived walking him, he made him walk
(instead of riding in a vehicle).’ ()
 Activity incorporates in some Athabaskan languages 159

c. middle of transitive
hok’e-d-o-l-ho ɬ ‘He is making himself walk.’
Similar examples are shown in () through (), with a lexically intransi-
tive verb ((a) forms), a causative/transitive ((b) forms), and a middle of the
transitive ((c) forms). These middles are reflexives of some sort, either direct
reflexives (, ) or self-benefactives ().
() a. dee-kk’onh ‘It ignited, it caught on fire.’ ()
b. yedee- ɬ -kk’onh ‘He burned it.’
c. hodee-l-kk’onh ‘He burned himself, he burned up.’
() a. le-do ‘He is sitting, he is staying.’ ()
b. yee-tl-do ‘He is causing her to sit, stay, he is
keeping him.’
c. eeyet hok’edeedle-do ‘He allows himself to stay there.’
() a. le-t’aa ‘It fried, roasted; softened in the sun,
heat.’ ()
b. ye-tl-t’aa ‘He roasted it, fried it.’
c. nelaan yenhedee-l-t’aa ɬ ‘They are all cooking meat for
themselves.’
These examples show that middles built on ɬ transitives have surface voice/
valence l, indicating transitivity and middle marking.
As noted in section .., the activity incorporate construction occurs
with l voice/valence. As discussed in section .., the activity incorporate
construction is built on a lexically intransitive verb of motion or body stance.
The presence of the transitiving part of the voice/valence marker suggests
that the activity incorporate construction itself is in fact transitivised from
an intransitive lexical entry, with the activity incorporate a direct object, and
the middle marking piece of the voice/valence marker perhaps due to shared
event properties. Thus the activity incorporate construction is transitive,
built from an intransitive lexical entry, with the activity incorporate filling
the role of the direct object. No other direct object is possible (e.g. a verb like
(b) with an activity incorporate as well, meaning something like ‘walk him
singing’ is not possible because there are two direct objects).
The analysis of the activity incorporate as a direct object provides an
account for the observation that the incorporating frame must be intransitive
(section ..). As noted in the previous paragraph, only a single direct object
is possible. Recall from section .. that lexically transitive motion class
verbs exist. As noted in section .., these verbs, called classificatory, place
restrictions on the type of object possible. For instance, the Koyukon basic
lexical entry O+G+ɬ+kooɬ ‘carry flat, flexible, fabric-like object’ (p. )
demands that the object be flat and flexible; the entry O+ɬ +taa ‘handle,
160 Rice

carry singular animate object (living or dead)’ (p. ) requires an animate
object; the entry O+G+Ø+ton ‘carry rigid, elongated, stick-like object’ (p.
) requires a rigid object; and so on. Thus, while lexically transitive verbs
of the motion class exist, they require a direct object that meets semantic
restrictions imposed by their lexical entry, and the activity incorporates do
not meet these restrictions. The activity incorporate construction occurs
with only a limited set of verbal categories as possible incorporating frames,
verbs of motion and verbs of stance, and thus transitive verbs of other cat-
egories are not possible candidates for the incorporating frame. Since there
can be only one direct object, it is not possible to create the activity incorp-
orate construction with a lexically transitive motion verb, as the verbs in
the appropriate category place demands on the object. Verb words such as
‘carry animate object singing’ or ‘carry flat fabric-like object crying’ are thus
not possible.

. Other incorporation constructions

In this section I consider briefly structures that are similar to the activity
incorporate construction in some ways but differ in others. First are cases
where an activity incorporate is present, but the semantics of the construc-
tion differs from that of the activity incorporate construction as discussed so
far. Recall that Jetté introduced the term ‘combined action’ for the simultan-
eous performing of two actions. In a different type of case with an activity
incorporate, the incorporate further defines the action specified by a verb of
locution.
() Verbs of locution: Koyukon
a. yetlaa ‘barking’ ()
ɬ eekaa yetlaa-de-l-ghus
dog bark-qualifier-.S-l voice/valence-holler
‘The dogs are barking, are making a clamor barking.’ (, )
b. se ɬ ‘shouting, noise, howling’ ()
se ɬ -he-do-l-ghos
shout-.PL.S-qualifier/aspect-l voice/valence-holler
‘They were shouting.’ ()
() Verbs of locution: Ahtna
a. tsagh, tsax ‘crying’ ()
tsax-d-o-l-ghos
cry-qualifier-.PL.S-l voice/valence-make noise
‘You pl. are crying.’ ()
cf. del-ghos ‘They are having fun, shouting.’ ()
 Activity incorporates in some Athabaskan languages 161

b. ts’uux ‘shriek’ ()

da-ts’uux-gh-a-l-yae ɬ
above-shriek-qualifier-aspect-l voice/valence-talk
‘They (baby birds) are sitting up above shrieking.’ ()
While with verbs of motion and body stance, two simultaneous activities
are expressed, with verbs of locution, there is a single activity, with the
incorporate providing more detailed information about the type of locu-
tionary act.
This construction further differs from the activity incorporate construc-
tion in that the qualifier gh is not present in Koyukon. In addition, while it
has the voice/valence marker l, as in the activity incorporate construction,
this is the marker that is lexically part of the verb with the verb found in
most of the examples, and thus it is not possible to tell if it is present because
of its lexicalised status or because of the construction. However, the example
in (b) also has l, and this verb has null voice/valence lexically, suggesting
that the use of l voice/valence is productive with the locutionary verb as
incorporator.
Another example is found in the Koyukon dictionary of a verb word with
an incorporate with this type of meaning. Jetté and Jones identify a motion
stem ggok ‘sg, dual runs’. There is a progressive form gguyhtl, the form that
serves as an incorporate:
() a. gguyhtl ‘running’ ()
b. gguyhtl-h-o- ɬ -de ɬ tl
running-PL-progressive-voice/valence-PL. go
‘They are running along, they are racing.’ (, )
This example is similar to those with locution verbs: the incorporate delin-
eates the type of going, not a distinct simultaneous activity. In addition,
the qualifier gh is not present. Finally, this verb word has a different voice/
valence marker than occurs in the activity incorporate construction, ɬ rather
than l. I have translated the stem as ‘plural go’, after Jetté and Jones. There
is an identical verb theme which Jetté and Jones (: ) translate as ‘plu-
ral objects move independently, fall’, and it can be transitivised, meaning
‘throw plural O, handle pl. O quickly’. Perhaps this example means ‘they are
handling the run’ or something similar. (The third trapping that is found in
the activity incorporate construction, the prefix d, is not expected since this
incorporate is not an oral activity (see section ..).)
A further type of incorporation exists that is worthy of mention, with a
meaning similar to that of the activity incorporate construction, but distinct
other properties. This type is illustrated in (), with the incorporated noun
geɬtl ‘hook’.
162 Rice

() a. taa-ge ɬ tl-’e-tl-do

water-hook-aspect-voice/valence-singular, dual sits, stays, dwells
‘She is sitting angling, fishing.’ (, )
b. taa-ge ɬ tl-’e-tl-taanh
water-hook-aspect-voice/valence-singular, dual reclines, lies down
‘He is lying down fishing with a hook; He is reclining, relaxing
while fishing with a hook and line.’ (, )
Like the activity incorporates, these examples indicate simultaneous events.
They are composed of a body stance verb and an incorporate. The examples
are distinct from the activity incorporate construction in that they lack the
gh qualifier that is common in the activity incorporate construction and they
have the voice/valence marker ɬ rather than l.
The forms discussed in this section are similar to the activity incorporate
construction in some ways, with an incorporate that provides information
about an event, one that is either simultaneous with the event specified by
the verb (activity incorporate construction, ()) or elaborates on the event
specified by the verb (). The activity incorporate construction is distinct
from the others in meaning in some cases, as well as the morphological trap-
pings that accompany the construction. Further study of forms such as those
discussed in this section is in order.

. The activity incorporate construction in the

complex predicate typology
Having described the Koyukon and Ahtna activity incorporate construc-
tion, I now address the construction in terms of the typology of complex
predicates. Baker and Harvey (this volume) propose two types of com-
plex predicate constructions, a merger structure and a coindexation struc-
ture. As described by Baker and Harvey, merger involves a combination of
predicates yielding a construction with the range of aspectual structures
found in monomorphemic predicates, and they group merger constructions
and monomorphemic predicates together as simplex event constructions.
The constraints involved in simplex event constructions are aspectual, with
major predicate functions (CAUSE, BECOME, MOVE, BE) occurring only once in
the construction. Complex predicates such as ‘be sitting laughing’ and ‘go


These verb words differ from the activity incorporate construction in another way. Usually
a preverb forms a semantic unit with the stem (e.g. in a verb word such as na-c’uudyiis- de-
l-yaa ‘He returned whistling’, or ɬ u-hdogh-d-a-l-yaa ɬ ‘He is walking around while making
a potlatch speech’, the preverb (na ‘back’, ɬ u ‘around’) and the verb stem (yaa, yaa ɬ ‘go’)
are a unit semantically). In ‘sit hooking’ and ‘lie hooking’ in (), the preverb taa ‘in water’
(Jetté and Jones : ) refers to the position of the hook rather than to where the sub-
ject of the verb is positioned.
 Activity incorporates in some Athabaskan languages 163

along singing’ are not possible simplex event structures. Thus, the activity
incorporate construction of Ahtna and Koyukon, with two events specified,
is not a merger structure.
The alternative structure proposed by Baker and Harvey is a coindexation
structure. Coindexation constructions are multi-predicate constructions and
not subject to the constraints that exist on simplex constructions. The Ahtna
and Koyukon constructions are coindexation structures, and in general have
properties of these, with the verb stem being one predicate and the activity
incorporate the second.
One type of coindexation structure is the serial verb construction, and the
activity incorporate construction is similar to the serial verb construction
in some ways. In work on the typology of serial verbs, Aikhenvald (: )
outlines a number of properties of such constructions. She says:
A serial verb construction (SVC) is a sequence of verbs which act together
as a single predicate, without any overt marker of coordination, subordin-
ation, or syntactic dependency of any other sort. Serial verb constructions
describe what is conceptualized as a single event. They are monoclausal;
their intonational properties are the same as those of a monoverbal clause,
and they have just one tense, aspect, and polarity value. SVCs may also
share core and other arguments. Each component of an SVC must be able
to occur on its own.
The activity incorporate construction bears parallels to the serial verb
construction: there is no overt marker of coordination, subordination, or
syntactic dependency (section ...); the events are conceptualised as a
single event in that simultaneity is an important characteristic of the con-
struction; the construction is monoclausal, with phonological properties of
a single phonological word (section ...); there is one marking for subject
and aspect (sections ..., ...). The components can appear on their
own, with one a verb and the other a noun.
The construction is similar to a serial verb construction in other ways.
Many languages have what Aikhenvald (: , –) calls an asymmetrical
serial verb construction, a construction in which one of the verbs is from a
grammatically or semantically restricted class (e.g. a motion or posture verb).
This contrasts with a symmetrical serial verb construction where the verbs
are from semantically and grammatically unrestricted classes. Asymmetrical
constructions, Aikhenvald says, ‘denote a single event described by the verb
from a non-restricted class. The verb from a closed class provides a modifi-
cation specification: it is often a motion or posture verb expressing direction’
p. ). The activity incorporate construction is asymmetrical, with the verb
of the motion or posture class and the noun of the class of oral activities.
Both the verb and the noun are restricted, but in different ways.
The construction differs from a serial verb construction in several ways.
Most obviously is the fact that it does not involve two verbs, but a verb and
164 Rice

a noun. This has interesting consequences that differentiate the serial verb
construction and the activity incorporate construction. Aikhenvald notes
that in asymmetrical serial verb constructions, the ‘minor’ verb is from a
closed class and the ‘major’ verb is from an open class and functions as the
head of the construction, determining the transitivity of the construction as
a whole (: ). This is not true of the construction under study. First,
the minor verb counterpart is the head, and the major verb counterpart is a
noun. Aikhenvald notes a tendency, in asymmetrical constructions, for the
minor verb to undergo grammaticalisation in a variety of ways, developing
into tense/aspect and mode, directionals, and adpositions, among others
(pp. –). This has not occurred in the Athabaskan family. This can be
attributed to the fact that, while the activity incorporate construction resem-
bles the serial verb construction, it is composed of a verb and a noun. The
verb is a light verb in content, but it nevertheless is the sole verb, and is
retained as such, without another obvious candidate to be the head.
The constraints on the serial verb construction are similar to those in
other languages. First, as Aikhenvald discusses (: ), the most frequent
verbs in a serial verb construction involve basic verbs of motion, posture,
and stance. Thus, the verbs of the activity incorporate construction reflect a
cross-linguistic tendency in the similar serial verb construction. In terms of
the nouns, Aikhenvald (p. ) notes ‘combining verbs into an SVC may turn
out to be unacceptable if they do not match a “recognizable event-type”’ and
serial verb constructions ‘must relate only events which are somehow con-
ceived as notably more commonly associated together in experience or those
events which form a culturally important concatenation of events’. The choice
of incorporates of oral activity as the major verb in a serial verb construction
is common cross-linguistically, and thus it is not surprising that this restric-
tion is found in the languages under study. The absence of some potential
possibilities – for instance, forms such as ‘walk drinking’ and ‘go coughing’,
might be due to difficulties of performing such events simultaneously, and
may, perhaps, be attributable to cultural reasons. For instance, Jetté and
Jones (: ) list a noun ‘sneezing’. Jetté and Jones (: ), citing Jetté,
note the following of sneezing in Koyukon: ‘Sneezing is very decidedly omin-
ous. From the left nostril, it is an ill omen … But from the right nostril, it is
a presage of food and plenty.’ While one can imagine ‘entering sneezing’ or
‘sitting sneezing’, perhaps sneezing while engaged in something else is not
seen as a recognisable event sub-type by speakers of Koyukon. Alternatively,
this verb may not be considered to be a very good example of oral activity, so
it might be ruled out by the constraints discussed earlier.

. Summary and conclusion

The activity incorporate construction is a complex predicate involving
a verb and an incorporated noun of oral activity. The construction forms
 Activity incorporates in some Athabaskan languages 165

a type of coindexation structure in the Baker and Harvey typology, with

constraints on the predicates similar to those found on serial verb con-
structions in many languages. The incorporating frame, or verb, is a light
verb that indicates motion or stance. The incorporated element, an activity
noun formed by a non-productive conversion process, is restricted in that
it relates to an intransitive predicate of oral activity. The combination of
these is one that is likely to be of the relevance that Aikhenvald argues to
characterise a serial verb construction. The study is based on dictionaries,
and shows both the strengths of the dictionaries, and the limitations, par-
ticularly in determining just what absence means, and whether it has to do
with formal constraints or lack of plausibility under real-world and cultural
conditions.
In a recent study on serial verb constructions in Lakota, a Siouan language
of North America, de Reuse observes that heavily polysynthetic language
families, including Athabascan, do not ‘need verb serialization much, since
its functions can be carried out by affixation’ (: ). While the activ-
ity incorporate construction is not a serial verb construction, it is similar,
and it is thus interesting that it exists at all, given de Reuse’s claim, since, as
we have seen, there are alternative strategies available to express a meaning
similar to that found in this construction.

This work is funded by the Canada Research Chair in Linguistics and

Aboriginal Studies to Keren Rice. Thank you to Mengistu Amberber, Brett
Baker, Henry Davis, and Mark Harvey for helpful comments, and to the
participants at the conference for feedback.

References
Aikhenvald, A. Y. . Serial Verb Constructions in Typological Perspective. In
Aikhenvald, A. Y. and Dixon, R. M. W. (eds) Serial Verb Constructions; A
Crosslinguistic Typology, –. Oxford University Press.
Arce-Arenales, M., Axelrod, M., and Fox, B. . Active Voice and Middle
Diathesis: A Cross-linguistic Perspective. In Fox, B and Hopper, P. (eds)
Voice: Form and Function. Amsterdam: John Benjamins.
Axelrod, M. . The Semantics of Time: Aspectual Categorization in Koyukon
Athabaskan. Lincoln: University of Nebraska Press.
Baker, B. and Harvey, M., this volume. Complex Predicate Formation.
Butt, M., this volume. The Light Verb Jungle: Still Hacking Away.
de Reuse, W. . Serial Verbs in Lakota (Siouan). In Aikhenvald, A. Y. and Dixon,
R. M. W. (eds) Serial Verb Constructions. A Ccross Linguistic Typology, –.
Oxford University Press.
Jelinek, E. and Willie, M. . Psych Verbs in Navajo. In Jelinek, E., Midgette, S.,
Rice, K., and Saxon, L. (eds) Athabaskan Language Studies. Essays in Honor of
Robert W. Young. Albuquerque: University of New Mexico Press.
166 Rice

Jetté, J. . On the Language of the Ten’a. Microfilm AL:–, JOPA, Foley
Library, Gonzaga University.
Jetté, J. and Jones, E. . Koyukon Athabaskan Dictionary. Fairbanks: Alaska
Native Language Center.
Kari, J. . Athabaskan Verb Theme Categories: Ahtna. Fairbanks: Alaska Native
Language Center Research Papers no. .
. Ahtna Athabaskan Dictionary. Fairbanks: Alaska Native Language Center.
. Summary of the Contents and Other Conventions of the KAD. In Jules
J. and Jones, E. Koyukon Athabaskan Dictionary xci–xciv. Fairbanks: Alaska
Native Language Center.
Rice, K. a. Morpheme Order and Semantic Scope. Word Formation in the
Athapaskan Verb. Cambridge: Cambridge University Press.
b. Voice and Valency in the Athapaskan family. In Dixon, R. M. W and
Aikhenvald, A. Y. (eds) Case Studies in Transitivity, –. Cambridge:
Cambridge University Press.
. On the Evolution of Activity Incorporates in Athapaskan Languages.
Diachronica : –.
Rice, K. and Saxon, L. . Comparative Athapaskan Syntax: Arguments and
Projections. In Guglielmo, C. and Kayne, R. (eds) Handbook of Comparative
Syntax. Oxford: Oxford University Press.
Smith, C. . Aspectual Categories in Navajo. International Journal of American
Linguistics : –.
Speas, M. . Phrase Structure in Natural Language. Dordrecht: Kluwer.
Thompson, C. . The Low Topicality Prefix k’i- in Koyukon. Studies in Language
:  –.
. The Na-Dene Middle Voice: An Impersonal Source of the D-element.
International Journal of American Linguistics ,: –.
Wilhelm, A. . Telicity and Durativity: A Study of Aspect in Dene Suɬiné
(Chipewyan) and German. New York: Routledge.
7 Warlpiri verbs of change and
causation: the thematic core*

MARY LAUGHREN

. Introduction
The morphosyntactic, semantic, and phonological properties of Warlpiri
verbs have been investigated by a number of scholars including Hale (,
), Hale, Laughren, and Simpson (), Harvey and Baker (),
Laughren (, ), Legate (, , ), Levin (), Nash (,
), Reimer (, , ), and Simpson (, ). This study
will focus on that part of the verbal complex which I will refer to as the
‘thematic core’. It minimally consists of a thematic verb (V) which may be
augmented by a non-inflecting preverbal (PV) element of a class which,
following Nash , I will refer to as ‘lexical’: [(PVlexical])=V], or by PVs
derived from other phrasal categories. I will argue that the thematic core of
the larger verbal constituent ‘maps onto’ an event structure which represents

* This work was supported by The University of Queensland Special Studies Program .
The opportunity to collect additional Warlpiri data came from fieldwork support from
ARC Linkage Grant LP . This study had the benefit of feedback on preliminary
presentations to the Workshop on Complex Predication and the Coverb Construction at the
Annual Meeting of the Australian Linguistic Society in July , in a Linguistics Seminar
at Macquarie University in October , and at the Australian Languages Workshop,
University of Sydney (Pearl Beach) in March . Thanks also to the very helpful feed-
back from an anonymous reviewer and the editors of this volume. This investigation would
not be possible without the contribution of over  Warlpiri speakers to the documentation
of their language over the past  years, and to the many linguists who have pooled their
collective findings. I would like to dedicate this chapter to the memory of the late Professor
Ken Hale who first inspired me to think more deeply about Warlpiri verbs all those years
ago and to the late Kay Napaljarri Ross who devoted countless hours over some  years
trying to help me to recognise their many faces. Shortcomings in the account given here
are of my making.

Warlpiri is an Australian language spoken in the Tanami Desert region of the Northern
Territory. The data is taken from the collection of Warlpiri oral and written texts compiled
as part of the Warlpiri Dictionary Project, which includes Professor Ken Hale’s exten-
sive collection of Warlpiri fieldnotes. It also includes data from the Warlpiri Dictionary
(Laughren et al. ) and from fieldwork carried out by Laughren –. An exten-
sive bibliography of linguistic studies on Warlpiri can be found at the website maintained
by David Nash: www.anu.edu.au/linguistics/nash/aust/wlp/wlp-lx-ref.html.

The verbal lexicon of Warlpiri is limited to around  verb stems (Nash ). As would
be expected, the meanings associated with Warlpiri thematic verbs are more general than
their translation equivalents in languages such as English with its thousands of basic verbs.
Finer shades of meaning are expressed by PVs.

167
168 Laughren

its predicate argument structure (PAS) and its Aktionsart properties. This
study investigates the relationship between the PV and V forms which
instantiate the thematic core, from the perspective of their individual and
combined contribution to the underlying event structure. It also compares
the event structure of ‘heavy verbs’ (HV) with their ‘light verb’ (LV) homo-
phones, and also with synonymous PV-LV complex verbs.
A question addressed in a number of relatively recent studies of similar
complex verbs in a wide range of languages including Persian (e.g. Karimi-
Doostan , Megerdoomian , Folli, Harley, and Karimi , Goldberg
, Ghomeshi and Massam  inter alia), Malagasy and Tagalog (e.g.
Travis a, b, ), is whether the complex predicates are formed
in the lexicon in a distinct word-formation component, or whether they are
formed in the syntax. Like their Persian counterparts, Warlpiri complex
verbs present a mix of idiosyncratic form to meaning mapping and system-
atic form–meaning associations. They also show evidence of syntactic inde-
pendence between the PV and V such that each element is visible to syntactic
processes they may participate in independently of the other. I will argue for
a level of lexical event structure or L-Syntax in the sense of Hale and Keyser
(, ), distinct from that of S-Syntax structure. An S-Syntax represen-
tation may map onto more than one event structure, e.g. intransitive monadic
stance verbs may be mapped onto a causative event structure or a stative one;
only those semantic elements of event structure which map onto elements of
syntactic structure are treated as entailments, whereas elements of a predi-
cate’s event structure which are not mapped onto elements of syntactic struc-
ture can only be interpreted as being presupposed. As documented for other
languages, Warlpiri distinguishes predicates which express resultant states,
such as rdilyki ‘broken’ which presupposes a change of a state brought about
by some external cause, from those such as ngurrju ‘good’, which just express
a state without presupposing it to be the result of a process of CHANGE. Both
rdilyki and ngurrju nominals may combine with a verb to form a complex
PV-V verb that expresses both CHANGE STATE or CAUSE CHANGE OF STATE
events. However, the role of the verbs which combine with each type of predi-
cative PV is different. In the case of rdilyki type PVs, the CAUSE CHANGE OF
STATE event structure is part of the lexical entry of the PV, and the role of
the lexical thematic V which combines with the PV is to express particular
components of this eventive L-Structure, such as CHANGE or CAUSE CHANGE
which are mapped onto the syntactic structure underpinning the surface
PV-V constituent. The ngurrju type PVs, on the other hand, only project a
stative L-Structure. This stative structure may be embedded in an aspectu-
ally dynamic eventive structure. The verbs which typically mark the map-
ping between the eventive L-Structure and and S-Structure are drawn from


The analysis of other complex verb phenomena such as Germanic verb plus particle com-
binations have engendered similar debate (e.g. see chapters in Dehé et al. (eds) ).
 Warlpiri verbs of change and causation: the thematic core 169

a set of syntactic dependent light verbs. Thus Warlpiri distinguishes between

lexical Vs which actualise CHANGE and CAUSE CHANGE functions in the event
structure of a ‘result of change’ predicate, and syntactic LVs which map onto
S-Structure positions corresponding to an aspectually dynamic event struc-
ture which incorporates that of a stative nominal.
My analysis of the Warlpiri data presented in this chapter is more in
line with the analysis of similar PV-V complex verbs in modern Persian
given by Karimi-Doostan (), who places their formation partially in
the L-Syntax and partially in the S-Syntax, as opposed to the more recent
analyses of Megerdoomian () and Folli, Harley, and Karimi () who
argue that complex verb formation is syntactic with surface forms the result
of the mapping of phonological forms onto components of syntactic struc-
ture in a spellout phase. By rolling Hale and Keyser’s L-Syntax into a single
S-Syntax component, the ability to account for mismatches between syn-
tactic and semantic structure disappear and the price to be paid is a more
and more elaborate syntactic representation which may still fail to capture
syntactically relevant semantic contrasts.
This study concentrates on complex verbs (and their simplex HV syno-
nyms) which express a notion of change: change of stance, change of location,
and change of state. Both inchoative and corresponding causative complex
verbs are investigated. The exploration of a range of lexicalisation patterns
(in the sense of Talmy ) found in both semi-productive and produc-
tive PV-LV combinations reveals evidence for distinctions in cause event
type: ‘external’ ( John stood the vase on the table) versus ‘internal’ ( John stood
up) (see Levin and Rappoport Hovav ( and ) for a detailed discussion
of this distinction), and ‘eventive’ ( John hammered it flat) versus ‘situational’
(The sun burned it black) cause. The latter distinction, I will argue, is aspec-
tual. Aspectual properties have been shown to be crucial to an understand-
ing of the syntax of predicate argument stucture as reviewed by Arad (),
and they play a crucial role in the formation of complex verbs (Folli, Harley,
and Karimi a, Megerdoomian , Travis a). Warlpiri PV-LV
combinations are sensitive to whether the end state expressed by the PV is
momentary or enduring, or whether the PV itself presupposes a change of
state in some entity or not. As in Persian complex verbs, the Warlpiri the-
matic V marks the argument structure of the complex verb with the strict
division between transitive and intransitive verb forms seen with HVs also
applying to LVs.


Megerdoomian () argues for a unitary ‘agent’ subject of the cause predicate while Arad
() distinguishes these two types of causing arguments. Arad calls our eventive cause
‘agentive’ and our situational cause ‘eventive’. I would argue that while this distinction
must be made in L-Structure, it does not need to be made at S-Structure. The inability,
for example, to interpret an instrumental phrase with a situational cause as opposed to an
event one, would derive from the differing L-Structures, not from their intrinsic under-
lying S-Structure.
170 Laughren

Some studies of complex verbs consisting of a PV and V treat the V as

uniformly ‘light’. The Warlpiri data show a continuum from HV to LV.
Verbs which only seem to express the rather abstract predicates such as
CHANGE and CAUSE need to be distinguished from those which express addi-
tional ‘manner’ features of both the CAUSE and CHANGE event. And these
lexical thematic verbs must be distinguished from syntactic thematic verbs.
Some verbs, I will argue, only check Aktionsart features within the thematic
core structure. Among these are simplex ‘stance’ verbs in their copula use
with syntactically independent nominal predicates.
As stated already, I assume a level of event structure or conceptual struc-
ture (Jackendoff , Laughren , Levin and Rappaport , , ,
Pustejovsky , , inter alia) which is independent of particular syntactic
categories. I also assume that the mapping between event structure and syntac-
tic structure is constrained by properties of syntactic categories. Languages,
including Warlpiri, may differ in the type of syntactic category – nominal,
verbal, adjectival – which maps onto components of the same event structure.
The event structure acts as a locus for the expression of syntactically relevant
meaning components – an interface between the full semantic representation
and its syntactic expression. The event structure I propose follows a long tra-
dition of predicate decomposition. It also incorporates the Aktionsart features
which underpin the Vendler () type classification. As well as interacting
with clause-level aspectual features, it has been claimed that these lower-level
Aktionsart features play an important role cross-linguistically in the map-
ping between semantic and syntactic representations (e.g. Pustejovsky ,
Ramchand , Tenny , , chapters in Tenny and Pustejovksy a,
Van Valin , among others). In the event structure I propose, the relation-
ship between a predicate and its ‘subject’ argument is mediated by Aktionsart
features: the subject argument is in the specifier position of an Aktionsart
phrase (AKTP) whose complement is the relevant predicate phrase (PredP)
governed by the features in the head of the AKTP.
The subject matter canvassed in this chapter will be presented in the follow-
ing way. Section . sets out some of the basic facts about the morphological and
syntactic properties of Warlpiri verbs – or more precisely, verbal complexes,
establishing the criteria for distinguishing the ‘thematic core’ from the rest


Travis (a) proposes an Aspect Phrase between her VP (equivalent to vP in mini-
malism literature) and VP structure which represents a conjunction of semantic and syn-
tactic properties in a single structure reminiscent of the Generative Semantics tradition.
Megerdoomian () takes this further in arguing for a number of aspect heads within the
verbal projection (which involves more than one verbal head), such that an aspect phrase
links each verb phrase to the one above it. My approach is a little different in that I am
assuming that the relationship between an argument and its predicate is mediated by an
aspectual head, such that the argument is in the specifier position of the Aktionsart (or
VP internal aspect) phrase. Hoeskstra () explored the relation between Aktionsart and
predication in his study of resultative predicates and causative constructions. Some of his
insights inform this study.
 Warlpiri verbs of change and causation: the thematic core 171

of the verbal complex. Section . investigates the relationships between the
elements which make up the ‘thematic core’ of verbal constituents, limiting the
study to a stance/location/state predicate in the scope of CHANGE and CAUSE
functions. Sets of verbal constituents with varying degrees of form–meaning
transparency and productivity are examined. Section . proposes a represen-
tation of the event structure expressed by the simplex and complex verb forms
seen in section ., and shows how this event structure relates to the semantic
properties of verbal, nominal, and postpositional predicates. Section . dis-
cusses some syntactic properties of the PV-V complexes discussed in previous
sections, showing a contrast between compound PV-V structures and phrasal
ones. A summary of findings is presented in section ..

. Structure of Warlpiri verbal complex

Before launching into a discussion of the interaction between the semantic
and morphosyntactic properties of Warlpiri verbs, it is important to set out
some basic facts about their formal properties and to distinguish between
different types of verbs which may be found in a single verbal constituent.

.. Inflecting verb

What I refer to as a verb in this paper is a member of a closed class of bound
morphemes which may host a suffix which marks tense or mood contrasts
thus creating a finite inflected verb, or which may host a suffix which derives
non-finite stems: infinitive (VINF) and participial (PRT). Verb stems fall into
one of  main conjugation classes according to the formal properties of the
set of inflectional and derivational suffixes they host. The minimal stems (or
roots) of conjugation , , and a verbs are polysyllabic, while conjugation
b, , and  verbs have monosyllabic stems. Conjugation  verb roots belong
to two paradigms (a and b) depending on their stem form: stem = root+nja
in a, stem = root in b. By comparison with the other conjugations, each of
the conjugation  paradigms is defective. Warlpiri has no default conjugation
akin to English ‘weak’ verbs or French er verbs. Nor is there any synchronic
process of conversion or ‘zero’ derivation to create verbs from other word or
stem classes. There is no productive verb formation, only productive com-
plex verb formation involving the association of a non-verbal PV with a V.
Partial verbal paradigms in () to () illustrate some of these formal prop-
erties of Warlpiri verbs.


A noun may also be derived from a verbal stem by application of the nomic suffix (similar to
English -er derived nouns) (see Nash ).

This parallels the behaviour of certain cognate verbs reconstructed in non-Pama-Nyungan
Arnhemland languages by Green () (see also Alpher, Evans, and Harvey ).

The full range of Warlpiri verbal inflections classified into conjugation paradigms is given
in Appendix .
172 Laughren

() Conjugation  verb forms:

karri- ‘stand’ bound stem = root
karri- ‘stand-PRT’ uninflected bound
stem (participle PRT)
karri ‘stand-NPST’ uninflected word
karri-mi ‘stand-NPST’ inflected word
karri-nja- ‘stand-INF’ inflected bound stem
(infinitive VINF)
karri-nya ‘stand-PRSTV’ inflected word
(presentative)
karri-ja ‘stand-PST’ inflected word
karri-ji ‘stand-FUT’ inflected word
karri-ya ‘stand-IMP’ inflected word
karri-yarla ‘stand-IRR’ inflected word

() Conjugation  verb forms:

paka- ‘hit’ bound stem = root

paka-rni- ‘hit-PRT’ inflected bound stem
paka-rni ‘hit-NPST’ inflected word
paka-rni-nja- ‘hit-INF’ inflected bound stem
paka-rni-nya ‘hit-PRSTV’ inflected word
paka-rnu ‘hit-PST’ inflected word
paka-ku ‘hit-FUT’ Inflected word
paka-ka ‘hit-IMP’ inflected word
paka-karla ‘hit-IRR’ inflected word


Abbreviations used in glossing Warlpiri examples:  = FIRST PERSON,  = SECOND PER-
SON,  = THIRD PERSON, ALLAT = ALLATIVE, ANAPH = ANAPHOR, AUX = AUXILIARY, CS =
CHANGED STATE, DAT = DATIVE, DIR = DIRECTIONAL, DU = DUAL, ERG = ERGATIVE, EXCL =
EXCLUSIVE, FUT = FUTURE, IMP = IMPERATIVE, IMPF = IMPERFECTIVE, INCEP = INCEPTIVE,
INCH = INCHOATIVE, INCL = INCLUSIVE, INF = INFINITIVE, IRR = IRREALIS, LOC = LOCATIVE,
NPST = NON-PAST, O = OBJECT, PL = PLURAL, PR = PRIOR STATE, PRS = PRESENT, PRSTV =
PRESENTATIVE, PRT = PARTICIPLE, PST = PAST, PURPD = PURPOSEFUL DISPLACEMENT, REFL =
REFLEXIVE, S = SUBJECT, SG = SINGULAR, SS = SAME STATE, SUBJCOMP = SUBJECTIVE COM-
PLEMENTIZER, VINF = INFINITIVAL VERB.

The inflected and uninflected non-past forms for conjugation  verbs seem to be in free
variation in all dialects. The inflected -mi form derives from an historic potential or future
form maintained in Warlmanpa (Nash ).
 Warlpiri verbs of change and causation: the thematic core 173

() a. Conjugation a verb forms:

yi.nja- ‘go & give’ bound stem
yi.nja-ni- ‘go & give-PRT’ inflected bound stem
yi.nja-ni ‘go & give-NPST’ inflected word
yi.nja-ni-nja- ‘go & give-INF’ inflected bound stem
yi.nja-nu ‘go & give-PST’ inflected word
yi.nja-nku ‘go & give-FUT’ inflected word
yi.nja-nka ‘go & give-IMP’ inflected word
yi.nja-nka.rla ‘go & give-IRR’ inflected word
(No PRESENTATIVE form)

b. Conjugation b verb forms:

(No PARTICIPLE form)

yi/yu- ‘give’ bound stem = root

yi-nyi ‘give-NPST’ inflected word
yi-nja- ‘give-INF’ inflected bound stem
yu-ngu ‘give-PST’ inflected word
yu-ngku ‘give-FUT’ inflected word
yu-ngka ‘give-IMP’ inflected word
yu-ngka.rla ‘give-IRR’ inflected word
yu.nga-nya ‘give-PRSTV inflected word

For all verbs, with the exception of conjugation b forms, the bound parti-
ciple form is phonologically equivalent to the non-past form, and constitutes
the infinitive stem. In the polysyllabic root conjugations  and , and in con-
jugation  (with its sole member, nga- ‘eat/drink’), the participial form also
constitutes the stem for the presentative inflection (used in similar contexts
to the English There VERBS xNP, There xPRON VERBS). Monosyllabic
root conjugations b and , on the other hand, have distinctive presentative
forms involving the historically augmented stems, e.g. yu.nga- ‘give’ b and
ya.na- ‘go’ , which host the presentative suffix nya.
The non-finite VINF and PRT forms which constitute the stems of inflected
verbs are in complementary distribution. The PRT serves as the stem for
inflections such as ‘presentative’ (PRSTV) and ‘infinitive’ (INF) and hosts the

The full stop/period in yi.nja and elsewhere marks an etymological morpheme boundary
which is synchronically inactive.
174 Laughren

defective inceptive aspectual verb -nji(na)-. The VINF, on the other hand,
behaves like a nominal stem. It may host postpositions and/or case-marking
suffixes. As we will see in section . it may also function as a PV with both
thematic and aspectual verbs.

.. Thematic verbs versus aspectual verbs

Thematic verbs express argument-taking predicates and once inflected may
constitute what I will refer to as an independent HV as seen in (a–c), or as
a LV in (a’–c’).
() a. karri - STAND (x) a′. pata-karri FALL(x)
b. paka- HIT (x,y) b′. rdilyki-paka- HIT & BREAK(x,y)
c. yi- ~ yu- GIVE c′. pina-yi~yu- SHOW, TEACH
(x,y,z) (x,y,z)

In (a′–c′) the thematic LVs contribute to the expression of argument-

taking predicates along with a non-inflecting preverbal constituent (PV),
forming a morphologically (and syntactically) complex verbal thematic
core.
As seen in the studies of Persian PV-LV combinations referred to in sec-
tion ., the LV marks the argument structure or valency of the complex
predicate. The examples in () show that homophonous and semantically
related HV and LVs have the same valency even though they may differ in
other elements of their meaning. As we will see in sections . and ., the
meaning contribution of an LV tends to differ more from its HV counterpart
in intransitive PV-V combinations, than in transitive ones.
Aspectual verbs combine with a thematic or another aspectual verb in
either its VINF or PRT form as determined by the aspectual verb. Aspectual
verbs do not express argument-taking predicates; they serve to modify
mainly spatio-temporal properties of the situation or event denoted by the
thematic core of the verbal constituent they are part of. Unlike the rather
selective relationship between a PV and thematic verb within the thematic
core, the combination of a thematic verb with an aspectual verb is totally
productive.


Inceptive -nji- is the only Warlpiri verb to have retained the formal distinction between
a perfective (unmarked) and imperfective (marked by a stem augmenting suffix -na)
verb form retained in all other languages of the Ngumpin-Yapa group (McConvell and
Laughren ) as a productive part of their verbal morphology. This verb is also defective
and the failure of its inflectional paradigm to match that of any other verb has led me to
place it in a sixth conjugation (Appendix , Table B).

In Lander River Warlpiri (Willowra), pina-yinyi ‘teach, show’ and yinyi ‘give’ both have
three arguments, but they express them differently: NPSUBJ show NPDO PPINST (like English
‘X furnish Y with Z’); NPSUBJ give NPDO NPIOdat.
 Warlpiri verbs of change and causation: the thematic core 175

Two of the Warlpiri aspectual verbs, ya- and parnka-, are homophonous
with thematic HV counterparts, ya- ‘go, leave’ illustrated in (a, e) and
parnka- ‘move rapidly’. When used as an aspectual verb, -ya- must com-
bine with the VINF form of the immediately preceding verb and it expresses
a notion that I label as ‘path’. It may infer associated motion (VERB while
moving along), or a progressive aspect (VERBING through time) or it may
infer a linear spatial configuration of multiple entities (VERB arrayed along
a path) (see Simpson ). Another aspectual verb is the inceptive -nji- (go
and VERB) which has no thematic verb homophone. This inceptive verb
must be suffixed to the participial form of the immediately preceding verb,
be it thematic as in (b, d) or aspectual as in (e). Note that (e) contains
both the aspectual ‘path’ verb ya- and the inceptive nji-. The ‘purposeful
displacement’ aspectual verb parnka- is like aspectual ya- in that it encliti-
cises to the VINF form of the preceding verb as shown in (f).
() a. karri-nja-] = ya-ni
stand-INF] =PATH-NPST ‘stand along’
b. karri-]- nji-ni
stand(PRT)]-INCEP-NPST ‘go and stand’
c. paka-rninja-] = ya-ni
hit-INF] = PATH-NPST ‘hit along’
d. paka-rni-]- nji-ni
hit-PRT] –INCEP-NPST ‘go and hit’
e. paka-rninja-]= ya-ni]- nji-ni
hit-INF] = PATH-PRT]-INCEP-NPST ‘go and hit along’
f. ma-ninja=parnka-ja
get-INF]=PURPD-PST ‘went expressly to get
(something)’
The VINF and PRT verb forms are never free-standing; they must be incor-
porated into a larger constituent. In complex verbal constituents such as
those in (), it is the final or rightmost verb which inflects to mark finite or
non-finite properties; the other verb forms are invariant. This structure is of
course analogous (although as a mirror image) to the English complex of aux-
iliary verb(s) and lexical verb in which only the initial or leftmost verb may
inflect for tense values while the form of each following verb is determined
by the category of the preceding verb. The thematic verb is always the first
verb in a Warlpiri verbal complex.

As when person goes out hunting for game, more than one act of animal killing
(= ‘striking’) typically occurs – person goes and kills one animal then continues along and
kills another and so on.
176 Laughren

The contrast between ‘-’ and ‘=’ in () represents the contrast in the for-
mal relationship between the VINF and following verb and the PRT and fol-
lowing verb. While no element may intervene between the PRT and inceptive
nji-, as shown by (d, f), the relationship between the VINF and following
verb is much looser: they may be separated by directional enclitics (DIR) and/
or by the auxiliary (AUX) complex of tense/aspect morpheme and pronom-
inal clitics (Hale, Laughren, and Simpson , Laughren ) as shown in
(b, e).
() a. luwa-rninja=ya-ni=rra ka
pelt-INF=PATH-NPST=THITHER AUX:PRS

b. luwa-rninja=rra ya-ni ka
pelt-INF=thither PATH-NPST AUX:PRS
‘It (e.g. lightning) is striking (it) as it moves away
(from speaker).’
c. luwa-rni-nji-ni =rra ka
pelt-PRT-INCP-NPST = THITHER AUX:PRS
‘(He) is going off to shoot.’
d. *luwa-rni=rra -nji-ni ka
pelt-PRT= thither-INCP-NPST AUX:PRS

e. wari-rninja=rra ka ya-ni
coil_around-INF=THITHER AUX: PRS go:PROG-NPST
‘(It) is coiling around (it) as it moves away.’
f. *wari-rni ka -nji-ni
coil_around-PRT AUX:PRS -INCP-NPST

.. Independent verbs versus dependent verbs

While most LVs are homophonous with independent HVs, there is also a
small subset of verb forms which only operate as LVs. These are dependent
light verbs.
Aspectual verbs are always dependent verbs even though they may be
homophonous with heavy and/or light thematic verbs, e.g. ya-ni, parnka.
Thematic ya- operates as an HV in (a), while aspectual LV -ya- figures in
(b). In (c) aspectual LV -ya-ni combines with thematic HV ya-ninja in its
VINF form. In (d) -ya- operates as a thematic LV in combination with the
invariant lexical PV tarda.


Directional deictic enclitics are =rra ‘thither’, =rni ‘hither’ and =mpa ‘across, by’.
 Warlpiri verbs of change and causation: the thematic core 177

() a. Ya-ni ka=lu. ‘They go.’ (Independent thematic HV)

go-NPST AUX=PL.S
b. Karri-nja=ya-ni ka-lu. ‘They are standing along.’ (Dependent aspectual LV)
stand-INF=PATH-NPST AUX=PL.S
c. Ya-ni.nja=ya-ni ka=lu. ‘They are going along.’ (Thematic HV= Aspectual LV)
go-INF=PATH-NPSTAUX=PL.S
d. Tarda=ya-ni ka =lu. ‘They are landing/sitting.’ (Dependant thematic LV)
PV=go-NPST AUX=PL.S

Thematic verbs fall into both classes: independent and dependent.

Independent verbs used with a PV may be semantically ‘heavy’ or ‘light’.
Dependent verbs are always semantically ‘light’. The ‘productive’ dependent
LVs -jarri- and ma- in (a, b) express only the Aktionsart and PAS proper-
ties of the complex predicate. It is the preverbal nominal predicate wiri ‘big’
which expresses the state of affairs; the dependent LVs signal that this state
of affairs is to be interpreted as the result of CHANGE, and the contrasting
forms mark whether a causing event is entailed (i.e. with -ma-) or not (i.e.
with -jarri). Unlike the English verbs get and make as used in the phrasal
constructions translating (a, b), neither jarri- nor ma- has a correspond-
ing independent HV form. These ‘syntactic’ verbs – or verbalisers – map
directly onto syntactic ‘heads’ and do not project an intrinsic lexical event
structure.
() a. Wiri-jarri-ja ‘(It) got bigger.’ (PV-Dependent verb)
big-INCH-PST
b. Wiri-ma-nu. ‘(It/he/she) made (it) bigger.’ (PV-Dependent verb)
big-CAUSE-PST
c. *jarri-ja
d. # ma-nu.

.. Verbal constituent

... Thematic structure
The minimal or simplex verbal constituent consists of one inflected thematic
HV such as ya-ni in (a) or the verbs in (a–c). A thematic LV may be


Warlpiri independent ma- verb ‘get, obtain, pick up’ may be the historical source of the
independent verb. Both are conjugation  verbs and mark transitive predicates. For con-
venience’ sake I label these verbs as inchoative and causative, but these ‘meanings’ are
not a direct property of these verbs. The meaning derives from the combination of the
Aktionsart features of the PV with the requirement that PV-V thematic core must be
dynamic, and the PAS, signalled by the LV form.

(d) is grammatical interpreted as HV meaning ‘get/pick up-PST’ (He/she/it got (it)).
178 Laughren

augmented by a non-inflecting preverbal element as in (a'–c'). The PV in

(a'–c') belongs, along with the thematic verb it combines with, to the the-
matic core of the complex verbal constituent. The thematic core is the obliga-
tory ‘inner’ part of the thematic structure component of a verbal complex;
this thematic structure component may be augmented by outer modifiying
elements in what I call the thematic periphery.
The class of PV exemplified in (a') and in () was classified by Nash
() as ‘lexical’ because it tends to combine with a thematic LV in an oper-
ation of limited productivity and semantic transparency. Nash () also
distinguishes a limited class of what he calls ‘semi-productive’ PVs includ-
ing the non-productive negative kanginy(pa) ‘fail to perceive’, which forms
a semantically opaque complex verb with the thematic LV pinyi in (a), or
modifies a PVlexical=V complex in (b).
() a. kanginy-pinyi ‘fail to recognise’
b. kanginy-[purda-nyanyi] ‘fail to hear’
c. purda-nyanyi ‘hear, listen to’
d. nyanyi ‘see, look at, find, look for’
e. parlu-pinyi ‘see, look at, find’
f. milya-pinyi ‘recognise’
g. pinyi ≠ ‘perceive’ (= ‘strike, hit, bite,
sting, attack’)
Used as an independent HV as in (g), pinyi cannot denote an act of per-
ception; its typical interpretation is as a dyadic ‘impact-by-contact’ verb as
indicated by the list of English glosses in parentheses in (g), although it
also has a semantically ‘lighter’ do/make interpretation in certain contexts
not discussed herein. It is only the presence of one of a small set of lexical
PVs such as kanginy-, milya, and parlu with which pinyi forms a complex
verbal constituent in (a, e, f) which results in the formation of a complex
verb denoting an act of perception. The contribution of pinyi to (a, e, f) is
to provide a legitimate V host for the expression of tense/mood inflection,
while it also signals that the PV-pinyi complex verb expresses a transitive


I retain Nash’s use of the term ‘lexical’ PV without wanting to claim that this class of
PVs necessarily combines with a LV in the lexicon as opposed to the syntactic compo-
nent. Nash () also used this term to distinguish basic PVs from derived forms. As will
be shown in section ., both ‘basic’ and derived PVs show the same syntactic behaviour
with respect to the LV they combine with. As Nash’s classification of PVs is well known in
Warlpiri linguistics, I retain its use.

I consider kanginy to also belong to the ‘lexical’ PV class.

As an independent verb as in (g), pinyi is semantically quite ‘light’ compared with other
‘impact-by-contact’ verbs which encode specific information about the nature of the
impact. This will be pursued in section ..
 Warlpiri verbs of change and causation: the thematic core 179

two-place predicate. It is kanginy- which forces the ‘fail to perceive’ mean-

ing in (a). In (b) it is the PV purda- which serves to specify the type of
perception act, while the V nyanyi expresses a general notion of ‘percep-
tion’ in combination with the PV, although as an independent HV in (d)
nyanyi has the narrower interpretation ‘perceive with sight/eye’. The PV’s
parlu and milya – like kanginy in combination with pinyi – express the ‘per-
ception’ element of meaning, while pinyi indicates the transitive properties
of the complex predicate.
The PV-LV combinations which form the thematic core of a verbal
complex vary along a productivity scale from being ‘fixed’ expressions
to wholely productive. Similarly they also range along a semantic trans-
parency scale from opaque to transparent where the meaning of the com-
plex constituent derives in a straightforward way from the sum of the
meaning associated with each part. We can distinguish two types of PV
which combine with a LV to form the thematic core: underived or basic
PVs which belong to Nash’s ‘lexical’ and ‘semi-productive’ PV classes,
and derived PVs. Examples of both types will be examined in sections
.–..

... Thematic periphery

The thematic core of a verbal constituent may be further augmented and
modified by a closed set of ‘outer’ non-inflecting preverbal or coverbal words
which Nash () places in one of three distinct classes: quantificational,
adverbial, and dative adjunct. These sit outside of the inner thematic core as
shown in ().
() PVquantificational = PVadverbial = PVdative adjunct = [PVsemi-productive = [PVlexical = [V-]]]
PERIPHERY OF THEMATIC STRUCTURE [THEMATIC CORE]

As noted by Nash (), the association of outer PVs is totally productive,

whereas the relationship between the inner ‘lexical’ and ‘semi-productive’
PVs and the inflecting verb is restricted semantically, morphologically,
syntactically, and phonologically. PVs in the thematic core are distin-
guished syntactically from the three types of PV found in the thematic
periphery:

(i) PV in thematic core may not be postposed to the thematic V; PV in the-

matic periphery may be pre- or postposed to the thematic V;

In () all verbs are in their non-past (NPST) form which is also their citation form.

Hale (), Nash (), Simpson (), and Simpson and Bresnan () clearly show
that different syntactic relationships are set up between non-verbal constituents associated
with the lower VP than with adjunct dative/applicative constituents associated with the
upper VP, especially in terms of pronominal agreement and control structures.
180 Laughren

(ii) order of PVs within the thematic core is fixed; order of PV in the the-
matic periphery is variable.
The complex verbal constituents in () illustrate the varying order of
peripheral PVs relative to each other and to the (strictly ordered) thematic
core contained inside square brackets.
() a. muku= [ yiily-nga-] rnu jurnta PVquantificational=[PVlexical-V-] INFL PVdative adjunct

‘used up all on (someone)’

b. muku=jurnta= [walku-jarri-] ja PVquantificational=PVdative adjunct=[PVlexical-V-] INFL

‘all ran out on’

c. [ yiily-nga- ] rnu muku [PVlexical-V-] INFL PVquantificational
‘used up all’
d. yarda=pina= [ ya- ] nu PVquantificational=PVadverbial=[V-] INFL

‘go back again’

e. jurnta= [tarda-ya- ] nu pina PVdative adjunct =[V-] INFL PVadverbial
‘sat/landed back away (from)’

In a finite clause, the thematic core, represented by the lower verbal

constituent (VC) in Figure , along with the outer PVs in the higher VC
in Figure ., lie within the scope of verbal inflectional categories – tense
and mood – and in turn in the scope of a directional clitic which indicates
the spatial orientation of the event or situation expressed by the rest of the
complex verbal constituent relative to the speaker (Nash , Laughren
). Figure . shows a structure in which the deictic Directional Phrase
(DIRP) dominates a three-part structure: inflectional (either with finite or
non-finite values), aspectual, and thematic. The thematic core of the ver-
bal constituent expresses the complex of predicate argument relations which
are represented in the hierarchical event structures proposed in section ..
In Figure ., I use the symbol VC (verbal constituent) rather than VP as a


Many inner PVs have two allomorphs, a consonant final root form (e.g. kanginy) which
cannot constitute an independent phonological word, and an augmented root-pa form (e.g.
kanginypa) which can enter a looser phrasal relationship with the following verb; some
adverbial PVs have similar dual forms. The word-like adverbial PV can be postposed to the
verb unlike inner word-like PVs (see further discussion in section .).

Directional clitics may only be hosted by an element of the verbal constituent (dominated
by upper VC in Figure .), but there is some variation in its placement as illustrated by
(i)–(iv), in which the adverbial PV or the V may host rra ‘thither’ irrespective of the rela-
tive order of PV and V (see Laughren () for more details).
i. Pina ya-nu=rra. back go-PST=thither ii. Yanu=rra pina.
iii. Pina=rra ya-nu. iv. Ya-nu pina=rra.
‘Went back (away from speaker’s location).’

No Voice Phrase is included in Figure . because voice alternations (passive, antipassive,
middle, etc.) are not found in finite clauses. I don’t exclude, however, the possibility that
Voice may play a part in Warlpiri verbal and clausal grammar.
 Warlpiri verbs of change and causation: the thematic core 181

Spatial deixis structure

DI R P

MoodP
Inflectional structure
TenseP

IncepP (-nji)

PathP (-ya-) Aspectual structure

PurpDisP (-parnka-)

VC2 ((PVs), VC1-) Thematic & Aktionsart

structure
Thematic periphery

VC1 ((PV), V-)

Thematic core

Figure . Morphosyntactic structure of the Warlpiri inflected verbal

constituent (finite).

theory-neutral term for a constituent to be interpreted as a place-holder for

the phrasal categories which dominate the ‘core’ and ‘periphery’ components
of thematic structure realised within the verbal complex. I am assuming that
the surface ordering of inflectional morphemes, which presents the mir-
ror image of the scopal relationships shown in Figure ., derives through
successive checking of features in each immediately c-commanding inflec-
tional head. Phrasal constituents move into specifier positions as features are
checked.
The morphologically complex verbal constituent in () illustrates the
structure in Figure . in which the thematic material is embedded in the
aspectual (Purposive Displacement, Path, Inceptive) and inflectional (Tense,
Mood) structures.
() [[[[[Pina- [ pata- [karri-nja]]] -ya-ni] -nji] -ni] -rra]
back-down-stand-INF-PATH-PRT-INCP-NPST-THITHER
[DIR [ INFL[[ASPECT [THEMATIC STRUCTUREPVADV [THEMATIC COREPVLEX-[LV-INF]]]
LV-PRT]-LV] T/M] DIR]
‘falling as (it) moves back (where it came from)’ (e.g. of rain)
Alternative analyses of the surface structure of the verbal complex in
(), which allow for the variant positions of the PV constituents, will be
182 Laughren

entertained when I return to a discussion of the morphosyntactic properties

of the complex verbal constituent in section ..
The main focus of this study is on the event structure of verbs of CHANGE,
both causative and inchoative, expressed by the thematic core elements rep-
resented by the lower VC in Figure .. Phonological reflexes of the argu-
ments in the corresponding event structure are not present in the surface
verbal constituent, which is not to deny the presence of argument posi-
tions within the syntactic structure represented by the VC constituents in
Figure .. In finite clauses, arguments are realised as pronominal enclit-
ics in the higher AUX constituent headed by functional categories marking
clausal aspect and tense, as opposed to the verbal aspectual and inflectional
categories represented in Figure .. The values associated with the AUX
functional categories must be compatible with the values associated with the
corresponding lower verbal inflectional categories, and with the predicate
argument structure defined in the thematic structure of the verbal complex.
Modal features are expressed in COMP. COMP and AUX morphemes form a
word-like constituent. Additional properties of arguments may be expressed
by syntactically optional nominal phrases linked by case to a particular
argument position. These case phrases may be in any position relative to
the inflected verbal constituent represented by Figure ., hence the much
discussed free word order of Warlpiri (e.g. Hale , Jelinek , Laughren
, Legate , Nash , Simpson ).

. Meaning–form mapping within the inner thematic core

This section examines a series of thematic core structures corresponding to
either V, VINF-V, or PVlex-V. My aim is to determine the respective roles of
each constituent, starting with complex verbs in which an infinitival verb
form (VINF) functions as a PV in combination with a following V. In section
.., I compare stative uses of STANCE verbs with their CHANGE STANCE and
CAUSE CHANGE STANCE counterparts. In section .., I compare these with
CHANGE OF LOCATION V and PV-V constituents. I return in section .. to
STANCE predicates and compare PV-V forms with the VINF-V and V forms
seen in section .. and with the PV-V CHANGE LOCATION verbs.

.. VINF -V verbal complex

The first class of predicates I examine feature ‘stance’ verbs. Consider the
data in ().


INF indicates the infinitival verbal inflection, whereas VINF indicates a verb form con-
sisting of a stem and the infinitival suffix -nja.

These ‘stance’ verbs correspond to a subset of Levin’s English ‘verbs of spatial configur-
ation’ (Levin ).
 Warlpiri verbs of change and causation: the thematic core 183

() a. [karri ]-mi (stand-NPST) ‘stand, stop’ (x)

b. [karri-nja- [ pardi]]-mi (stand-INF-rise-NPST) ‘stand up, arise’ (x)
c. [karri-nja-[ yirra]] -rni (stand-INF-put-NPST) ‘put in standing
position’ (x,y)
d. [karri-nja]-parnka (stand-INF-run-NPST) ‘go purposefully
-mi to stand’ (x)
e. [karri-nja]-ya-ni (stand-INF-go-NPST) ‘stand along’ (x) 

In (a–e) the thematic core of the verbal constituent falls within the outer-
most square brackets. The HV karri in (a) is directly inflected for tense and
heads a verbal projection, whereas in its VINF form in (b–c) it constitutes
a derived PV which combines with the underscored thematic V. These the-
matic Vs mark the contrast in predicate argument structure (PAS) associated
with each of these verbal constituents. In (d, e) the HV karri- in its VINF
form constitutes the thematic core. The inflected aspectual verbs, parnka-
and ya-, are outside the thematic core as indicated by their non-enclosure
in square brackets – they freely associate with the thematic core and make
no contribution to its PAS. These aspectual verbs express event modifying
aspectual values: ‘purposeful displacement’ in (d) and ‘path’ in (e).

... Thematic versus aspectual verbs

One may ask whether the dichotomy between thematic and aspectual
domains is as clear-cut as presented thus far. For example, (a) is aspectu-
ally ambiguous. While Warlpiri verbs do not express individual-level stative
predicates (Hale , Simpson ), karri- has several possible interpret-
ations, as set out in (), in which ‘>’ is to be read as ‘changes to’. Only one
of these readings corresponds to a stage-level predicate.
() karri-
(a) stage-level predicate [be in standing position (x)] ‘stand’
(b) negative achievement predicate with polar opposite pre-
supposition (underlined): [move (x) > (NOT move (x)) & (be
standing(x))] (=cease/not continue to move & come to be in
standing position) ‘halt, come to a standstill, stop’


The sequence of INF-PATH /-nja-ya-/ is pronounced with long [a:] due to elision of /y/,
e.g. /k’arrinj’aani/.

From () onwards no distinction will be made between different types of constituent-
internal morpheme boundaries, marking all with a hyphen, unless otherwise indicated.

The meanings of karri- set out in () are not exhaustive. The ‘copula’ use of stance verbs
is discussed in section ..
184 Laughren

(c) negative stage-level predicate: NOT [(NOT move (x)) &

(stand (x)) > move (x) & NOT stand (x)] (=continue to/not
cease to stand; not change from stationary standing position)
‘remain standing, stay, wait’
For clarity of exposition, it is useful to distinguish syntactic aspect from
lexical aspect or Aktionsart (Tenny and Pustejovsky b). I will use
(verbal) aspect for the semantic properties marked by the aspectual verbs
briefly discussed in section ., and Aktionsart for the aspectual proper-
ties integral to the verbal thematic core. This distinction is useful even
though Aktionsart features interact with syntactic aspectual categories
in ways that have been extensively documented in other languages. For
example, the addition of a ‘path’ aspectual verb to a semelfactive as opposed
to an achievement predicate produces different interpretations, i.e. iterative
sub-events as opposed to durative event, while the addition of an inceptive
verb forces an homogenous predicate to be viewed as a succession of stages,
focusing on the initiation stage.
Unlike other ‘stance’ HVs such as nguna- ‘lie’, nyina- ‘sit’, parntarri- ‘crouch
over’, which express an additional positive achievement reading, ‘x come
to be in a lying/sitting/crouching position’ [NOT lie/sit/crouch (x) > lie/sit/
crouch (x)], karri- does not permit this reading even where embedded under an

INCEP verb or an imperative mood. To express a positive achievement reading
‘move into a standing position’ karri- combines with a ‘path’ or ‘change loca-
tion’ verb pardi- ‘rise’, with the resultant stance expressed by the de-infinitival
PV karrinja as in (b). The verb pardi- in (b) is not a pure ‘change’ verb as
it also expresses semantic information about the nature of the PATH (upwards)
travelled by (part of) x (relative to other parts of x) in order to achieve a standing
position. We can compare pardi- ‘move upwards/out of’ in (b) with wanti-
‘move downwards/into’ in (b) to see the expression of this semantic contrast
of the PATH orientation, an observation we will return to in section ...
While nguna- ‘lie’ (like nyina- ‘sit’) may be ambiguous between stage-level
predicate (constituting with its argument a situation in my terminology) and
achievement (event involving a change of stance/position), this ambiguity can
be resolved by a complex construction parallel to that seen in (b) (stance
INF-path V). Interestingly, the compound pattern for the change of stance
reading seen in (b) and (b) is rarely employed with nyina- ‘sit’, although
I have also recorded nyina-nja-wanti ‘sit down’.

With respect to Warlpiri, verbal aspect is distinct from clausal aspect expressed by auxil-
iary morphemes in finite clauses.

Warlpiri stance verbs nyina- ‘sit’ and nguna ‘lie’ differ from neighbouring Western Desert
language simplex stance verbs which do not allow a positive achievement meaning; they
must form compounds with a ‘change of location’ verb, e.g. Ngaanyajarra nyina ‘be seated’,
nyina-kati ‘come to be seated’; ngarri ‘be lying’, ngarri-kati ‘lie down’ (Glass and Hackett
). Note that ‘move into standing position’ paka- is distinct from ‘be standing, stop,
wait’ ngara. (Levin and Rappaport () comment on the variation in lexical forms for
this class of verbs cross-linguistically.)
 Warlpiri verbs of change and causation: the thematic core 185

() a. nguna ‘be in lying position’/‘come to be in lying

position’/‘continue to lie’
b. nguna-nja- [wanti] -mi ‘move into lying position, typically by
downward movement’
c. nguna-nja- [ yirra] -rni ‘cause to come to be in lying position’
The interpretation of the VINF stance verb preposed to the transitive ‘put’
verb yirra-rni in (c) and (c) embeds the achievement ‘change of stance’
within the structure projected by the CAUSE predicate expressed by yirra-rni
(see Carter () for a relevant discussion of the relationship between CAUSE
and CHANGE). The expression of the semantic relationship between the pre-
verbal VINF and the finite verb in (b–c) and (b–c) is fairly transparent.
These combinations of VINF-V are also productive; the infinitive stance verb
expresses the resultative complement of the CHANGE predicate expressed by
verbs pardi- in (b) and wanti- in (b), or of the CAUSE CHANGE predicate
complex expressed by yirra-rni ‘put’ in (c) and (c). In (b) and (b)
the change of stance is internally controlled, whereas in (c) and (c) it is
externally controlled.

... Contrasting properties of HV and LV homophones

The HV wanti- ‘fall’ is not a hypernym of the complex ngunanja-wanti
in (b). As an independent verb, wanti- ‘fall’ (move downwards under
force of gravity) contrasts with jiti- ‘descend, climb down/out of’ (move
downwards under one’s control). Although ngunanja-wanti in (b) typi-
cally denotes an event in which some entity moves into a lying position
under its own control while independent wanti- does not imply controlled
motion, but rather presupposes some unnamed external cause, ngunanja-
jiti- (containing the controlled downward motion verb jiti-) is not used as
a synonym for (b). The verb pardi- ‘rise’ exhibits the same behaviour;
as a non-control directed motion HV it contrasts with the controlled
directed motion HV warrka-rni ‘climb up, get (up) into’. In (b) (and in
many other INF-pardi complex verbs), pardi functions within a controlled
motion PV-LV (where the de-infinitival PV expresses the end stance).


The stance verb parntarri- ‘crouch over’ may also combine with –wanti to express the
CHANGE STANCE meaning, although, like nguna- ‘lie’, parntarri- may express both ‘be in
STANCE’ and ‘CHANGE STANCE’ meanings.

Independent yirra-rni means ‘x cause y to come to be at some location’ and also ‘x keep y at
a location’; similarly ngunanja-yirrarni can also mean ‘cause to continue to lie at’.

In preserving the CHANGE component in the scope of CAUSE I agree with Levin and
Rappaport () and Megerdoomian () and disagree with Folli, Harley, and Karimi
(), who argue for the non-existence of a CHANGE event in the scope of a CAUSE predicate.
Finnish (Pylkkänen ) has overt ‘cause’ and ‘change’ morphemes in verbal complexes.

PV-wanti complex verbs such as yakarl-wanti ‘duck down’ also express internally caused
change of stance/location events.

This notion of control is extensively discussed by Hoekstra (), Levin and Rappaport
Hovav (), and Smith (), inter alia.
186 Laughren

As previously documented for Warlpiri, and for other languages, the

relationship between form and meaning for an independent HV and its
homophonous dependent LV counterpart may differ, as in these cases.
One of our tasks is to discern in which cases the control element is present
and in which it is not, and how it should be represented. While signalling
a change of stance involving directed motion (‘up’ vs. ‘down’) in VINF-V
complex verbs, pardi and wanti also signal that the change is not externally
caused – there is no event external to the change event which acts as the
cause of change – a feature of ‘achievements’ as opposed to ‘accomplish-
ments’ in Vendler’s () terms. By contrast, the presence of the V yirra-
rni in the VINF-V complex in (c) and (c) signals an external causing
event distinct from the change event, but information about the nature
of the cause event is not encoded. These VINF-yirrarni forms express
‘accomplishments’.
The VINF stance predicates in (b, c) behave like lexical PVs and not like
outer periphery adverbial PVs such as warru in (c, d) – syntactically and
semantically. The VINF cannot be postposed to the inflected verb as in (a,
b), whereas an adverbial PV may be preposed (c) or postposed (d) to the
inflected verb.
() a. *pardi-mi nguna-nja cf. (b)
b. *wanti-mi nguna-nja cf. (b)
c. warru nguna-mi ‘lie around’
d. nguna-mi warru ‘lie around’
While (c, d) refer to a situation in which an entity is in a lying stance
while also being in the spatial configuration expressed by the PV warru
(i.e. be around AND be in lying configuration) , (b) [karri-nja- [ pardi ]] -mi
(stand-INF-rise-NPSJ) ‘stand up’, cannot refer to a standing type of rising
(e.g. ≠ move upwards while in a standing position), nor does (b) refer to
simultaneous lying and falling (e.g. ≠ move downwards while in a lying/
horizontal position). As shown by the outer adverbial PV warru in (c,
d), outer PVs modify the situation or event expressed by the inner the-
matic core (represented by the independent HV nguna-mi ‘lie’ in (c, d))
whereas, inner thematic core PVs express the end state of some process
of change. The event structure of CHANGE OF STANCE predicates seems to
parallel that of CHANGE OF LOCATION predicates. In section .. we will


Given the ability of outer preverbs to occupy pre- or post-verbal positions, coverb might
seem a better term for them. However I will continue to use the term ‘preverb’ for all of
these non-inflecting elements which lie within the orbit of the thematic verb, since it is
well established in the Warlpirist literature.

These notions can be expressed in biclausal structures, e.g. [nguna-nja-karra] ka wanti
[lit. [lie-INF-SUBJCOMP] AUX:PRS fall:NPST] ‘he falls while lying’ (Hale , Laughren ,
Simpson and Bresnan ).
 Warlpiri verbs of change and causation: the thematic core 187

examine some PV-V combinations which express CHANGE OF LOCATION

predicates.

.. Semantically transparent PV lexical-LV complex verbs

Consider the contrasting pair of complex verbal constituents in () which
have the structure PVlexicalLVthematic and again in () where the thematic core
elements are in the scope of the aspectual LV -ya-.
() a. pata-[karri ] -mi ‘fall, drop’ (x) (synonymous with [wanti ] -mi)
down-stand-NPST
b. pata-[kiji ] -rni ‘throw down; (make) drop’ (x,y)
down-throw-NPST

() a. pata-[karri-nja]-ya-ni ‘fall (while going ) along’ (x) / ‘be falling (x)’
b. pata-[kiji-rninja]-ya-ni ‘throw down/ drop (while going) along’ (x,y)

The relationship between the lexical PV pata in () and () and the follow-
ing thematic LV karri/kiji- is less semantically transparent and productive
than the VINF-LV combinations in () and (), although the morphosyn-
tactic construction is basically of the same type. Clearly (a) is not a type of
‘standing’ stance; the role of the LV karri is mainly syntactic. It marks the
monadic diathesis of the complex verb, and formally contrasts with dyadic
kiji-rni in (b), which marks a causative predicate. Recall that the HV karri,
as the sole member of the thematic core, as in (a), cannot express a posi-
tive CHANGE STANCE (‘move into standing position’) meaning, so it seems
surprising that in the complex constituent in (a) and (a) the LV karri
expresses this CHANGE OF LOCATION meaning. In fact (a) cannot have a sta-
tive reading such as ‘be down, as a result of falling’. It is the PV pata which
indicates the change of location along a path whose endpoint is lower than
the starting point. We conclude then that the CHANGE OF LOCATION meaning
with its dynamic Aktionsart feature derives from the PV (a) and (a). The
PV pata has not been recorded as an independent predicate of the nominal
category. The role of the LVs karri and kiji- is to indicate the mapping
of event structure components of the CHANGE LOCATION and CAUSE CHANGE
OF LOCATION event structures onto syntactic structure projected by verbal
categories.


Even if LV karri in (a) and (a) has a more general meaning like existential or locational
‘be’ associated with its cognate in other Ngumpin-Yapa languages as observed by Nash
(), this does not explain its use as a ‘change’ verb in these CVs.

Pata may have been borrowed from a Western Desert language in which it is a verb root.
Pintupi pata-ngara- ( pata-stand) ‘fall’ (Hansen and Hansen : ) but zero-derived
transitive pata-lku ( pata-FUT) ‘shake out’. The transitive pata-lku in Ngaanyatjarra is
glossed ‘drop’ and ‘spread out on (ground/bed)’ (Glass and Hackett : ).
188 Laughren

Pata-karri in (a) is synonymous with the HV wanti ‘fall’. Ideally,

whatever semantic representation we have for the complex verbal constitu-
ent pata-karri we want to apply to the HV wanti for the relevant range of
senses, even though the surface morphosyntactic properties of each dif-
fers. Unlike pardi and wanti, which express a directed path, ‘upwards’
versus ‘downwards’, whether used as HV or LV, the LV karri can be used
in a PV-V structure which expresses either upward (e.g. rduyu-karri ‘rise
(of dust, smoke, fog)’ or downward movement (as in the case of pata-
karri). These directional components of meaning derive from the PV
predicate, not from the verbal predicate in these cases. Unlike pardi and
wanti (as opposed to yirrarni ) in examples in () and () which express
direction of movement and indicate a monadic PAS, karri in (a) and
(a) (as opposed to kiji- in (b) and (b)), only signals the monadic
PAS expressing the CHANGE LOCATION component of event structure
associated with pata.
The complex pata-kiji- ‘drop, throw down’ in (b) can be viewed as a
semantically restricted version of independent V kiji- ‘throw’, which denotes
an event type which involves some entity in an action which results in a change
of location of some other entity to a location distinct from that of the causing
entity (compare with take/bring), with pata specifying that the end point of
the trajectory is lower than the starting point. The argument-taking proper-
ties of the complex pata-kiji- are marked by the LV kiji-, while the PV pata
designates specific properties of the PATH travelled by the moving entity.
Unlike an adverbial PV such as warru in (c, d), but like the VINF PV
seen in section .., pata cannot be postposed to the verb and hence belongs
within the verb’s thematic core. Note also that its semantic interpretation
is limited to specifying the relative location of some entity resulting from a
change of location. Pata cannot be interpreted in combination with karri or
kiji- to mean something like x stand/be down or x throw y while y is down.
Thus it seems that thematic aspectual – or Aktionsart – features must be
part of the semantic, and also syntactic, representation of such PV-LV com-
plexes (see Megerdoomian () and Travis (a, b, ) for pro-
posals along these lines).
Examples (a, b) parallel those in (d, e) in which the thematic core is
placed within the scope of an overt aspectual verb (‘path’ ya-ni in (e) and
(a, b); ‘purposeful displacement’ parnka in (d)). These aspectual verbs


While the Warlpiri pata-kijirni combination may seem analogous to the English V + par-
ticle combination throw down, the former lacks the full semantic transparency of the latter.
While throw down contrasts in meaning with put down, place down, etc., in Warlpiri, the
manner contrast between the simplex verbs yirrarni ‘put’ and kijirni ‘throw’ is neutralised,
with kijirni being the lexically specified V to combine with pata. There are many transitive
PV-V combinations in which yirrarni does contrast with kijirni in the expression of man-
ner, but this is not one of them.
 Warlpiri verbs of change and causation: the thematic core 189

play no part in expressing the predicate argument properties of the complex,

although they interact with the Aktionsart features of the thematic core and
hence indirectly affect predicate entailments.
The verb contrast in (a, b) is semi-productive; pairs where PV-karri
entails a ‘change of location’ (presupposing but not specifying externally
caused movement) while –kiji- or yirra- express ‘externally caused move-
ment along path’ are not rare (see Appendix  for further examples). In
(a) karri shares the CHANGE LOCATION meaning with the LV pardi ‘rise’
in the complex verbal constituent in (b) and wanti ‘fall’ in (b), but karri
does not specify the orientation of the relevant PATH unlike pardi and wanti
which do (both in complex PV-V and simplex V usage). Recall that while
the VINFstance-pardi/wanti constituents express self-controlled change of
stance, pata-karri expresses uncontrolled (or externally caused) change of
location.
Even relatively ‘transparent’ PV-V combinations such as those in ()
and () may appear semantically idiosyncratic, but only in the sense that
LV karri in (a) and (a) does not express the same range of meanings
as HV karri in (a–e). The limited productivity of these PV-V combin-
ations contrasts with the free association of outer adverbial PVs with the
thematic core (whether constituted by a HV or a PV-LV) and their pre-
dictable compositional meanings. However, contrasting pairs of PV-karri
(change location) and PV-kiji-/yirra- (cause change of location) do form a
semantically coherent subset of ‘semi-productive’ change location verbs in
which the verb signals which components of event structure are projected
onto syntactic structure.
Although pata-karri and the HV wanti- are synonymous, the combin-
ation *ngunanja-[pata-karri] is not allowable as a synonym of ngunanja-
wanti (b) ‘lie down’. As discussed already, HV wanti- and LV wanti in
complex verbs with preverbal stance VINF are not semantically equiva-
lent: pata-karri is a synonym of the former, but not of the latter, which
forms a predicate with an internal controlling cause. There is also a struc-
tural reason for the unacceptability of *ngunanja-[pata-karri]. As a derived
PV, VINF is only found in the immediate preverbal position within the
thematic core, even though VINF may be separated from either a thematic
or aspectual V by AUX and directional enclitic as seen in section . and
section ..


Other spatial adverbial PVs include pina ‘back, again’; jaala ‘to and fro’; wapirdi/nganjini
‘on arrival’ (see Nash ). Adverbial PVs must be distinguished from the ‘adverbial’ use
of nominal phrases which are case-marked (Ergative or Absolutive depending on the scope
of the adverb and the diathesis of the verb in its clause) and which may occupy any position
within a clause. Unlike adverbial PVs, adverbial nominals are not part of the verbal con-
stituent represented in Figure .. They constitute case-marked NPs which may occupy any
position in a clause relative to the verbal complex; their scope is indicated by their case.
190 Laughren

.. PV-V ‘stance’ predicates

Unlike the VINF-V change of stance verbs discussed in .., the complex
verbs in (a, b) are made up of synchronically unproductive PVlexical-V
combinations. (a, b) express some of the same meanings as that expressed
by the HV in (c), as discussed briefly in section ....
() a. [tarda-ya] -ni ‘sit’ (x)  a'. [[tarda] -yirra] -rni ‘put in sitting position’
b. [ pirri-ma] -ni ‘sit’ (x) b'. [[ pirrily]-yirra] -rni ‘put in sitting position’
c. [nyina] -mi ‘sit’ (x) c'. [[nyina-nja] -yirra] -rni ‘put in sitting position’

The data in () suggest that it may be somewhat problematic to specify

which element of the complex verb, PV or LV, determines the valency of the
verbal complex. In (c) the HV nyina- expresses the monadic predicates
‘be seated; come to be seated/perched/landed; be/stay in process of sit-
ting’, hence it seems reasonable to conclude that the verb not only expresses
the meaning of this monadic predicate since the addition of a CAUSE event
predicate which presupposes a dyadic structure must be overtly marked by
the addition of the ‘put’ LV yirra- to the PV derived from the stance VINF
nyina-nja as in (c'), a pattern already seen with the stance verbs karri
‘stand’ and nguna ‘lie’ in () and (). The PV-LV forms in (a, b) appear
to express the same meaning as the HV nyina in (c) and have the same
valency; however they contain different LVs.
As a HV, ya-ni ‘go/come’ is a monadic thematic CHANGE OF LOCATION
verb, but in (a) dependent thematic LV -ya-ni does not mean ‘go’; in fact
while the CHANGE OF STANCE reading ‘become seated’ is possible, so is the
‘negative achievement’ non-change reading or internally controlled cause
reading ‘maintain self in seated stance’. In (a) both the thematic and
Aktionsart properties appear to derive from the combination of the PV and
the LV; the LV -ya-ni marks this predicate as monadic ‘sit’, since it con-
trasts with dyadic yirra-rni in (a′) expressing the external ‘cause to sit’
verb. In (b), we find an LV ma-ni which is homophonous with the dyadic
transitive LV seen in (b) (to be examined in more detail in section ...).
However in the synchronically semi-productive combination with the PV
pirri in (b), -ma-ni fails to mark a dyadic argument structure; pirri-ma-ni


Tarda-yani is used mainly in the Willowra Lander River dialect as the ‘normal’ verb for
‘be, sit, perch, land, wait, stay’. Interestingly, tarda-yani and pirri-mani do not have the
full range of meanings associated by simplex nyina (see ...).

Pirrily-yirrarni is not a commonly used verb. It has been recorded from a single speaker in
the Warlpiri corpus, so for most speakers the causative counterpart of pirri-mani is (c′).

The Warlpiri dictionary (Laughren et al. ) identifies at least six mani LVs distinguished
by their meaning and diathesis; -mani LV of sound emission is always monadic/intransi-
tive while most other mani LVs form dyadic/transitive predicates. However, in addition to
some monadic change stance verbs, LV-mani is also found in semelfactive complex verbs of
‘light emission’, such as maarr-mani ‘flash’, mirirl(ki)-mani ‘glisten, glint’.
 Warlpiri verbs of change and causation: the thematic core 191

is monadic. For most Warlpiri speakers there is no corresponding ‘cause to

sit’ verb of the form pirri-LV, however, the verb pirrily-yirrarni ‘seat some-
one’ does occur in the Warlpiri corpus. PVs pirri and pirrily are likely to be
etymologically related, although the addition of ly to a root does not reflect a
productive morphological pattern in Warlpiri. However, there is diachronic
evidence for CVCV vs. CVCVly- PV variants in Warlpiri.
Given the set of synonyms in (a–c), one might expect that ‘causative’
verbs could be formed on the nyinanja-yirra- pattern, where the stance verb
in its VINF form acts as a PV which combines with the dyadic causative verb
yirra-. However such forms are not found – perhaps blocked by the pres-
ence of PV-yirra- forms, e.g. *[tarda-yaninja]-yirra-, *[pirri-maninja]-
yirra-. This may seem like a collocational restriction on this particular verb,
which would pertain to a level of vocabulary insertion or spellout, and not
to the semantic–syntactic interface. However, as I will argue in section .,
the LV yirra- introduces an external cause to a change predicate without
an internal cause. The monadic complex verbs in (a, b) always project an
internal cause argument, while the HV nyina, may or may not. The spatial
configuration of an inanimate entity can be expressed by nyina but not by
pirri-mani or tarda-yani without presupposing a change of stance. However
the dyadic verbs in (a′–c′) only project an external cause and suppress any
internal cause. Yirra- is associated with the CHANGE function as well as the
higher CAUSE function, hence it is not possible to have another ‘change’ verb
associated with the same event structure position.

.. Argument structure and verb selection

Warlpiri independent verbs are either transitive (in the sense that they licence
an ergative case-marked subject NP) or intransitive (having an unmarked
absolutive case NP subject), except for a pair of synonymous ‘burn/cook’
verbs janka and kampa and some special kinship register verb forms dis-
cussed by Laughren (). In the contrasting pairs of inchoative and causa-
tive PV-V pairs seen so far, the same strict dichotomy holds. However, the
picture is not always so straightforward. Let’s now consider the data in ().
() a. [nyunjurl-ya-ni] ‘spit curse at/on’ (x,y)
spit-go-NPST
b. [nyunjurl-ya-ninja]-ya-ni ‘spit curse at/on along’ (x,y)
spit-go-INF-PATH-NPST
c. nyinypa kiji-rni / kiji-rni nyinypa ‘spit’ [‘lit. spit propel/throw’]
spit:ABS throw-NPST / throw-NPST spit:ABS


See Appendix , Table C for other examples of monadic stance PV-mani verbs.
192 Laughren

While ya- as both HV and thematic LV typically marks a monadic predicate

as discussed in relation to (a), in the complex verbal constituent in (a,
b) it fails to do so. The verb nyunjurl-ya- ‘put curse on’ is dyadic and tran-
sitive. (b) illustrates the combination of two ya- LVs where the first in its
VINF form yaninja is thematic, and the second (inflected) V ya-ni is aspect-
ual. The PV nyunjurl- is etymologically related to the noun nyuny(pa)/
nyiny(pa) ‘spit’, but the meaning of the thematic core PV-LV complex in
(a, b) is unexpected, because of the mismatch between its valency and that
of the LV. The complex verb nyunjurl-ya- contrasts with (c), which con-
sists of two independent phrasal constituents: nyinypa ‘spit, saliva’ which
is the absolutive cased direct object NP of the inflected HV kiji-rni. These
constituents may occupy any position relative to each other within the same
finite clause.
However nyunjurl-ya- in (a, b) does not provide serious counter-
evidence to the claim that the valency associated with a complex verb is
marked by the LV. This exceptional transitive complex verb with the LV
ya- is likely to have been borrowed from neighbouring and genetically close
Warlmanpa, in which the ‘put’ verb is ya(n)-, which combines with PVs to
form complex verbs (David Nash , personal communication). While
Warlmanpa ya(n)- is a second conjugation verb (Nash ) in that lan-
guage, monosyllabic verbs can only belong to conjugations  to  in Warlpiri;
the presence of stem final n would signal a conjugation  verb leading to the
incorporation of a verb homophonous with conjugation  -ya- ‘go’ in all its
forms. Whatever the etymology of ya- in (a, b), this PV-V combination is
exceptional.
Another unexpected PV-LV combination is the intransitive juurl-pinyi
‘jump, leap, hop’ since pinyi as an HV is transitive and typically combines
with a PV to form a transitive complex verb. This verb only operates as an
internally caused change of spatial configuration verb, but even in this use
the presence of pinyi seems exceptional by comparison with its use in other
PV-LV forms as seen in () and in the forms to be discussed in section ...

. Event structure and the thematic core

Ideally the representation of the semantic properties expressed by the
thematic core of verbal constituents should capture their Aktionsart and
argument-taking properties as well as how subpredicates combine to cre-
ate the complex predicate we intuitively associate with a simple or com-
plex verbal constituent. In this section I will sketch out a representation of


In a finite clause, the transitive subject NP is marked by an overt ergative case suffix while
an intransitive subject or direct object NP is not overtly case-marked. Pronominal enclit-
ics have distinct subject and non-subject forms. Third person (singular) subject or direct
object has no overt enclitic form (Hale , ).
 Warlpiri verbs of change and causation: the thematic core 193

the semantic properties common to a number of predicates which express

STATES, CHANGE OF STATE, and CAUSE CHANGE OF STATE, as an event struc-
ture or conceptual structure (in the tradition of Hale and Keyser ,
, Jackendoff , , Laughren , , Levin and Rappaport
, Levin and Rappaport Hovav , , , inter alia). The event
structure I propose identifies argument-taking predicates and Aktionsart
features; the latter playing a crucial role in the argument-predicate
relationship.

.. Stance verbs reviewed

Let us return to examples of stance verbs seen in () and (), repeated here
as () and ().
() a. [karri]-mi ‘stand, stop, stay’ (x)
stand-NPST
b. [karri-nja- [pardi]] -mi ‘stand up, arise’ (x)
stand-INF-rise-NPST
c. [karri-nja-[yirra]] -rni ‘put in standing position’ (x,y)
stand-INF-put-NPST

() a. nguna ‘be in lying position/come to be in lying

lie.NPST position/continue to lie’ (x)
b. [nguna-nja- [wanti]] ‘move into lying position, typically by
lie-INF-fall.NPST downward movement’ (x)
c. [nguna-nja- [yirra]] -rni ‘cause to come to be/cause to stay in lying
lie-INF-put-NPST position’ (x,y)

Only one of the semantic representations associated with karri in (a) is

associated with karri-nja ‘stand-INF’ in (b, c), i.e. the stative ‘be in STAND-
ING stance’. Similarly for nguna-nja ‘lie-INF’ in (b, c). The ‘event’ struc-
ture underlying the stative predicate is shown in Figure ., in which a
Predicate Phrase (PredP) (unspecified for syntactic category) is in the scope
of Aktionsart features, (+/–stage and +/–dur(ative)) associated with the
head of an Aktionsart Phrase (AKTP). The ‘subject’ argument of the PredP is
in the Specifier position of the AKTP.
In (b, c) and (b, c) the stance situation is interpreted as the ‘end-
state’ of a CHANGE of stance. We can represent the meaning of (b) as an


My approach to the interaction between Aktionsart and PAS properties is like that of Borer
(), who argues that argument positions in syntactic structure are identified as speci-
fiers of aspectual phrases.
194 Laughren

SITUATION
AKTP1
AKT

Xi
AKT Predicate1 α
+/– stage
+/– dur
STANCE
LOCATION
STATE

Figure . Event structure of stative predicate

(Negative) AKTP3 ACHIEVEMENT

Xi CHANGEP
AKT
+/– dyn

AKTP2
AKT
CHANGE
Xi AKTP1

¬Predicate1 α AKT

AKT
+/– dyn Xi
AKT Predicate1 α
+/– stage STANCE
+/– dur – dyn
LOCATION +/– stage
STATE +/– dur STANCE
LOCATION
STATE

SITUATION SITUATION

Figure . Event structure of ‘change’ predicate

event involving change from one situation to its opposite situation: NOT in
STANDING stance (xi) CHANGE TO in STANDING stance (xi). As shown in
Figure ., the CHANGE predicate is a relationship between two situations
mediated by Aktionsart features. While the Aktionsart feature govern-
ing the PredP in each of the ‘situations’ involving the same argument may
be –dyn(amic), the feature governing the CHANGEP can be either +dyn or
–dyn (depending on the scope of negation). The argument represented by
X undergoes a change of stance (location or state) such that the predicate
holding of it in the post-change situation is presupposed to have not held
 Warlpiri verbs of change and causation: the thematic core 195

CAUSEP (ACHIEVEMENT)

Xi

CAUSE
(Negative) AKTP3 (ACHIEVEMENT)

CHANGEP
Xi

Fig. 7.3 'Change' event

Figure . Internally controlled cause change event

of X prior to the change. As we will see, meanings associated with identi-

cal forms may vary as a result of different scopes given to negation in these
structures.
A CHANGE OF STANCE resulting from an action initiated and controlled by
the entity which changes stance is expressed by the combination of VINF-
pardi in (b). This contrasts with (c), in which the change of stance results
from an action involving the ‘undergoer’ which is initiated by an entity other
than the undergoer. As discussed by Levin and Rappaport Hovav ()
with respect to English ‘verbs of spatial configuration’, independent Warlpiri
verbs like karri in (a) and nguna in (a) may be associated with a seman-
tic representation in which some self-controlling entity maintains self in the
spatial configuration designated by the verb, as represented in Figure .
where the change event is negated (X cause self NOT change from sitting
situation to non-sitting situation); or they may be associated with a stative
representation in which some entity or being happens to be in a specific spa-
tial configuration and hence not in the scope of a ‘change’ function (as shown
in Figure .). The verb nguna in (a) may also be interpreted in the same
way as (b), where the change of stance to the ‘lying’ position is internally
caused as represented in Figure ..
In (c) and (c) the role of causal agent of the change of stance event
is distinct from the role of undergoer of that change. The presence of yirra
‘put’ in the complex verb entails a causing event or situation which permits
the projection of an ‘agent’ argument in addition to the projection of an
‘undergoer’ argument.
In Figure ., the Y argument is represented as the specifier of the AKTP
that dominates the CauseP. The value of the +/–dyn feature in the head of
this AKTP indicates whether the cause is a dynamic event as in John burnt

This is similar to Pustejovky’s ‘transition’ between states of affairs or situations in which
predicates with polar values hold of the same entity (Pustejovky , ).
196 Laughren

AKTP5 (ACCOMPLISHMENT)

Yj
CAUSEP
AKT
+/– dyn
AKTP4 (EVENT/SITUATION)

Yj PREDP
AKT
+/– dyn CAUSE
+/– dur (Negative) AKTP3 (ACHIEVEMENT)

Xi CHANGEP
AKT

Fig. 7.3 'Change' event

Figure . Externally caused change event

the toast (i.e. John did something involving the toast which resulted in toast
becoming burnt) or a non-dynamic situation involving some relationship
between the ‘agent of change’ and the ‘undergoer of change’, as in The sun
burnt John (i.e. John became burnt as a result of contact with the sun which
has burning properties). These constructions have different syntactic prop-
erties, e.g. the ability to license an instrument phrase, which correspond to
differences in Aktionsart features which define the relationship between the
‘cause’ function and the ‘change’ function and consequently between the
‘agent’ and ‘theme/undergoer’ roles.
X and Y are variables associated with thematic roles determined by the
Aktionsart properties governing the predicate they c-command; they inherit
these thematic or ‘event’ roles as a function of the predicates which hold of them
and also of the Aktionsart features which mediate the relationship between an
argument position and a predicate, as shown in Figures .–..
The contrast between a self-controlling or internal cause and a non-
self-controlling or external cause is indicated by the contrast between the
individual X argument in the specifier of the CAUSE function, as shown
in Figure ., and the AKTP argument in the same position in Figure ..
When the ‘cause’ is a situation or event then the Predicate Phrase which
expresses this event must be in the scope of aspectual features (as opposed
to the individual defining pronominal features associated with the position
filled by X in Figure .). What clearly distinguishes these internal cause
self-controlling ‘agents’ as in verbs like tarda-ya- ‘X sit, sit down, remain
seated’ from their external cause counterparts such as nyinanja-yirra- ‘Y sit
X down’ is the absence of a causing event distinct from the actual event of
moving from non-seated to seated, or the situation in which one is seated.
 Warlpiri verbs of change and causation: the thematic core 197

The internal causing agent is represented in Figure . by X raised from its
position in the CHANGE event (Spec, AKTP) to the Specifier of the CAUSEP.
On the other hand, the external causing agent Y associated with complex verbs
such as nyinanja-yirra- moves from its original position as the Specifer of the
causing event phrase represented by the AKTP in Figure . to the Specifier
position in AKTP. Y must move into this higher specifier position to check the
AKT feature [+/–dyn] in the head of AKTP which determines its interpretation
as a participant in a causal ‘situation’ or a causal ‘event’. It must also move to be
in a c-commanding position relative to X, in the specifier of the lower AKTP.
I have claimed that the relationship between an argument and a predicate
is mediated by an Aktionsart head. This head has features traditionally asso-
ciated with Aktionsart such as +/– dynamic, +/– durative, +/–stage. Some
Aktionsart properties may derive from the relational structure linking predi-
cates. The feature telic would seem to be of this sort; a stative predicate can be
viewed as providing a boundary or endpoint for some ‘transition’ or process of
‘change’. These features are syntactically active (as well as semantically) in that
they interact with features associated with inflectional heads within the clausal
structure, as well as with the determiner structure of argument expressions
and with temporal expressions and so on. A full examination of the syntactic-
ally relevant functions of AKT features is beyond the scope of this study. Only
those aspects that impinge on the selection of verb forms in PV-V structures
expressing CHANGE and CAUSE functions are addressed in this study.

... Morphosyntactic reflexes of contrasting event structures

The predicate in the stative Aktionsart phrase (AKTP) in Figure ., may be
expressed by a constituent made up of a stance HV stem (e.g. nyina, karri,
nguna, parntarri) and its corresponding VINF form, or a PV root (e.g. pata-,
pirri-, tarda-), or a N[ominal] (to be discussed in section ..). The CHANGE
event structure shown in Figure . is only mapped onto a constituent con-
taining V, whether HV (e.g. nyina, nguna) or PVlexical/VINF-V constituents
(e.g. pata-karri, tarda-ya-, pirri-ma-, ngunanja-wanti, karrinja-pardi). A
subset of these HV and PV-V forms may also map onto the internal CAUSE
event structure in Figure .. However, the verb forms which map onto the
complex EXTERNAL CAUSE event structure in Figure . are distinct from
those which map onto the other event structures (unlike a permitted English
pattern which allows some identical ‘causative’ and ‘non-causative’ forms).
The CHANGE and CAUSE functions are only expressed by verbal constituents,
whether HV or PV-LV forms.

... Negation in the change stance event structure

The structure in Figure ., as well as representing the ‘change STATE/
STANCE/LOCATION’ event which may be expressed by a simplex or complex
verbal constituent, also allows us to represent the ‘halt’ and ‘stay’ mean-
ings expressed by stance verbs. A negative operator must have scope over
the presupposed prior predicate if the lower result of change predicate is
198 Laughren

positive as in (b); where a negative operator has scope over the lower
resultant, another negative has scope over the CHANGE event structure as
in (c). Synonymous verbs may express all three semantic representations
given as (a–c). Only nyina-, however, may express the spatial configuration
which is not internally caused or maintained, as in (a) which maps onto the
structure shown in Figure .. Apart from nyinanja-wanti ‘sit down’ in (b),
which only maps onto the structure in Figure . in which the resultant situ-
ation is not negated, the other three verbs (nyina, tarda-ya- and pirri-ma-)
may map onto Figure . with either pattern of negation. Unlike a verb such
as wanti ‘fall’ or its complex verb synonym pata-karri which map onto the
structure in Figure ., the verbs tarda-ya- and pirri-ma- always presuppose
internal control over the change event as shown in Figure .. This require-
ment does not limit them to animate subjects, since an aeroplane can be said
to pirri-ma- or tarda-ya- in the sense of (b), since it moves to the landed
position under the control of its own engine. However, these verbal constitu-
ents cannot be used of a leaf or some other object which falls to the ground
due to external forces over which it has no control.
() nyina-, nyinanja-wanti, tarda-ya-ni, pirri-ma-ni ‘sit’

a. seated (x) SITUATION ‘be in seated nyina-

position’
b. CAUSE (x, EVENT ‘assume seated nyina-, nyi-
(NOT seated position’ (sit, perch, nanja-wanti,
(x)) > seated land) tarda-ya-,
(x)) pirri-ma-
c. NOT CAUSE NegEVENT ‘not change from nyina-,
(x, (seated (x) seated position’ tarda-ya-,
> NOT seated (stay, wait, remain pirri-ma-
(x))) seated, sit)
The verb karri- ‘stand’ can be associated with the semantic representations
in (a–c).
() karri- ‘stand’

a. standing (x) SITUATION ‘be in standing

position’
b. NOT CAUSE (x, (standing NegEVENT ‘stay, wait, continue to
(x) > NOT standing (x))) stand’
c. CAUSE (x, (moving (x) > EVENT ‘halt, stop’
NOT moving (x)))


The meanings of karri shown in (b, c) may also be expressed by the complex verb jupu-
karri.
 Warlpiri verbs of change and causation: the thematic core 199

As both an independent or dependent verb, yirra- ‘put’ expresses the event

complex in Figure ., where some ‘agent’ Y engages in some action or is
in some situation involving a ‘theme’ X which causes X to change (or not to
change) its stance or location. The contrast between (a) and (b) can be
expressed by the presence or absence of a negative operator over the CHANGE
event as shown in Figure ..
() yirra-rni ‘put, keep’

a. Cause (EVENT/SITUATION COMPLEX ‘put’

(y), change location (x)) EVENT

b. Cause (EVENT/SITUATION COMPLEX ‘retain, keep at,

(y), not change location NegEVENT detain’ (cf. (c)
(x)) and (c)).

The subject of yirra- has the role of external causer of the change of loca-
tion of the theme, or of its failure to change location. The EXTERNAL CAUSE
in (b) may be a situation in which the existence of the external causer (y)
relative to the undergoer (x) is sufficient to prevent a change of location, e.g.
Ngapa-ngku=nganpa yirra-rnu (rain-ERG=PL.EX.OBJ put-PST) ‘The rain
made us stay put’.
As discussed in section ., when yirra- is used with an VINF stance verb,
the subject of the VINF-yirra- complex verb is understood as referring to the
participant in an unspecified event or situation which causes the change of
stance and whose role in the complex event is distinct from that of the par-
ticipant which undergoes or fails to undergo a change of location or stance.
The ‘logical’ subject of the stance VINF is interpreted as an entity which
changes stance, due to the manipulation of some external agent. Monadic
stance verbs fall into at least two classes: internally caused stance verbs and
not internally caused stance verbs (in the sense of Levin and Rappaport
Hovav ()). The latter (no internal cause) fall into Levin’s () ‘verbs
of existence’ class. Interestingly, only the simplex HVs operate as stative ‘no
internal cause’ or ‘existence’ verbs as shown in Table ., as well as operating
as (negative) ‘achievement’ verbs.
The data in Table . reinforce the observations made in section .,
that the PV/VINF-yirra- ‘cause change of stance’ verb only expresses an
externally caused change of stance. However, the PV-V combinations
which express an internally caused change of stance cannot be expressed
as PV-VINF preverbal elements in combination with the ‘put’ causative verb
yirra-, e.g. *[tarda-yaninja] -yirra-. While wardu-karri ‘lie prone’ may
refer to an act of self-manipulated change of stance or a self-maintained
stance, a reflexive clause using dyadic wardu-yirra- is typically used of
people prostrating themselves over the coffin of a deceased relative, thus
200 Laughren

Table .. Stance and spatial configuration verbs

English be in be in change into change into/

gloss configuration configuration configuration remain in
(internal (no internal (internal cause) configuration
cause) cause) (external cause)
sit nyina nyina nyina; nyinanja-
nyinanja-wanti yirrarni
tarda-yani tarda-yani tarda-yirrarni
pirri-mani pirri-mani (pirrily-yirrarni)
stand karri karri karrinja-pardi karrinja-yirrarni
lie nguna nguna nguna; ngunanja-
ngunanja-wanti yirrarni
crouch parntarri parntarri parntarrinja- parntarrinja-
over wanti yirrarni
squat mirdi. mirdi.jirrpijirrpi- mirdi.jirrpijirrpi-
jirrpijirrpi- wanti yirrarni
nyina
kneel mirdi. mirdi.pirrjipirrji- mirdi.pirrjipirrji-
pirrjipirrji- mani yirrarni
mani
lie prone wardu-karri wardu-karri wardu-yirrarni

forcing the external causer role and the changer of stance role to be linked
to the same referent via syntactic binding.

.. Change of state verbs

... Enduring changed state
Warlpiri has a set of ‘impaired integrity’ nominal predicates which include
rdilyki ‘broken’, larra ‘split, torn’, rdiirr(pa) ‘busted’, and rdilypirr(pa) ‘per-
forated’, which belong to a larger set of ‘result of change’ stative predicates.
These nominals combine as a PV with the monadic LV such as ya- ‘go’, in
order to express a CHANGE STATE event: e.g. x break/split/burst etc. These
same denominal PVs combine with a dyadic LV, e.g. pinyi, to express a dyadic
CAUSE CHANGE event complex. This class of nominal (and PV) predicate
presupposes a former ‘whole’ (unimpaired) state of its argument. Thus its
event structure is not that of a simple situation as shown in Figure ., but
rather like that of Figure ., where the entailed situation is presupposed to

The contrast between the syntactic reflexive construction with wardu-yirrarni and the
intransitive wardu-karri marks the same contrast of focus on the causing event as opposed
to the resultant stance as is expressed by the contrast between the English ‘They pros-
trated themselves over the coffin’ versus ‘They lay prostrate on the coffin’.
 Warlpiri verbs of change and causation: the thematic core 201

be the result of a CHANGE from a former situation where the predicate does
not hold. Stage-level nominal predicates such as rdilyki ‘broken’ contrast with
individual nominal predicates such as a colour-denoting terms, e.g. tiri-tiri
‘red’, whose meaning representation does not (although it can) presuppose
a former state of being of some entity without that colour property. Colour
predicates are typically non-dynamic individual-level predicates, although
they may be used as stage-level predicates.
While the use of yani and pinyi with denominal ‘impaired integrity’
PVs such as rdilyki form a semantically transparent constituent, this pat-
tern of complex verb formation appears to be only partially productive.
Other change state nominal predicates such as pukulyu ‘rotten, putrid,
bad (typically of meat)’ or yurnmi ‘ripe, cooked’, which we might expect,
on semantic grounds, to fall into the rdilyki class do not. As a preverb
expressing a changed state in a PV-LV verbal constituent, pukulyu and
yurnmi follow the productive ‘default’ pattern which involves the depen-
dent LV jarri and causative LV mani seen in () and to be reexamined in
section ....
CHANGE OF STATE verbs with impaired integrity predicates are typical
achievements since they involve an instantaneous transition between being
unimpaired and being impaired. The resultant state of impairment can be
conceived as ongoing (+durative). Why do these ‘impaired integrity’ predi-
cates contrast with verbs of change involving predicates such as pukulyu ‘rot-
ten’ and yurnmi ‘ripe, cooked’? What seems to differentiate these classes of
nominal predicate (i.e. rdilyki type versus the pukulyu type) is that the rdilyki
type is associated with an event structure involving a presupposed EXTER-
NAL CAUSE CHANGE function whereas the pukulyu predicate is not. Levin
and Rappaport Hovav () argued that ‘break’ verbs in their intransitive
form presupposed an external cause in their meaning representation which
was not projected into their syntactic representation. In languages such as
English and French, ‘break’ predicates are typically verbal as in (a), while
‘ripe’ predicates are typically adjectival as in (a), each with contrasting
Aktionsart properties. While in French (and other Romance languages),
intransitive active ‘break’ verbs are expressed by the reflexive intransitive
construction with the auxiliary être as in (b), the transitive causative con-
struction involves the simple verb form and the avoir auxiliary verb. On the
other hand, the inchoative ‘ripen’ is expressed by a non-reflexive verb form
as in (b), as befits an unergative ‘activity’ predicate while the causative
sentence in (c) has the same surface form as causative ‘break’ in (c),
past participle of a lexical verb with the accusative assigning auxiliary avoir.
However, while casser ‘break’ allows either an eventive or situational cause,
mûrir ‘ripen’ only allows a situational cause (i.e. relationship between sun and
apple) within the simple or non-periphrastic VP. An eventive cause can only
be expressed by introducing the verb faire which takes the expression of the
‘ripening’ event as its infinitival complement.
202 Laughren

() a. Le verre est cassé.

‘The glass is broken.’
b. Le verre s’est cassé.
the glass REFL is broken
‘The glass broke.’
c. Jean/le vent a cassé le verre.
‘John/the wind has broken the glass.’
() a. La pomme est mûre.
‘The apple is ripe.’
b. La pomme a mûri.
‘The apple has ripened’
c. Le soleil*Jean a mûri la pomme.
‘The sun/*John has ripened the apple’
‘The sun has ripened the apple.’ (≠ John has ripened
the apple.)
d. Jean a fait mûrir la pomme.
John has made ripen the apple
‘John has ripened the apple.’
In Warlpiri, both rdilyki ‘broken’ predicates and yurnmi ‘ripe’ predicates are
nominal. However, in addition to their selection of different LVs to form com-
plex inchoative and causative predicates, rdilyki nominals display other syn-
tactic properties which align them with dynamic verbal predicates as opposed
to stative nominal predicates such as yurnmi. A Warlpiri verb may be con-
verted into a resultant state predicate by putting it (in its VINF form) into a
phrase headed by the postposition warnu as illustrated in (c), in which the
complex verb rdilyki- [vinfkati-rni-nja] forms a resultant state predicate headed
by -warnu. Interestingly, nominal ‘eventive’ predicates of the rdilyki class
may also be directly suffixed by warnu as in (b, d) to create a resultant state
predicate. By comparison with the bare nominal predicate rdilyki in (a), the
warnu form focuses not just on the resultant state but signals it as the product
of some event – fully specified in (c) but not in (b or d).
() a. Kurlarda yali ka nguna rdilyki.
spear that AUX:PRS lie:NPST broken
‘That spear is (lying) broken.’
b. Kapi=rna ngurrju-ma-ni yali=ji kurlarda rdilyki-warnu.
FUT=SG.S good-CAUS-NPST that=TOP spear broken-result
‘I’ll fix that (having been) broken spear.’
 Warlpiri verbs of change and causation: the thematic core 203

c. Kurlarda yali ka nguna rdilyki-kati-rninja-warnu.

spear that AUX:PRS lie.NPST broken-run_over-result
‘That spear is (lying) having been run over and broken.’
d. Kurlarda yali ka nguna rdilyki-warnu.
spear that AUX:PRS lie.NPST broken-result
‘That spear is (lying) (having been) broken.’
By contrast, non-eventive predicates including yurnmi and pukulyu cannot be
used in the resultative warnu construction, as seen in (b). Only the verbal
forms with jarri or -mani may be governed by -warnu, as shown in (c).
() a. Miyi nyampu ka nguna yurnmi.
vegetable_food this AUX:PRS lie.NPST ripe
‘This food is (lying) ripe.’
b. Kapi=rna nga-rni miyi nyampu=ju yurnmi/*yurnmi-warnu.
FUT=SG.S eat-NPST food this=TOP ripe/ripe-RESULT
‘I will eat this ripe fruit.’
c. Kapi=rna nga-rni miyi nyampu=ju yurnmi-jarri-nja-warnu/
FUT=SG.S eat-NPST food this=TOP ripe-INCH-INF-RESULT/
‘I will eat this ripened fruit (which has become ripe/
yurnmi-ma-ninja-warnu
ripe- CAUS-INF-RESULT.
been made to become ripe).’
The addition of LV ya-ni ‘go’ to a PV derived from an impaired state
rdilyki nominal as in (a) actualises the event structure in Figure .,
in that the change of state is now entailed, rather than just presupposed.
The addition of V-pinyi to the rdilyki type PV exemplified by rdilyki-
pungu ‘broke’ in (b, c) signals that the distinct ‘agent’ and ‘undergoer’
roles defined in the specifier position of their respective A KTPs in the
event complex shown in Figure . are mapped onto positions in syn-
tactic structure. As this analysis would predict, the only way of express-
ing an internal causer reading meaning that ‘x causes its own breaking’,
equivalent to ‘x causes/maintains its own sitting/standing/lying’ (Table
., column ), has to be by syntactic binding in a reflexive construction
as in (b), where both the ‘agent’ and ‘theme/patient’ roles have the
same referent.
() a. Kurdu waku rdilyki-ya-nu.
child arm broken-GO-PST
‘The child broke (his) arm.’ = ‘The child’s arm broke.’
204 Laughren

b. Kurdu-ngku=nyanu waku rdilyki-pu-ngu.

child-ERG=ANAPH arm broken-CAUSE-PST
‘The child broke (his own) arm.’ (as in deliberate act)
c. Kurdu-ngku waku rdilyki-pu-ngu.
child-ERG arm broken-CAUSE-PST
‘The childi broke his/her*i/j arm.’ (someone else’s arm)
These change of state verbs contrast then with the intransitive stance verbs
which may express an event structure in which the CAUSING EVENT is an
integral component of the CHANGE STANCE event, that is the internal caus-
ation interpretation.
The nature of the causing event entailed by rdilyki-pinyi is not encoded,
as opposed to rdilyki-katirni ‘break-apply_downward_pressure_to’ in (c),
in which the nature of the causing event is specified. In this sense, pinyi
in (b, c) is a semantically ‘light’ verb; it expresses the cause function and
licenses the syntactic projection of the ‘agent’ argument. The role of yani
and pinyi as LVs is to signal how much of the event structure of eventive
rdilyki predicates is projected into the syntactic structure: CHANGE function,
or CAUSE and CHANGE. However LVs yani and pinyi as in () differ from
the default syntactic inchoative -jarri and causative -mani discussed in sec-
tion .... These verbalisers do not express an event structure associated
with a nominal or PV predicate which presupposes a CHANGE and/or CAUSE
CHANGE function; rather these dependent LVs are directly associated with
those elements of event structure which transform a stative predicate (of the
type shown in Figure .) into a dynamic one: achievement or accomplish-
ment. They combine with the stative event structure of the nominal predi-
cate they combine with to create a more complex event structure.

... ‘Default’ change and cause verbs

As pointed out by Hale () and Simpson (), Warlpiri verbs express
dynamic predicates, which include stage-level predicates but exclude individ-
ual-level predicates (Carlson , Kratzer ). Complex verbs expressing
CHANGE OF STATE may be formed by taking a stative nominal predicate which
expresses an ‘individual-level’ or ‘stage-level’ predicate and combining it as a
PV with a dependent LV jarri which expresses a change in some entity from
not having the property denoted by the denominal PV to having it. The PV is
interpreted as expressing the resultant situation, which is in the scope of the
CHANGE function, as shown in Figure .. The LV jarri marks this change of
event structure from ‘state’ to achievement’, as shown in (a, b).


Stative nominal predicates may be derived from verbs. Individual-level predicates are
encoded as nominals, as are some stage predicates. Most nominal predicates can function
as either type.
 Warlpiri verbs of change and causation: the thematic core 205

Notice, however, that PV-jarri verbs which express a change in some

attribute of some entity do not encode an internally caused or controlled
change of state. Nor do they permit a self-maintained-in-state-of-having
attribute interpretation (akin to the maintain-self-in-stance meaning asso-
ciated with the stance verbs we have discussed). In (a, b), the transition
from a situation in which a predicate does not hold of x to one where it does
hold, can be viewed as a process when in the scope of the aspectual LV -yani
‘path’. This aspectual use of -yani is different from that of thematic yani,
which forms achievement predicates with the ‘break’ denominal PVs such as
rdilyki-yani, where the transition is not a process.
() a. Kuyu ka [pukulyu-jarri-nja]-ya-ni.
meat AUX:PRS bad-INCH-INF-PATH-NPST
‘The meat is going bad.’
b. Wati ka [ngurrju-jarri-nja]-ya-ni.
man AUX:PRS good-INCH-INF-PATH-NPST
‘The man is getting better/becoming good.’
PV-mani CHANGE STATE verbs, which pair with PV-jarri CHANGE verbs, must
project a structure with an external cause argument position in addition to an
internal theme argument, since the cause event must be distinct from the
change event. Notice that like the PV-yirrarni and PV-pinyi expressions of a
cause function, the PV-mani verbs in () fail to specify any manner feature
of the cause event, other than the involvement of an agent and an undergoer.
() a. Kuyu ka [pukulyu-ma-ninja]-ya-ni jirri-ngki.
meat AUX:PRS bad-CAUS-INF-PATH-NPST maggot-ERG
‘Maggots are making the meat go bad.’
b. Wati ka [ngurrju-ma-ninja]-ya-ni nyanguwaji-rli.
man AUX:PRS good-CAUS-INF-PATH-NPST doctor-ERG
‘The doctor is making the man better.’
c. Mangarri-rli ka wati [ngurrju-ma-ninja]-ya-ni.
food-ERG AUX:PRS man good-CAUS-INF-PATH-NPST
‘Food is making the man better.’
Unlike the LV involved in change stance complex verbs, neither the incho-
ative nor causative dependent verbs can combine with a VINF preverbal
item: *nyina-nja-mani ≠ nyina-nja-yirrarni ‘cause to be seated’; *nyina-nja-
jarri ≠ nyina ‘come to be seated’. Given the range of preverbal elements both
-jarri and -mani may combine with, this restriction might seem rather sur-
prising. Even more so as jarri and mani productively create complex CHANGE
predicates by combining with various complex phrasal preverbal elements.
The VINF, however, does not constitute the right sort of phrasal category,
206 Laughren

perhaps because it retains its +dynamic Aktionsart feature. To combine

with jarri or mani, VINF must be embedded in a functional projection such
as a postposition phrase (PP), e.g. [wangka-nja-ku-wangka-nja-ku]-ma-ni
([speak-INF-PURPOSIVE] [speak-INF-PURPOSIVE]-ma-NPST) ‘make to speak’.
These reduplicated VINF-ku preverbal constructions are limited to -mani,
and the VINF is derived from an internally caused speech verb, wangka
‘speak’, yula ‘cry’. These verbs are associated with an external causer which
brings about a distinct event with an internally controlling cause.
PVs derived from a PP predicate may combine with either jarri or -mani
to form a complex verb such as [walya-kurra]-jarri ([earth-ALLAT]-jarri)
‘to land’ or [walya-kurra]-mani ([earth-ALLAT]-mani) ‘make to land’. These
verbs may have either a CHANGE LOCATION interpretation ‘to land/make to
land’ or a CHANGE STATE one, ‘become earth/make to become earth’, where
the state is an individual predicate. This latter meaning can also be expressed
without the use of the allative, by deriving the PV from the NP.
Both the PV-yani/-pinyi and PV-jarri/-mani verbs express a change to
a state of affairs which is typically understood to be potentially enduring or
ongoing. In Figure . this is represented by the choice of the + value for
the feature [dur(ative)] in the head of the Aktionsart phrase AKTP, which
immediately contains the stative predicate. However some states are merely
transitory, and hence [–durative].

... Transitory change of state predicates

Another type of change of state is expressed by PV-V combinations such as
kinyirr-ngarni or julurr-ngarni ‘give a start, be startled, involuntarily jump/
move with fright’. The event structure associated with these verbs is simi-
lar to that shown in Figure . since an external cause event is presupposed
as the trigger for the change of state in the undergoer. However, no agent of
cause is syntactically projected. Thus PV-ngarni verbs of this type behave
like rdilyki-yani ‘break’ verbs with a PV predicate which presupposes an
externally caused change of state, although it is not syntactically projected.
However, whereas the resultant state with rdilyki preverbs is [+durative],
with kinyirr- type PVs it is [–durative].
The dyadic transitive counterpart of these PV-ngarni verbs, where the
agent of the external cause event is syntactically projected, is formed by
dependent PV-jirrirni. Unlike the PV elements in the PV-yani/-pinyi and
PV-jarri/-mani verbs, which may express a stative predicate ‘x is in STATE’,


jarri and mani are also used to derive non-telic activity predicates which are not discussed
herein. A full investigation of the role of these verbalisers must be the topic of another
study.

Despite allowing an N or PP as PV, these verbs do not derive PVs with complex NPs,
e.g. *walya wiri-kirra-jarri/mani (earth big-ALLAT-jarri/mani) ≠ ‘(make to) come to be on
big ground’, *karnta wiri-jarri/mani (woman big-jarri/mani) ≠ ‘turn into a big woman’.
Reduplicated PVs are permitted.
 Warlpiri verbs of change and causation: the thematic core 207

Table . CHANGE of state and CAUSE change of state dependent verbs

STATE type CHANGE into [+dur] CHANGE into [+dur] CHANGE into
state impaired state [–dur] state

External cause -jarri -yani -ngarni

of change not ( pukulyu-jarri (rdilyki-yani ‘break’) (kinyirr-ngarni
projected ‘become rotten, rot’) ‘be startled’)
External cause of -mani (pukulyu- -pinyi (rdilyki-pinyi -jirrirni
change projected mani ‘make rotten, ‘break’) (kinyirr-jirrirni
rot’) ‘startle’)

lexical PVs such as kinyirr and julurr do not form independent stative predi-
cates. They only express a momentary non-durative transition from one
state to another followed by a cancellation or cessation of the momentarily
acquired state. That is, the state entered into as a result of some external
cause is presupposed to be quickly reversed or changed, and can only be
expressed by a PV predicate incorporated into a verbal constituent which
can express this aspectually dynamic event.
Table . summarises the types of stative, change state, and cause change
state verbal constituents investigated in our study so far.

.. Nominal predicates

Nominal predicates may be used in verbless clauses with an aorist (inherent
individual-level predicate) or present time (stage-level predicate) interpret-
ation as in (a), but must be associated with a verb if inflectional features
(TAM) are to be expressed. Typically a stance verb serves as a sort of default
stative ‘light’ verb in an ‘inflected’ clause, as in (b, c). Its role is syntactic
rather than thematic in this sense. It checks functional features in a verbal
complex (Figure .), but does not constitute the main predicate. The pres-
ence of the stative verb and the subsequent projection of the verbal com-
plex permits restrictions over the truth conditions of the associated nominal
predicate; its truth may be restricted to a period of time relative to the time
of the utterance, or modified by modal features. Although in (b, c) the
stance verb is not the primary thematic predicate, the choice of stance verb
is determined by the spatial configuration or some relevant property of the
entity referred to by the subject of the nominal predicate. Importantly, the


The momentary status of these PV-ngarni/jirrirni verbs might suggest that they should be
classed as semelfactive in contrast with the achievement verbs such as rdilyki-yani/pinyi
‘break’ (Smith ) However, Smith excludes all ‘change’ predicates from her semelfac-
tive class and they pattern aspectually with ‘break’ verbs rather than with true semelfac-
tives such as ‘flash’ maarr-mani.

See Dixon () for data on copula verbs in Australian languages.
208 Laughren

choice of copula verb is not determined by the nominal predicate, unlike the
relationship between a thematic PV predicate and the LV it associates with
to form a complex verbal predicate.
() a. Kuyu pukulyu/ngurrju/linji/yurnmi/wanka/wiri.
meat bad/good/dry/ripe/raw/big
‘(The) meat (is) bad/good/dry/ripe/raw/big.’
b. Kuyu ka pukulyu/ngurrju/linji nguna/*karri.
meat AUX:PRS bad/good/dry lie/*stand
‘The meat is (=lies) bad/good/dry.’
c. Kuyu=lpa pukulyu/ngurrju/linji nguna-ja/*nyina-ja.
meat=AUX:IMPF bad/good/dry lie-PST/*sit-PST
‘The meat was (lying) bad/good/dry.’
d. *Kuyu ka pukulyu-nguna/ngurrju-nguna/linji-nguna.
meat AUX:PRS bad-lie/good-lie/dry-lie
‘≠The meat is (=lies) bad/good/dry.’
‘≠The meat lies badly/well/dryly.’
The nominal predicate cannot be expressed as a PV within the verbal con-
stituent, hence the ungrammaticality of (d) with either the predicate of
‘meat’ or modifier of ‘lie’ interpretations. This behaviour of the stative nom-
inal predicates in (a–d) contrasts with the PV-LV forms in () and (),
in which the nominal predicate is converted to a PV which forms part of the
thematic core of the complex verbal constituent. I propose that the copula
use of stance verbs is basically aspectual in the sense that the stance verb
maps onto the syntactic category which matches the head of the AKTP which
governs an empty predicate (PredP) position. The external NP predicate is
linked to this position by case agreement, in the same way that nominal pred-
icates are linked to argument positions in the verbal complex. This structure
is shown in Figure .. The selectional restrictions imposed by the subject
on the verb operate in the usual way, since they are in a Specifier–Head rela-
tionship within the stative AKTP.
If in (b, c) the verb nguna were the main predicate, then it would have to
map onto the PredP position in the AKTP shown in Figure .. The nominal
predicates could not also be mapped onto this position – either directly or
indirectly via case-marking. The only possible interpretation of the nominal


Pukulyu translates ‘bad’ in the sense of ‘rotten’, ‘decayed’.

With different intonation and underlying syntactic structure, (a) can be a complex NP
meaning ‘bad/good/dry/cooked/raw/big meat’, in which the qualifying nominal is used
attributively as a non-dynamic individual-level predicate.

(b) with karri and (c) with nyina-ja are acceptable where kuyu is interpreted as ‘ani-
mal’ rather than ‘meat’.
 Warlpiri verbs of change and causation: the thematic core 209

AKTP1

AKT

Xi

AKT PredicateP
NP
+/– stage
+ dur

karri unfilled Case

nyina- predicate agreement
nguna position
parntarri
…

Figure . Copula verb and nominal predicate

predicates would be ‘adverbial’, modifying the manner of lying (nguna), or

as an attribute or predicate directly related to the subject, e.g. ‘rotten/good/
dry meat was lying (there)’ or ‘the meat was lying there being rotten/good/
dry’. This ‘secondary predicate’ interpretation of the nominal predicate is
not represented by the structure in Figure ., in which the verb is not pred-
icative but aspectual and the NP is the primary predicate. As primary predi-
cate, why can’t the predicative NP be directly associated with PredP in the
AKTP and then form a PV-V constituent with nguna? The failure of *ngur-
rju-nguna in (d) as either a stative ‘be good’ or dynamic ‘become good’
complex predicate contrasts with the complex verbal constituents PV-jarri
in (a, b), in which the PV is derived from an NP predicate which can only
be interpreted as being in the scope of a dynamic predication, and hence
associated with the PredP in event structure.
As mentioned already, it is only the simplex HV stance verbs which can
operate as copula verbs, not the complex ‘stance’ verbs such as tarda-yani
or pirri-mani. As we saw, these verbs are associated with a CHANGE func-
tion while the PV is associated with PredP. Given that the PVs such as tarda
and pirri are associated with the PredP, it is not possible to also link a nom-
inal predicate into that position. Only simple verbs have the capacity to map
onto the head of AKTP, and it is simple stance verbs which may check the
[–dynamic] feature. Where a stance verb is the main predicate then it maps
onto PredP as well as checking the governing AKT with its subject argument
as Specifier of AKTP, as in Figure .. In a change stance interpretation, the
verb maps onto PredP and also checks AKTP and the CHANGE function in the
same way as complex verbs.
While denominal PVs cannot incorporate into a stance predicate mean-
ing ‘be PV’ as seen in (d), lexical PVs are found in PV-V stance constitu-
ents which express the predicate expressed by the PV, e.g. lamurr-nyina ‘be
210 Laughren

round’, liirl-nyina ‘be white/shinny’, purlurn-nyina ‘be bulging/prominent’,

luurl-nyina ‘be raised (of surface)’, pirlpirl-nguna ‘lie scattered’, waraly-
karri ‘hang/dangle’, jaa-karri ‘be agape’, as pertaining to the referent of
the verb’s subject. These verbs do not entail a change of state, though they
may presuppose one. Unlike the denominal PVs of the rdilyki ‘broken’ and
ngurrju ‘good’ classes, these PVs do not also function as nominal predicates.
Nominal predicates are derived from these PV roots by the suffixation of a
predicative nominal formative -pari, e.g. liirl-pari ‘shiny, whitened’, luurl-
pari ‘raised’, etc.
These PVs only form a predicative constituent if either incorporated as
PV into a PV-V structure, or as the root of a -pari nominal. While a nominal
such as kardirri ‘white’ may form an independent predicate which cannot
be incorporated as PV in a PV-V constituent with a stative stance verb, syn-
onymous liirl- ‘white’ may do so. Once converted to a nominal predicate,
liirlpari ‘white’ can no longer serve as PV, e.g. *liirlpari-nguna. The distinc-
tion between liirl-, and jaa- PV forms on the one hand, and nominal predi-
cates ngurrju, linji, and kardirri on the other, is syntactic. Just what features
underlie this syntactic distinction is unclear.

.. Aspectual enclitics

Nominal predicates in verbless clauses such as (a) may have an individ-
ual-level or stage-level reading. However this ambiguity can be dispelled,
as shown in (a–d), by the use of aspectual enclitics: ‘individual-level’ is
marked by =puka or =pirdinypa, while ‘stage-level’ is marked by a mem-
ber of a set of enclitics =lku ‘changed state’, =wiyi ‘former state’, and =juku
‘unchanged state’.
() a. Kuyu nyampu pukulyu=pirdinypa/puka.
‘This meat (is the) rotten (one).’
b. Kuyu nyampu pukulyu=lku.
‘This meat (is) now rotten.’ (NOT rotten (meat) > rotten (meat))
c. Kuyu nyampu ngurrju=wiyi.
‘This meat (was) good before.’ (good (meat) > NOT good (meat))
d. Kuyu nyampu ngurrju=juku.
‘This meat is still good.’ (NOT ((good (meat) > NOT good (meat))))
It is evident that the interpretation of (b–c) fits the CHANGE event schema
in Figure . with the same variation in the scopal operations of negation
that we saw in the ‘event’ readings of stance verbs discussed in section ...

The contrast in syntactic behaviour between lexical PV predicates and nominal predicates
requires further research.

These enclitics may also be hosted by verbal predicates over which they have scope.
 Warlpiri verbs of change and causation: the thematic core 211

However, despite the interpretation of the predicative NPs in (b–d) relying

on this CHANGE event structure, these NP predicates cannot be expressed
within the morphosyntactic schema of the verbal complex in Figure ..
To be linked to the verbal complex, either the NP is incorporated as a PV
into the thematic verbal core headed by the inchoative jarri or the causative
-mani, as seen in () and (), or it is associated via case with the complex
headed by a ‘copula’ verb as discussed in section ...

.. The resultative construction

As already seen, the ‘impaired integrity’ PVs such as rdilyki ‘broken’, larra
‘split, torn’, rdiirr(pa) ‘busted’, rdilyirr(pa) ‘perforated’, etc. have homo-
phonous nominal counterparts. The aspectual interpretation of these terms
rests on a semantic schema in which the impaired integrity state is viewed as
the result of some externally caused change in the material integrity of the
entity it holds of. Other predicative nominals do not presuppose this seman-
tic ‘cause change’ schema, although, as seen in section .., they may be
placed within an inchoative or causative construction associated with ‘syn-
tactic’ LVs such as jarri and -mani, or may be in the scope of an aspectual
enclitic, such as those discussed in section .. which forces the property
they designate to be viewed as the result of change, or holding prior to
change, or holding in the absence of change.
Impaired integrity nominals differ from other predicative nominals with
respect to their ability to incorporate as PVs into the thematic core headed
by a verb (other than the ‘default’ inchoative jarri and causative mani), where
they are interpreted as resultant states. The verb they combine with may
specify the nature of the change event or of the causing event. Nominal pred-
icates which don’t presuppose this CAUSE CHANGE schema cannot be incor-
porated into a thematic core with a verb which specifies either the change
or the cause event. These contrasting behaviours of nominal predicates are
demonstrated by examples in () and ().
() a. Wati-ngki ka paka-rni kuturu ngurrju-karda.
man-ERG AUX:PRS chop/hit-NPST club good-transl
‘The man is chopping the club to make it better.’
b. *Wati-ngki ka ngurrju-paka-rni kuturu.
man-ERG AUX:PRS good-chop/hit-NPST club
≠ ‘The man is chopping the club to make it better.’/≠ ‘[…] is
chopping the club well’.
() a. Wati-ngki ka rdilyki-paka-rni kuturu.
man-ERG AUX:PRS broken-chop/hit-NPST club
‘The man is breaking the club by chopping it.’
212 Laughren

b. Yarla-pama-rlu ka parrka rdilypirr-yarlki-rni.

yam-grub-ERG AUX:PRS leaf perforated -bite-NPST
‘Yam-grubs bite holes (in) the leaves.’
c. Rdilypirrpa ka muku yarlki-rni watiya rdilypirrpa-karda,
hole AUX:PRS PV:ALL bite-NPST tree perforated-TRANSL
‘It bites holes in the whole tree to make it (full of) holes
kuja=ka=rla marlaja=lku linji-jarri.
COMP=AUX:PRES-DAT RELATED _ TO = CS dry-INCH:NPST
which then dies (=dries) as a result of its (action).’
We have already seen in section .. that impaired integrity nominal predicates
such as rdilyki combine as PV with a verb to form CHANGE STATE (PV-yani)
and CAUSE CHANGE STATE (PV-pinyi) complex verbs, whereas other nominal
predicates unassociated with this presupposed CAUSE-CHANGE event structure
form inchoative and causative PV-LV complexes with other ‘light’ verbs. In
section .. we also saw how rdilyki nominals are verb-like in forming result-
ant state predicates with -warnu, whereas other nominal predicates do not.
In (a), rdilyki in combination with the verb paka-rni ‘chop, hit’ expresses
the resultant state (as we’ve seen it do in combination with LV pinyi), while
paka-rni expresses features of the external causing event (man chops the
club). The verb paka-rni can be replaced by any transitive verb which can
be interpreted as a ‘causing event’ and the referent of the direct object of the
verb is also the entity which undergoes the change of state. A parallel example
is given in (b) with the PV rdilypirr ‘perforated’ and the V yarlkirni ‘bite’.
To be interpreted as a ‘causing event’ the verb must have the right semantic
features. For example, a transitive perception verb cannot combine with a
rdilyki PV to create a complex predicate meaning ‘see/hear/smell/touch
and break’ (e.g. *rdilyki-nyanyi [lit. broken-see/look]), even though Warlpiri
perception verbs have both stative (e.g. ‘see’) and activity (e.g. ‘look’, ‘watch’)
interpretations. Perception verbs, unlike ‘impact’ verbs such as pakarni ‘hit’
and yarlkirni ‘bite’, fail to project an event structure which can be the event-
ive ‘subject’ of a CAUSE predicate in the representation in Figure ..
On the other hand, if we take a nominal predicate from the non-rdilyki
class, such as ngurrju ‘good’ in (b), and treat it as a PV with an eventive
verb such as paka-rni, the PV fails to be interpreted as a resultant state
brought about by the event expressed by the verb it combines with. In
order to express this idea, the nominal end-state predicate is embedded in

In Warlpiri plants are not said ‘to die’. They are said to ‘become dry’. Similarly, one does
not ‘kill’ a tree.

Nyanyi ‘see, look at/for, watch’, purda-nyanyi ‘hear, listen to’, parnti-nyanyi ‘smell, smell
out/for’. etc.

How to capture these event structure differences is not clear, but it would seem to derive
from contrasting Aktionsart features.
 Warlpiri verbs of change and causation: the thematic core 213

a constituent headed by karda; this forces a potential resultant state inter-

pretation as in (a). However, the karda constituent cannot be incorpo-
rated into the verb (*ngurrju-karda-pakarni). NP karda never becomes part
of the ‘thematic core’ of a verbal constituent, even though it appears to map
onto the same semantic change event structure as the PV rdilyki. However,
the NP-karda constituent which operates as an independent phrase outside
the verbal constituent does not presuppose an external CAUSE, it is purely
inchoative. The event structure of the NP-karda phrase is equivalent to the
CHANGE state event in Figure . which excludes the CAUSE predicate; karda
maps onto the CHANGE predicate which selects a situation as a changed state
of affairs. Thus the presence of karda forces a dynamic interpretation of a
basically stative predicate. However, unlike the PV rdilyki in (a), which
entails that the club comes to be broken (by some external force specified by
the verb pakarni), the ngurrju-karda phrase in (a) does not entail that the
club becomes good/better. This may be the intended or possible outcome,
but the change of state is not necessarily entailed. Thus the event structure
associated with the NP-karda phrase is linked to the event structure of the
verb by another functional head which is the locus of a modal ‘goal’ or ‘pos-
sible outcome’ feature. The event structure of the verbal constituent and
that of the -karda constituent are not linked such that if the verbal predicate
is entailed, then the NP predicate is automatically entailed (or vice versa);
nor are they linked at the level of surface morphosyntactic structure, since
the NP-karda phrase is not incorporated into the verbal constituent.
Sentence (c) provides an example of a rdilyki nominal embedded in a
-karda phrase: rdilypirrpa-karda (perforated-karda) ‘to become perforated’
used as an adjunct to the structure projected by the HV yarlkirni ‘bite’.
Notice that in the PV-LV rdilypirr-yarlkirni (‘perforate-bite’) in (b), the
PV consists of the root (rdilypirr), whereas the nominal rdilypirripa (root-pa)
is used in (c), both as an independent absolutive NP meaning ‘hole/per-
foration’ and as an NP in the karda-headed phrase, with the predicative
‘with hole, holey, holed, perforated’ meaning.
A rdilyki type N/PV presupposes as part of its meaning a change in the
material integrity of the entity it is predicated of; thus its semantic represen-
tation includes the CHANGE event complex schema in Figure . as part of its
presuppositonal structure. By incorporating the PV into the inner thematic
core of the verbal constituent headed by ya-, the presuppositional structure is
syntactically realised. The verb ya- as in (a) does not introduce the change
schema into the event structure but signals its mapping into the syntac-
tic structure, as -ya- must be the spellout of some verbal feature associated
with the head of a verbal projection; the PV-LV constituent encodes an event
(change of state), not just a resultant state, as in (b) and also in (c), where
the absolutive NP rdilyki is external to the verbal constituent nguna ‘lie’.

Simpson () refers to -karda as the ‘translative’, following the nomenclature of a similar
Finnish suffix.
214 Laughren

() a. Karli rdilyki-ya-nu.

boomerang broken-GO-PST
‘The boomerang broke.’
b. Karli rdilyki.
‘The boomerang is broken.’
c. Karli ka rdilyki nguna. (*rdilyki-nguna)
boomerang AUX:PRES broken lie-NPST
‘The boomerang is lying broken.’
Although an impaired integrity NP presupposes a change of state, rdilyki in
(a, b) cannot operate as an independent predicative nominal with a change
of state event interpretation.
() a. Wati-ngki ka kuturu paka-rni rdilyki.
man-ERG AUX:PRS club chop-NPST broken
‘The man is chopping/striking a club (that is) broken.’ / ‘[…] a
broken club.’
≠‘The man is breaking a club (by) striking/chopping it.’
b. Wati-ngki ka kuturu paka-rni rdilyki=lki.
man-ERG AUX:PRS club chop-NPST broken=CS
‘The man is chopping/striking a club (that is) now broken.’
≠’The man is now breaking a club (by) striking/chopping it.’
c. Karli rdilyki-paka-rnu.
boomerang broken-hit-PST
‘(He) struck the boomerang and broke (it).’
In (a, b), rdilyki can only be interpreted as a stative predicate hold-
ing of kuturu ‘club’. Even if we add the changed state (CS) clitic =lki to
rdilyki, as in (b), this predicate will not be interpreted as referring to
a state resulting from the act of striking the club. In (a, b), rdilyki is
interpreted as predicated directly of the referent of the Direct Object NP
kuturu ‘club’ without this relation being mediated by the verb. In order
to achieve a causal link between the ‘strike’ event and the ‘broken’ state
either the stative rdilyki must be embedded in the phrase projected by
-karda, as in () and (), or must be in the thematic core of the verbal
complex as a PV, as in rdilyki-paka-rnu seen in (c). Syntactically, case
agreement is required between the predicate rdilyki and its logical sub-
ject kuturu in (a, b); both are in the phonologically null absolutive case.
Paka- ‘hit, chop, strike’ lacks any CHANGE OF STATE entailment; it only
entails a type of contact between one entity and another. A possible effect
can only be presupposed, not entailed. In order to achieve the resultant
 Warlpiri verbs of change and causation: the thematic core 215

state interpretation as an entailment, rdilyki must be incorporated into the

verbal constituent headed by paka-rni as a PV where it does not need to be
case-governed, as in (c).

.. ‘Light’ verbs versus ‘not-so-light’ verbs in PV-V thematic

complex verbs
The PV-V rdilyki-yanu [lit. broken-go:PST] in (a) only entails that the
boomerang changed state from being not broken to being broken. It does
not entail, like its HV counterpart, that the boomerang went (or moved
along a path), although it might be argued that the process of breaking
entails a metaphorical path, the transition from being unbroken to being
broken. The verb expresses the CHANGE predicate, checking the Aktionsart
features which assure an event reading for (a), in contrast with the sta-
tive reading in (b, c). In (), however, both the fact that the boomerang
broke and that it fell is entailed, which suggests that the verb wanti ‘fall’
may function like paka-rnu ‘hit’ in (c), in that both verbs express features
of the causing event. Since -wanti only allows a single argument to be
syntactically projected, karli must be identified as the subject of both the
‘breaking’ and ‘falling’ events. The breaking event can only be interpreted
as the result of the falling event – not as an independent contemporaneous
event.
() Karli rdilyki-wanti-ja.
boomerang broken-fall-PST
‘The boomerang broke as a result of falling.’
Similarly in (c) (repeated here as (a)), the ‘striking’ event expressed by
paka-rnu is not independent of the ‘breaking’ event, the latter is the result
of the former. In both () and (a) the V is used with the same meaning
as when used as a HV. Similarly for the examples in (c, d). However
pinyi in (b), like yani in (a), does not specify the nature of the causing
event, only the fact that it involves both the undergoer of the change of
state as its logical object and an ‘agent’ role which is not bound to that of
its object.
() a. Karli rdilyki-paka-rnu.
boomerang broken-chop-PST
‘He struck the boomerang and broke it.’


Note that I don’t assume that a ‘causing event’ interpretation must be projected onto a syn-
tactically ‘transitive’ structure with separate ‘agent’ and ‘undergoer’ argument positions,
and hence expressed by a constituent containing a transitive verb. Rdilyki-wanti ‘fall and
break’ differs from ngunanja-wanti ‘lie down’ in that the cause is eventive in the former
case (Figure .), but not in the latter (Figure .).
216 Laughren

b. Karli rdilyki-pu-ngu.
boomerang broken-CAUSE-PST
‘He broke the boomerang.’ (He acted on the boomerang and it
broke.)
c. Karli rdilyki-katu-rnu.
boomerang broken-exert_pressure_on-PST
‘He stood on/ran over the boomerang and broke it.’
d. Karli rdilyki-luwa-rnu.
boomerang broken-strike_with_missile-PST
‘He broke the boomerang when he struck it (e.g. with another
boomerang).’
The verb -pungu in (b) marks the projection into the syntax of an exter-
nal cause event, but does not specify the nature of that event. All verbs in
() project an external causer into the event structure, a role that is dis-
tinct from the undergoer of change. The intransitive verb wanti in (),
on the other hand, does not project an external causer argument and can
only be interpreted as involving the same entity as the one that undergoes
the change to being ‘broken’. However, unlike yani and pinyi, verbs such as
wanti in (), and paka-, kati-, and luwa- in () all specify aspects of the
cause event. These differences, I suggest, are encoded within the PredP in
the causing event AKTP in the specifier of the CauseP in the event complex
in Figure .. These distinctions are not syntactically relevant, only seman-
tically relevant. What is syntactically relevant is that the head of this PredP
is mapped onto a V position in syntactic structure. The nature of the event is
recoverable from the semantic information associated with the lexical entry
for this verb root.
Verbs such as yani and pinyi are semantically ‘lighter’ than wanti in ()
and paka-, kati-, and luwa- seen in () (which have a specified ‘manner’
component). In this class of PV-V construction, however, they behave differ-
ently from what I have referred to as the default or ‘syntactic’ LVs -jarri and
mani. LVs such as yani and pinyi are lexically selected in that they may only
form PV-V (CAUSE) CHANGE STATE expressions with a restricted class of PV –
including those expressing ‘impaired integrity’ predicates. The same goes
for -ngarni and –jirrirni, whose selection is tied to the Aktionsart features of
the PV predicate they combine with.

.. Verb as an Aktionsart head

Although rdilyki nominals in (a,b), repeated as (a,b), presuppose a
change of state, they only operate as eventive change of state predicates if
incorporated as PV-V complex as in (a-d).
 Warlpiri verbs of change and causation: the thematic core 217

() a. Wati-ngki ka kuturu paka-rni rdilyki.

man-ERG AUX:PRS club chop-NPST broken
‘The man is chopping/striking a club (that is) broken.’ / ‘[…] a
broken club.’
≠ ‘The man is breaking a club (by) striking/chopping it.’
b. Wati-ngki ka kuturu paka-rni rdilyki=lki.
man-ERG AUX:PRS club chop-NPST broken=CS
‘The man is chopping/striking a club (that is) now broken.’
≠ ‘The man is now breaking a club (by) striking/chopping it.’
The nominal predicate rdilyki ‘broken’ in () is predicated directly of the NP
kuturu ‘club’ in a syntactic relationship marked by case agreement (unmarked
absolutive). It refers to the state of the club quite independently of the striking
event expressed by the verb paka-rni. The ‘striking’ event can be interpreted
as a telic achievement (single blow to club) or as a non-telic activity (multiple
blows to club). The accomplishment reading entailing the change of state (from
not broken to broken) is expressed when a rdilyki predicate combines with an
appropriate verb to form a PV-V constituent. In order to be in the scope of the
appropriate Aktionsart features, rdilyki must combine with a dynamic verb, be
it a LV like yani or pinyi (discussed in section ..) that expresses no infor-
mation about the manner of the causing event, or a HV such as pakarni in ()
which does. We have already seen in section .. that in these PV-V com-
plexes only dynamic predicates of the rdilyki class are selected as PV by these
verb types. Thus there appears to be a compatibility requirement between
the intrinsic Aktionsart features of the PV and the V it combines with. The
dynamic meaning of externally caused change of state that is lexically encoded
by a rdilyki class predicate is only actualized if it is placed in a larger constitu-
ent where it is in the scope of the appropriate Aktionsart features. This larger
constituent contains either a verbal or a postpositional head such as -warnu
discussed in section ... (see examples in () and ()).
These facts raise the question as to whether V (or v) heads in syntactic
structure may be reduced to Aktionsart heads in the sense that only the
aspectual features associated with them are relevant or visible to the syntac-
tic component. A related issue is whether these verbal heads are governed by
separate Aktionsart heads associated with features which must ‘agree’ or at
least be compatible. I will not pursue these issue here. Clearly the relationship
between PV and V is mediatedby Aktionsart features.

.. Form–meaning combinations in inchoative and causative

PV-V constituents
Some of the facts presented in this section provide evidence for a model of
language in which the event structure which defines the thematic functions
218 Laughren

which determine the PAS properties of a predicate associated with a lex-

ical form is distinct from the syntactic structure into which the predicate is
mapped. We have examined a collection of semi-productive PV-V complex
verbs expressing CHANGE and CAUSE CHANGE functions with varying degrees
of semantic transparency, as well as the more productive ‘default-like’ pat-
terns. The Warlpiri data reveal that the relationship between which verb
forms are used in combination with which PVs to form the inner the-
matic core of a complex verbal constituent is not random. While some
semi-productive patterns might perhaps be usefully thought of synchronic-
ally as idiomatic collocations, or collocation patterns, we have seen that other
patterns correlate with identifiable features. The set of verbs which license
the syntactic expression of an external ‘agent’ are distinct from those which
do not; verbs which actualise the CHANGE and CAUSE functions presupposed
by a dynamic predicate of the rdilyki class differ from those which add these
components of event structure to a stative predicate of the yurnmi/ngurrju/
pukulyu class. Verbs which express change into a [–dur] transitory state dif-
fer from those which express change into a [+dur] state.
While factors derived from the pre-history of certain forms is undoubt-
edly relevant to the mapping between phonological form and pieces of syn-
tactic structure which are mapped onto a particular event structure, some
consistent patterns of mapping to phonological form have been revealed. We
have seen four semantically ‘light’ verbs which mark the syntactic projection
of an external causer participant as the subject of the complex verb: -pinyi
with ‘impaired integrity’ PVs (although it is not limited to this class of PV);
-yirrarni with INF/PV which express an change of stance or spatial config-
uration; jirrirni with a PV expressing a momentary change of psychological
state which may involve internal physical change (kinyirr-jirrirni ‘startle’,
miil-jirrirni ‘shock’); -mani which combines with a PV or XP predicate to
express an externally caused change of state, activity, or location.
Verbs which do not express an external cause of change show more vari-
ation. An INTERNAL CAUSE of CHANGE predicate takes a number of forms. A
simplex HV (karri, nyina, nguna, parntarri) may express internally caused
maintenance or change of stance (nyina, nguna, parntarri). Internally con-
trolled maintenance or change of stance may also be expressed by a PV-LV
complex (wardu-karri, tarda-yani, pirri-mani). Internally caused change
of stance is also expressed by a complex VINF-LV predicate (karrinja-pardi,
ngunanja-wanti, parntarrinja-wanti). We have seen that as independent
verbs, neither the directed motion verbs pardi ‘rise’ and wanti ‘fall’ nor the
stance verb karri ‘stand’ may express an internally caused change of loca-
tion. Furthermore, independent verbs which express internally caused

Patterns akin to the use of English ‘drive’ as a syntactic causative verb which only com-
bines with impaired behavioural or cognitive state phrases: drive mad/silly/crazy/to
drink/insane/out of one’s mind/bonkers/nuts versus *sad/*happy/*charming/*calm/*bad
/*good/*sick/*sane (Richard Carter, personal communication ).
 Warlpiri verbs of change and causation: the thematic core 219

change of location (warrkarni ‘ascend’, jitirni ‘descend’) which correspond

to non-internally controlled verbs ( pardi ‘rise’ and wanti ‘fall’), are not used
with a stance VINF to create an internally caused change of stance complex
verb. Where the internally caused change of stance is expressed by a PV-V
combination, the PV expresses the resultant stance while the verb expresses
the change. The choice of directed motion verbs such as pardi and wanti,
rather than the corresponding control verbs warrkarni and jiti, in verb
complexes which express a self-initiated change of stance, suggests that
it is not possible to encode one internal cause within another another (*x
causes self to cause self to change). Such constraints cannot be semantic,
they must be syntactic. This does not necessarily provide evidence for the
position that the event structure of the type in Figures .–. is constructed
in the syntax (cf. Megerdoomian (), Folli, Harley, and Karimi ()),
although it would be possible to translate our event structures into syntactic
trees by substituting syntactic categories for the semantic labels I have used.
However, I have tried to show in this section that identical components of an
event structure can be realised by different syntactic projections, e.g. NP or
PV may express identical predicates, and a change event may be expressed
by a V or PP form. However, the aspectual features which mediate between
a predicate and its argument determine the possible semantic–syntactic
pairings.
Contrasting aspectual features of PV predicates are also overtly marked
by the choice of V in complex thematic core structures. Whereas PV-ya-
expresses a CHANGE predicate which presupposes an unexpressed external
cause (rdilyki/larra/rdiirr/rdilypirr-yani) where the resultant state is ongoing
(+DUR), it is expressed by PV-ngarni where the resultant state is not ongoing
(–DUR), but only momentary ( julurr-/kinyirr-ngarni). The CHANGE event
expressed by PV-jarri can be seen as a default option – it applies where other
marked lexical options fail to apply. However its PV must be derived from a
phrasal category such as NP or PP.

. Morphosyntactic behaviour of Warlpiri complex verbs

A verbal constituent, no matter how complex, can behave as a single syntactic
constituent as tested by its ability to precede the Auxiliary (AUX) complex
(Hale , , Laughren , Nash , Simpson ), as shown in (a)
where the quantificational PV yarda ‘more, again’ and the complex thematic
core rdilypirr-panturnu ‘pierce hole in’ precede the AUX. It may ‘split’ into
two constituents, one preposed to AUX and the other immediately postposed


Semantically equivalent English verbs startle, frighten, scare are typically transitive; the
external causer of the experience is Subject of the active voice form while the experiencer
is Object. The passive construction must be used to suppress the obligatory projection into
the syntactic representation of the ‘causer’ argument: *He startled ≠ he was startled.
220 Laughren

to AUX, as in (b) (Laughren () refers to this as auxiliary stranding).

In (b), the quantification PV yarda ‘more, further, again’ precedes AUX,
while the complex thematic verb wantiki-ma-nta follows AUX.
() a. Yarda rdilypirr-pantu-rnu=lpa=lu.
more hole-pierce-PST=AUX:IMPF=PL.S
‘They were piercing holes (in it) again.’
b. Yarda=rla wantiki-ma-nta.
more=AUX:DAT wide-CAUSE-IMP
‘Widen it more (to get to it).’
As mentioned in section ., while outer PVs such as yarda may also follow
AUX, leaving the rest of the verbal complex in preAUX position, a thematic
core PV-LV combination may also be ‘split’ by AUX, but the order must be
maintained, so that PV always precedes LV. Also discussed in section . was
the ability of directional enclitics (DIR) to intervene between the PV and V,
as in (a). In this case, the PV=DIR constituent is expressed as a phono-
logical word distinct from the following verb word (Pentland and Laughren
). PV=DIR may also be followed by AUX, which is then followed by the
inflected V as in (b).
() a. [lawa]=rra ma-ni ~ [lawa-ma-ni]=rra
nothing= DIR CAUS-NPST

‘cause to not be present, make (it) go away, disperse, empty, get rid of …’
b. [pirda]=rra=lpa jarri-nja-ya-nu ~ [[pirda-[jarri-nja]]-ya-nu]=rra=lpa
sated=DIR=AUX:IMPF INCH-INF-PATH-PST [pirda-[jarri- nja]]=rra [ya-nu] =lpa
‘(he) was getting full’ [pirda- [jarri-nja]]=rra=lpa [ya-nu]

The independence between the PV and the thematic V (underlined) in ()

with respect to the placement of DIR and/or AUX was argued by Laughren
() to derive from the ability of the PV to occupy a high specifier position
in the phrase projected by an inflectional head (tense, mood) as a result of
phrasal movement from a lower PREDP position within the VC (as in Figure
.). It was argued that the PV (or other phrasal constituents) may move


If yarda in (b) were an adverbial phrase independent of the verb wantiki-manta, it would
have to be case-marked with Ergative, i.e. *yarda-ngku. The construction in (b) can
be compared with one using an adverbial such as maya ‘more’ in Maya-ngku=rla wan-
tiki-manta. ‘Keep on widening it to (get) it’. It is not possible to have both mayangku and
the verb preposed to AUX, unlike PV yarda and the V panturnu in (a), *Maya-ngku
panturnu=lpa=lu. Either mayangku or panturnu can be preposed to AUX =lpa=lu, but
not both.

In the Warlpiri data corpus, clause-initial yarda is overwhelmingly found in the AUX-
straddling construction with AUX immediately following yarda and preceding the rest of
the verbal complex.
 Warlpiri verbs of change and causation: the thematic core 221

up into the Spec of DIRP, or alternatively, that the whole verbal complex
(Figure .) moves into this position. The variable placement of DIR = rra
as in () provides a useful diagnostic of surface syntactic constituent struc-
ture. A satisfactory account of the syntactic properties of complex verbs
must explain their single constituent-like behaviour, while at the same time
allowing for the limited phrasal movement of a PV (or thematic VC con-
stituent) relative to other components of the verbal complex, especially in
finite clauses. (See Legate () for an alternative analysis.)

.. VINF-V verbal constituent

For the less productive VINF-V or PV-V thematic core constituents we might
expect to not see the syntactic freedom displayed in () and (). However,
the data do not bear out this expectation, at least not in an obvious way. In
() we see VINF-V constituents in which a DIR enclitic separates the VINF
from the V. In () VINF is separated from V by the AUX.
() a. karri-nja]=rni pardi ‘stand up hither’’
stand-INF=DIR rise:NPST
b. karri-nja]=rni yirra-rni ‘put in a standing position hither’
stand-INF=DIR put-NPST
() a. karri-nja]=lpa pardi-ja ‘(he) was standing up’
stand-INF=AUX:IMPF rise-PST
b. karri-nja]=ka pardi ‘(he) is standing up’
stand-INF=AUX:PRS rise:NPST
Notice that this syntactic behaviour is what we find when a thematic verb is
in the scope of the aspectual PATH verb yani, or PURPOSEFUL DISPLACEMENT
verb parnka. The preceding (thematic) V must be in the VINF form and it
may constitute an independent phrase which hosts DIR and/or it may be
preposed to AUX. Examples of the separation of the thematic structure from
the aspectual are given in ().
() a. [pirda-jarri-nja]=rra ya-nu =lpa
sated-INCH-INF=THITHER PATH-PST =AUX:IMPF

‘He was getting full.’

b. [pirda-jarri-nja]=rra=lpa ya-nu
sated-INCH-INF=THITHER=AUX:IMPF PATH-PST

‘He was getting full.’

c. Ngapiri ka [karri-nja]=rra ya-ni karru-ngka.
Red.Gum AUX:PRS stand-INF=THITHER PATH-NPST creek-LOC
‘The Red Gums are standing/growing along the creek.’
222 Laughren
d. [Nya-nja]=rni=lpa ya-nu.
look-INF=HITHER=AUX:IMPF PATH-PST

‘(He) was looking as he came.’

e. [Nya-nja]=rni parnka ka=lu.
look-INF=HITHER PURPD:NPST AUX:PRS =PL.S

‘They are coming to see/visit (him).’

These observations lead us then to ask whether the verbs such as pardi in
(a) and (a,b) or even yirra- in (b) actually belong to the thematic core
as we have defined it, or whether they constitute another type of aspect-
ual verb such as ya- in (d) and parnka in (e). Unlike these aspectual
verbs, thematic verbs such as pardi in (a) and () do not combine freely
with preverbal infinitive verbs as do the PATH verb ya- or the PURPD ‘go to’
verb parnka. We have seen that the VINF=V combinations of the karrinja-
pardi type are extremely constrained. As we have already noticed, pardi in
karrinja-pardi ‘stand up’ contrasts with wanti in ngunanja-wanti ‘lie down’
in its spatial entailment – part of body of X rises. It also determines the
PAS of the complex verb while aspectual verbs play no role in determining
argument structure, which is why nyanja ‘see-INF’ in (d, e) maintains its
dyadic status.

.. Lexical PV-V constituent

So what about the inner thematic core verbal constituent with an underived
lexical PV (in the sense of Nash )? Does the PV behave like an XP which
can separate from the thematic verb or is it more tightly related to the verb
as in a compound constituent? In the barely productive PV-V collocation
pirri-mani ‘sit, land, perch’, the PV can be separated by a directional enclitic
(a, b) and may be further separated from the verb by the auxiliary com-
plex, as in (b) in which AUX ka=lu intervenes between the PV pirri and
the V ma-ni.
() a. pirri]=rni ma-ni ‘sitting (facing/coming) hither’
seated=DIR V-NPST
b. pirri]=rra ka=lu ma-ni ‘they are sitting thither’
seated=DIR AUX=PL.S V-NPST

Like the VINF-V cases, which are only semi-productive at best, these PV-V
combinations are visible to the S-Syntax. Like VINF, the PV may have some
very limited syntactic independence by optionally moving into the specifier
of DIRP and/or moving into the pre-AUX position.
‘Impaired integrity’ denominal PVs of the rdilyki class may also be sepa-
rated from the verb by DIR and aspectual enclitics, as shown in (a), and
also by AUX as in (b).
 Warlpiri verbs of change and causation: the thematic core 223

() a. rdilyki]=rra=lku ya-ni

broken=DIR= CS V-NPST
‘now breaks off’
b. Rdilyki-rdilyki]=lpa=lu pu-ngu.
broken-broken=AUX:IMPF =PL.S V-PST
‘They were breaking it (into pieces).’
What we have seen so far is that the thematic core may separate from the
following aspectual verb as in (), or PVs may separate from the rest of the
verbal complex, including inner thematic core PVs which may separate from
their thematic V as in () and (). In section .. we see another PV-V
construction which does not permit movement of the PV away from the V,
and in which the PV appears to behave like a Xº category rather than an XP.
The examples of AUX-straddling by components of a verbal complex
reviewed above differ from examples of discontinuous case phrases (KP) as
in (a).
() a. Kuyu-ku ka=rla marlu-ku yura-ka-nyi.
animal-DAT AUX:PRS=DAT kangaroo-DAT sneaking-go-NPST
‘He is sneaking up on the (animal) kangaroo.’
b. Kuyu marlu-ku ka=rla yura-ka-nyi.
animal kangaroo-DAT AUX:PRS=DAT sneaking-go-NPST
‘He is sneaking up on the (animal) kangaroo.’
c. *Kuyui ka=rla marlui-ku yura-ka-nyi.
Although it is possible to have two contiguous nominals within the scope of
the case on the final member as in (b), it is not possible to have the uncase-
marked nominal in pre-AUX position while the case-marked nominal follows
AUX as shown in (c). The PV elements which precede AUX followed by
the remainder of the verbal constituent are syntactically licensed whereas
uncase-marked nominals are not. AUX-straddling by verbal constituents
is only licensed where they are in the scope of the finite verbal and clause
level inflectional categories, which may be why, once COMP is filled, AUX-
straddling is no longer valid (Laughren ). Both the discontinuous KP
construction in (a) and AUX-straddling by verbal constituents are syntactic
processes and are not in any way governed by prosodic factors.

.. Limits on PV independence

Many Warlpiri PVs have a consonant final root form as well as a vowel final
augmented form derived from the consonant final root by the addition of pa

(c) is only grammatical on a reading where kuyu and marlu refer to different individuals,
e.g. an animal is sneaking up on the kangaroo.
224 Laughren

or in some cases ki/ku. The consonant final PV cannot be separated from

the V with which it forms a complex constituent. This PVroot-V combination
has the prosodic properties of a word–word compound in which the PV and
V maintain their underlying prosodic word structure, i.e. the PV has a min-
imum of two vocalic morae (CVVC, CVCVC), both PV and V are stress-
bearing, but the PV lacks the requirement of an independent word to end in
a vowel. Another phonological requirement of Warlpiri words is that they are
consonant-initial. However where the initial consonant of the V is a glide y
or w, it is typically elided so that the final consonant of the PV is resyllabi-
fied to form the onset of the verbal word. Some relevant examples are given
in ().
() a. wuruly-yani [wu-ru-lya-ni] a′. wurulypa=rra yani
hidden-go
b. nyunjurl-yani [nyu-nju-rla-ni] b′. nyunjurlpa=rni yani
curse-go
c. rdiirr-wanti [rdii-rran-ti] c′. rdiirrpa=rra wanti
busted-fall
a″. *wuruly=rra yani
b″. *nyunjurl=rni yani
c″. *rdiirr=rra wanti
It would seem as though the speaker has two options: to use a phrasal or
non-phrasal root PV. Even PVs with a unique vowel final form may be
used in the ‘compound’ construction or in the ‘phrasal’ construction. The
speaker’s choice may be determined by pragmatic requirements; certainly
the preverbal element can be focused and fronted in the phrasal construc-
tion in a way that is not possible within the compound structure. It may be,
of course, that the PV in the phrasal construction always moves out of the
inner thematic core in order to be in the scope of some focus operator, and
that for some PVs there is a distinct in focus as opposed to out of focus phono-
logical form. While the PV-V constituents in (aʺ–cʺ) are phonologically
ill-formed since illicit consonant clusters result from the encliticisation of
the DIR clitics, the motivation for the dual PV forms is not phonological.
The motivation is syntactic since only the vowel final phrasal forms may
precede AUX in AUX-stranding constructions, irrespective of the conson-
ant clusters which might result from the sequences formed by the PV-final
C and the AUX-initial C. The vowel final form of many PVs may also be
categorised as nominal and may constitute an NP independent of a V, as
we saw to be the case with rdilyki in section .. The consonant final form,
on the other hand, is not phrasal. The non-phrasal PV may result from the
incorporation of the PV head (in its root form) of the PREDP in the stative
structure shown in Figure ., into the head of the higher verbal predicate
 Warlpiri verbs of change and causation: the thematic core 225

associated with a CHANGE or CAUSE function, whereas the phrasal PV results

from the movement of the lowest PREDP into a specifier position within the
inflected verbal phrase. Head movement involves the root form of the pre-
verbal constituent, while movement from complement to specifier involves a
phrasal PV category.
While the various ‘motion’ or ‘stance’ verbs which combine with a PV to
form CHANGE stance/location/state verbal constituents may combine with
the root form, the productive intransitive verbaliser jarri only combines
with augmented non-root PVs. One might expect this to be also true of the
productive causative mani. However mani does combine with root forms,
e.g. karaly-mani ‘make smooth’, laarr-mani ‘split/section’, minyminy-mani
‘dampen’, pinpin-mani ‘make thinner’, rdangkarl-mani ‘shorten’, rdily-
pirr-mani ‘make hole in, perforate’, yurnuly-mani ‘muddy’, rarraly-mani
‘straighten, smoothe out’, etc. This structural contrast between jarri and all
other verbs, including causative -mani, calls for further investigation.
As shown by the distribution of rdilypirr(pa) ‘hole, with hole(s)’ in (b,
c), only the augmented nominal form may be incorporated into a karda
phrase (*rdilypirr-karda), the root form being limited to the PV function.
In neither the pre jarri or karda environment is the ungrammaticality of
the consonant final root PV due to phonological constraints, as sequences
of apico-alveolar tap or trill rr (or most lateral consonants) followed by j or k
are permitted.

.. Syntactic or lexical causatives

A distinction has long been made between a syntactic causative construction,
as in the English periphrastic John made Paul (feel) sad, and a lexical causa-
tive expressed within one morphological word, as in John saddened Paul.
Carter () noted that while it is possible to embed a causative clause
inside a syntactic causative, as in Caesar made the consul make the general
make the soldiers sack the city, no single verb encodes a CAUSE (CAUSE) rela-
tion where both causer arguments would be external causers, as in *Caesar
sacked the general the city, meaning that ‘Caesar made/had the general sack
the city’. Only a single external cause predicate may be lexically encoded,
or more precisely expressed by a monoclausal constituent. The Warlpiri
causative PV-V constituents may be expressed as a tightly bound compound
word or they may also be expressed as a combination of morphophonological


A causative form rdilypirr-(y)inyi is also used; HV yinyi ‘give’.

While I have classified both dependent -ma- and -jarri as ‘syntactic’ verbs, it may be the
case that only -jarri has this property, -ma- being a lexical verb.

Syntactic causative constructions are not necessarily periphrastic in the sense of being con-
structed out of several morphophonological words as in English. Languages may encode
causative constructions in a single morphologically complex phonological word (see Travis
(b) for Malagasy and Tagalog examples).
226 Laughren

words, in what was referred to in section .. as the phrasal construction.

Like lexical causatives, the Warlpiri PV-V or V verbal constituent can-
not embed an externally caused predicate inside another externally caused
predicate, e.g. *karrinja-yirrarninja-yirrarni/*karrinja-yirrarninja-mani
(≠to make to stand something up) or *rdilyki-pinja-mani (≠to make x break
y). As noted in section ., unlike causative yirra-, causative ma- cannot
have an uninflected VINF in its preverbal position. However, as observed by
Nash (), yirra- can not host a transitive VINF, i.e. one with an external
causer argument. The Warlpiri thematic core, be it a single morphophono-
logical word or several, can only express a single external CAUSE predicate
conforming to the semantic representation in Figure .. This constituent is
monoclausal. Even where a causative event schema is expressed by a simplex
verbal constituent, e.g. maja-rni ‘straighten, stretch’, mirri-rni ‘erase’, purra-
‘burn, cook’, it is not possible to embed this verb as a VINF preverb within a
thematic core headed by a verb which expresses a CAUSE predicate. Thus it
is not a constraint on morphological complexity which excludes the possibil-
ity of having an external cause VINF combine as PV with an external cause
V, rather the constraint derives from interface constraints in mapping event
structure onto syntactic categories.

. Summary of findings and concluding remarks

This study has distinguished a number of different verb types: thematic,
aspectual, independent, and dependent. Thematic verbs can have an HV use
when used as the sole exponent of the inner thematic core and also when
used in combination with a PV where the V expresses a causing event, while
the PV expresses the changed (or maintained) state, stance, or location.
Thematic verbs can also be used as LVs in combination with a PV, such that
their semantic contribution to the complex predicate is underspecified by
comparison with their HV use, while their role in licensing the syntactic pro-
jection of arguments defined by the event structure remains constant. The
inherently dependent verbs – those that require a PV – are either aspectual
(e.g.-njini) or if thematic, then ‘light’ verbs whose selection is determined by
Aktionsart properties of the PV predicate and PAS properties of the complex
predicate. Only simplex stance verbs (i.e. without a PV) may operate as sta-
tive predicates. Only this class of verb may also operate as a copula; in this
role I have argued that the verb does not map onto the predicative PredP
component within the event structure as shown in Figure ., but rather
onto the AKTP which dominates an empty PredP position linked to an overt

These simplex external cause verbs have no cognate internal cause counterparts; internal
causation can only be expressed by a syntactic reflexive construction, e.g. Jakamarra-rlu
maja-rnu kurlarda (Jakamarra- ERG straighten-PST spear: ABS) ‘Jakamarra straightened a
spear’, vs. Jakamarra-rlu=nyanu maja-rnu (J-ERG =ANAPH straighten-PST) ‘Jakamarra
stretched/straightened up’.
 Warlpiri verbs of change and causation: the thematic core 227

NP predicate via case-marking as in Figure .. While additional research is

required to pinpoint the nature of the default verbalisers jarri and -mani, I
have argued that these are also Aktionsart verbs in that they mark a relationship
between an Aktionsart head in Event Structure and a V head position in syn-
tactic structure. I have classified them as non-lexical syntactic verbs. They are
true ‘light’ verbs in that they do not specify any properties of a CHANGE or
CAUSE CHANGE event. I have shown that in inner thematic core PV-V combina-
tions there is a spectrum of ‘verbal weight’ measured in semantic terms. The
dependent verbs are at the extreme ‘light’ end of this spectrum.
The Warlpiri verbal constituent is underlyingly complex in its morpho-
logical, syntactic, and semantic structure. The verbal constituent is a phrasal
complex consisting of a thematic core which must contain a verb and may
additionally contain a preverbal constituent (which may be morphologically
complex). The thematic structure may be augmented by an outer periphery
consisting of ‘outer’ preverbal constituents which modify core elements in a
variety of ways not addressed in this study. The thematic structure compo-
nent of the verbal complex is in the scope of aspectual verbs, and in a finite
clause, inflectional tense and mood categories realised as suffixes on the
rightmost verb. The inflected constituent is in the scope of the deictic DIR
category, realised as an enclitic to some phrasal component of the inflected
verbal constituent.
This study focused on the composition of the thematic core and par-
ticularly on stance and state predicates, achievement predicates expressing
change of stance, location, and state, and accomplishments with a causal
event component. It is evident that there is no simple correspondence
between the morphological form and the event structure of the inner the-
matic core of a verbal complex. Synonymous predicates may be expressed
by simplex V- or complex PV-V forms, e.g. wanti- and pata-karri ‘fall’. The
same verb form may map onto different event structures (or components of
event structure) with different Aktionsart properties, as illustrated by certain
stance verbs, both simplex and complex. Homophonous verbs may have dif-
ferent semantic as well as morphosyntactic properties depending on whether
they are dependent or independent, e.g. yani as a thematic independent verb
(GO(x)); -yani as a thematic dependent verb (CHANGE STATE (x)) where the
state is necessarily conceived as the result of some external cause change;
-yani as an aspectual PATH verb which has scope over the situation or event
expressed by the thematic core.
In VINF/PV-V structures, the CAUSE and CHANGE functions are expressed
by a number of different verbs. While there are some idiosyncratic or seem-
ing exceptional combinations, the choice of verb was shown to be determined
by systematic factors. These factors pertain to both the event structure
underpinning the PAS and Aktionsart properties of the verbal constituent,
and to the manner in which these event components are mapped onto the
syntactic structure.
228 Laughren

This study represents a still preliminary exploration of Warlpiri verbal

predicates. Some questions that have been raised remain to be answered,
such as the exceptional character of the verbaliser -jarri in not allowing root
PVs. However, it indicates that further investigation of verbal predicates
would shed more light on the nature of lexical organisation and particu-
larly on the interface between the meaning of predicates and the morpho-
logical and syntactic structure of their formal expression. The Warlpiri data
examined in this study reveal some striking similarities with thematic PV-V
structures documented in a range of languages such as Modern Persian
with respect to the role of these constituents in the aspectual and thematic
structure of complex verbs. We have also seen that the universally found dis-
tinction between result of event ‘broken’ type predicates and non-eventive
‘good’ or ‘ripe’ predicates is also attested in Warlpiri, even though both are
expressed by nominal categories. Their different underlying event struc-
tures – which include Aktionsart features – determine the syntactic verbal
structures in which they may appear.

References
Alpher, B., Evans, N., and Harvey, M. . Proto-Gunwinyguan Verb Suffixes.
In Evans, N. (ed.) The Non-Pama-Ngungan Languages of Northern Australia:
Comparative Studies of the Continent’s Most Linguistically Complex Region,
–. Canberra: Pacific Linguistics.
Arad, M. . VP-Structure and the Syntax-Lexicon Interface. PhD dissertion,
University College London. (Distributed by MITWPL, MIT Occasional
Papers in Linguistics .)
Borer, H. . The Projection of Arguments. In Benedicto, E. and Runner, J.
(eds) Functional Projections. University of Massachusetts Occasional Papers
: –. Amherst, MA: University of Massachusetts.
Carter, R. . Some Constraints on Possible Words. Semantikos : –.
Carlson, G. . A Unified Analysis of the English Bare Plural. Linguistics and
Philosophy : –.
Dehé, N., Jackendoff, R., McIntyre, A., and Urban, S. (eds) . Verb-particle
Explorations. Berlin; New York: Mouton de Gruyter.
Dixon, R. M. W. . Copula Clauses in Australian Languages: A Typological
Perspective. Anthropological Linguistics , : –.
Folli, R., Harley, H., and Karimi, S. . Determinants of Event Structure in
Persian Complex Predicates. Lingua : –.
Ghomeshi, J. and Massam, D. . Lexical/syntactic relations without projection.
Linguistic Analysis : –.
Glass, A. and Hackett, D. . Ngaanyatjarra and Ngaatjatjarra Dictionary. Alice
Springs, NT: IAD Press.
Goldberg, A. . Optimizing constraints and the Persian complex predicate.
Annual Proceedings of the Berkeley Linguistic Society : –.
Green, R. . Proto Maningrida within Proto Arnhem: Evidence From Verbal
Inflectional Suffixes. In Evans, N. (ed.) The Non-Pama-Ngungan Languages of
 Warlpiri verbs of change and causation: the thematic core 229

Northern Australia: Comparative Studies of the Continent’s Most Linguistically

Complex Region, –. Canberra: Pacific Linguistics.
Hale, K. L. . Person marking in Walbiri. In Anderson, S. R. and Kiparsky,
P. (eds) Festschrift for Morris Halle, –. New York: Holt, Rinehart &
Winston, Inc.
. Some Essential Features of Warlpiri Verbal Clauses. In Swartz, S. (ed.)
Papers in Warlpiri Grammar: In Memory of Lothar Jagst. Work Papers of SIL-
AAB, Series A vol. , –. Berrimah, NT: SIL-AAB.
. Warlpiri and the Grammar of Non-configurational Languages. Natural
Language & Linguistic Theory , : –.
Hale, K. L. and Keyser S. J. . On Argument Structure and the Lexical
Expression of Syntactic Relations. In Hale, K. and Keyser, S. (eds) The View
From Building , –. Cambridge, MA: MIT Press.
. Prolegomenon to a Theory of Argument Structure. Cambridge MA: MIT
Press.
Hale, K. L., Laughren, M., and Simpson, J. . Warlpiri Syntax. Section .
In Jacobs, J., von Stechow, A., Wolfgang Sternefeld, S., and Vennemann,
T. (eds) Syntax. Ein internationales Handbuch zeitgenössischer Forschung /
An International Handbook of Contemporary Research. Halbd. , –.
(Handbücher zur Sprach- und Kommunikationswissenschaft series.) Berlin;
New York: Walter de Gruyter.
Hansen, K. and Hansen, L. . Pintupi/Luritja Dictionary, rd edn. Alice
Springs, NT: I.A.D. Press.
Harvey, M. and Baker, B. . Vowel Harmony, Directionality and Morpheme
Structure Constraints in Warlpiri. Lingua : –.
Hoekstra, T. . Aspect and Theta Theory. In Roca, I. (ed.) Thematic Structure: Its
Role in Grammar, –. Berlin; New York: Foris Publications.
Jackendoff, R. . Semantic Structures. Cambridge, MA: MIT Press.
. Parts and Boundaries. Cognition :  –.
Jelinek, E. . Empty Categories, Case, and Configurationality. Natural Language
and Linguistic Theory :  –.
Karimi-Doostan, M-R. . Light Verb Constructions in Persian. PhD dissertation,
University of Essex.
Kratzer, A. . Stage-level and Individual-level Predicates. In Carlson, G. and
Pelletier, F. (eds) The Generic Book, –. Chicago: University of Chicago
Press.
Laughren, M. . Towards a Lexical Representation of Warlpiri Verbs. In Wilkins,
W. (ed.) Thematic Relations, –. New York: Academic Press.
. Secondary Predication as a Diagnostic of Underlying Structure in Pama-
Nyungan languages. In Roca, I. (ed.) Thematic Structure: Its Role in Grammar,
–. Berlin; New York: Foris Publications.
. What Warlpiri ‘Avoidance’ Registers do with Grammar. In Simpson,
J., Nash, D., Laughren, M., Austin, P., and Alpher, B. (eds) Forty Years
On: Ken Hale and Australian Languages, – (Pacific Linguistics ).
Canberra: Australian National University.
. Syntactic Constraints in a ‘Free Word Order’ language. In Amberber, M.
and Collins, P. (eds) Language Universals and Variation, –. Westport,
CT: Praeger Publishers.
230 Laughren

Laughren, M. and Warlpiri Lexicography Group. . Warlpiri-English Encyclo-

paedic Dictionary. Electronic files, The University of Queensland.
Legate, J. A. . Warlpiri: Theoretical implications. PhD dissertation, MIT,
Cambridge, MA. (MIT Working Papers in Linguistics).
. The Morpho-semantics of Warlpiri Counterfactual Conditionals. Linguistic
Inquiry : –.
. Warlpiri and the Thoery of Second Position Clitics. Natural Language and
Linguistic Theory : –.
Levin, B. . On the Nature of Ergativity. PhD dissertation, MIT, Cambridge,
MA. (MIT Working Papers in Linguistics).
. English Verb Classes and Alternations: A Preliminary Investigation. Chicago:
University of Chicago Press.
Levin, B. and Rappaport, M. . What to Do With θ-roles. In Wilkins, W. (ed.)
Thematic Relations, –. New York: Academic Press.
Levin, B. and Rappaport Hovav, M. . Wiping the Slate Clean: A Lexical
Semantic Exploration. Cognition : –.
. A Preliminary Analysis of Causative Verbs in English. Lingua : –.
. Unaccusativity. Cambridge, MA: MIT Press.
. Argument Realization. Cambridge: Cambridge Unviersity Press.
McConvell, P. and Laughren, M. . The Ngumpin-Yapa Subgroup. In Bowern, C.
and Koch, H. (eds) Australian Languages: Classification and the Comparative
Method, – (Current Issues in Linguistic Theory). Amsterdam; Philadelphia:
John Benjamins Publishing Company.
Megerdoomian, K. . Beyond Words and Phrases: A Unified Theory of Predicate
Composition. PhD dissertation, University of Southern California.
Nash, D. . Preliminary Vocabulary of the Warlmanpa Language. TS. pp.,
Department of Linguistics and Philosophy, MIT, Cambridge MA.
. Warlpiri Preverbs and Verb Roots. In Swartz, S. (ed.) Papers in Warlpiri
Grammar: In Memory of Lothar Jagst. Work Papers of SIL-AAB, Series A vol.
, –. Berrimah, NT: SIL-AAB.
. Topics in Warlpiri Grammar (Outstanding Dissertations in Linguistics).
New York; London: Garland Publishing Inc.
. Warlpiri Verb Roots in Comparative Perspective. In Bowern, C., Evans, B.,
and Miceli, L. (eds) Morphology and Language History: In honour of Harold
Koch, (Current Issues in Linguistic Theory ), –. Amsterdam: John
Benjamins Publishing Co.
Pentland, C. and Laughren, M. . Distinguishing prosodic word and phono-
logical word in Warlpiri: prosodic constituency in morphologically complex
words. In Mushin, I. (ed.) Proceedings of the  conference of the Australian
Linguistic Society <http://ses.library.vsyd.edu.au/handle//>.
Pustejovsky, J. . The Syntax of Event Structure. Cognition : –.
. Events and the semantics of opposition. In Tenny, C. & Pustejovsky, J.
(eds), a, –.
Pylkkänen, L. . Introducing Arguments. PhD dissertation, MIT, Cambridge, MA.
Ramchand, G. . Aspect and Predication: The Semantics of Argument Structure.
Oxford: Clarenden Press.
 Warlpiri verbs of change and causation: the thematic core 231

Riemer, N. . Verbal Polysemy and the Vocabulary of Percussion and Impact in
Central Australia. Australian Journal of Linguistics : –.
. Les nominaux dans les complexes verbaux du warlpiri. Actances (publication
of CNRS research group Rivaldi-GDR : Relations intercatégorielles: les
variations aspecto-temporelles et les structures diathétiques) : –.
. The Semantics of Polysemy: Reading Meaning in English and Warlpiri
(Cognitive Linguistics Research, ). Berlin; New York: Mouton de Gruyter.
Roca, I. M. (ed.). . Thematic Structure: Its Role in Grammar (Linguistics mod-
els series ). Berlin; New York: Foris/Walter de Gruyter.
Simpson, J. . Warlpiri Morpho-Syntax. A Lexicalist Approach. (Studies in
Natural Language and Linguistic Theory Volume ) Dordrecht; Boston;
London: Kluwer Academic Publishers.
. From Common Ground to Syntactic Construction: Associated Path in
Warlpiri. In Enfield, N. J. (ed.) Ethnosyntax: Explorations in Grammar and
Culture, –. Oxford: Oxford University Press.
Simpson, J. and Bresnan, J. . Control and Obviation in Warlpiri. Natural
Language & Linguistic Theory .:  –.
Smith, C. . The Parameter of Aspect. Dordrecht: Kluwer.
Talmy, L. . Lexicalization Patterns: Semantic Structure in Lexical Forms. In
Shopen, T. (ed.) Language Typology and Syntactic Description : Grammatical
Categories and the Lexicon, –. Cambridge: Cambridge University Press.
Tenny, C. . Grammaticalizing Aspect and Affectedness. PhD dissertation, MIT,
Cambridge, MA.
. The Aspectual Interface Hypothesis. In Sag, I. and Szabolcsi, A. (eds)
Lexical Matters, –. Stanford, CA: CSLI Publications.
Tenny, C. and Pustejovsky, J. (eds) a. Events as Grammatical Objects: The
Converging Perspectives of Lexical Semantics and Syntax. Stanford, CA: CSLI
Publications.
b. A history of events in linguistic theory. In Tenny, C. and Pustejovsky, J.
(eds) a, –.
Travis, L. a. Event Structure in Syntax. In Tenny, C. and Pustejovsky, J.
(eds) Events as Grammatical Objects: The Converging Perspectives of Lexical
Semantics and Syntax –. Stanford, CA: CSLI Publications.
b. The l-syntax/s-syntax Boundary: Evidence from Austronesian. In Ileana,
P., Phillips, V., and Travis, L. (eds) Formal Issues in Austronesian Linguistics,
–. Dordrecht, The Netherlands: Kluwer Academic Publishers.
. Agents and Causes in Malagasy and Tagalog. In Erteschik-Shir, N. and
Rapoport, T. (eds) The Syntax of Aspect: Deriving Thematic and Aspectual
Interpretation, –. New York: Oxford University Press.
Van Valin, R. D. . Semantic Parameters of split Intransitive. Language
: –.
Vendler, Z. . Verbs and Times. In Vendler, Z. Linguistics in Philosophy, –.
Ithaca, NY: Cornell University Press.
 Appendix 1: Warlpiri inflectional
paradigms by conjugation class

Table A. Warlpiri independent verb conjugations

  a b  
NON-PAST (a) wangka paka-rni kanja-ni nga-rni ya-ni
‘speak’ ‘hit’ ‘take’; ‘eat’ ‘go’;
yinja-ni ma-ni
‘give’; ‘get’;
pinja-ni ji-ni
‘strike’ ‘scold’
NON-PAST (b) wangka-mi ka-nyi ‘take’
yi-nyi ‘give’
pi-nyi
‘strike’
PAST wangka-ja paka-rnu kanja-nu ka-ngu nga-rnu ya-nu
FUTURE wangka-ji paka-ku kanja-nku ka-ngku nga-lku ya-nku
IMPERATIVE wangka-ya paka-ka kanja-nka ka-ngka nga-nja ya-nta
IRREALIS wangka- paka- kanja- ka-ngkarla nga- ya-ntarla
yarla karla nkarla njarla
PRESENTATIVE wangka- paka-rni. ? ka-nga.nya nga-rni. ya-na.nya
nya nya nya
INFINITIVE wangka- paka-rni. ?kanja- ka-nja- nga-rni. ya-ni.nja-
nja- nja- ni.nja- nja-
PARTICIPLE = wangka- paka-rni- kanja-ni- nga-rni- ya-ni-
NON-PAST (a)

NOMIC (=N) wangka- paka-rnu kanja-nu ka-ngu nga-rnu ya-nu

ngu
Number of c.  c.  : ka - ‘take’;  : ya- ‘go’;
members: yi - ‘give’; ma - ‘get’;
pi - ‘strike’ ji- ‘scold’

Notes:
a. The Warlpiri non-past (b) forms for conjugation  and b verbs derive historically from
Ngumpin-Yapa potential or future mood forms. Future forms shown in Tables A and B are
not used in Eastern Warlpiri dialects (Lander and Hansen River). Warlpiri non-past (a) and
past forms (with the exception of conjugation  -ja inflection) derive from non-finite participial
forms, Warlpiri having lost the historical Ngumpin-Yapa past tense -nya on conjugation b
verbs. Other Ngumpin-Yapa languages have present forms based on imperfective or progres-
sive aspect marked forms.
b. Two stance verbs whose stems end in -na, namely nyina ‘sit’ and nguna ‘lie’, have alternative
imperative and irrealis forms: nyina-ya(rla) ~ nyina-ka(rla).

232
 Warlpiri verbs of change and causation: the thematic core 233

Table B. Warlpiri dependent verb conjugationsa

  a/b    (defective)

NON-PAST (a) -jarri -jirri-rni -ra-nja-ni - nga-rni -ma-ni -nji-ni ‘go &
‘become; ‘cause’; ‘put’b; ‘move’ ‘cause; V’
do’ -pirri-rni -ka-nja-ni make’; -nji-na-ni
‘gather’; ‘move’; -ma-nic ‘go & V
-parri- -yi-nja-ni ‘sound’; reiteratively’
rni ‘act; ‘cause’; -ji-ni ‘put
remove’ -pi-nja-ni via mouth’
‘cause/effect’
NON-PAST (b) -jarri-mi -ra-nyi
-ka-nyi
-yi-nyi
-pi-nyi
PAST -jarri-ja -jurru- -ra-nja-nu/ -nga-rnu -ma-nu -nju-nu
rnu -ra-ngu -nji-na-nu
FUTURE -jarri-ji -jirri-ki -ra-nja-nku/ -nga-lku -ma-nku -nji-nki
-ra-ngku
IMPERATIVE -jarri-ya -jirri-ka -ra-nja-nka/ -nga-nja -ma-nta -nji-nta ~
-ra-ngka -nji-ngka
IRREALIS -jarri- -jirri- -ra-nja- -nga-nja- -ma-ntarla -nji-nta-
yarla karla nkarla rla rla ~ nji-
-ra-ngkarla ngka-rla
PRESENTATIVE -jarri- -jirri-rni. N/A -nga-rni. -ma-na. N/A
nya nya /-ra-nga. nya nya
nya
INFINITIVE -jarri- -jirri-rni. -ra-nja-ni. -nga- -ma-ninja- - nji-ni-nja-
STEM nja- nja- nja- rninja- -nji-na-
/-ra-nja- ni-nja-
PARTIC. = -jarri- -jirri-rni- -ranja-ni- -nga-rni- -ma-ni - nji-ni-
NON-PAST (a) /N/A -nji-na-ni-
NOMIC -jarri- -jurru- -ranja-nu/- -nga-rnu -ma-nu -nju-nu
ngu rnu ra-ngu -nji-na-nu
Number of      
verbs

Notes:
a. The dependent verbs are distributed among the same five conjugations as the independent
inflecting verbs, plus a sixth containing the defective aspectual verb -nji-.
b. Dependent -ra- is only found in Eastern dialects, especially Lander River, in two verbs: mapa-
ra-nyi ‘anoint, paint, rub with’ (mapa-rni in other dialects) and yurlpa-ra-nyi ‘send’ ( yilya-mi
in other dialects).
c. The Warlpiri dictionary (Laughren et al. ) distinguishes six verbalisers -ma-ni on the
basis of meaning and argument structure.
 Appendix 2: Sample of complex verbs

Table C. Stance verbs

PV (stance) VINF PV/VINF-V PV/VINF-V Gloss

(stance) (CHANGE) (CAUSE
(CHANGE))
nyina-nja -wanti -yirrarni sit
nguna-nja -wanti -yirrarni lie
karri-nja -pardi -yirrarni stand
parntarri-nja -wanti -yirrarni crouch over
wapa-nja -pardi ? get up and
move
wardu -karri -yirrarni lie prone
tarda -yani -yirrarni sit, perch,
land
yakarl- -wanti (-kijirni) duck down;
(turn over)
mirdi-jirrpijirrpi -wanti -yirrarni kneel
pirri -mani ? sit, perch, land
pirrily- ? -yirrarni sit/seat
mirdi-pirrjipirrji -mani ? kneel
parntarr- -mani ? land in flock
marrayangkarra -mani ? land in flock
parrparr- -mani ? shake/ quiver
(one’s body)
julurr-  -mani ? give sudden
jump


“?” in a cell indicates a gap in the Warlpiri data base rather than confirmation that no form
exists.

Used with V -ngarni and -jirrirni to express uncontrolled body movement (see Table F).

234
 Warlpiri verbs of change and causation: the thematic core 235

Table D. Change location verbs

PV (location) PV-V (CHANGE) PV-V (CAUSE Gloss

(CHANGE))

jakurr(pa) -pardi mani/yirrarni move up onto/put onto

jikirr- -ngarni ? move upwards, climb,

jump over
karaly(pa) -ngarni; -wanti ? skid, slip; slip over
lakarn -pardi/-yani -mani/-pinyi detach, peel away
lurlurl -karri -kijirni fall scattered (leaves,
berries)
paarr(pa) -pardi -jirrirni take off in flight
pata -karri -kijirni fall, drop
piraly(pa) -pardi -jirrirni soar/move up through air
pirlpirl(pa) -karri/ wanti -kijirni fall drop by drop
puurl(pa) -ngarni ? set off/ start out
rangkarr(pa) -ngarni/-kanyi dawn
rdululu -ngarni -yilyami (send) suddenly emerge from and
scatter; send scattering
rdulypa-rdulypa -ngarni ? stream/swarm out of
rdurruly(pa) -pardi -parrirni strip away from
rduul(pa) -pardi -jirrirni burst out of/force to burst
out of
rduyu -karri -yirrarni rise/raise (of smoke)
rikal(pa) -pardi -jirrirni move up and away from
wari -yani/-ngarni -yirrarni move up/onto
wilypi -pardi -mani move out of
wira -ngarni ? take off into air
wirily(pa) -ngarni; -wanti ? move rapidly out/away
from; slip over
wirrinti -ngarni ? gush/squirt out from
yakarra -pardi -mani get up/out of/wake up
yalyi -pardi -mani detach from, take out from


Parri-rni is a dependent verb with very restricted distribution as a causative ‘removal’
verb.
236 Laughren

Table E. Change state [+dur] verbs (external cause)

PV (impaired PV-V (CHANGE) PV-V (CAUSE Gloss

state) (CHANGE))

lakarn(pa) -yani/-pardi -pinyi peel off

larra -yani/-ngarni -pinyi/-mani split, crack
palu pali- (=V) -pinyi die, kill,
extinguish
pangkirri ? -pinyi slash, split
panpan(pa) ? -mani split down the
middle
raa -yani -pinyi open, clear
rdawilyiwilyi ? -kanyi reduce to pieces
rdiirr(pa) -yani -pinyi burst open
rdilyki -yani -pinyi break
rdilypirr(pa) -yani -yinyi/-mani perforate
tuurl(pa) -yani -pinyi come/pull apart
yalyi -yani -pinyi break off from

Table F. Change state [–dur] verbs (external cause)

PV (momentary PV-V (CHANGE) PV-V (CAUSE Gloss

state [reaction]) (CHANGE))

julurr(pa) -ngarni -jirrirni give a jump; cause to jump with

fear, shy away
kinyirr(pa) -ngarni -jirrirni be startled; startle
miil(pa) -ngarni -jirrirni be shocked; shock
8 Complex predicates in Wambaya:
detaching predicate composition
from syntactic structure

RACHEL NORDLINGER

. Introduction
In this chapter I compare and contrast two complex predicate constructions
in Wambaya, a moribund non-configurational, non-Pama-Nyungan lan-
guage of north-central Australia. The first of these – the associated motion
construction – is the Wambaya reflex of the original proto-Mirndi verb-
coverb construction, still present in Jaminjung (Schultze-Berndt ). In
modern-day Wambaya this presents as a series of portmanteau directional +
tense/aspect/mood suffixes on the second position auxiliary, in conjunction
with a lexical main verb, as in ():, 
() a. Gannga mirnd-amany.
return .DU.INC.S-PST.TWD
‘We came back.’
b. Gannga mirnd-any.
return .DU.INC.S-PST.AWY
‘We went back.’
c. Gannga mirnd-a.
return .DU.INC.S-PST
‘We returned.’


I wish to extend my heartfelt appreciation and gratitude to my Wambaya teachers and
friends: Molly Grueman (deceased), Mavis Hogan (deceased), Minnie Nimarra (deceased), and
Judy Holt. Many thanks also to Brett Baker, Mark Harvey, and Mengistu Amberber for inviting
me to participate in the workshop on Complex Predicates at the ALS Conference in July ,
at which this work was first presented, and to Brett Baker, Mark Harvey, and an anonymous
reviewer for detailed comments on this chapter leading to many substantial improvements.
Unfortunately, I remain responsible for any errors or infelicities.

All Wambaya examples are taken from Nordlinger (a) and my own fieldnotes.

Abbreviations used in examples are: A ‘transitive subject’, ACC ‘accusative’, ALL ‘allative’, AWY
‘direction away from deictic centre’, COMIT ‘comitative’, DAT ‘dative’, DU ‘dual’, ERG ‘ergative’,
EXC ‘exclusive’, FOC ‘focus’, FUT ‘future tense’, HAB ‘habitual’, INC ‘inclusive’, LOC ‘locative’, M
‘masculine’, NOM ‘nominative’, NP ‘non-past’, OBL ‘oblique’, PL ‘plural’, PR ‘present tense’, PROG
‘progressive’, PST ‘past tense’, RDP ‘reduplicated’, S ‘intransitive subject’, SG ‘singular’, TWD ‘dir-
ection towards deictic centre’.

237
238 Nordlinger

It is noteworthy that in this respect Wambaya differs from related languages,

such as Jaminjung (Schultze-Berndt ), which only show a two-way con-
trast since the finite verb (corresponding to the associated motion marker
in Wambaya) is obligatory and so there is no Jaminjung equivalent for (c)
(Mark Harvey, personal communication).
The second complex predicate construction is a serial verb construction,
in which two lexical verbs (one of them yarru ‘go/come’) combine to form
a single semantic predicate, as in (a). Examples (b) and (c) demonstrate
that each of the elements in the serial construction in (a) is able to function
as a clausal predicate in its own right also.
() a. Gayini g-a yarru ginkanyi nanganangali?
who (NOM) .SG.S-PST go this.way sneak
‘Who went sneaking off this way?’
b. Gayini g-a yarru ginkanyi?
who (NOM) .SG.S.PST go this.way
‘Who went this way?’
c. Gayini g-a ginkanyi nanganangali?
who (NOM) .SG.S-PST this.way sneak
‘Who snuck this way?’
Although these two complex predicate constructions are structurally dis-
tinct, they share some interesting similarities. For example, as shown in the
contrast between (a) and (), the semantic contribution of the associated
motion marker appears different depending on the nature of the main verb.
When the main verb is a motion verb (i.e. a verb encoding locomotion along
a path), as in (a), the associated motion marker adds only the direction of
the motion event. When the main verb is a non-motion verb as in () and (),
however, the associated motion affix encodes a sequential motion event ‘go/
come and VERB’ (Nordlinger ).
() Bungmanyi-ni gin-amany yanybi.
old.man-ERG .SG.M.A-PST.TWD get
‘The old man came and got her.’
() Nayirrundurna irri-n gajurru. Gajurru
women (NOM) .PL.S (NP)-PROG dance dance
ng-uba irraga-yili.
.SG-NP.AWY .PL.OBL-COMIT
‘All the women are dancing. I’m going to go there and dance with them.’

Note however, that Jaminjung has a wide range of finite verbs that buru ‘return’ can co-occur
with, so it is likely that there are different semantic distinctions that can be made in Jaminjung,
but not in Wambaya (see Schultze-Berndt () for discussion).
 Complex predicates in Wambaya 239

In these examples, the associated motion marker does not encode the direc-
tion of the main verb event: () does not mean ‘get this way’ and () does not
mean ‘dance away’. Rather, the associated motion marker encodes a motion
event that is temporally prior and sequential to the main verb event – ‘come
and (then) get’, ‘go and (then) dance’, respectively.
A very similar contrast is found with the motion verb yarru (glossed ‘go’,
but actually neutral as to deixis) when used in serial verb constructions.
When combined with another motion verb, as in (a), the resulting semantics
is one of simultaneous motion; a single event. When combined with a non-
motion verb, on the other hand, the resulting combination usually encodes a
sequential motion event, as in ():

() Yarru ngurl-aji lingba-lingba.

go .DU.EXC.S-HAB.PST swim-RDP
‘We used to go and swim.’

These two Wambaya constructions raise a number of issues that this

chapter seeks to address. First, how are associated motion construc-
tions such as in () and () to be integrated into a theory of complex
predicates, if at all? Second, how is the interaction between motion and
non-motion predicates to be accurately reflected in a model of complex
predication? And third, what are the implications for theory arising from
the existence of two very different syntactic constructions exhibiting the
same types of semantic predicate composition, as we find in the Wambaya
data?
My aim in this chapter is to provide an analysis of these complex predi-
cates in Wambaya that suggests answers to these questions. The analy-
sis presented here draws on earlier work within the Lexical-Functional
Grammar (LFG) tradition, particularly Butt (, , ), Andrews
and Manning (), and Broadwell (). I argue that the Wambaya
data supports previous work on complex predicates (e.g. Butt , ,
Alsina , but cf. Baker ) that argues for the separation of predicate
composition from syntactic structure, as is inherent in a formal model of
grammar such as LFG (Bresnan ). In particular, the Wambaya data
show that: (i) the same processes of predicate composition can apply to two
very different syntactic structures and; (ii) the same syntactic structure
can be subject to different processes of predicate composition. Thus we
need to be careful to allow a many-to-many relationship between syntac-
tic structures and processes of predicate composition in complex predicate
formation.


This verb means ‘to play or bathe in the water’ (cool off, wash, etc.); i.e. ‘to bogey’ in Aboriginal
English/Kriol. It does not refer to motion along a path by way of swimming, for which waja-
ngarnja would be used instead.
240 Nordlinger

. A closer look at the data

Before presenting my analysis of complex predicates in Wambaya, I will
examine the properties of each construction type in more detail.

.. Associated motion constructions

As shown in (), the associated motion construction involves the presence
of a directional marker on the auxiliary (contrasting ‘go/away’ and ‘come/
towards’), in combination with a main lexical verb. For the Wambaya speak-
ers who I worked with, the presence of a lexical verb was obligatory; the
auxiliary (with or without directional marker) could not form a predicate on
its own:
() a. Yarru ng-uba.
go .SG.S-NP.AWY
‘I’ll go (away).’
b. Yarru ng-ulama
go .SG.S-NP.TWD
‘I’ll come.’
c. *Ng-uba.
.SG.S-NP.AWY
‘I’ll go (away).’
d. Baraj-bali gun-uba irra yabu
old.man-PL(ACC) .SG.M.A-NP.AWY .PL.ACC take
‘He’ll take all the old men (away).’
Since Wambaya is a non-configurational language with grammatically
unconstrained word order (except for the auxiliary, which must always be
in second position (Nordlinger b)), the lexical verb and the associated
motion marker need not be contiguous in the clause (d) and nor must they
appear in any fixed order with respect to each other (d vs. a–b).
As discussed above, the semantic contribution of the associated motion
marker appears different depending on the nature of the main verb. When
combined with a motion verb, the associated motion marker adds the direc-
tion of the motion event, since motion verbs in Wambaya are direction-
neutral. When the main verb is a non-motion verb, however, the associated


The associated motion construction is discussed in some detail in Nordlinger (), on which
this section draws heavily.

This is in contrast to the claims of Chadwick (: ), who gives (c) as grammatical. This
was strongly rejected by my consultants, and may reflect a difference between the language of
the speakers that Chadwick and I each worked with.
 Complex predicates in Wambaya 241

motion affix encodes a sequential motion event ‘go/come and VERB’. Thus,
when the main verb is a motion predicate, the resulting complex predicate
encodes a single event with the apparent contribution of the associated motion
marker being directional information. On the other hand, when the main
verb is a non-motion predicate, the resulting complex predicate encodes two
sub-events: one of motion + direction (encoded by the associated motion
marker) and the other encoded by the main verb. Broadwell () discusses
a similar distinction in Choctaw associated motion constructions, although
in Choctaw the two meanings are encoded by distinct directional particles.
Following Broadwell (), I will refer to these two functions as the ‘single
event’ function and the ‘dual event’ function respectively.
At this point it is important to consider whether associated motion con-
structions should be treated as complex predicate constructions at all. Butt
() defines a complex predicate as having the following properties:
() Definition of a complex predicate (Butt : )
• the argument structure is complex (two or more semantic heads contrib-
ute arguments);
• the grammatical functional structure is that of a simple predicate. It is
flat: there is only a single predicate (a nuclear PRED) and a single subject
[i.e. it is monoclausal – RN];
• the phrase structure may be either simple or complex. It does not neces-
sarily determine the status of the complex predicate.
Thus, we can define complex predicates as monoclausal structures in which
information from two or more semantic heads (i.e. predicating elements)
is contributed to a single predicate argument structure. It is clear that the
Wambaya associated motion construction is monoclausal: it contains only a
single lexical verb, a single subject, and the associated motion marker can-
not constitute a clausal predicate on its own. Perhaps less immediately clear
is whether this construction satisfies the first condition in (), namely that
there is a complex argument structure with information contributed by
more than one predicating element. In other words, to what extent can the
associated motion marker be considered a semantic predicate? Obviously we
cannot use syntactic predication as a diagnostic here, since the associated
motion marker is never found as the sole predicate of the clause (although, as
noted above, this is found in Chadwick’s () description of the language,
and in the neighbouring language Jingulu (Pensalfini )).
However, it is quite clear that the associated motion marker is a semantic
predicate, nonetheless. First, when combined with non-motion verbs in the
dual event function, it clearly contributes a motion predication that cannot
be associated with the main lexical verb (as in ()). Second, the fact that its
function is sensitive to the predicate semantics of the lexical verb it combines
with – adding only a path when combined with a motion verb, and a distinct
242 Nordlinger

motion sub-event elsewhere – shows it to be operating at a level of predicate

argument structure, as will be shown in more detail below.
The associated motion affix contributes only a single argument (the
‘mover’), which is always fused with the highest argument of the lexical verb.
Thus, the lexical verb is clearly the head for the purposes of determining
transitivity and all other aspects of argument structure. This is seen clearly
in () and (), in which the associated motion marker is combined with a
transitive and intransitive lexical verb respectively, with the corresponding
clauses having the same transitivity as the lexical verb.
An analysis of the associated motion construction in Wambaya must,
therefore, be able to account for the following features of the construction:
Features of the associated motion construction to be accounted for:
(i) The associated motion marker cannot form a complete predicate on its
own (c).
(ii) The function of the association motion marker is sensitive to the pred-
icate semantics of the lexical verb that it combines with (() vs. ()).
(iii) With motion verbs, the associated motion marker adds the path for
the motion; with non-motion verbs, the associated motion marker pro-
vides a temporally prior and sequential motion event (+ path).
(iv) The agent of the motion event encoded by the associated motion
marker is always identical to the highest argument of the lexical verb.
(v) Argument structure and transitivity are always determined solely by
the lexical verb.

.. Serial verb constructions

Wambaya also has complex predicates in the form of serial verb constructions.
Serial verb constructions (SVC) are not frequent in Wambaya discourse, but
exist in sufficient quantity in natural, spontaneous speech to justify their analy-
sis as a robust construction type in the language. There are a few different
types of serial verb construction, corresponding to the common SVC types
found cross-linguistically (e.g. Durie ), namely postural serialisations (),
manner serialisations (), cause-effect serialisations (), and motion serialisa-
tions () (see Nordlinger, in prep. for detailed discussion). Each of the verbal
elements of these serialised constructions can function as independent clausal
predicates in non-serialised constructions (see () above for an example).
() Darridarri irri-n garranbi.
be.in.a.line .PL.S (NP)-PROG stand
‘They’re sitting/standing in a line.’

Butt (: ff) seeks to draw a distinction between serial verb constructions and complex
predicates (i.e. light verb constructions). I return to this point in section ..
 Complex predicates in Wambaya 243

() Gaj-ba ng-u ganjimi manganyma mama.

eat-FUT .SG. A-FUT finish tucker(ACC) this (ACC)
‘I’m going to eat all of this tucker.’
() Warima ng-a yana jinbarrardi
hold .SG.A-PST this (ACC) squash
‘I squashed it by holding it.’
() Yarru ngurl-aji lingba-lingba.
go .DU.EXC.S-HAB.PST swim-RDP
‘We used to go and swim.’
We will focus here on motion serialisations. All of the motion serialisations
in the corpus involve the general motion verb yarru. Usually yarru comes
first in the verb sequence (), but it need not (–). Consistent with the
general ‘free word order’ properties of Wambaya, the verbs need not be con-
tiguous in the clause.
() Gulugbi ng-u ngawurniji yarru.
sleep .SG.S-FUT .SG.NOM go
‘I’ll sleep while going (on the bus).’
() Gayini g-a yarru ginkanyi nanganangali?
who (NOM) .SG.S-PST go this.way sneak
‘Who went sneaking off this way?’
() Ganinggiji gi-n yarru!
be.close .SG.S (PR)-PROG go
‘He’s coming close!’
As with the associated motion affixes, the contribution of the motion verb
yarru in these serialised constructions differs depending on the verb it com-
bines with. With motion verbs yarru always specifies concurrent motion,
functioning to highlight the motion component, e.g. ‘go sneaking’ vs.
‘sneak’. With non-motion verbs, the function of yarru is usually to indicate
sequential motion (‘go and (then) swim’, ‘come and (then) be close’), but
unlike associated motion affixes, it can have a simultaneous motion mean-
ing with non-motion verbs also, as in () ‘sleep while going’. Out of con-
text, () is therefore ambiguous between ‘go and sleep’ and ‘sleep while
going’. Somewhat unusually for serial verb constructions (e.g. Durie ),
the ordering of the two verbs need not be iconic with the ordering of the
sub-events: in () yarru is second in the sequence, despite the fact that the

It is logically possible that the motion event could follow the non-motion event (i.e. ‘sleep and
go’). However I have no examples in the corpus of a yarru serialised construction with this
interpretation, and it is a question I neglected to test explicitly while in the field.
244 Nordlinger

motion component necessarily precedes the ‘being close’ component of

the complex predicate. Anti-iconicity is a characteristic feature of Wambaya
serial verb constructions (see Nordlinger, in prep).
Yarru serialisations are clearly structurally distinct from associated
motion constructions, since they involve two lexical verbs (in addition to the
auxiliary). However, the semantics of the composite predicates share strong
similarities across the two construction types. Indeed, yarru serial verb
constructions as in () and () above can always be paraphrased with an
associated motion construction, as in (′) and (′):

(′) Gayini g-amany ginkanyi nanganangali?

who (NOM) .SG.S-P.TWD this way sneak
‘Who snuck off this way?’
(′) Ganinggiji g-ulama!
be.close .SG.S-NP.TWD
‘He’s coming close!’
The serialised construction in () can also be paraphrased with an associ-
ated motion construction as in (′), but in this case the meaning can only
be of two sequential sub-events (since this is the only option available for
associated motion constructions containing non-motion verbs).
(′) Gulugbi ng-uba ngawurniji.
sleep .SG.S-NP.AWY .SG.NOM
‘I’ll go (away) and sleep.’
As far as I can determine, the only semantic difference between (′) and
(′) and their serialised equivalents is that the associated motion construc-
tions explicitly encode deixis (i.e. motion away vs. motion towards), whereas
yarru in the serial verb constructions is deixis-neutral, leaving the direction
of motion to be determined from the context.
Butt (: ff) argues that serial verb constructions should be excluded
from the definition of complex predicates. While acknowledging that there
is substantial overlap between her complex predicates (i.e. light verb con-
structions) and serial verb constructions, she argues that they differ on two
parameters identified as characteristic of serial verbs by Durie (e.g. ),
namely: (a) serial verbs may allow the single tense/aspect feature to be
encoded on all verbs in the construction, whereas complex predicates do
not allow multiple encoding of tense/aspect/mood; (b) complex predicates
do not encode the instrumentality, resultativeness, or benefaction expressed
by serial verbs in quite the same way (Butt : ). However, serial verb
constructions come in many different forms cross-linguistically, and the two
properties characteristic of (some) serial verb constructions mentioned above
are irrelevant to Wambaya, in which serial verb constructions generally
don’t involve multiple encoding of tense/aspect/mood nor express notions
 Complex predicates in Wambaya 245

of instrumentality, resultativeness, or benefaction. In a model of grammar

that separates syntactic configuration from predicate argument structure (as
in LFG, for example), there is no reason a priori to assume that all serial
verb constructions (i.e. all monoclausal constructions consisting of multiple
lexical verb co-heads) should compose their predicate argument structure
in identical ways or conversely, that they should undergo processes of predi-
cate argument structure distinct from those of other complex predicate
types. In Wambaya, we find that the yarru serial verb constructions show
near-identical behaviour, in terms of predicate argument composition, to
associated motion constructions. It therefore seems pertinent to treat them
similarly in formal terms, in order to account for these similarities. Below
I will present an analysis of these constructions in LFG which allows the
same processes of predicate composition to apply across both associated
motion and yarru serialisation constructions.

. Analysis of complex predicates as Predicate Composition

In this section, I present my analysis of these complex predicates in
Wambaya, drawing on work within the LFG framework by Alsina (,
, ), Butt (, , ), Andrews and Manning (), and
Broadwell (). I will assume the analysis of Wambaya phrase structure
presented in Nordlinger (b), in which the basic clause has the structure
in (), for example ().

()
IP

( FOC)

NP I

N I S

bungmanyi-ni gin-amany

old.man-ERG 3.SG.M.A-PST.TWD V

yanybi

get


I return to the issue of whether my analysis should be extended to all serial verb constructions
in Wambaya in section ...
246 Nordlinger

The pertinent features of this structure are as follows. Firstly, the simple
clause has two co-heads (heads are indicated by ↑ = ↓): I and S. S is an exo-
centric category (see Bresnan ), which in Wambaya has a flat structure
containing the verb(s) and any number of NPs (including none, as in ()) in
any order, thereby accounting for the non-configurational properties of the
language. Within S, functional annotations (i.e. head or argument/adjunct
functions) are freely assigned. The exact number of NPs possible in any S
is constrained by the argument structure of the verb, and their grammatical
functions are determined by their case-marking – see Nordlinger (b)
for detailed discussion. Since I is the locus of the auxiliary and V is the
head of S, I and V (via S) are co-heads of the clause, meaning that their
lexical information is unified into the same clausal f-structure (). This
accounts for the fact that they both contribute predicate information. Note
also that, following standard LFG assumptions, the unexpressed object in
() is not represented in the c-structure, but is contributed to the f-structure
via information associated with the verb (see Austin and Bresnan () and
Nordlinger (b) for discussion).
The c-structure in () corresponds to the f-structure in (). The informa-
tion associated with I and V (via S) map to the outer f-structure in (), which
is the f-structure of the clause. Following Alsina (), I assume that the PRED
information in this f-structure is essentially a pointer to a-(rgument) struc-
ture, the level at which predicate argument relations are encoded and mapped
to both the semantics and the syntax (i.e. f-structure). The discussion about
predicate fusion below relates to this part of the grammatical structure.

 PRED ‘come-get’< ... > 

TENSE PAST 
 
  PRED ‘old man’ 
 PERS 3 
  
SUBJ  NUM SG 
()   
 GEN MASC  
  
CASE ERG 
  PRED ‘P
P RO ’

  PERS 3  
OBJ   
  NUM SG  
   
 CASE ACC  


There is in principle no restriction that there be only one V in S, which allows for serial con-
structions, providing the PRED features can unify through Predicate Composition, as we will
see below.

The auxiliary is constrained to always being in second position in the clause. Where there is
no NP in initial FOC position, it undergoes a process of prosodic inversion (Halpern ) to
appear after the first constituent of S (see Nordlinger (b) for further discussion).

The annotation ↑=↓ in () indicates that the information associated with the lower node is
unified at f-structure with the information associated with the higher node (i.e. specifying the
‘head’ relation). The standard algorithm (Kaplan and Bresnan ) which unifies functional
 Complex predicates in Wambaya 247

The possibility of clausal co-heads lends itself naturally to an analysis of com-

plex predicates of various types. Since the associated motion affixes attach to
the auxiliary in I, it is not surprising that they should form a complex predi-
cate with the verb, since both co-head the clause. Serial verb constructions are
accounted for by allowing S to contain multiple instances of V, hence multiple
co-heads within S, providing their PRED information can successfully combine
at the level of a-structure (see below). This analysis also allows for a combina-
tion of these two complex predicate types, namely multiple Vs in S (a serial
verb construction) in addition to an associated motion marker on the auxiliary,
in I. We will see an example of this type of construction in section ..
Following Alsina () and especially Butt (, , ), I assume
that complex predicates involve predicate composition at the level of argu-
ment structure, which I take to be elaborated along the lines of Jackendoff’s
() Lexical Conceptual Structure. The process of complex predicate
formation is triggered at a-structure by the presence of a transparent event
argument (Butt, Isoda, and Sells ) (represented as {}Et in the figures
below). A lexical item containing a transparent event in its argument struc-
ture is an incomplete predicate (Alsina ) – it cannot function as a clausal
predicate without having undergone predicate composition to form a com-
plex predicate with another lexical item. Following Butt (, , ),
we can define two processes by which predicate composition occurs: Event
Fusion and Argument Fusion. Event Fusion unifies the information within
two events (cf. ‘merger’ (Baker and Harvey, this volume)), while Argument
Fusion coindexes two arguments so that only one argument is available for
linking to grammatical functions (cf. ‘coindexation’ (Baker and Harvey, this
volume)). These two processes have been shown to account for a wide range
of complex predicate formation in the world’s languages (see, for example,
the work by Butt (), Wilson (), and Broadwell (), among others).
Below we will see that they also provide a natural account of complex predi-
cate formation in Wambaya. Moreover, the detachment of argument struc-
ture (where predicate composition occurs) from c-structure (the level of
‘surface’ syntactic structure), that is assumed in the framework of LFG,
allows straightforwardly for the fact that two very different types of syntac-
tic structure can undergo the same processes of predicate composition.
Following Butt (, , ), Wilson (), and Baker and Harvey
(this volume), I adopt a version of Jackendoff’s () theory of Conceptual
Semantics, using his Lexical Conceptual Structure (LCS) to construct an
information annotated at c-structure to produce a well-formed f-structure disallows the unifi-
cation of PRED information. In order to account for complex predicates, Alsina () redefines
↑ = ↓ such that it allows for the composition of PREDs (through the composing of their argu-
ment structures, as outlined below), and the unification of all other types of information. I will
follow this analysis here also.

Note, however, that nothing in my analysis hangs on the notion that Predicate Composition
occurs at the level of a-structure. It could be straightforwardly translated into a model that
represents LCS information (and Predicate Composition) directly at f-structure (e.g. Wilson
), or into the more articulated model of Andrews and Manning ().
248 Nordlinger

elaborated form of a-structure. Jackendoff proposes that LCSs have distinct

‘action’ and ‘thematic’ tiers – the former encoding information about Actor–
Patient relations (marked with AFF), and the latter encoding the main
information about the semantic content of the lexical item (decomposed into
functions such as CAUSE, GO, BECOME, etc.). Butt () argues for the
need to also add an aspectual tier to account for the role of inceptive, dura-
tive, and completive aspect in complex predicate formation in Urdu/Hindi.
In the discussion below, I will only represent the parts of the LCS relevant
to the analysis of Wambaya complex predicates. In most cases, this will con-
sist of just the thematic tier.

.. Associated motion

Associated motion affixes have two lexical entries – one reflecting their
‘single event’ function and the other reflecting their ‘dual event’ function.
Beginning with the first of these functions, associated motion markers can
be analysed as having the (simplified) LCS in ().
() ‘Single-event function’:
a. ‘-amany’:

{ Event GO ([ Thing ] ,  Path TO [ Place )}

HERE]  ET

b. ‘-any’:

{ Event GO ([ ] , 
Thing Path TO [ Place THERE ] 
 )} ET

As indicated by { }ET, associated motion markers are incomplete predicates,

accounting for the fact that they cannot function alone as clausal predicates,
but must combine with another predicate (i.e. the lexical verb) to form a
complex predicate. Apart from this, the LCS is that of a standard motion
verb ( Jackendoff ).
The presence of the transparent event argument triggers predicate fusion,
namely the LCS of the associated motion marker can merge with another
LCS with which it can unify. This is effected via the process of Event Fusion
(e.g. Butt : ), which unifies the information contained within two
events. Since unification is involved, it follows that the two events must
be unifiable, in this case meaning that the LCS of the associated motion


Square brackets [ ] are used to mark individual events; braces { } denote sets of events (as in
()), unless they are marked with ET,{ }ET, in which case they denote a transparent event, as
discussed below.
 Complex predicates in Wambaya 249

markers in () can only combine with motion predicates (e.g. ), which are
also GO events.

()  Event GO ([Thing ] , [ Path ])

Since both the associated motion markers and the motion verbs are GO
events, once Event Fusion has occurred, the net effect of the associated
motion marker is to provide a value for the Path argument. Hence their
apparent function as directional markers with motion verbs:

()  Event GO ([Thing ] ,  Path TO [ Place HERE] )

 
Clearly Event Fusion is not possible with non-motion verbs, since these will
not contain a GO predicate and so unification will fail. Thus, we account for
the fact that associated motion markers do not function as directional mark-
ers with non-motion verbs.
However, as we saw in section ., associated motion markers can co-
occur with non-motion verbs, but in this case form a complex predicate
specifying two sub-events – a motion event and a non-motion event – as in
() repeated from above.
() Bungmanyi-ni gin-amany yanybi.
old.man-ERG .SG.M.A-PST.TWD get
‘The old man came and got her.’
As well as allowing for the encoding of two sub-events, the LCS of these com-
plex predicates also needs to specify that the motion event occurs prior to the
event encoded by the lexical verb. I propose that the LCS corresponding to the
‘dual event’ function of associated motion markers is as in (). In this struc-
ture, more elaborated than those for the ‘single-event’ function, we have a the-
matic tier which is a set of two events (following Andrews and Manning’s ()
set-based approach to symmetric serial verb constructions) and a temporal tier,
which specifies the temporal relationship between the two events (following
Broadwell’s () analysis of dual-event directional particles in Choctaw).


The prediction made by this analysis is that associated motion markers in single event func-
tion should also be able to unify with non-motion events that also contain a GO predicate in
their LCS; verbs such as ‘take’ or ‘throw’, for example. Unfortunately my corpus does not
contain the data to test this prediction conclusively, but there is one example (given in d),
which suggests that this is indeed possible, at least with the verb yabu ‘have/take’ (i.e. ‘take
away’ not ‘go and take’).

Note that the correspondence between the events in the temporal tier and the events in the set
is effected through the subscripted variables x and y associated with each event in the set, and
in the temporal tier.
250 Nordlinger

() ‘Dual-event function’:

a. ‘-amany’:

{
 
  Eventx GO ([Thing D ] ,  Path TO [ Place HERE] ) ,  Eventy AFF

  { ([ ] , )}
D
ET
}
TEMP : Eventx < Eventy 

b. ‘-any’:

{
 
{
  Eventx GO ([Thing D ] ,  Path TO [ Place THERE] ) ,  Eventy AFF

  ([ ] , )}
D
ET
}
TEMP : Eventx < Eventy 

In these lexical entries, then, we find a single LCS that specifies two sub-
events – Eventx and Eventy. Eventx is the motion event and Eventy is a trans-
parent event argument which triggers complex predicate formation. The
highest argument of Eventy (specified by AFF ([ ], ) on the action tier) is
identified with the agent of the motion event. The temporal tier specifies that
the motion event must temporally precede the transparent event (Broadwell
). In this case, the transparent event is an argument of the associated
motion marker, and thus the a-structure of another predicate (namely, the
verb) is simply substituted for the transparent event argument. Thus, one
event is embedded within another event, and so it is not Event Fusion, but
Argument Fusion that is the applicable process for predicate composition.
Argument Fusion will ensure that the highest argument of the embedded
event is coindexed with the argument of the matrix event (indicated with α
in ()), making it invisible for linking purposes (e.g. Butt : ). Thus,
the argument array of the resulting complex predicate is identical to that of
the lexical verb. The LCS of the complex predicate in () is therefore as
follows:

()



(
CAUSE [
D
] , BE ([ ] , AT [D ])) 
 Eventx GO ([Thing D ] ,  Path TO [ Place HERE] ) ,

 



( D
 Eventy AFF [ ] , )  

TEMP : Eventx < Eventy 


Note that Broadwell () treats the whole superordinate event as the transparent event,
rather than just the second sub-event as I do here. As far as I can tell, there are no major
empirical or theoretical implications arising from this difference, although my approach is
more consistent with Butt’s () definition of Argument Fusion, which applies when the
transparent event is an argument of another event.

Note that Argument Fusion in Butt’s () terms actually involves the coindexation of the
highest embedded argument with the lowest matrix argument (p. ). However, since the
 Complex predicates in Wambaya 251

The fact that associated motion markers can have two very different
functions depending on the lexical verb that they combine with is thus cap-
tured by the availability of two different processes for predicate composition
at a-structure – Event Fusion and Argument Fusion. Since predicate com-
position takes place at the level of a-structure, it operates independently of
surface syntactic structure, meaning that a single construction at c-structure
(namely associated motion marker + lexical verb) can correspond to two dif-
ferent types of predicate composition, resulting in two different construc-
tional meanings. This highlights the importance of a model of complex
predicate formation that clearly separates syntactic structure from predicate
composition (as in the work by Alsina (e.g. , , ), Butt (e.g. ,
, ), Wilson (), Andrews and Manning (), and others). In
the next section this point is further reinforced by the fact that very differ-
ent syntactic construction types can, in the one language, undergo the same
types of complex predicate formation at a-structure.

.. Analysis of yarru serial verb constructions

Above we saw that the interaction of the associated motion markers with
motion and non-motion verbs can be accounted for using the mechanisms
of Event Fusion and Argument Fusion, well established in the complex
predicate literature. We saw that the same syntactic structure can result in
two types of complex predicate – one formed by Event Fusion and one by
Argument Fusion – arguing strongly for the need to separate the processes
of complex predicate formation from syntactic structure. This point is rein-
forced by the fact that yarru serial verb constructions, while completely dis-
tinct from associated motion constructions syntactically, also enter into the
same two types of predicate composition.
Recall that yarru in serial verb constructions shows a similar interaction
with motion and non-motion verbs as the associated motion affixes do. When
combined with a motion verb it serves to further specify the motion compo-
nent of a single event (); whereas with non-motion verbs it may encode a
distinct motion sub-event (), or not ().
() Gayini g-a yarru ginkanyi nanganangali?
who (NOM) .SG.S-PST go this.way sneak
‘Who went sneaking off this way?’

matrix event in our case always has just a single argument, this distinction is not relevant to
the present discussion.

In fact, this analysis does not rule out the possibility that a motion verb could combine with
the associated marker in dual event function (i.e. filling the transparent event slot). This would
produce meanings like ‘go and run’ (rather than ‘run away’), for example. I have no examples
of motion verbs in such dual event structures, but also don’t have the negative evidence to
rule such possibilities out completely. It may well be that there is a strong preference for
interpreting motion verbs with associated motion markers as single event structures, ruling
out dual event structures on pragmatic grounds.
252 Nordlinger

() Gulugbi ng-u ngawurniji yarru.

sleep .SG.S-FUT .SG.NOM go
‘I’ll sleep while going (on the bus).’
() Yarru irr-a ngajbi nanawulu ilarra-wulu
go .PL.S-PST see this.DU.ACC eaglehawk-DU(ACC)
‘They went and saw the two eaglehawks.’
There are however, two differences in the behaviour of yarru and the associ-
ated motion markers. Firstly, yarru can function as a clausal predicate in its
own right (), and so its participation in a complex predicate is optional, not
obligatory (as it is for the associated motion markers).
() Injani ny-u yarru ?
where .SG.S-FUT go
‘Where are you going?’
Secondly, yarru encodes no temporal specification in combination with non-
motion verbs. Whereas combining an associated motion marker with a non-
motion verb always yields a complex predicate in which the motion sub-event
precedes the non-motion sub-event (i.e. ‘go/come and VERB’), when yarru
is combined with a non-motion verb, the motion sub-event in the resulting
complex predicate can be sequential to the non-motion sub-event (e.g. ‘go
and see’ in ()) or concurrent with it (e.g. ‘sleep while going’ in ()).
We can thus provide a similar analysis for yarru as for the associated
motion affixes, with only two minor differences, discussed below. The basic
lexical entry (modified from () above) for yarru contains the LCS in ().

() { Event (

GO [Thing ] , [ Path ])} ( ET )

Here yarru is only optionally specified as a transparent event. This accounts

for the fact that yarru can both occur as an independent motion predicate in
its own right (), and undergo Predicate Composition to form a complex
predicate with a motion verb (as in ). Since both yarru and the motion
verb (e.g. nanganangali in ()) contain GO predicates, they can unify via
the process of Event Fusion, exactly as we saw with the associated motion
affixes in section .. above.

In the LCS for nanganangali ‘sneak’, I follow (for convenience) Jackendoff’s (: )
treatment of predicates encoding both manner of motion and motion along a path as being
formed by an adjunction rule that results in a composite structure containing both GO and
MOVE functions. The distinction between GO activity verbs and MOVE activity verbs is
important in Wambaya since only the former count as motion verbs for the purposes of the
complex predicate constructions discussed here (cf. Baker and Harvey (this volume) who
revise Jackendoff’s framework to treat all activity predicates (whether motion or otherwise)
as involving the single function MOVE).
 Complex predicates in Wambaya 253

 (
GO [Thing ]D , [ Path ] ) 

()

AFF [ ] ( )
D 

 
 Event  BY MOVE sneak ([D ])  

As with the dual-event associated motion affixes, yarru also has an alterna-
tive argument structure, in which it allows for a second sub-event (). The
only difference here is that there is no temporal ordering specified for the
two events (and therefore no temporal tier specified here). In this case, yarru
can combine with non-motion verbs (e.g. ‘sleep’ in ()) via the process of
Argument Fusion, as we saw in () above. The complex predicate ‘go-sleep’
in () thus has the a-structure in ().

() {
  Eventx GO Thing D , Path
  ([ ][ ]) , { Eventy AFF ([ ] , )}
D
ET
}
()
  (
 Event MOVE sleep  Thing  
  )

()


 Eventx GO ([Thing D ] , [ Path


]) , 
(
MOVE sleep  Thing 
D
)



(D
 Eventy AFF [ ] , )  

This treatment of yarru serialisations as complex predicates runs counter
to the arguments made by Butt (: ff) that serial verbs and complex
predicates (or, light verbs) are fundamentally distinct. In Butt’s terms, they
are distinguished primarily by the fact that light verbs contain a transparent
event argument at a-structure, and thereby subcategorise for an argument-
taking predicate in order to form a complete predicate in the syntax, while
serialising verbs do not (p. ). While I agree that there are significant
differences between robust, productive serial verb constructions and the
light verb constructions discussed in detail by Butt, the analysis presented
here for yarru serialisations in Wambaya can be justified on a number of
grounds. First, as has been shown in the discussion above, the interaction
between yarru and the second lexical verb in yarru serialisations is virtu-
ally identical to the interaction between the two predicates in non-serialised
complex predicates, namely the associated motion constructions. The fact
that the same processes of predicate composition are found in these two
254 Nordlinger

construction types argues strongly for an analysis in which they are treated
as undergoing the same processes at argument structure. Furthermore,
motion serialisations in Wambaya are very restricted – only ever involving
the motion verb yarru – thus suggesting that they are not as fully productive
in their combinatory possibilities as motion serialisations in most serialising
languages. Yarru is unique among the Wambaya motion verbs in being able
to participate in these serialising constructions. For this reason, an analysis
that treats yarru as having alternative argument structures containing trans-
parent event arguments is plausible: clearly yarru is special in this respect.
In fact, it may well be that yarru was in the process of developing a function
as a light verb, accounting for its behaviour in serial verb constructions and
reflected in its optional transparent event argument (e.g. ).
I do not wish to claim, therefore, that all serial verb constructions – not
even all of those that exist in Wambaya – should or could be appropri-
ately analysed according to the analysis presented for yarru serialisations
above. Indeed, the fact that yarru serialisations differ from other serial verb
constructions in this way reinforces the central point of this chapter: that
features of syntactic structure (i.e. whether a construction consists of mul-
tiple lexical verbs, none of which are dependent upon or complements of
another) need to be clearly distinguished from properties of semantic and/
or argument structure (i.e. whether the interaction of multiple predicates
in a single clause involves predicate composition). Yarru serialisations are
serial verb constructions in the syntax, but differ from these by undergoing
processes of predicate composition in the argument structure like ‘regular’
complex predicates (in this case, associated motion constructions).

. Combining associated motion with yarru serialisation

The analysis presented above accounts for the single-event and dual-event
functions of motion-based complex predicates in Wambaya by exploiting
the two processes of predicate composition well established in the LFG lit-
erature on complex predicates: Event Fusion and Argument Fusion. These
two processes of predicate composition interact with the two lexical entries
of associated motion markers and the motion verb yarru to yield the vari-
ous complex predicates described and exemplified in section . above.
Furthermore, this analysis illustrates the need to separate surface syntactic
structure from predicate composition, thereby allowing for a single syntactic
structure to correspond to two different types of complex predicate forma-
tion at argument structure, and conversely, for two different syntactic struc-
tures to undergo the same process of complex predicate formation.
A remaining question is whether it is possible to combine associated motion
markers with yarru serial verb constructions. The model presented here, with
its notion of co-heads at the level of c-structure, and alternative lexical entries
for both associated motion markers and yarru, allows in principle for various
 Complex predicates in Wambaya 255

combinations of these constructions types. For example, a single event associ-

ated motion marker could combine with dual event yarru via Event Fusion,
yielding the argument structure in (). Since this argument structure con-
tains a transparent event argument (by virtue of the dual event lexical entry of
yarru), this could in principle combine with a non-motion verb via Argument
Fusion, resulting in a complex predicate containing an associated motion
marker and a serial verb construction, as in the hypothetical ().

() -any + dual function yarru (hypothetical)

{
  Eventx GO Thing D ,  Path TO Place THERE   ,  Eventy AFF
  ([ ]  [ ]  )  { ([ ] , )}
D
ET
}
() ‘go away and sleep’ (hypothetical)



  Eventx GO ([Thing D ] ,  Path TO [ Place THERE] ) ,

 (
MOVE sleep  Thing 
D
)
 


( D
 Eventy AFF [ ] , )  


Unfortunately, perhaps due to the relatively small number of serial verb con-
structions in the corpus, examples combining associated motion markers
with serial verb constructions are very difficult to find. However, there is one
example (from a text) which suggests that this is indeed possible:
() Di-didija wurl-any yarru.
RDP-hold.on.hip .DU.S-PST.AWY go
‘They went away carrying (their children) on their hips.’
Didija does not usually combine with associated motion markers in their
single event function. Thus, the most plausible analysis of this construction
is that it consists of a dual event serial verb construction ‘di-didija yarru’
‘go while holding on hip’, combined with a single event associated motion
marker. The (simplified) lexical entries for the three predicate items are
given in (a–c). Yarru and didija combine via Argument Fusion to form
the complex predicate in (). The process of Event Fusion then unifies the
transparent event of the associated motion marker with the GO event in (),
yielding () as the argument structure for the complete complex predicate.

() a. yarru

{
  Eventx GO Thing D , Path
  ([ ][ ]) , { Eventy AFF ([ ] , )}
D
ET
}
256 Nordlinger

b. didija


 Event
CAUSE ([ ] D
, BE ([ ] , AT HIP (D ) ))
c. -any (single function)

{ Event GO ([ Thing ] ,  Path TO [ Place THERE] )} E T

() di-didija yarru



 CAUSE
]) , 
([ ] D
, BE ([ ] , AT HIP (D ) )) 
  Eventx GO ([Thing D ] , [ Path  

  Eventy AFF [ ( ]
D
, )  

() -any + di-didija yarru


 


D
(
CAUSE [ ] , BE ([ ] , AT  HIP (D ) ) ) 
  Eventx GO ([Thing D ] ,  Path TO [ Place THERE]  ) ,

 
 


(D
 Eventy AFF [ ] , )  

. Conclusion
In this chapter I have provided an analysis for two complex predicate con-
structions in Wambaya within the flexible theoretical framework of LFG.
Irrespective of the particular theoretical framework chosen, the Wambaya
data argue strongly for the need to separate issues of ‘surface’ syntactic
structure from those of complex predicate formation (see also Butt ).
In Wambaya, a single syntactic construction (e.g. the associated motion con-
struction) corresponds to two different types of complex predicate formation
at argument structure (i.e. involving Event Fusion and Argument Fusion).
Moreover, a single process of predicate composition (e.g. Argument Fusion)
applies to two different types of syntactic construction (i.e. associated motion
constructions and yarru serialisations).
The analysis provided here has built on a large body of existing work on
complex predicates within the LFG framework (e.g. Alsina, , Butt ,
Wilson , Broadwell ). The fact that the Wambaya data can also be
accounted for within this same general model of complex predicate forma-
tion is a strong testament to its cross-linguistic viability. Furthermore, the
fact that this model can be extended to account for motion serialisations in
Wambaya has important implications for future research into the integration
 Complex predicates in Wambaya 257

of serial verb constructions into a general framework of complex predicate

formation.

References
Alsina, A. . Predicate Composition: A Theory of Syntactic Function Alternations.
PhD dissertation, Stanford University.
. The Role of Argument Structure in Grammar. Stanford: CSLI Publications.
. A Theory of Complex Predicates: Evidence from Causatives in Bantu and
Romance. In Alsina, A. et al. (eds) –.
Alsina, A., Bresnan, J., and Sells, P. (eds). . Complex Predicates. Stanford: CSLI
Publications.
Andrews, A. and Manning, C. . Complex Predicates and Information Spreading
in LFG. Stanford: CSLI Publications.
Austin, P. and Bresnan, J. . Non-configurationality in Australian Aboriginal
Languages. NLLT : –.
Baker, M. . Complex Predicates and Agreement in Polysynthetic Languages. In
Alsina, A. et al. (eds) –.
Baker, B. and Harvey, M., this volume. Complex Predicate Formation.
Bresnan, J. . Lexical-Functional Syntax. Oxford: Blackwell.
Broadwell, G. A. . Choctaw Directionals and the Syntax of Complex Predication.
In Butt, M. and King, T. (eds) Argument Realization. Stanford: CSLI
Publications.
Butt, M. . The Structure of Complex Predicates in Urdu. PhD dissertation,
Stanford University.
. The Structure of Complex Predicates in Urdu. Stanford: CSLI Publications.
. Complex Predicates in Urdu. In Alsina, A. et al. (eds) –.
Butt, M., Isoda, M., and Sells, P. . Complex Predicates in LFG. MS, Stanford
University.
Chadwick, N. . The West Barkly Languages: An Outline Sketch. In Wurm, S.
(ed.) Australian Linguistics Studies C-. Canberra: Pacific Linguistics.
Durie, M. . Grammatical Structures in Verb Serialization. In Alsina, A. et al.
(eds) –.
Halpern, A. . On the Placement and Morphology of Clitics. Stanford: CSLI
Publications.
Jackendoff, R. . Semantic Structures. Cambridge, MA: MIT Press.
Kaplan, R. and Bresnan, J. . Lexical Functional Grammar: A Formal System of
Representation. In Bresnan, J. (ed.) The Mental Representation of Grammatical
Relations, –. Cambridge, MA: MIT Press.
Nordlinger, R. In preparation. Serial Verbs in Wambaya: Free Word Order and
Anti-iconicity. MS, University of Melbourne.
a. A Grammar of Wambaya, Northern Territory (Australia). Canberra: Pacific
Linguistics.
b. Constructive Case: Evidence from Australian Languages. Stanford: CSLI
Publications.
. Wambaya in Motion. In Simpson, J. et al. () Forty Years On: Ken Hale
and Australian Languages, –. Canberra: Pacific Linguistics.
258 Nordlinger

Pensalfini, R. . A Grammar of Jingulu, an Aboriginal Language of the Northern

Territory. Canberra: Pacific Linguistics.
Schultze-Berndt, E. . Simple and Complex Verbs in Jaminjung. Nijmegen: MPI.
Wilson, S. . Coverbs and Complex Predicates in Wagiman. Stanford: CSLI
Publications.
 Compound verbs and ideophones in
Wolaitta revisited

AZEB AMHA

. Introduction: overview of the phenomenon

Complex predicates, subsuming what are known as particle+‘say’ verbs,
compound verbs, and ideophonic constructions are widely attested in the
Ethiopian Cushitic, Omotic, and Semitic languages. They are reported
for Amharic (Leslau , ; see also Amberber, this volume), Awngi
(Hetzron ), Bench (Rapold ), Hamar (Lydall ), Qafar
(Hayward ), Somali (Dhoorre and Tosco ), and Wolaitta (Adams
, Lamberti and Sottile , Amha , Amha and Dimmendaal a,
Amha, this volume), among others. Appleyard () and Cohen, Simeone-
Senelle and Vanhove () have highlighted the importance of the con-
struction for (historical)-comparative studies of Afroasiatic as it can account
for a number of innovations in the verbal system of Cushitic, Omotic, and
Semitic languages.
The construction involves two predicative elements, which could be
labelled temporarily: P(redicate)  + P(redicate) . P is often represented by
a verbal form with restricted inflectional possibilities, e.g. by the converb
(also known as gerundive), ideophonic verb, or a derived verbal stem. In
some languages, e.g. Qafar (Central Cushitic), P may be a noun, an adjec-
tive, or a postpositional phrase (cf. Hayward ). P is a fully inflecting
verb if the complex predicate is the head of a clause. The focus in the pre-
sent contribution will be on complex predicates in which P is represented
by a verb. For this reason, we will use henceforth the representation V+V
instead of P+P. The cover term ‘complex predicate’ is used here to indicate
that the various constructions we are dealing with comprise two predicative


I would like to thank an anonymous reviewer for helpful comments and suggestions. I am
grateful to Felix Ameka, Gerrit Dimmendaal, Maarten Mous, and the editors of the present
volume, Mengistu Amberber, Bret Baker, and Mark Harvey for their comments and sug-
gestions on earlier versions of the chapter. I am indebted to Ato Gebremichael Kuke, a
native speaker of Wolaitta and a tireless contributor to its standardisation, for sharing his
knowledge and collection of books and stories. Remaining errors are my responsibility. The
present contribution emerged from a project on ‘Two Modal Categories in Omotic: the
declarative and interrogative’ at the LUCL, Leiden University. I would like to thank the
Netherlands Science Foundation (NWO) for financial support.

259
260 Amha

elements a combination of three predicates is rare, only attested in ideo-

phonic constructions).
The type of complex predicates we focus on is illustrated in (–), with
data from Wolaitta, (also written as Wolaytta), an Omotic language spoken
in southwest Ethiopia by over . million people. Fleming () classifies
Wolaitta as a member of the North Ometo group of the Omotic language
family.
() haʔʔí-kka kasé biráta-n ʔolláa book-ídí
now-INCL earlier iron.bar-LOC/INST hole:M:ACC dig-SS:A:CNV
kess-í ʔekk-iísi 
take.out-SS:A:CNV take-MSG:PF
‘Again, he dug a hole with that same iron bar and escaped.’
[Wolayttatto Leemisuwa, p. ]
In () the final, main predicate comprises two verbs: kess- ‘take out’ and
ʔekk- ‘take’ to express ‘escape’. The first verbal constituent (V) is a depend-
ent verb, which cannot form a sentence on its own. It is marked by the short-
ened form (i.e. -í ) of the same-subject anterior converb marker -ídí (see
section .. on allomorphs and distribution of the converb in Wolaitta). The
V, ʔekk- is fully inflected for subject agreement, tense, (positive) polarity,
and (affirmative declarative) mood. The values of the latter-mentioned ver-
bal categories are shared by the V and V. The converb (V) is not morpho-
logically marked for tense, negation, or mood; it only shows partial subject
agreement (cf. section ..). Complex predicates such as the underlined
form in () are distinguished from a modifying converb plus main verb con-
struction in that in complex predicates nothing can intervene between V and
V, whereas Objects and other modifying elements can intervene between a
modifying converb and a main verb (cf. section ..). Moreover, complex
predicates express a single event whereas the combination of a modifying
converb plus a main verb expresses distinct events which take place sequen-
tially or simultaneously. Generally, complex predicates have a high expressive

Unless otherwise indicated, data on Wolaitta are based on the author’s native speaker intui-
tions and on data collected during fieldwork in Ethiopia. Other sources of data are non-
academic books written in the Wolaitta language. These include (i) Wolayttatto Leemisuwa
[‘Wolaitta Proverbs’], compiled by Getadrew Talachew and Tsegaye Ammenu, and pub-
lished in  by the Ethiopian Languages Academy; (ii) two books with a collection of
children’s stories: Taani Wolqaama [‘I am powerful.’], first printed in , by Qale Hiwot
Church, and Wolaytta Haysiyaa [‘Wolaitta tales’] published in  by Getachew Talachew.
Data from these works are cited by giving the book title and page number; (iii) Textbooks
on the Wolaitta language (grades two to seven), published in  by the Education Bureau
of the Southern Nations, Nationalities, and People’s Regional State. These are cited by
an abbreviated title and page number (e.g. G: , i.e. Wolaitta language textbook grade
two, page ). Full reference to these works is given in the References section. The author
adapted the transcription of these sources to broad phonological transcription and added
tone-accent marking.

In the examples in this chapter, complex predicates are underlined.
 Compound verbs and ideophones in Wolaitta revisited 261

power compared to single-predicate verbs that express related meanings.

For instance, the sentence in () is used in the following context: someone
was locked up in a house which was subsequently set on fire. The compound
kess- ʔekk- better expresses the drama of escape in this case than using for
example, bet- ‘go away to avoid danger/punishment’.
In example (), the complex predicate is the main verb of the clause.
Complex predicates may also be (co)subordinate to other verbs. The lat-
ter may be a dependent or main verb, a single-predicate or a complex verb.
Example () illustrates these possibilities. The sentence contains five clauses,
comprising four dependent clauses and a final matrix clause represented by
a complex predicate gufant-í ʔagg-iísi ‘he made fall forward’. Two of the four
dependent clauses are headed by complex predicates (underlined).
() mus-ídá godáa yé- ɗɗ í ʔek’k’-í
rot-PF:REL wall:M:ACC step.on-SS:A:CNV stand-SS:A:CNV
ment-ídí sóo gel-ó-saa-ra har-eé-ssi
break-SS:A:CNV home enter-REL-place-INST donkey-FEM:GEN-DAT
t’ólb-iyá-n kíy-í ʔek’k’-ídí
side-M:ACC-LOC go.out/climb-SS:A:CNV stand-SS:A:CNV
gufant-í ʔagg-iísi
fall.forward:CAUS-SS:A:CNV give.up-MSG:PF
‘(The hyena) broke the rotting wall by pushing it hard (and) the moment it
entered into the house it climbed (i.e. hit or pushed) the side of the donkey
hard and made it fall forward.’
[Taani Wolqaama, p. ]
The event expressed by the first complex predicate, yeɗɗ- ʔek’k’- takes
place prior to that expressed by the immediately following anterior con-
verb: mentídí ‘having broken’. Similarly the clause headed by the second
complex predicate, kíy- ʔek’k’-, is anterior to that expressed by the main
verb: gufantí ʔaggiísi ‘he made (it) fall forward’. The first two complex pred-
icates have the same V : ʔek’k’-. The latter verb indicates that the action
expressed by V is done forcefully. The verb ʔagg- ‘give up’, the V of the last
complex predicate in (), is productively used to indicate that the realisation
of the event expressed by V is ‘immediate’ (see section ... on the semantic
contribution of V).
Complex predicates in examples () and () have functional and struc-
tural similarities with verbal ideophones, which are also formed by com-
bining two predicative elements. An example of this is given in (), in


The term ‘cosubordination’ is used when the relation between two or more clauses in a
sentence is neither subordination nor coordination. A typical case is a chain of clauses with
converbs (cf. Van Valin and LaPolla ).
262 Amha

which the V is the ideophone t’úlʔu ‘fall, of something small’ and the V is
the verb g- ‘say’. The construction is a more expressive counterpart of the
verbal lexeme kund- ‘fall’. In the context of (), using the ideophone, the
speaker emphasises that she found it ridiculous that she failed the test on
her own mother tongue.
() kúnd-étt-íyo t’úlʔug-aási
fall-NMZ-F:ACC IDEOsay-SG:PF
‘I failed’ [when talking about a test on the Wolaitta language] G: 
Both complex predicates and ideophonic constructions are frequent
in narratives and conversation. The type of constructional meanings
expressed by V+V composites is varied. However, two salient and wide-
spread ones are: (i) a high degree of semantic specificity or expressiveness
about the event expressed by V; or (ii) the opposite of this, namely denot-
ing that the event expressed by V is not fully or completely carried out.
The following characterisation of the phenomenon in Qafar (Cushitic)
clearly summarises this:
VCP [verbal compound particle – AA] verbs may carry a distinct sense of a
lessening of the full vigor or expected duration of the activity/event denoted
by the base verb and for this reason they have sometimes been referred to as
‘Diminished Action’ forms … But this would not always prove an apt label,
for sometimes the difference of meaning conveyed by use of a VCP verb
would better be described as ‘emphatic’ or ‘dramatic’; compare the choice of
‘to pop up’ instead of ‘to appear’ in English. (Hayward : –)
Some languages use distinct morphological means to mark the distinction
between the two meanings of complex predicates mentioned above. For
example, in Amharic different verb stems are used for the ‘intensive’ and
‘attenuative’ meanings (cf. Appleyard ).
As already mentioned, V is an ideophone or a basically open class of ver-
bal lexemes given semantic compatibility with V. V, however, comprises a
restricted class. The most widely attested V verbs are translational equiv-
alents of the pair: SAY and DO/MAKE, respectively used in intransitive
and transitive complex predicates, e.g. in Wolaitta (Adams , Lamberti
and Sottile , Amha and Dimmendaal a), Hamar (Lydall ),
and Amharic (Leslau ). In Bench the same V màk is used in intransi-
tive and transitive contexts, thus translated as ‘say/do’ in Rapold (). A
number of languages including Tigre and Tigrinya (Ethio-Semitic), Bilin,
and Xamtanga (Cushitic), use the causative form of SAY to represent the
transitive V. In Oromo the verb for PUT is used whereas the V in Qafar is
translated as ‘DO, MAKE, PUT’ in Hayward (). In Somali the lexeme
for GIVE is used as a V component of ideophonic constructions (Dhoorre
and Tosco ).
 Compound verbs and ideophones in Wolaitta revisited 263

Another commonly attested V pair involves the directional verbs

COME-GO, which take semantically predictable V components such as the
motion verbs TAKE, RUN, FLY, SWIM, etc. This is the case in Omotic
languages, e.g. Haro (Hirut Woldemariam ), Wolaitta (Adams ), and
in Cushitic languages, e.g. in Awngi (Hetzron ), Oromo (Griefenow-
Mewis and Tamene Bitima ), as well as in Ethio-Semitic languages such
as Amharic and Tigrinya (Hetzron ).
In Wolaitta, V comprises the above-mentioned pairs plus several other
verbs, each of which contribute distinct semantic modification of manner
or duration or which add various senses of affective meaning to the event
expressed by V, as will be discussed in detail in sections .–.. In the
existing literature on other Omotic languages, we did not find a system as
extensive in the use and variety of V verbs as that reported for Wolaitta. For
example, work in progress by the present author on Zargulla (East Ometo,
Omotic) shows that either productively derived ‘intensive’ verb stems or
ideophonic verbs may be used as V to form complex predicates with hiyy-
‘say’ (intransitive) or his- ‘make say’ (transitive). However, apart from the
latter verbs, Zargulla has only a handful of other verbs which serve as V
components. These include ʔol- ‘give up, cease’, yew- ‘come’, and kess- ‘go
out’. It seems that the Cushitic language Kamabata has quite a number of
complex predicate forms comparable to those attested in Wolaitta (Yvonne
Treiss, p.c.). Future text-based comparative research among Omotic and
Cushitic languages is required to isolate language-specific and widespread
features of complex predicates.
The present work is based on two earlier studies which dealt with com-
plex predicates in Wolaitta in some detail: Adams () and Amha and
Dimmendaal (a). The aim is to expand coverage of the topic by includ-
ing data recently acquired. The overall data are examined in light of current
theoretical and typological–comparative studies. Contrary to earlier sug-
gestions, I argue that V verbs in Wolaitta are not auxiliary verbs. A con-
sequence of this position is to address the nature of semantic composition
between V and V.

. Complex predicates and the role of the converb in Wolaitta

With the exception of a few examples involving ideophones in which three
predicates are used, complex predicates in Wolaitta comprise two verbal ele-
ments: V+V. The construction is similar to compounds in that often the
composite meaning of V and V is not the sum of the component parts.
However, it does not involve base-form compounding. Rather, the complex
predicate contains a partially inflected V constituent and a V which may be
fully inflected or morphologically marked for (co)subordination. The V is
the converb form of any verb or an ideophonic verb, both of which are mor-
phologically dependent. They cannot head a clause on their own and they do
264 Amha

not inflect for verbal categories such as tense, aspect, and illocutionary force.
V is from a restricted class and it may take a full range of verbal inflec-
tional affixes if the complex predicate is the head of the clause, or it may
be marked by morphemes indicating dependency relations if the complex
predicate is not the head of the independent clause. That the V is always
a dependent form is one of the defining features that distinguish complex
predicates in Wolaitta (and related languages) from serial verb constructions
(cf. Aikhenvald : ). In serialising languages none of the component
verbs is formally marked as main or dependent verb.
In the next section we briefly introduce the morphosyntax of the converb
in Wolaitta to serve as a background for the discussion of converb-based
complex predicates in section ...

.. The structure of the converb

The converb in Wolaitta is a dependent verb form which gets tense–aspect,
mood, and modality value from the independent verb in the clause. There
are two types of converbs in this language: the simultaneous and the an-
terior/general converb, each of which is further distinguished for switch
reference. Same-subject converbs (both anterior and simultaneous) distin-
guish the gender and number of the subject but not person (cf. Amha and
Dimmendaal b). Examples in (–) illustrate the use of the anterior
converb wurs-ádá ‘having finished’ as head of the dependent clause ʔoós-
úwa wurs-ádá, which is followed by the inflecting main clause headed by
the verb šemp- ‘rest’. With the exception of third person masculine sin-
gular, all singular subjects are identified in the anterior converb clause by
affixing the morpheme -ádá to the verb root, as in (). The third person
masculine singular and plural subjects are marked by -ídí (cf. () and ()
respectively). The two variants of the same-subject anterior converb, -ádá
and -ídí are glossed differently using subscripted numbers: SS:A:CNV and
SS:A:CNV respectively.
() ʔoós-úwa wurs-ádá šemp-aúsu
work-M:ACC finish-SS:A:CNV rest-FSG:IPF
‘Having finished work, she is resting.’
() ʔoós-úwa wurs-ídí šemp-eési
work-M:ACC finish-SS:A:CNV rest-MSG:IPF
‘Having finished work, he is resting.’


The pragmatically unmarked order of constituents within a clause is SOV (with some flex-
ibility for focus purposes); lexical and clausal modifiers precede the head; the language has
only postpositions. The numerous inflectional and derivational markers both in the nomi-
nal and verbal morphology are exclusively marked by suffixes.
 Compound verbs and ideophones in Wolaitta revisited 265

Table .. Converb markers in Wolaitta.

Anterior converb Simultaneous converb

S-Subject D-Subject S-Subject D-Subject

,  & FSG -ádá -(í)n(í) -aíddá -íšin(i)
Plural and -ídí -iíddí
MSG

() ʔoós-úwa wurs-ídí šemp-oósi

work-M:ACC finsih-SS:A:CNV rest-PL:IPF
‘Having finished work, we are resting.’
As illustrated in (–), the converb shows partial agreement with the subject
in number or gender but it does not have the full range of inflectional pos-
sibilities as main verbs. Main verbs in Wolaitta show subject agreement for
each person, number, and gender and they also indicate tense and aspect
distinctions as well as distinctions in mood and negation.
In the examples in (–), the agent of the transitive converb wurs- ‘finish’
and the subject of the intransitive (main/matrix) verb šemp- ‘rest’ is the
same. When the subject of the converb and the main verb are different, the
converb morpheme -ín(í) is used, as in (). The different-subject anterior
converb is invariable for the number or gender of the subject.
() né ʔoós-úwa moor-ín ʔetí bor-étt-ídosona
SG:SBJ work-M:ACC spoil-DS:A:CNV PL:SBJ criticise-PAS-PL:PF
‘You having done the work incorrectly, they were criticised for it.’
The simultaneous converb has a similar number-gender and same- and
different-subject distinction as the anterior converb (illustrated in –).
Table . represents the converb marking morphemes in Wolaitta. The same-
subject anterior and simultaneous converbs are partially similar. Likewise
the two different-subject converbs are formally similar.
Besides adverbial subordination illustrated above, the same-subject con-
verbs are used as depictive secondary predicates, expressing optional parti-
cipant-oriented physical state or posture (cf. Amha and Dimmendaal ).
Another frequent use of the converb involves its occurrence as V of com-
plex predicates. Of the four sub-types of the converb in Table ., the same-
subject anterior converb is the most frequently used for this purpose. The
same-subject simultaneous converb is also used as V of a complex predicate
in a very restricted sense. It is only attested in complex predicates in which
the V is a durational or motion verb, e.g. y- come, b- ‘go’, or gel- ‘ enter’. An
example of this is given in (). Because of its relatively limited usage, the
simultaneous converb will not be further discussed in the present chapter.
266 Amha

() ʔó naag-iíddí ʔútt-ída ʔasa-t-í wot’-iíddí

MS:ACC wait.for-SS: S:CNV sit-PF:REL person-PL-NOM run-SS: S:CNV
gákk-ídí he kakáa-ppe ʔoík’k’-ídí kess-ídósona
arrive-SS:A:CNV that cliff: ACC-ABL hold-SS:A:CNV take.out-PL:PAST:AFF:DCL
‘The people who have been waiting for her arrived running and they held her and took
her out of the cliff.’
[Taani Wolqaama p. ]

As illustrated in example (), the different-subject anterior converb can

be directly affixed to a verb root, just like the same-subject anterior con-
verb. However, whereas the latter is regularly used as the V of a com-
plex predicate, no complex predicate is attested in which V is represented
by a verb root directly marked by the different-subject anterior converb
marker -ín(i). However, there is a special negative construction (identi-
fied as ‘the negative converb’ in Lamberti and Sottile ()), which com-
prises two verbal constituents: V can be represented by any verbal lexeme,
marked by the third person masculine singular negative morpheme -énna
plus-in(i). The V must be the verb ʔagg- ‘give up, cease’, also inflected in
the negative but the person, number, and gender value of V corresponds
to that of the subject of the main clause. Example () illustrates that V and
V inflect in the same way (morphemes highlighted). The subject in () is a
(covert) inanimate noun (ʔíra ‘rain’) and controls/triggers the third person
masculine singular agreement on the verb which is the default agreement
for inanimate subjects. In example (), on the other hand, the subject of
the main clause is a feminine singular noun. In this case V takes the third
person masculine negative form, whereas V takes the feminine negative
form.
() bukk-énna-n ʔagg-énna
rain-MSG:PRES:NEG:-DS:CNV give.up-MSG:PRES:NEG
‘It will most likely rain.’
[lit. ‘Without it raining, it will not give up.’]
() wónta k’aar-íya nu ʔeéssaa
earlier monkey-F:NOM SG:POS honey:ACC
mogol-énna-n ʔagg-úkku
spoil-MSG:PRES:NEG:-DS:CNV give.up-FSG:PRES:NEG
‘It is probable that that monkey spoiled our honey.’
Contrary to the analysis of Adams (), in the present chapter the above-
mentioned negative construction is not treated on a par with the affirmative
converb-based complex predicates illustrated in examples (–) above (see
 Compound verbs and ideophones in Wolaitta revisited 267

section .. for discussion). Accordingly, the discussion in section .. will
only address the same-subject anterior converb.

.. Converb-based complex predicates

A V converb in complex predicates is morphosyntactically distinct from
that used as a modifying adverbial clause discussed in section ... First,
converb-based complex predicates contain only two verbs whereas a series
of three or more clauses headed by a converb or another dependent verb
may be used in clause chains. Consider the example in () from Adams
(: ) in which four converbs occur consecutively; of these one (i.e.
gid-ídí ‘having become’) has a clause-internal modifier phrase (ʔoiddú ʔura
‘four person’).
() he pars-úwa ʔúy-í káll-ídí ʔoiddú
that beverage-M:ACC drink-SS:A:CNV be.satiated-SS:A:CNV four
ʔura gid-ídí dend-ídí
man:ACC become-SS:A:CNV get.up-SS:A:CNV
‘Having drunk that beverage, having been satiated, having become as
fat as four men and risen …’
Second, in complex predicates the converb markers -ídí and -ádá must
respectively be shortened as -í and -á (as in the examples –). Modifying
converbs, on the other hand, often take the long converb markers -ádá or
-ídí, but when two or more modifying converbs occur in sequence it is
possible that all but the last converb in the chain take the short form of the
converb. In (), for example, there are three consecutive converbs. The
first two (t’aaf- ‘write’ and ʔ ekk- ‘take’) are marked by the short converb
(-í ), but it is possible to alternate this by the long converb marker (-ídí )
and this will not result in any grammatical or semantic difference in the
sentence.


The transcription and glossing in Adams () is adapted to match the one followed by the
current author, e.g. long vowels are represented by repeating the vowel rather than using
diacritics and tone-accent marking is added. Translation is as in the original work.

The verb ʔekk-‘take’ is often used as V to form complex predicates with motion verbs
such as y- ‘come’, b - ‘go’, and paall- ‘fly’. In (), however, we do not analyse ʔekk- ‘take’
and ʔeh- ‘bring’ as a complex predicate because the verb ʔeh- ‘bring’ itself is a contracted
form of the complex predicate: ʔekk- ‘take’ + y- ‘come’, which is still used alternatively
to ʔeh-. Corresponding to this there is the verb ʔef- ‘take away’, a contracted form of the
complex ʔekk- ‘take’ and b- ‘go’. It may be suggested that the two complex predicates are
fully lexicalised and are treated like simple verbs to be used as V components. However,
the function of complex predicate such as ʔekk- y- ‘bring’ and ʔekk- b- ‘take away’ is
to express the direction of motion and since this meaning is already expressed in the
lexemes ʔeh- ‘bring’ and ʔef- ‘take away’, we prefer to analyse the sequence ʔekk- ‘take’
ʔeh- ‘bring’ in () as representing two converb clauses rather than a single clause with a
complex predicate.
268 Amha

() t’aaf-á geét-ett-idoo-ga t’aaf-í

write-SG:IMP say:RDP-PAS-PF:REL-M:NMZ:ACC write-SS:A:CNV
ʔekk-í ʔeh-iídí šeeɗɗ -iísi
take-SS:A:CNV bring-SS:A:CNV deliver-MSG:PF
‘He wrote (grudgingly) what he was ordered to write (i.e. a punitive
work), he took it, brought and submitted it (to the teacher)’ (G: )
In contrast, if the short converb markers (-í or -á) on the V of complex
predicates are replaced by their respective full converb markers, the result-
ing construction expresses two sequential events. In some cases, e.g. ʔútt-í
c’eegg-iisi (in example ), a sequential interpretation, i.e. ‘having sat, he
became old’, is somewhat odd and would be pragmatically unacceptable.
Finally, the converb and V of a complex predicate form a closely bound
syntactic unit and object nouns and other complements cannot intervene
between them. This inseparability restriction does not apply to a modify-
ing converb and an inflecting main verb. This is illustrated in (), in which
ʔasáa ‘person’ intervenes between the converb gukkídí ‘having spilled’ and
the main verb meés(i) ‘he/it eats’:
() tamá-n wott-ín pent-ída haattáy … gukk-ídí
fire-LOC put-DS:A:CNV boil-PF:S:REL water-M:NOM…spill-SS:A:CNV
ʔasáa m-eés
person:ACC eat-MSG:PRES:AFF:DCL
‘Boiling water (and other liquids) on fire boil over and burn people.’(G: )

Similarly in (), the object noun ʔubbábaa intervenes between the converb
form of the complex predicate čaddí t’eell- ‘examine, discuss’ and the main
verb tobbeés(i) ‘discuss, deliberate’:
() dereé ʔakeéka-n č’add-í t’eell-ídí
village.people:NOM careful-LOC stab/crush-SS:A:CNV see-SS:A:CNV
ʔubbábaa tobb-eés
every.thing:ACC deliberate-MSG:IPF
‘The villagers discuss everything, having carefully examined
(the subject matters of their meeting).’ (G: )
In (), the consecutive dependent verbs waass- ‘cry’, b- ‘go’, and súg- ‘push’
are each marked by full converb markers, whereas gel- ‘enter’, which is the
V of the immediately following complex predicate gel- ʔagg-, is and can only
be marked by the shortened -í. In this context the complex predicate: gel-í
ʔagg-ana (enter-CNV + give up-FUT) expresses the action of ‘entering into
the house’ and would involve ‘speed and unexpectedness’.
 Compound verbs and ideophones in Wolaitta revisited 269

() nu k’aála-n nu k’aála-n waass-iíddí b-iídí

PL:GEN word-LOC PL:GEN word-LOC cry-SS:S:CNV go-SS:A:CNV
he ʔifitt-áa súg-ídí gel-í ʔagg-ana
that door-ACC push-SS:A:CNV enter-SS:A:CNV give.up-FUT
yáa-g-iísi
that-say-MSG:PF
‘(The donkey) said (to the cat, chicken, and dog) “we will go each cry-
ing/shouting in our respective language and push that door (open) and
enter (the house) quickly”.’ (Taani Wolqaama, p. )

The order of the constituents of converb-based complex predicates in

Wolaitta is fixed: converb(V)-V. In most cases the main lexical meaning
of the complex predicate is contributed by the converb. V adds semantic
modification including manner, direction, intensity, certainty, and
duration, and/or grammatical information such as transitivity about the
situation expressed by the V (converb). There are also complex predi-
cates which do not show this semantic structure because the meaning of
the complex structure is not strictly compositional (see section .. on
the role of V).
Tense–aspect or mood is marked only once, on the V , if the complex
predicate is the head of the main clause. However, bearing verbal inflec-
tion is not the defining property of V verbs. Like monoverbal predi-
cates, the complex predicate can be nominalised or it can be put into a
subordinate relation to a main verb, in which case the morphemes which
indicate nominalisation or clausal dependency are marked on the V of
the complex predicate. This is illustrated in (), which contains two
juxtaposed sentences. The first complex predicate ʔútt-í c’eegg- ‘get old
without doing anything’ is the verbal head in the first sentence and is
marked for aspect and subject agreement. The complex predicate yé ɗɗ-a
ʔek’ k’- ‘to do a physical activity energetically’ in the second sentence,
is an adverbial clause that modifies the main verb goyy-. Accordingly,
the V of yé ɗɗ - ʔek’ k’- is affixed with the same-subject anterior converb
marker -ádá.
() ta naʔáa naʔá-y galbá -n ʔútt-í c’eegg-iisi
SG:GEN child child:NOM mat-LOC sit-SS:A:CNV get.old-MSG:PF
táání yéɗɗ-á ʔek’k’-ádá boóra goyy-aísi
SG:NOM step_on-SS:A:CNV stand-SS:A:CNV ox:M:ACC plough-SG:IPF
‘My grandson has become old without doing anything. I (still)
plough the land energetically (as a young and strong person would
normally do).’
270 Amha

Table .. V verbs in Wolaitta in Adams (: –).

Aspect type V lexical meaning (‘aspectual’) meaning in CPs

 ʔagg- ‘give up’ Immediacy

 beʔ- ‘see’ Trial
 ʔer- ‘know’ Experiential
 ʔekk- ‘take’ Certainty
 ʔak’- ‘spend the night’ Agreement aspect
 han- ‘happen’ Imminence
 bess- ‘show’ Obligation
 wóɗɗ - (itr.) ‘climb down’ Suddenness
ʔekk- (tr.) ‘take’
 péʔ- ‘spend the day’ Expectation
 (repetition of (lexical meaning of Intensification
converb) converb)
 ʔaggénna ‘it will not cease’ Probability
 g- ‘say’ Decisive/compulsory aspect
 wott- ‘put down’ Unconcern/precedence/
preparation
 ʔútt- ‘sit’ Durative, expectancy

.. Inventory and classification of V verbs in converb-based

complex predicates
The number of verbs used as V of complex predicates is restricted. Adams
() includes fourteen V verbs (see Table .), which he labelled as Aspect,
Aspect, etc. However, in his discussion Adams mentions another verb that
is not included in this list, namely báy- ‘be lost, disappear’, noting that this
is an example of ‘other auxiliary verbs … (that) seem to cover a wide area in
the manner of action and have to be glossed differently in different contexts’
(p. ).
Based on various morphosyntactic criteria, Amha and Dimmendaal
(a) exclude four of the above fourteen Vs from their own list. These are
han- ‘happen’, bess- ‘show’, ʔaggénna ‘he/it will/does not give up/cease’, and
‘repetition of converb’ as V verbs. Below these four cases will be examined
and the reason(s) for their exclusion is/are justified.
The construction with the first two V verbs that are excluded, i.e. han-
‘happen’ and bess- ‘show’, is structurally similar. Thus the two will be dis-
cussed together. Adams treats the verbs han- ‘happen’ and bess- ‘show’ in
() and (), taken from Adams (: ), as V verbs which respectively
express constructional meanings imminence and obligation.
 Compound verbs and ideophones in Wolaitta revisited 271

() sóo gel-aíčč-a ʔíra-y bukk-aná(-w)

home enter-ASP-SG:IMP rain-M:NOM rain-FUT-DAT
han-eési
happen-MSG:PRES:AFF:DCL
‘Get into the house quickly, for it is about to rain.’
() naʔá-y ba aawáa k’aálaa síy-ána-w bess-eési
child-M:NOM his father word:ACC hear-FUT-DAT show-MSG:IPF
‘The child ought to listen to his father’s words.’
The underlined predicates in () and () have the structure Vroot-FUT-
DAT + inflected V pattern. The reasons for not treating these as com-
pound verbs include that the Vroot-FUT-DAT constituent is a nominalised
clause which is derived from an independent (main) clause inflected for
future tense. It is thus different from the complex predicate types we focus
on in the present chapter, namely those involving two verbal constituents,
the first one of which is not inflected for tense–aspect and polarity but only
partially for subject. The fact that () and () are nominalised clauses is
clear from the fact that the future tense verb in these examples can head
a clause on its own. Compare the main verbs bukk-aná ‘it will rain’ and
síy-ána ‘I/we/he/she etc. will hear’ with bukk-aná-w and síy-ána-w in
() and () respectively. Any of the three allomorphs of the dative case-
marking morphemes, i.e. -u/w, -ssi, and -yyo, can be used in this construc-
tion. Second, the examples in () and () and related constructions must
occur with Vs that form a semantically coherent class of verbs including
kóyy- ‘want, seek’, košš- ‘be necessary, be wanted’, and dandaʔ- ‘be able’,
which comprise verbs that are cross-linguistically widely used as modal
verbs, which express the obligation, necessity, probability, etc. of the event
expressed by V. These are better analysed as special constructions with
predictable meanings. Third, some of the Vroot-FUT-DAT + V construc-
tions allow modifiers between the two verbal components, whereas this is
not possible in the converb-based complex predicates we discussed in sec-
tion ... Compare example () with () and (), in which the modifiers
loytí ‘do well, properly’ and dáro ‘a lot, much, very’ intervene between the
dependent and main verbs.
() naʔá-y ba ʔaawáa k’aálaa síy-ána-w
child-M:NOM his father word:ACC hear-FUT-DAT
loyt-í bess-eési
make.good-SS:A:CNV show-MSG:IPF
‘It is highly proper that the child ought to listen to his father’s words.’

() naʔá-y ba ʔaawáa k’aálaa síy-ána-w dáro

child-M:NOM his father word:ACC hear-FUT-DAT a.lot
272 Amha

košš-eési
be.necessary-MSG:IPF
‘It is highly necessary that the child listens to his father’s words.’
The verb han- ‘be, happen’ in example (), is interpreted by Adams as an
aspectual category expressing imminence. This verb is semantically differ-
ent from the modal verbs such as košš- ‘be necessary’ and dandaʔ- ‘be able’.
However, the use of han- in () can be linked to modality in that, as a predic-
tion of what will surely happen in the immediate future, it can be interpreted
as an expression of ‘strong possibility’. Amha and Dimmendaal exclude the
verb bess- ‘show’ (in example ) on the grounds that there it combines with a
clausal complement. Distinct from this, we find bess- as V in converb-based
complex predicates, in which the constructional meaning ‘do lightly, of event
expressed in V’ is expressed. This is illustrated in (–).
() yaá-t-ídí wolakk-í bess-í-nne
that-do-SS:A:CNV mix-SS:A:CNV show-SS:A:CNV-COORD
ʔagg-í g-eettees
give_up-SS:A:CNV say-PL:IPF
‘We mix (the ingredients) lightly like this and further we leave
(the recipe) just like that.’
() he tókk-ídó-ge búlʔ-í búlʔ-í
that plant-PF:REL-NMZ:NMZ dig-SS:A:CNV dig-SS:A:CNV
bess-ídó-ge ʔiččášu laítta-ra woozz-eés
show-PF:REL-NMZ:NOM five year-INST be.ripe-MS:PF
‘(The ensete) which is planted becomes ripe in five years, (the soil
that is around it) being regularly lightly moved.’
For the above-mentioned reasons we propose to treat the Vroot-FUT-
DAT + V construction distinct from the complex predicate with V (same-
subject anterior converb) + V pattern.
The third case Amha and Dimmendaal (a) exclude from their list
of Vs is the repetition/reduplication of the V (converb), illustrated in the
underlined form in example (), from Adams (: ).
() mára-y baassá badaláa m-í m-iídi
calf-M:NOM baassa:POS corn-M:ACC eat-SS:CNV eat-SS:CNV


The verbal form yáat- ‘do like that’ is a combination of the deictic proclitic yáa- ‘that’ and
the verb root ʔoott- which is reduced to -t. The parallel intransitive verb yáan- ‘be like that’
comprises yáa- ‘that’ and the reduced form of the verb han- ‘happen’.

A highland plant known by the scientific name ensete ventricosum.
 Compound verbs and ideophones in Wolaitta revisited 273

káll-iísi
be.satiated-MSG:PF
‘The calf, having gorged himself on Bassa’s corn, was satiated.’
The reason for the exclusion is that reduplication of a constituent for the
expression of pluractionals is not unique to complex predicates. It is used
with adjectives to express plurality, as in (). Moreover, the function of ver-
bal repetition is not restricted to expressing ‘intensity’. It may be used to
express a distributive action as well, as illustrated in (). Moreover, redu-
plication can also be used for expressing (non-intensive) iterative actions, see
búlʔ-í búlʔ-í ‘dig soil regularly’ in () above.
() c’íma c’íma ʔasa-t-i šiik’-ídí.…
old old person-PL-NOM gather-SS:A:CNV
‘Several old people gathered and …’
() naa-t-á t’eég-í t’eég-í miiššáa
child-PL-M:ACC call-SS:A:CNV call-SS:A:CNV money:M:ACC
ʔim-iísi
give-MSG:PF
‘He called the children apart (each individually or in groups of two or
more) and gave them money.’
Moreover, unlike complex predicates which involve only two constituents,
converbs may be repeated three or four times to express the intensive or dis-
tributive. Finally, Wolaitta has a productive derivational affix -eret- express-
ing intensive or verbal plurality. For these reasons reduplication should not
be considered as a special type of complex predicate construction.
The fourth construction which Amha and Dimmendaal exclude from
their list of V verbs is ʔaggénna ‘he/it will/does, not give up/cease’. Example
() from Adams (: ) illustrates its use.
() wontó ʔetí gákk-énna-ni
tomorrow PL:NOM arrive-MSG:FUT:NEG-DS:A:CNV
ʔagg-énna gišša-u giig-iss-á
give.up-MSG:FUT:NEG reason-DAT be.prepared-CAUS-SG:IMP
‘Prepare things, because they will probably arrive tomorrow.’
In its usage in (), ʔaggénna ‘he/it will/does not give up/cease’ must be
preceded by a dependent clause derived from the third person masculine
singular negative declarative main clause form of any verb. In () this is
gákkénna ‘he/it will/does not arrive’, which is subordinated by suffixing
the different-subject anterior converb marker n(i). Amha and Dimmendaal
() claim that structurally the different-subject converb marker is parallel
274 Amha

to sentential complementisers. It occurs in the same slot in which comple-

mentisers such as -dani ‘in order to’ occur. Compare the verb gákk-énna-ni
‘without arriving’ in example () with gákk-énna-dan ‘in order not to
arrive’ and gákk-énna ‘he does/will not arrive’. Irrespective of the person,
number, or gender of the subject, the verbal form that precedes ʔagenna
‘he/it will/does not give up/cease’ is the third person masculine singular
form. The corresponding feminine form ʔaggúkku ‘she will/does not give
up’, or any of the other forms in the future negative paradigm, cannot be
used in this context. The literal translation of the underlined forms in ()
is: ‘he will/does not give up without arriving’. This is a special construc-
tion type involving two verbs inflected for tense and negative polarity. It is
structurally different from the kind of complex predicates we discussed in
section .. and in the present chapter it is not treated on a par with the
other verbs in Table ..
While excluding four of the fourteen Vs in Adams (), discussed
above, Amha and Dimmendaal add a large number of V verbs which were
not included in the earlier work. Some of these have a regular meaning asso-
ciated with them when they are used as V. Others form a fixed collocation
with a specific V converb and they express a unique and often unpredictable
meaning. For this reason Amha and Dimmendaal (a: –) classi-
fied complex predicates into two types:
(i) semantically asymmetrical compounds, which ‘involve freely generated
(V) converbs, given compatibility with the meaning of (V), which are
drawn from a closed list’;
(ii) semantically symmetrical compounds, involving unique verb combin-
ations because of which it is not possible to generalise the semantic
contribution of V on its own. One has to interpret each V+V com-
bination separately, even in cases in which the same V is used, as in
ʔek’k’- ‘stand up’ in the following two complex predicates: wóɗɗ-
ʔek’k’- [‘descend’ + ‘stand up’] ‘turn to a previous state, do again of an
activity’ vs. yéɗɗ- ʔek’k’- [‘step on, stamp’ + ‘stand up’] ‘do a physical
activity vigorously’.
The list of asymmetrical compound verbs in Amha and Dimmendaal
(a: ) includes ten of the fourteen given in Adams (), see Table
. above, plus the ten V verbs shown in Table .. The V and V in the
symmetrical compound verbs are shown in Table ..
The list of symmetrical complex predicates in Table . is not complete.
The following combinations are attested in text recently examined.
() kay- m-
tend to fire- eat-
‘become old, live long decently’
 Compound verbs and ideophones in Wolaitta revisited 275

Table .. Asymmetrical V verbs in Wolaitta (Amha and Dimmendaal a: ).

V lexical meaning constructional meaning

báy- ‘disappear’ counter expectation
digg- ‘remove something, forbid’ irreversible state (tr.)
kícc- ‘remove oneself’ irreversible state (itr.)
haík’k’- ‘die’ extreme degree (itr.)
wor- ‘kill’ extreme degree (tr.)
ʔiiss- ‘insist, persist’ continuity of negatively perceived
state, e.g. feel pain, make noise
kaall- ‘follow’ continuity/progress of positively
evaluated action, e.g. do work, study
šaác’c’- ‘spend the season/year’ duration
t’eell- ‘look at, examine’ adversive
wur- ‘be finished’ near-complete action

Table .. Symmetrical set in Wolaitta (Amha and Dimmendaal a: ).

converb (V) V meaning of complex

predicate
baizz- ‘sell’ m- ‘eat’ sell and squander
baizz- ‘sell’ ʔekk- ‘take’ sell (for one’s benefit)
bak’k’- ‘slap’ ʔoík’k’- ‘hold’ be too tight, stick to
something
k’at’t’- ‘twist’ zaar- ‘return’ capture a person by
misleading him/her
miic’c’- ‘laugh’ kaaʔ- ‘play’ make fun of someone
sing- ‘smell’ gákk- ‘arrive’ find intuitively or by
surprise
ʔaáɗɗ- ‘pass’ wóɗɗ- ‘descend’ turn over (intransitive)
ʔaatt- ‘let pass’ yégg- ‘add, drop’ turn over (transitive)
ʔekk- ‘take’ y- ‘come’ bring
ʔekk- ‘take’ b- ‘go’ take away

[the verb kay- means ‘to warm oneself by sitting close to the hearth/
campfire and keep it burning; inhale smoke of medicinal or sweet-
smelling plant(product)’]
276 Amha

() mokk- wull-

receive.guest- collapse-
‘disappear or hide after having been seen briefly’
() mokk- ʔekk-
receive.guest take
‘to welcome a guest, to host’
() wott- ʔekk-
keep/put- take-
‘loot/rob, ejaculate’
() kíy- wóɗɗ-
come.out- descend-
‘arrive in a place unexpectedly’
() wóɗɗ- ʔek’k’-
descend- stand.up-
‘turn to a previous state, do again of an activity that is supposed to
have been completed’
() yéɗɗ- ʔek’k’-
step.on- stand.up
‘do a physical activity vigorously’
() naak’k’- ʔoik’k’-
insert/exploit hold
‘have a good grip on something, do a handwork well’
() yigg- t’eéll-
measure see
‘be thoughtful, consider different sides of an issue’
() t’aát’- waat’-
wrap- bind.yoke-
‘conclude/summarise a speech, story, or an argument’
() ʔeét’t’- kiy-
burn go.out
‘be aflame, catch fire quickly; be very expensive of price’
() ʔeett- kess-
make.burn take.out
‘ignite, set fire to; make very expensive, of price’
 Compound verbs and ideophones in Wolaitta revisited 277

A striking absence in the list of verbs in Tables ., ., and . and in
the examples in (–) is the verb ʔimm- ‘give’, which is used in many seri-
alising languages and in languages with compound verbs to form complex
predicates (Bril : , Denis Creissels, p.c.).
Concerning the syntactic unity of complex predicates in Wolaitta, we
mentioned earlier that no other element can intervene between the two con-
stituents. Prosody is another feature that provides evidence of the syntac-
tic unity of complex predicates. Wolaitta is a tone-accent language. Words
in isolation have lexically determined tone patterns. However, when two or
more lexical items are put together to form phrases or compound words,
there are predictable patterns of reduction of accentual prominence. In
nominal compounds, for example, the reduction affects the rightmost con-
stituent (cf. Amha ). The same is observed in complex predicates, since
the lexical high tone of V is reduced and the prominent (acoustically salient)
tone-accent is that of the V. The writing behaviour of native speakers also
points to the prosodic unity just mentioned. Adams (: ) notes: ‘edu-
cated Wolaittas who are learning to write their own language, consistently
try to write the V [i.e. converb + V – AA] as one word, not realising that
because each verb can conjugate and has its own high pitched stress there
are two words involved, not just one’.
In fact, the reduction in the phonetic realisation of the high pitch/inten-
sity in the second verb could be one of the motivating factors for the speak-
ers’ writing of V+V as one word. The other factor could be the semantics
of the complex predicates. That is, complex predicates express a single
event and this may eventually lead speakers to treat the two components
like single words and merge the tone-accent pattern of the components
into a single prosodic unit. In school textbooks and in other published
documents there is inconsistency in writing complex predicates. The same
complex predicate may be written as one word in one place and as two
words elsewhere. Some of the entries of complex predicates in a Wolaitta–
Amharic dictionary published in  are written as a single word, e.g.

ʔaggibeʔiis [give.up + see] ‘he gave up completely’. Others are written as
two words, e.g., ʔaatti yeggiis [let.pass + drop, add] ‘he turned something
upside down’; still others are written with a hyphen separating the two
constituents, e.g. mokki-wulliis [receive.guest + collapse, fall] ‘disappear’.
The dictionary makers, all of whom are native speakers, do not explain
their writing convention.

. Ideophone-based complex predicates

Ideophones in Wolaitta are characterised by their highly specific and/
or expressive semantics and special morphological shape, e.g. use of

Ethiopian Languages Research Centre. . Wolayttatto Qaalatu Amaaratto Birshettaa
[Wolaitta Amharic Dictionary]. Addis Ababa: Institute of Ethiopian Studies.
278 Amha

reduplication and/or predictable pattern of vowel combinations in a word.

There are two types of ideophones in the language: adjectival and verbal. In
the present study we examine only verbal ideophones briefly, to draw atten-
tion to their structural and functional parallel with the converb-based com-
plex predicates discussed in section ...
Verbal ideophones constitute a complex predicate together with V verbs
g- ‘say’ and ʔoott- ‘do’. They may or may not involve reduplication, depend-
ing on the duration of the event expressed. Punctual events or onset of dura-
tional states or activity are often expressed by non-reduplicated CVC(C)V
forms. Consider the reduplicated and non-reduplicated ideophones in ()
and (), which are immediately followed by the verb g- ‘say’.
() kúrúru g- ‘to move, of small children, and as an insult for a short
adult person’
sáláli g- ‘to move in a slow, easy manner’
c’úrúru g- ‘to spill, of liquid coming out in a slow consistent manner’
(Amha : )
() láwhu g- ‘to leave or pass by quickly’
šóttu g- ‘to stand up abruptly’
jóllu g- ‘to become motionless and absent minded’ (Amha : )
Verbal ideophones involve sound symbolism: those with high vowels are
associated with smallness/lightness and those with mid vowels are associ-
ated with bigness/heaviness.
() túlku g- ‘to break easily, suddenly of something small’
tólku g- ‘to break easily, suddenly of something big’ (Amha : )
Ideophones are not inflected for verbal categories such as tense, negation,
and mood. The latter are marked on the V. Transitivity is distinguished by
alternating the V verbs: ʔoott- ‘do’ indicates that the ideophone is transitive
whereas g- ‘say’ indicates that it is intransitive. The following are examples:
() ʔissí-to wottáa díʔʔi g-iídí ʔubb-áa laagg-ídí
one-ORD race IDEO say-SS:A:CNV all-M:ACC chase-SS:A:CNV
simm-í ʔagg-ókkoní yáa-g-iísi
return-SS:A:CNV give.up-PL:IPF:Q that-say-MS:G:PF
‘He said “Don’t we all run at once and chase them away and
return?”’
(Taani Wolqaama, p. )
() guútta moóre ʔašo cácácácáca ʔoott-ádá tamá-n
little fat meat:ACC IDEO do-SS:A:CNV fire-LOC
 Compound verbs and ideophones in Wolaitta revisited 279

t’iit’t’-aásu
roast-FS:G:PF
‘She roasted a little fat meat on the fire.’
As can be observed from (), the semantic content of an ideophone can
simply be imitation of a sound during an activity. However, most ideophones
are not sound ‘imitation’ or ‘onomatopoeic’ forms. In general, ideophonic
forms express the event as well as information on the manner or duration of
the event. Most ideophones have corresponding ‘regular verbs’ with some-
what similar lexical content but lacking the effect of ideophones. Compare:
láwhu g- ‘leave/pass by quickly’ and b- ‘go’, šóppu g- ‘to spill (of liquid) at once,
in large quantity’ and gúkk- ‘spill’. Apart from indicating transitivity and ver-
bal inflectional categories such as tense–aspect, the lexical semantic contri-
bution of the V components g- ‘say’ and ʔoott- ‘do’ is not directly apparent.
However, the latter verbs are essential to identify the transitivity of the ideo-
phonic construction.
Like converb-based complex predicates, the ideophone and the V g- or
ʔoott- cannot be separated by other constituents. Moreover, the ideophone is
rarely extracted from the composite and used on its own.

. Transitivity and complex predicates

In the present context we use ‘transitivity’ in connection only with the pos-
sibility of a verb to take one, more, or no Direct Object, although the Object
nouns may or may not be overt and some verbs can be characterised as labile.
In Wolaitta, transitive verbs may be derived from intransitive ones by suf-
fixing the productive causative/transitive morpheme -is. With a number of
verbal lexemes non-productive final consonant alternations (or perhaps petri-
fied morphemes) distinguish intransitive/middle, transitive/causative, and
ditransitive verbs, e.g. gufann- ‘fall forward’ vs. gufant- ‘make fall forward’,
dagamm- ‘be scared’ vs. dagant- ‘scare someone’, in which the transitive verb
ends in -t. But there are a few examples in which the reverse of this, i.e. the
intransitive/middle form ends in -t and the corresponding transitive form
ends in -y: yiillot- ‘be angry’ vs. yiilloy- ‘make angry’, ʔámott- ‘crave, long
for (certain food)’ vs. ʔámoy- ‘make long for food’. In others the intransi-
tive form ends in a velar consonant whereas the transitive counterpart ends
in -ss: yeekk- ‘cry’ vs. yeess- ‘make cry’, t’uúk’k’- ‘explode’ vs. t’uuss- ‘make
explode’. Finally, there are a few that contrast only in ejective vs. non-ejec-
tive voiceless palatal consonant alternations: laac’c’- ‘lick’ vs. laacc- ‘make
lick’ and miic’c’- ‘laugh’ vs. miicc- ‘make laugh’. Some of these alternations
can be explained by phonological processes resulting from affixation of the
causative morpheme -is.
Of the twenty asymmetrical V verbs in Table . and ., twelve are
intransitive as they normally occur without a direct object when used
280 Amha

independently. The remaining eight take a direct object noun when used in
single-predicate clauses.
In complex predicate clauses, intransitive V verbs often combine with
intransitive V and transitive Vs combine with transitive V. However, this
is not a strict rule since a mismatch in transitivity is also attested as in (),
taken from Adams (: ), in which V is transitive and V intransitive.
The same V is used in () in which V is intransitive.

() m-í bay-iísi

eat-SS:A:CNV disappear-MSG:PF
‘He ate it carelessly.’

() ʔek’k’-í bay-iísi

stand-SS:A:CNV disappear-MSG:PF
‘He was just standing (not working).’

The verb bay- ‘disappear’ is one of the frequently used asymmetrical V

verbs and it contributes the meaning ‘counter expectation’ in such construc-
tions. In five short texts (all children’s stories) we found fourteen utterances
with complex predicates taking bay- ‘disappear’ as V. Of these, eleven have
intransitive V, matching the intransitive V bay- ‘disappear’. Two of these
are given in (–).
() daafur-ídí ʔetí naʔáa-kko gákk-íyo wode naʔaí
be.tired-SS:A:CNV PL:NOM boy-DIRC arrive-IPF:REL time child:M:NOM
ʔek’k’-í báy-ídí miic’c’-eés
stand-SS:A:CNV disappear-SS:A:CNV laugh-MS:IPF
‘Tired, when the people got there to rescue him, the boy was simply standing
there and laughing (although he had made an alarm call earlier).’
(Taani Wolqaama, p. )

() b-á t’aát-étt-ádá zinʔ-á

go-SS:A:CNV curl-REFL-SS:A:CNV lie.down-SS:A:CNV
bay-aasu
disappear-FS:PF
‘She went and lied down, curled up (instead of seeking solution for
her problem).’
The remaining three have transitive V verbs, two of the three involve the
same V, namely ʔagg- ‘give up, cease’; one of the two cases is given in
(). The third involves the verb ʔol- ‘throw’ (). The examples (–),
as well as the combination m- + bay- [‘eat’ + ‘disappear’] ‘eat carelessly’ in
example (), suggest that in the case of asymmetrical complex predicates,
 Compound verbs and ideophones in Wolaitta revisited 281

the transitivity value of V overrides that of V since in all attested cases of

mismatch the compound has the same transitivity value as V.

() godareé túma-kka g-iídí ʔagg-í bay-iis

hyena-M:NOM true-INCL say-SS:A:CNV give.up-SS:A:CNV disappear-MSG:PF
‘The hyena gave up saying “that is true, indeed” (although one would
expect that the hyena would know that what was told to him, that donkeys have
no heart, is not true).’ (Taani Wolqaama, p. )
() kútto-y k’int’áll-íya bak’k’-í-nne
chicken-M:NOM side.of.face-M:ACC slap-SS:A:CNV-COORD
yiic’oy-ídí ʔol-í bay-iísi
make.feel.dizzy-SS:A:CNV throw-SS:A:CNV disappear-MSG:PF
‘The rooster slapped (the man) hard, causing him feel dizzy and fall.’
(Taani Wolqaama, p. )
Similarly, in the symmetrical compounds, it is possible for the composites
to have different values for transitivity. However, the common pattern is for
V and V to match in transitivity. The list in Table . and the examples in
(–) comprise twenty symmetrical complex predicates. Out of these, V as
well as V of sixteen complex predicates have the same value for transitivity
(i.e. they are either transitive or intransitive) and the complex predicate the
two components form has a transitivity value identical to its component parts,
i.e. it is transitive if both components are transitive and intransitive if both are
intransitive. The remaining four complex predicates are the following: sing-
gakk- [‘smell’ + ‘arrive’] ‘find intuitively or by surprise’, mokk- wull- [‘receive.
guest’ + ‘collapse’] ‘disappear having been briefly seen’, ʔekk- y- [‘take’+
‘come’] ‘bring’, and ʔekk- b- [‘take’ + ‘go’] ‘take away’. In these four cases,
V is intransitive whereas V is transitive; and the derived complex predicate
is also transitive just like V. Thus, as we said for asymmetrical compounds,
where there is mismatch between the transitivity of V and V, the complex
predicate has the same value for transitivity as that of V.
For some of the constructional meanings, in fact, there are intransitive
and transitive V pairs. This is the case with the V verbs wóɗɗ- ‘descend’
and ʔekk- ‘take’, both of which express ‘suddenness’. However, ʔekk- com-
bines with transitive V (), whereas wóɗɗ- combines with intransitive ones
(). The two examples are from Adams (: ).
() táání ʔakeek-énna-ni kana-y táná
SG.SBJ be.attentive-NEG:IPF-DS:A:CNV dog-M:NOM SG.OBJ
sáʔʔ-í ʔekk-iísi
bite-SS:A:CNV take-MSG:PF
‘When I was not expecting it, the dog suddenly bit me.’
282 Amha

() ʔetí dap-íšini wodoró-y duut’t’-í woɗɗ-iísi

PL.SBJ stretch-DS:S:CNV rope-M:NOM break-SS:A:CNV descend-MSG:PF
‘While they were stretching the rope, it suddenly broke.’
In some sentence pairs the V is transitive, but alternating intransitive and
transitive Vs denotes subtle differences in meaning. For example, as simple
verbs, the pairs: ʔútt- ‘sit’ and wott- ‘put, set down, place’ are semantic-
ally related but they differ in transitivity. According to Adams (), as
V verbs the two have different meanings: wott- ‘put, set down’ expresses
‘unconcern/precedence/preparation’ whereas ʔútt- ‘sit’ expresses ‘durative,
expectancy’. However, it seems that ‘preparation, durative, expectancy’
characterise both verbs but the translation differences could be attributed
to their difference in valence. The reason for this claim is that, in all its
attestations so far, complex predicates with ʔútt- as V are reflexive/middle
in which the subject is both the agent of the event expressed by V and at
the same time the participant affected by it. This is illustrated in (), in
which the transitive cognition verb beʔ- ‘see’ in V position goes with the V
ʔútt- ‘sit’. In the judgment of the present author, the sentence is unaccept-
able if V is replaced by wott- ‘put, set down’ (this was not tested with other
native speakers). In contrast, in () the V can be either ʔútt- ‘sit’ or wott-
‘put, set down’. With ʔútt- ‘sit’ as V the ‘preparation, expectancy, duration,
etc.’ in example () affects the third person masculine singular subject.
The subject is here an experiencer waiting and anticipating the results of
the action he took. In contrast, with wott- as V the subject (in ) has an
agent or causative role.
() sintaá-ra deʔ-íya-ge gaamm-úwa demm-íbeénna
in.front-INST exist-IPF:REL-M:NMZ:NOM lion-M:ACC find-MSGPF:NEG
šin guyyeé-ge beʔ-í ʔutt-iisi
ADVST behind-M:NMZ:NOM see-SS:A:CNV sit-MSG:PF
‘The person (walking) in front did not see the lion, but the one (walking) behind
him saw the lion (and was prepared to escape but he did not warn his friend).’
(Taani Wolqaama, p. )

() sígaa-ssi šoroo-t-á šiišš-á wott-aasu

reconciliation:ACC-DAT neighbor-PL-ACC gather-SS:A:CNV put-FSG:PF
‘She (the wife) had gathered the neighbours to reconcile her with her husband.’

Example (), with a di-transitive verb kunt- ‘fill’, is taken from Adams
(: ); the translation is as in the original work. In this sentence, too,
if the V wott- ‘put, place down’ is replaced by ʔútt- ‘sit’, the interpretation
would be that the subject is ‘expecting’ or testing whether he would indeed
‘feel fear in the dark’.
 Compound verbs and ideophones in Wolaitta revisited 283

() t’úmaa-yyo yáyy-ennaa-dani lambá

darkness:ACC-DAT fear-MSG:FUT:NEG-COMP kerosene:ACC
kuraaz-iya-ni kunt-í wott-iisi
oil.lamp-ACC-LOC fill-SS:A:CNV put-MSG:PF
‘In order to not fear the dark, he took the precaution to fill the oil lamp
with kerosene.’

The semantically related verb pair haík’k’- ‘die’ and wor- ‘kill’ further illus-
trate the importance of matching in transitivity between V and V. Example
() illustrates that the intransitive V miic’c’- ‘laugh’ takes haík’k’- ‘die’
as V.

() ʔí miic’c’-í haik’k’-eesi

MSG:NOM laugh-SS:A:CNV die-MSG:IPF
‘He laughs/is laughing a lot.’

In contrast, the corresponding transitive verb miicc- ‘make laugh’ forms a

complex predicate with wor- ‘kill’, as in (). The verbs haik’k’- and wor-
are productively used as V constituents of complex predicates and their
function is characterised as ‘expressing an extreme degree for some action
or process expressed by V’ in Amha and Dimmendaal (a: ), from
which the examples in (–) are taken:

() ʔí ʔas-áa miicc-í wor-ees

MSG:NOM people-ACC laugh:CAUS-SS:A:CNV kill-MSG:IPF
‘He makes people laugh a lot.’
() ʔí ba bóllot-iyo ʔit’t’-í haik’k’-eesi
MSG:NOM LOG mother.in.law-F:ABS hate-SS:A:CNV die-MSG:IPF
‘He hates his mother-in-law badly.’

Transitivity concord is important in the symmetrical set of complex predi-

cates as well. In examples () and (), semantically related but formally
distinct intransitive and transitive V verbs wóɗɗ- ‘descend’ and yégg- ‘drop,
climb down, add’ form unique combinations with corresponding intransitive
and transitive verbs: ʔaáɗɗ- ‘pass’ and ʔaátt- ‘let pass’ (see also the list
in (–)).
() hargánc-íya zinʔó-sa-ppe ʔaá ɗɗ-á woɗɗ-aasu
patient-F:NOM lie.down-place-ABL pass-SS:A:CNV descend-FSG:PF
‘The patient turned over in her bed.’
284 Amha

() mac’c’aas-íya ʔoíttáa ʔaatt-á yegg-aasu

woman-F:NOM bread:ACC let.pass-SS:A:CNV drop-FSG:PF
‘The woman turned over the (flat)bread (in the baking pan).’ (Amha
and Dimmendaal a: )
Finally, as discussed in section ., the selection of a V is the only clue as
to the transitivity of an ideophonic construction: an ideophone + g- ‘say’
combination is always intransitive whereas ideophone + ʔoott- ‘do’ is always
transitive.
The above-mentioned cases of matching transitivity of V and V are com-
parable to what Bril () refers to as ‘transitive concord’ in the Oceanic
language Nêlêmwa. This language has morphemes indicating transitivity.
In ideophonic constructions, the choice of V helps identify transitivity
because the form of the ideophone itself does not alter in intransitive and
transitive constructions. In the case of converb-based complex predicates,
however, matching in transitivity among the verbs is the norm, with just a
few cases violating this tendency. Alternating V verbs (see the lists in Tables
. and .) in asymmetrical compounds do not turn a transitive verb into an
intransitive one or the other way round. Rather, each V has its own seman-
tic import. We have an example where a passive derivational morpheme is
affixed to the V component of the complex predicate, thereby affecting argu-
ment structure (). In this example, the transitive V verb m- ‘eat’ takes the
passive derivational suffix -ett (it is the lexical property of this minimal verb
root form C- to reduplicate the derivational suffix). The V used in combin-
ation with this derived form is ʔagg- which is also transitive.
() ʔetá ʔoík’k’-ana g-iíddí c’iimmá haatta-n
PL:NOM catch-FUT say-SS:S:CNV sea:GEN water-LOC
gel-ídí m-eét-ett-í ʔagg-iis
enter-SS:A:CNV eat-RDP-PAS-SS:A:CNV give.up- SG:PF
‘Intending to catch them, (the baboon) fell in the lake and drowned.’
[lit. ‘it was eaten by water’]
We have shown in sections . and . that converb- and ideophone-based
complex predicates in Wolaitta are not used to introduce non-subcategorised
arguments. The exception we mentioned was ʔekk- ‘take’, the converb form
of which is used to introduce an instrumental noun. However, in this case,
ʔekk- is not part of a complex predicate because it is used without V and its
tone-accent is not phonetically reduced.
() č’ač’č’ap-ídí k’aatt-oós míttaa ʔekk-ídí
sprinkle-SS:A:CNV stir-PL:PRES:AFF:DCL wood:ACC take-SS:A:CNV
‘Having sprinkled (the mixture of spices and butter-milk over the
ensete) we mix it with wood.’
 Compound verbs and ideophones in Wolaitta revisited 285

() mel-ída-ga zazzár-iya ʔekk-ídí zazzar-ídí

dry-PF:REL-M:NMZ:ACC sieve-M:ACC take-SS:A:CNV sieve-SS:A:CNV
‘That which became dry, we sieve with a sieve and …’
The verb ʔekk- and its object noun are used as alternative expressions to
noun + the instrumental or comitative case-marker -ra or noun + the loca-
tive marker -n, which can also be used to mark the instrumental. Replacing
this N+Affix form by a Verb+Noun construction may be in order to upgrade
the grammatical role of the instrument noun into a core constituent.

. The status of V verbs

Adams (:) analysed V verbs as auxiliaries. He notes: ‘[F]or each of
these verbs there exists a homophonous verb that occurs alone, as the sole
head element of a verb phrase, and has its own independent (lexical) mean-
ing, quite distinct from the grammaticalised “aspectual” meaning expressed
by the auxiliary verb counterpart.’ Amha (: ) refers to V as ‘an adver-
bial modifier or an auxiliary verb’ without stating the motivation for the ter-
minology, whereas Amha and Dimmendaal (a: ) write that V verbs
‘are best analysed as grammaticalized verb forms as a result of their frequent
occurrence in the main verb slot and their semantic interaction with the pre-
ceding verb’ but they do not specify the lexical status of V verbs as ‘auxil-
iary verbs’.
Treating V verbs as auxiliaries would not be correct despite differences
among V verbs in the level of grammaticalisation. There are formal/pho-
nological, structural or grammatical, and semantic reasons. Each of these is
considered in turn.
All of the V verbs are used as a single-predicate verb with their basic lex-
ical meaning. Phonologically, the verbs are exactly the same when they are
used on their own and when they function as V verbs in complex predicates,
i.e. there is no formal reduction in the segmental make-up of the lexeme.
The only difference is the reduction in the phonetic prominence of the tone-
accent of the V in complex predicates. However, this affects phrases and
compounds in general, including N + N compounds, and it cannot be used
as a criterion to distinguish main and auxiliary verbs. Analysing formally
identical verbs both as major verbal lexemes as well as auxiliaries would
result in unnecessary duplication of predicate forms in Wolaitta. Besides,
V verbs in Wolaitta constitute a relatively large number of verbs whereas
clear cases of auxiliaries in genetically or areally related languages are few in
number.
For example, the Ometo languages Zayse and Zargulla have auxiliary or
copula verbs, e.g. yéne/yéšše ‘be (present/past)’, and the negative auxiliary
baáʔa (past tense). These are used on their own (only) in nominal clauses;
otherwise they are used in combination with lexical verbs to express aspect
286 Amha

or negation, as with the verb ʔač’č’- ‘cut’ in ʔač’č’-aá-tte yéne ‘he/she, etc.
is cutting’, ʔač’č’-aá-tte yéšše ‘he/she, etc. was cutting’, ʔač’č’-aá baáʔa
‘he/she does not cut’, which contrast with the simple predicate verb in
the past: ʔác’c’átteséne ‘he cut’. Wolaitta has zero copula in the affirma-
tive; some of the functions expressed by auxiliaries in Zargulla (e.g. past/
present progressive and negation) are denoted in Wolaitta through verbal
affixes. There are a few cases of multi-verb constructions which can argu-
ably be analysed as auxiliary constructions. For example, some speakers
use the existential verb deʔ- in combination with another lexical verb to
express the progressive, e.g. k’oid-aídda daisi ‘I am counting’ [comprising
k’oid- ‘count’ and a reduced form of deʔ- ‘exist’]. This usage is alternative
to verbal affixes -aisi (first person), -aása (second person), etc., which can
be directly affixed to the lexical verb to mark the present tense or pro-
gressive aspect: k’oid-aísi ‘I count/am counting’, k’oid-aása ‘you count/
are counting’.
There are a number of structural reasons for not analysing V as auxiliary.
First, two verbs that belong to the class of V verbs in Wolaitta can combine
with each other to form a complex predicate that can, on its own, express
a state of affairs. That is, Vs greatly contribute to the semantic content of
an utterance. For example, the verbs ʔagg- ‘give up’ and báy- ‘disappear, be
lost’ are among the group of verbs that are used as V verbs. In (), how-
ever, the two verbs are combined to form a complex predicate. Two or more
auxiliaries may combine in languages that have auxiliaries as sub-types of
predicates, e.g. Dutch. However, these would still need another lexical verb
to form a complete utterance.
() ʔasá-y ʔaš-úwa šaakk-íyo-ga ʔagg-í
people-M:NOM meat-ACC divide-IPF:RELM:NMZ:ACC give.up-SS:A:CNV
bay-ídí keett-áa tamáa toiss-aná-w
disappear-SS:A:CNV house:GEN fire-ACC extinguish-FUT-DAT
wottá-n b-iís
race-LOC go-MSG:PF
‘The people gave up dividing the meat and ran to extinguish the fire which
caught the house.’

Second, one and the same verbal lexeme may be used as part of V as well as
V. This is illustrated in (), in which the verb g- ‘say’ occurs twice. Apart
from complex predicates involving ideophones we did not encounter com-
plex predicates comprising three verbal constituents.


The singular noun ʔasá ‘person’ is often used with plural reference (cf. the plural
form: ʔasatí ‘the people’) and triggers third person masculine singular agreement on the
verb.
 Compound verbs and ideophones in Wolaitta revisited 287

() wónta-y [waáʔʔi g]-ií g-iisi

day.light-NOM IDEO say-SS:A:CNV say-MSG:PF
‘It became daylight, just like that.’
In its first occurrence (in square brackets), g- is a co-predicate with the
ideophone waáʔʔi and provides the lexical meaning: ‘become (day)light’. In
its second occurrence, it is a V with the constructional meaning ‘decisive-
ness, compulsory action’. The third structural reason involves argument
structure. One of the general criteria for distinguishing between auxiliaries
and inflecting verbs in complex predicates is that auxiliary verbs lack argu-
ment structure whereas V components of complex predicates have argument
structure (cf. Baker and Harvey, this volume). That V verbs in Wolaitta have
argument structure is evident from the selection of certain V verbs accord-
ing to the valence of V. Alternating transitive and intransitive V may result
in distinct interpretations of the same construction, as we showed in section
.. In ideophonic constructions, transitivity of V is crucial in interpreting
the construction as transitive or not (cf. section .). Finally, V+V (both of
the symmetrical as well as asymmetrical type) as a unit get subordinated to
another inflecting single or complex predicate verb; they can be nominalised
or function as heads of a relative clause, whereas auxiliaries do not.
V verbs contribute to the semantic content of the complex predicate.
Unlike auxiliaries, their contribution in this regard is not limited to facilitat-
ing the tense–aspect, mood, or modality interpretation of V. These are spe-
cial constructions used mainly for discourse/pragmatic effects; they are not
grammatically determined, obligatory usages. As mentioned earlier, for most
of the complex predicates there are corresponding single-predicate verbs.
Complex predicates express both the event as well as the manner, direction,
duration, intensity, etc., within the verbal domain, whereas some languages
tend to do this by using adjunct adverbial expression. It is to be noted that
Wolaitta has very few manner adverbs.
It seems that a compositional analysis would better capture the function
of complex predicates in Wolaitta, rather than treating them as auxiliaries
(cf. Baker and Harvey, this volume). Wolaitta complex predicates, discussed
in sections .–., have a number of structural parallels to what Baker and
Harvey refer to as ‘merger constructions’ in which ‘the predicate information
from the contributing constituents merges where they have common concep-
tual structure. This method produces predicate structures whose range classes
with the range of predicate structures found in monomorphemic predicates
(p. ). In the present contribution we did not attempt to methodically test the
application of these ideas to the Wolaitta data. The work is limited to outlining
the morphosyntax and some aspects of the semantics of complex predicates in
the language. More research is needed on conceptual structure and lexicalisa-
tion patterns of simple and complex predicates in Wolaitta in order to better
understand the nature of the semantic composition of V+V.
288 Amha

A pertinent but difficult question in this regard is whether there is a

principled correlation between the lexical specification of a certain verbal
lexeme and the specific interpretation it contributes when used as V in
complex predicates. In some cases the semantic motivation for the selection
of V is directly apparent from the meaning of the verb when it is used in
single-predicate clauses. Note, for example, the verbs ʔak’- ‘spend night’
and péʔ- ‘spend day’ contributing the meaning ‘durational’ when used as
V constituents. On the other hand, why certain semantic interpretations
are attributed to some V verbs is difficult to specify. For example, why is
it that just g- ‘say’ and not any of the other Vs that can express ‘decisive-
ness’? Or, how is the sense of ‘precedence, preparedness’ is imparted by the
V pairs ʔútt- ‘sit’ and wott- ‘put, set down’ on V related to the use of these
same verbs as independent verbal heads? These are questions for further
research.

. Conclusion
Complex predicates are used in a number of Cushitic, Omotic, and Ethio-
Semitic languages. In the present study we discussed the phenomenon in
Wolaitta (Omotic). In this language, V of complex predicates may be rep-
resented by converbs or ideophonic verbs, both of which have restricted
inflectional possibilities. Unlike Amharic and Tigrinya (Ethio-Semitic),
Qafar (Cushitic), and Zargulla (Omotic), in Wolaitta there are no product-
ive derivational stems which serve exclusively as V components of complex
predicates. In comparison with the former languages, Wolaitta employs
a larger number of V verbs which are productively used to add different
semantic senses to the event expressed by the V. Complex predicates formed
by using such V verbs with predictable semantics are analysed as asymmet-
rical complex predicates. In addition to these, there are complex predicates
in which the V+V combination is fixed. Such complex predicates appear
to be more lexicalised. The latter are described as symmetric compound
verbs in earlier work. Despite their difference in the lexical–grammatical
cline, both asymmetrical and symmetrical compounds are formed using the
same morphosyntactic strategy. Both manifest the same construction type
involving composition or merger. Moreover, the function of the merger in
both types is to denote complex activities or events that could be viewed or
perceived as one event, e.g. an activity and the manner in which it is carried
out. Future research is needed to fully understand semantic composition in
different types of complex predicates in Wolaitta.

References
Adams, B. . A Tagmemic Analysis of the Wolaitta Language. PhD dissertation,
University of London.
 Compound verbs and ideophones in Wolaitta revisited 289

Aikhenvald, A. Y. . Serial Verb Constructions in Typological Perspective. In

Aikhenvald, A. Y. and Dixon, R. M. W. (eds) Serial Verb Constructions: A
Cross-Linguistic Typology, –. Oxford: Oxford University Press.
Amberber, M., this volume. The Structure of the Light Verb Construction in
Amharic.
Amha, A. . Tone-accent and Prosodic Domains in Wolaitta. Studies in African
Linguistics , :–.
. Ideophones and Compound Verbs in Wolaitta. In Kilian-Hatz, C. and.
Voeltz, F. K. E. (eds) Ideophones, –. Amsterdam/Philadelphia: John
Benjamins.
Amha, A. and Dimmendaal, G. . Secondary Predicates and Adverbials
in Nilotic and Omotic: A Typological Comparision. In Himmelmann,
N. P. and Schultze-Berndt, E. (eds) Secondary Predication and Adverbial
Modification: The Typology of Depictives, –. Oxford University Press.
a. Verbal Compounding in Wolaitta. In Aikhenvald, A. Y. and Dixon, R.
M. W. (eds) Serial Verb Constructions: A Cross-Linguistic Typology, –.
Oxford: Oxford University Press.
b. Converbs in an African Perspective. In Ameka, F., Dench, A., and Evans,
N. (eds) Catching Language: The Standing Challenge of Grammar Writing,
–. Berlin: Mouton de Gruyter.
Anderson, G. D. S. . Auxiliary Verb Constructions. Oxford University Press.
Appleyard, D. . The Verb ‘to say’ as a Verb Recycling Device in Ethiopian
Languages. In Zaborski, A. (ed.) New Data and New Methods in Afroasiatic
Linguistics: Robert Hetzron in Memoriam, –. Wiesbaden: Harrassowitz
Verlag.
Baker, B. and Harvey, M., this volume. Complex predicate formation.
Bril, I. . Complex Verbs and Dependency Strategies in Nêlêmwa (New
Caledonia). In Bril, I. and Ozanne-Rivierre, Françoise (eds) Complex Predicates
in Oceanic Languages, –. Berlin; New York: Mouton de Gruyter.
Cohen, D., Simeone-Senelle, M-C., and Vanhove, M. . The Grammaticalization
of ‘say’ and ‘do’: An Areal Phenomenon in East Africa. In Güldemann, T. and
von Roncador, M. (eds) Reported Discourse: A Meeting Ground for Different
Linguistic Domains, –. Amsterdam/Philadelphia: John Benjamins.
Dhoorre, C. S. and Tosco, M. . Somali Ideophones. Journal of African Cultural
Studies.  –.
Fleming, H. . Omotic Overview. In Bender, M. L. (ed.) The Non-Semitic
Languages of Ethiopia, –. East Lansing: African Studies Center,
Michigan State University.
Griefenow-Mewis, C. and Bitima, T. . Coordination of Actions in Oromo by
Means of Lengthening the Final Vowel of the Conjugated Verb. Paper pre-
sented at The Third World Congress of African Linguistics, Leipzig.
Hayward, R. J. . Compounding in Qafar. Paper Presented at the Twelfth
International Conference of Ethiopian Studies, Michigan State University,
East Lansing, Michigan, – September .
Hetzron, R. . The Verbal System of Southern Agaw. University of California
Near Eastern Studies, vol. . Berkeley, CA: University of California Press.
. Ethiopian Semitic: Studies in classification. Manchester: Manchester
University Press.
290 Amha

Lamberti, M. and Sottile, R. . The Wolaytta Language (Studia Linguarum

Africae Orientalis, ). Köln: Rüdiger Köppe Verlag.
Leslau, W. . The Influence of Cushitic on the Semitic Languages of Ethiopia: A
Problem of Substratum. Word , :  –.
. Reference Grammar of Amharic. Wiesbaden: Harrassowitz Verlag.
Lydall, J. . Having Fun with Ideophones: A Socio-linguistic Look at Ideophones
in Hamar, Southern Ethiopia. In Yimam, B., Pankhurst, R., Chapple, D.,
Admassu, Y., Pankhurst, A., and Teferra, B. (eds) Ethiopian Studies at the End
of the Second Millennium. Proceedings of the XIVth International Conference of
Ethiopian Studies, – November, , Addis Ababa University, Ethiopia. vol.
III, –. Addis Ababa: Institute of Ethiopian Studies.
Rapold, C. . Towards the Grammar of Benchnon. PhD dissertation, Leiden
University.
Talachew, G. . Wolaitta Haysiya [Wolaitta Tales] . Addis Ababa: Ethiopian
Languages Research Centre, Addis Ababa University.
Talachew, G. and Amenu, T. . Walayttatto Leemisuwa: Yewolaytigna
Missaliyawi Nigiggiroch Kenneamarigna Tirgumachew. Addis Ababa: Ethiopian
Languages Academy.
Van Valin, R. D. and LaPolla, R. J. . Syntax: Structure, Meaning and Function.
Cambridge: Cambridge University Press.
Wolaitta Qale Hiwot Church. . Taani Wolqaama: Wolaitta Haisiya. [No
publisher]
Woldemariam, Hirut. . The Grammar of Haro with Comparative Notes on the
Ometo Linguistic Group. PhD dissertation, Addis Ababa University.
10 The structure of the light verb
construction in Amharic*

MENGISTU AMBERBER

. Introduction
The aim of this chapter is to provide a detailed structural analysis of the light
verb construction in Amharic. In particular, the light verb construction will
be systematically compared with its morphologically simpler counterpart in
order to reveal the formal and semantic differences between the two types of
predicates. The key assumption of our analysis is that the light verb construc-
tion is derived through employing exactly the same independently motivated
morphosyntactic machinery. We show that the derivation of the light verb
construction does not require any additional mechanism over and above what
is required in the derivation of morphologically simple predicates.
In order to investigate the properties of the light verb construction, it is
important to start with a general understanding of the organisation of the
verb system in the language. As a Semitic language, Amharic employs the
so-called root-and-pattern morphology as its main word-formation strategy.
Root-and-pattern morphology is characterised by a root which consists of
consonantal radicals and a pattern which comprises consonantal positions
and vowels. In general, the roots encode lexical meaning, whereas the pat-
terns encode grammatical meaning. For example, the verb səbbərə ‘he broke
(TR)’ consists of the triradical root
sbr ‘break’ and the pattern CəC C əC-
encodes the perfect conjugation. The verbal noun of
sbr ‘break’ is məsbər ‘to
break’ which is formed by attaching the prefix mə- to the pattern -CCəC.
The overall predicate structure for a main clause predicate can be repre-
sented as follows: (a) perfect: [verb + subject + (object)]; (b) imperfect: [sub-
ject + verb + (object) + allə + subject]. Note that the (compound) imperfect
involves the use of the verb allə ‘be’, which is itself marked for person, num-
ber, and gender. This can be seen in () where examples of the (a) perfect
and (b) compound imperfect are shown.

* I would like to thank Azeb Amha, Baye Yimam, Brett Baker, Mark Harvey, Mary
Laughren, and an anonymous reviewer for useful comments and discussion. Of course, all
the usual disclaimers apply. Some sections of this chapter are slightly revised versions of
material that appear in Amberber (b) and Amberber (, Ch. ).

Abbreviations used.  = st PERSON,  = nd PERSON,  = rd PERSON, ACC = ACCUSATIVE,
ANTC = ANTICAUSATIVE, ATT = ATTENUATIVE, AUX = AUXILIARY, CAUS = C AUSATIVE, COMP

291
292 Amberber

() a. səbbər – əčč – (ɨw)

cut.PF-F-(MO)
‘She broke (it).’
b. tɨ -səbr-(əw)- all -əčč
F-break.IMP-(MO)-AUX-F
‘She breaks/will break (it).’
Thus in the perfect, subject agreement is expressed by a suffix, whereas in
the imperfect it is expressed by a combination of a prefix and a suffix.
Given its root-and-pattern morphology, it is not surprising that the
Amharic verb can be classified on the basis of the number of consonan-
tal radicals. Thus, there are biradicals, triradicals, quadriradicals, and
pluriradicals.
() Biradical ( radicals:
sm) səmma ‘hear’
Triradical ( radicals:
sbr) səbbərə ‘break’
Quadriradical ( radicals:
mskr) məsəkkərə ‘testify’
Pluriradical ( or more radicals: (tə)-brəkərrəkə ‘tremble’

brkrk)
The verbs are further classified into major conjugational classes (conj cl). For
example, triradical verbs are classified into three major classes traditionally
known as Type A, Type B, and Type C. This classification is based mainly on
the vocalic pattern of the stem and gemination, that is, whether or not the
second radical of the root is geminated throughout the conjugation (Leslau
: ff). In all three types of classes, the second radical is geminated in
the perfect.
() a. Perfect
Type A səbbərə ‘break (TR)’
Type B fəlləgə ‘search’
Type C marrəkə ‘cause to surrender’
b. Imperfect
Type A yɨsəbral ‘break (TR)’
Type B yɨfəllɨgal ‘search’
Type C yɨmarrɨkal ‘cause to surrender’

= COMPLEMENTISER, CV = COVERB, DEF = DEFINITE, F = FEMININE, GER = GERUND, IMP

= IMPERFECTIVE, IMPER = IMPERATIVE, INCH = INCHOATIVE, INT = INTENSIVE, INTR =
INTRANSITIVE, M = MASCULINE, NEG = NEGATIVE, O = OBJECT, PART = PARTICLE, PASS
= PASSIVE, PF = PERFECT, PL = PLURAL, POSS = POSSESSIVE, RECIP = R ECIPROCAL, REFL =
REFLEXIVE, S = SINGULAR, TR = TRANSITIVE.
 The structure of the light verb construction in Amharic 293

In Type A verbs, such as səbbərə ‘break (TR)’, the second radical is gemi-
nated in the perfect only (a). In Type B verbs, such as fəlləgə ‘search’, the
second radical is geminated in all forms. In Type C verbs, such as marrəkə
‘cause to surrender’, the vowel a occurs after the first radical.
In terms of derived transitivity, there are derivational prefixes (anticausa-
tive, passive-reflexive, causative) that attach to a basic stem to derive a verb
with certain lexical semantic and morphosyntactic properties. The morpho-
logical passive is formed by attaching the prefix t(ə)- to the verb stem. This
is realised as tə- before consonants and as t- before vowels. Thus, consider
the examples in ():
() a. aster dɨ nggay-u-n wərəwwər -əčč
A. stone- DEF-ACC throw.PF-F
‘Aster threw the stone.’
b. d ɨnggay-u tə-wərəwwər -ə
stone-DEF PASS-throw.PF-M
‘The stone was thrown (by someone).’
The same formal strategy, the prefix t (ə)-, is employed to derive the incho-
ative or anticausative. The term ‘anticausative’ is employed here to refer to a
derived intransitive form which is non-passive in its morphosyntactic prop-
erties (see Haspelmath () for a similar use of the term). Consider the
examples in ():
() a. səbbərə ‘break’ tə-səbbərə ‘break (INTR)’
b. kəffətə ‘open’ tə-kəffətə ‘open (INTR)’
c. bəttənə ‘scatter’ tə-bəttənə ‘scatter (INTR)’
d. ləwwət’ə ‘change’ tə-ləwwət’ə ‘change (INTR)’
The anticausative simply derives intransitive verbs. The object argument of
the transitive verb becomes the subject and there is no implicit agent (unlike
in the passive). The prefix t(ə)- is also used to derive the reflexive. Self-
grooming verbs such as ‘wash’ and ‘shave’ are marked by the reflexive mor-
pheme, shown below in ():
() a. aster t – at’t’əb-əčč
A. REFL-wash.PF-F
‘Aster washed herself.’
b. ləmma tə – lač’č’ə
L. REFL-shave.PF.M
‘Lemma shaved himself.’
294 Amberber

Regarding valency-increasing derivations the most noteworthy is the causa-

tive. There are two main types of causative prefixes: a- and as- as exempli-
fied in ():
() a. mət’t’a ‘come’ a-mət’t’a ‘bring’
b. k’wərrət’ə ‘cut’ as-k’wərrət’ə ‘make x cut y’
In general, the causative prefix a- is attached to intransitive stems whereas
the prefix as- is attached to both intransitive and transitive stems.

. The light verb construction

In the previous section we looked at the morphological encoding of transi-
tivity through the use of various derivational prefixes. In the typical case, a
derivational prefix (anticausative, passive-reflexive, causative) is employed to
derive a verb with certain lexical semantic and morphosyntactic properties.
Thus, transitivity can be indicated by marking the verb morphologically.
There is another type of transitivity encoding strategy that involves the use
of independent verbs. Consider the following examples:
() a. k’ɨbe – w k’əllət’-ə
butter – DEF melt.PF-M
‘The butter melted.’
b. k’ɨbe – w k’ɨllɨt’t’ al -ə
butter – DEF melt.CV.INT say.PF-M
‘The butter melted.’
Both (a) and (b) have essentially the same meaning, although (b) includes
a slightly different meaning component, as we shall see shortly. Notice, how-
ever, that formally, the two constructions are different. In (a) the predicate
is morphologically simple, whereas in (b) it is complex. The predicate in
(b) consists of two independent morphological items: the non-finite form
k’ɨllɨt’ followed by the finite verb alə ‘say’. Tense–aspect and agreement
are marked on the finite form. Another verb which is productively used as
the finite verb component of a complex predicate is the verb adərrəgə ‘do’,
‘make’.
This type of complex verb formation, also known as a ‘compound’ or
‘composite’ verb, is very productive in Amharic. The construction is also
quite common in other Ethiopian Semitic languages including Giiz, as the
following examples show (based on Dillmann () as quoted in Hetzron
(:)):
() a. oho bəhil ‘to obey’
b. ɨnbɨyə bɨhil ‘to refuse’
 The structure of the light verb construction in Amharic 295

Gutt (: ) has shown that in Silte (East Gurage) there are many
compound verbs ‘consisting of an uninflected morpheme and an inflected
helping verb like baala “say”, mañe “build” (mainly intransitive), ašeʡan ̴e
“do, make” (mainly transitive)’. Examples are shown below in ():
() a. buube baala ‘flee in fright’
b. buube aše ‘cause to flee in fright’
In Tigré, according to Raz (: ), the coverbal element may or may not
exist as an independent lexical form and the finite verb, the second element,
is limited to three verbs: bela ‘to say’, wada ‘to do, to make’, and gaʔa ‘to
become’.
() a. bə əl bela ‘to pardon, to excuse’
b. bərəf wada ‘to clear off’
c. hən gaʡ a ‘to become speechless’
The equivalent of the verb ‘to say’ is also used in Tigrinya to form a com-
pound verb. Kogan (: ) describes the coverbal element in Tigrinya as
a ‘quasi-verbal element consisting of two or more radicals’. This verbal elem-
ent combines with the verb bälä ‘to say’ and its causative variant ʔabbälä.
The following are some examples of compounding (from Kogan : ):
() a. ʢaw ʡabbälä ‘to raise one’s voice’
b. təx bälä ‘to be straight’
c. bədəd bälä ‘to get up, to rise’
Compound verbs are also common in Cushitic and Omotic (see Amha:
this volume) and attested in some Nilotic languages (see Armbuster ).
Consider the following examples from Oromo (cf. Bender et al. : )
where the verb jeccu ‘say’ is used to form compound verbs:
() a. ol jeccu ‘get up’ [lit. ‘say “up”’]
b. c’al jeccu ‘be quiet’ [lit. ‘say “ssh”’]
c. tole jeccu ‘assent, agree’ [lit. ‘say “yes”’]
This type of verb formation is so widespread across Ethiopian languages
that some scholars have identified it as a feature of the Ethiopian language
area. Indeed, Ferguson (: –) found compound verbs in thirteen of
his sample of sixteen indigenous languages. He describes the compound
verb as follows:
In addition to many simple verbs consisting of a verb stem with inflectional
elements, there are many verbs consisting of a noun-like or interjection-
like ‘coverb’ plus a semantically colourless auxiliary, commonly the verb
‘to say’.
296 Amberber

Hetzron () attributes the presence of compound verbs in Ethiopian

Semitic to an early Cushitic (probably Agaw) influence. According to
Hetzron (: ), in Southern Agaw the verb as ‘to say’ ‘is used for com-
posite verbal expressions in which a nominal or an unanalysable word is
combined with nəŋ and used always with it’. Some examples from Southern
Agaw are given below in (), (from Hetzron : ):
() a. yaƔas nəŋ ‘to agree’
b. ənga nəŋ ‘to refuse’
́
c. əwnəs nəŋ ‘to believe’
d. déss nəŋ ‘to be happy’
What we refer to here as a light verb construction is found productively in
many typologically and genetically diverse languages. A construction that is
similar to the Ethiopian composite verb is common in Australian languages,
particularly in the non-Pama-Nyungan group of languages (see the chapters
by Baker and Harvey; Laughren; and Nordlinger, in this volume; see also
Butt ()).
In languages like Amharic, the light verb construction is often used in
addition to simple verbs, that is, for most of the composite verbs there are
corresponding simple verbs. Compare (a) and (b) below:
() a. t’ərmus -u sɨbbɨrr al -ə
glass-DEF break.CV say.PF-M
‘The glass broke (in a smashing manner).’
b. t’ərmus – u tə – səbbər -ə
glass-DEF ANTC/PASS-break.PF-M
‘The glass broke.’
Although the same truth-conditional meaning is encoded by both the light
verb construction in (a) and the simple verb in (b), it does not mean
that they are identical. In Amharic, the light verb construction often has
an additional meaning specification. In most triradical roots, the coverb

In a critique of Ferguson’s notion of an Ethiopian language area, Tosco (: ) argues
that the compound verb formations are ‘calqued from Cushitic and are, therefore a very
strong indicator of Cushitic influence’.

It is interesting to note that a light verb construction seems to introduce a specific mean-
ing element that is left unspecified in the associated simple verb construction. Thus, as
Huddleston and Pullum (: ) pointed out, in English ‘[t]he use of a light verb and
noun tends to yield a significant increase in syntactic versatility over that of the associated
verb construction’. They further point out that the light verb construction ‘generally allows
for dependents to be added to the noun, allowing a considerably greater range of elabora-
tion by modifiers and determiners’. For example, consider the contrast between He gave a
scream and He screamed. Huddleston and Pullum (: ) argue that the construction He
gave a scream ‘is not fully equivalent to He screamed, for it involves a necessarily quite short
and continuous event, while He screamed is not so restricted, covering also cases where the
 The structure of the light verb construction in Amharic 297

typically appears in two different templates, identified in the literature

(Beyene ) by the terms INTENSIVE and ATTENUATIVE respectively. For
a typical triradical verb with the pattern CəCCəCə, the two templates are
as follows:
() a. CɨC CɨCC Intensive
b. CəC əCC Attenuative
The Intensive versus Attenuative distinction correlates with the presence or
absence of a range of related meanings including the intensity of the event. A
more idiomatic translation of the verb in sentence (a) below is not simply
the neutral ‘break’. It may encode a range of meanings including the manner
of the event that would normally be expressed by the use of adverbs such as
‘suddenly’ or ‘completely’.
() a. t’ərmus -u s ɨbb ɨrr al -ə
glass-DEF break.CV.INT say. PF-M
‘The glass broke (suddenly).’
b. t’ərmus -u səbərr al -ə
glass-DEF break. CV.ATT say. PF-M
‘The glass broke (slightly, somehow).’
The coverb can also occur in a reduplicated template with a ‘distributive’
meaning, as shown in ():
() sɨbɨrbɨrr al -ə
break.CV.INT say.PF-M
‘It broke into pieces.’
The shape of the coverbal element is predictable only when it is derived from
verbs or, in some cases, from other lexical categories. When the coverb is an
ideophonic element, it can occur in a variety of different often idiosyncratic
forms, as shown in ():
() a. k’uč’č’ al- əčč
sit.CV say.PF-F
‘She sat down.’
b. bɨdɨgg al -əčč
rise.CV say.PF-F
‘She rose up.’
screaming is prolonged and intermittent’. The same is true with the pair she kissed him
and she gave him a kiss. The construction She gave him a kiss ‘specifies a single kiss where
[She kissed him] does not, and hence could apply to a situation where she covered him with
kisses’.
298 Amberber

The coverb can be an onomatopoeic word, as shown in the following

examples:
() a. zɨnab – u t’əbb t’əbb al -ə
rain-DEF drip.drip.CV say. PF-M
‘The rain dripped.’
[lit. ‘The rain said: “t’əbb t’əbb” ’]
b. gomma -w sit’it’t’ al -ə
tyre-DEF squeak.CV say. PF-M
‘The tyre squeaked.’
[lit. ‘The tyre (of a car) said: “sit’it’t.” ’]
It is important to note that while most light verb constructions have a simple
verb variant, there is no one-to-one correspondence between the two types
of predicates. For example, there are some gaps in the simple verb inven-
tory that are filled by the light verb construction. For instance, the state of
‘being quiet’, ‘be silent’ is expressed by the light verb construction zɨmm
alə, for which there is no (idiomatically) equivalent simple verb. Therefore,
the Amharic complex predicate, while not the only way of forming a verbal
predicate, is a productive strategy of expanding and elaborating the verbal
inventory.
In an important study of composite verbs across Ethiopian languages,
Appleyard () proposes a useful typology of composite verbs which rec-
ognises three types on the basis of the nature of the coverbal element: (a)
ideophone particle + say verbs; (b) quotative particle + say verbs; and (c)
derivational particle + say verbs. Some of the ideophonic particles (‘coverbs’
hereafter) are clearly onomatopoeic but others are not. An ideophonic coverb
is defined as an element ‘not derived from any other item in the language,
neither noun, verb nor particle, nor can it normally be extracted from the
composite and used in isolation’ (Appleyard :). In this type, Appleyard
includes coverbs which are clearly onomatopoeic, e.g. hɨk’k’ alə ‘hiccup’, and
non-onomatopoeic ideophones such as zɨmm alə ‘be quiet’, b ɨk’k’ alə ‘appear
suddenly’, and kəff alə ‘be high’. However, as Appleyard (: ) suggests,
it is at times difficult to distinguish coverbs which are ‘truly imitative’ from
those which are not.
In the quotative coverb, ‘the uninflected particle element is a form that
may enter into other syntactic contexts, such as a noun, an interjection or
any other free-standing form in the language’ (Appleyard : ). Coverbs
of this type include ɨšši alə ‘agree’ (literally ‘say OK’), and ɨ mbi alə ‘disagree’,
‘refuse’. Appleyard suggests that there is a connection between this type of
composite verb and the fact that in most of the Ethiopian languages the verb
‘say’ is used to introduce both direct and indirect speech, sometimes as a
complementiser (see also Amberber , a).
 The structure of the light verb construction in Amharic 299

Table .. Transitivity of the light verb construction

Basic stem (transitive) Complex predicate

Intransitive Transitive
[COVERB.INT. + alə] [COVERB.INT. + adərrəgə]

səbbərə ‘break’


dəffək’k’ ‘immerse’


gət’t’əmə ‘join’


gəmməsə ‘cut in two’


k’ərrəfə ‘peel off bark’


fət’t’ərə ‘create’ *

gəddələ ‘kill’ *

gəlləs’ə ‘explain’ *

The third type identified by Appleyard (: ) is the derivational co-

verb, which is characterised by its derivation ‘from an existing verb root by
means of one or more predictable morphological processes’. Interestingly,
Appleyard notes that on the evidence available so far this type of ‘composite’
verb is productive only in three languages, Amharic and Tigrinya (Semitic)
and Qafar (Cushitic).

. Transitivity and the light verb construction

An important question is whether or not it is possible to predict the tran-
sitivity of the light verb construction from the transitivity of the light verb
itself. Table . summarises some of the possibilities available for some
transitive verbs. For simplicity, only the distribution of the intensive coverb
is indicated (
= the composite verb is possible; * = the composite verb is
not possible).

One problem with Appleyard’s typology is its inability to distinguish between quota-
tive particles and ideophones. Appleyard (: ) argues that the particle ɨšši in ɨšši alə
‘agree’ [lit. ‘say OK’] is a quotative particle because the particle ɨšši can stand alone, for
example as a reply to a request. While the ability of a coverb to occur on its own can
be a useful diagnostic test to determine its non-ideophonic status, it is not sufficient.
Ideophonic coverbs such as zɨmm in zɨmm alə or kəff in kəff alə can occur by themselves
in some contexts. Thus, for example, kəff can occur on its own as an imperative kəff!
‘move up’. Hence, the occurrence of a coverb in isolation does not distinguish between
ideophones and quotatives. Consequently, it is not clear why kəff should be analysed as
an ideophone whereas ɨšši is analysed as a quotative coverb. See Amberber () for
a representative list of coverbs where a distinction is made between underived coverbs
and derived coverbs. The former includes ideophones and what Appleyard () calls
‘quotation’ particles.
300 Amberber

Notice that a coverb derived from a transitive verb can almost always
occur in the adərrəgə ‘make’ composite verb. This is what we would expect
if the coverb retains the transitivity value of the basic verb. However, note
that in some cases a coverb based on a transitive verb can also occur with the
light verb alə ‘say’.
When we consider the lexical semantics of the basic verbs we find that the
distribution is not completely arbitrary. The class of coverbs that can occur
in the alə ‘say’ composite verb are generally the ones that are derived from
verbs that encode an event that can come about without the involvement of
an external causer. This is clearly the case with the verb fət’t’ərə ‘create’: its
coverb variant ( f ɨt’t’ɨrr) cannot occur with the light verb alə to derive an
intransitive predicate. As shown in (c) below, a well-formed intransitive
form can be derived by the regular morphological strategy, i.e. by attaching
the prefix tə- to the verb.
() a. fət’t’ərə ‘create’
b. *f ɨt’t’ɨrr alə ‘be created’
c. tə-fət’t’ərə ‘be created’ [passive reading]
Furthermore, it appears that the presence or absence of external causation is
relevant to the choice between alə ‘say’ and adərrəgə ‘make’, ‘do’ for at least
some basic transitive verbs.
Some intransitive verbs lack a morphological transitive variant in a- as
they have a suppletive transitive form. The verb wəddək’ə ‘fall’ (a) is a case
in point. This is an unaccusative verb and would be expected to be transitiv-
ised with the causative prefix a-. However, the resulting form *a-wəddək’ə is
ill-formed, presumably because the verb already has a suppletive transitive
(c). Notice that a coverb based on this verb can take alə ‘say’ in the com-
posite verb, as shown in (d):
() a. wəddək’ə ‘fall’
b. *a-wəddək’ə ‘cause to fall’
c. t’allə ‘drop’ (= ‘cause to fall’)
d. wɨddɨk’k’ alə ‘fall (suddenly)’
Particularly interesting are forms which may have different transitivity
frames. For example, the form t’əbbək’ə has two meanings: (a) ‘be tight’
(intransitive) and (b) ‘guard’ (transitive). A coverb derived from the intransi-
tive t’əbbək’ə can occur both with alə ‘say’ and adərrəgə ‘make’ in the light
verb construction, (a). A coverb based on the transitive sense, ‘to guard’,
however, can occur only with adərrəgə, as shown in (b):
() [ + alə] [ + adərrəgə]
a. t’əbbək’ə ‘be tight’


b. t’əbbək’ə ‘to guard’ *

 The structure of the light verb construction in Amharic 301

The presence of the verb adərrəgə ‘make’, ‘do’ with a coverb based on
a stative intransitive implies the presence of an additional (causative) argu-
ment. Thus, consider the following contrast in ():
() a. ayn-u f ɨt’t’ɨt’t’ al -ə
eye-DEF pop.CV.INT say.PF-M
‘His eyes popped.’
b. ayn-u-n f ɨt’t’ɨt’t’ adərrəg -ə
eye-DEF-ACC pop.CV.INT make.PF-M
‘He popped his eyes.’
c. *ayn-u-n f ɨt’t’ɨt’t’ al -ə
eye-DEF-ACC pop.CV.INT say.PF-M
The distinction between the two types of intransitive verbs, the stative
(unaccusative) verb versus the activity (unergative) verb, is also relevant
here. Indeed, it appears that coverbs related to unergative intransitives do
not normally occur with the verb alə ‘say’ but rather with the verb adərrəgə
‘make’, ‘do’ in the light verb construction, as can be seen in the following
examples:
() a. č’əffərə ‘dance’
*č’ɨff ɨrr alə
č’ɨff ɨrr adərrəgə
b. zəffənə ‘sing’
*zɨff ɨnn alə
zɨff ɨnn adərrəgə
Note that with coverbs based on activity intransitives, the presence of the
verb adərrəgə ‘make’ does not imply that the composite verb is overtly tran-
sitive. Thus, a more accurate gloss for the verb adərrəgə here is simply ‘do’
rather than ‘make’. For example, the composite verb in (a) č’ɨff ɨrr adərrəgə
does not mean ‘x make y dance’ (with two arguments) but rather simply ‘x
do dancing’.
It is observed that coverbs related to genuine unergative verbs do not
normally occur with the verb alə ‘say’. Further examples are provided in
Table ..
Coverbs derived from intransitive verbs that encode events that are invol-
untary (as in ()) take the verb alə ‘say’:
() a. bərəggəgə ‘be startled’
b. dənəbbərə ‘bolt (horse)’
c. gwəməǰǰə ‘crave’, ‘desire’
302 Amberber

Table .. Coverbs derived from activity intransitives

Basic stem Complex predicate

Intransitive Transitive
[COVERB.INT. + alə] [COVERB.INT. + adərrəgə]

səggədə ‘bow in worship’ *

fokkərə ‘boast’ *

s’əlləyə ‘pray’ *

tərrətə ‘tell a tale’ *

Interestingly, coverbs that are related to verbs of temperature occur with alə
plus an object suffix, such as alə-w (say.PF.M-MO), when used as experien-
cer predicates. Notice that the object suffix cross-references the experiencer.
Thus, consider the following examples:
() a. bərrədə ‘be cold’ bɨrrɨdd alə-w
‘he is cold’
b. mok’k’ə ‘be warm’ mok’k’ alə-w
‘he is feeling warm’
(also: ‘he is tipsy’)
c. wəbbək’ə ‘be humid’ wɨbbɨk’k’ alə-w ‘he is feeling humid’
The same is true with experiencer coverbs derived from bound forms, that
is, forms which cannot occur independently. This is exemplified in ():
() a. *gərrəmə ‘be amused’ gɨrrɨmm al-ə-w ‘he is amused’
b. *č’ənnək’ə ‘worry’ č’ɨnnɨk’k’ al-ə-w ‘he is worried’
Likewise composite predicates such as dəss alə- ‘be happy’ involve the
object suffix on the verb alə as in (), literally ‘it says to him: “dəss” ’:
() *dəss ‘be happy’ dəss alə-w ‘he is happy’
There is a class of verbs whose basic stem begins with the vowel a. These
verbs do not take the prefix a- to be transitivised. Since these verbs form
a phonological class, and not a lexical semantic one, we would expect
them to select either alə ‘say’ or adərrəgə ‘do’, depending on the lexical
semantics of the root. This is exactly what we find. Consider the follow-
ing examples:
() a. abbədə ‘be insane’
b. *a-abbədə
c. ɨbbɨdd alə ‘he became insane’
 The structure of the light verb construction in Amharic 303

() a. abbəsə ‘wipe’

b. *a-abbəsə
c. *ɨbbɨss alə
d. ɨbbɨss adərrəgə ‘he wiped (it) completely, suddenly’
() a. addəgə ‘grow’
b. *a-addəgə
c. ɨddɨgg alə ‘it grew’
As already mentioned, coverbs can be used as sources of other lexical cat-
egories. For example, there are a number of nominals that are clearly
formed from coverbs, as shown in (), where the nominalising suffix -ɨta is
employed (see Leslau : ; see also Kane ):
() Nouns derived from coverbs
a. dəss-ɨta ‘happiness’ (< dəss)
b. zɨmm-ɨta ‘silence’ (< zɨmm)
c. s’ət’t’-ɨta ‘quiet, clam’ (< s’ət’t’)
To summarise the discussion so far, we have shown that the light verb
construction has two components: a non-inflecting coverb and an inflected
light verb: alə ‘say’ or adərrəgə ‘make’, ‘do’. In most cases there is a tran-
sitivity distinction depending on the inflecting verb: alə ‘say’ often occurs
in an intransitive frame, whereas adərrəgə ‘make’, ‘do’ occurs in a transi-
tive frame. Two major types of coverbs are recognised: underived coverbs
(including ideophonic, quotative elements) and derived coverbs. The latter is
a productive strategy of complex verb formation as a coverb can be derived
from almost any verb.

. The syntagmatic relationship between the coverb and

light verb
In this section, we examine the nature of the syntagmatic relationship
between the coverb and the light verb. Consider, once again, an example of a
typical light verb construction in ():
() k’ɨbe – w k’ɨll ɨt’t’ al -ə
butter – DEF melt.CV.INT say.PF-M
‘The butter melted.’
The coverb k’ɨllɨt’t ‘melt.INT’ and the light verb al-ə ‘say’ do not appear to
form a single head or X category. The main evidence for this is that the
two forms can be separated from each other by other elements. In (), the
304 Amberber

complementiser ɨnd(ə) ‘that’ is prefixed to the light verb, thus occurring

between the coverb and the light verb:
() k’ɨbe – w k’ɨllɨt’t’ ɨnd- al -ə ayyə -hw
butter – DEF melt.CV.INT that-say.PF-M see.PF-S
‘I saw that the butter melted.’
In relative clauses, the relative clause marker (REL) is prefixed to the light
verb:
() k’ɨll ɨt’t’ y- al -ə-w k’ɨbe
melt.CV.INT REL-say.PF-M butter
‘the butter that melted’
This fact suggests that the coverb and the light verb do not form a single
morphological object. It is also difficult to analyse the coverb as some kind of
‘complement’ of the light verb (as in the standard analysis of Japanese light
verb constructions; Grimshaw and Mester ). A complement is normally
a theta-bearing argument selected by a verb. It is obvious that in the light
verb construction here the coverb is not an argument but rather part of the
predicate itself.
Thus, how should we characterise the formal morphosyntactic relation-
ship between the coverb and the light verb? The coverb and the light verb
are in a dependency relationship: the coverb does not normally occur inde-
pendently of the light verb, whereas the light verb (at least in the case of alə
‘say’) is a light verb only when it occurs with the coverb. A good piece of evi-
dence comes from wh-questions. A normal predicate can be wh-questioned
in-situ as follows:
() a. k’ɨbe – w k’əllət’-ə
butter – DEF melt.PF-M
‘The butter melted.’
b. k’ɨbe – w mɨn hon -ə
butter – DEF what become.PF-M
‘What happened to the butter?’
For independent reasons, the verb ‘to melt’, which is unaccusative, can only
be questioned with the help of the verb ‘become’, as in (b).
Now, surprisingly, in the light verb construction with ‘say’, the coverb
cannot be simply replaced by the wh-word as (b) shows:
() a. k’ɨbe – w k’ɨllɨt’t’ al -ə
butter – DEF melt.CV.INT say.PF-M
‘The butter melted.’
 The structure of the light verb construction in Amharic 305

b. k’ɨbe – w mɨn al -ə
butter – DEF what say.PF-M
* ‘What happened to the butter?’
[OK as: ‘What did the butter say?’]
Note that (b) is ill-formed only with the interpretation ‘what happened to
the butter?’ but is perfectly fine with the (unintended) interpretation ‘what
did the butter say?’. In other words, without the presence of the coverb, the
verb alə ‘say’ defaults into its ordinary (‘heavy’) use.
The only way of wh-questioning the predicate in the light verb con-
struction involves the use of the verb ‘become’, as in (), which is
exactly the same as wh-questioning the simple verb variant we saw above
in (b).
() k’ɨbe – w mɨn hon -ə
butter – DEF what become.pf-M
‘What happened to the butter?’
This is in marked contrast with the behaviour of the other common light
verb – adərrəgə ‘do/make’. First, consider how a morphologically simple
verb is wh-questioned:
() a. aster k’ɨbe – w-n a- k’əllət’-əčč – ɨw
A. butter-DEF-ACC CAUS-melt.PF-F-MO
‘Aster melted the butter.’
b. aster k’ɨbe – w-n mɨn adərrəg- əčč – ɨw
A. butter-DEF-ACC what do.PF-F-MO
‘What did Aster do to the butter?’
Now consider how the verb in the light verb construction is wh-
questioned:
() a. aster k’ɨbe – w-n k’ɨll ɨt’t’ adərrəg-əčč – ɨw
A. butter-DEF-ACC melt.CV.INT do.PF-F-MO
‘Aster melted the butter.’
b. aster k’ɨbe – w-n mɨn adərrəg-əčč- ɨw
A. butter-DEF-ACC what do.PF-F-MO
‘What did Aster do to the butter?’
Thus, (b) contrasts with (b) where the light verb alə ‘say’ cannot occur
without the coverb. This suggests that unlike the light verb alə ‘say’, the
light verb adərrəgə ‘do/make’ is ‘light’ in all of its occurrences.
306 Amberber

It also suggests that the light use of a verb is ‘read off’ from the entire
construction. For example, consider the use of the light verb take in English
(see also Huddleston and Pullum : –):
() i. a. We rested. b. What did we do?
ii. a. We took a rest. b. ?What did we take?
The construction in (, i. a) can be wh-questioned as in (, i. b). On the
other hand, the wh-question in (, ii. b), which retains the verb take appears
to favour the ordinary interpretation of take rather than its light verb inter-
pretation. As pointed out by an anonymous reviewer, the light verb construc-
tion also resists clefting: *was it a shower he took.

. The derivation of the light verb construction

In this section, we develop a formal analysis of the derivation of the light verb
construction. The key assumption of our analysis is that the light verb con-
struction is derived through employing the same independently motivated
structural configuration of predicate structure. In particular, we argue that
the derivation of the light verb construction does not require any additional
mechanism over and above what is required in the derivation of morphologi-
cally simple predicates.

.. Basic theoretical assumptions

Our assumption about the architecture of grammar is based on the theor-
etical framework of Distributed Morphology (DM) (Halle and Marantz
, Marantz , Harley and Noyer , Embick and Noyer ). In
Distributed Morphology there is only a single generative engine – namely
the syntax – and words are formed by syntactic operations (Merge, Move)
and as such ‘the principles of morphology are … to a large extent the princi-
ples of syntax’ (Embick and Noyer : ).
In addition to these syntactic operations, additional and language-specific
PF processes may ‘modify and elaborate syntactic structure in limited ways’
(p. ). Thus, in DM ‘some aspects of word formation arise from syntac-
tic operations such as head movement, which occur in the syntax proper,
while other aspects of word formation are accounted for by operations that
occur on the PF branch’ (p. ). In this approach, there is no independent
Lexicon in the traditional sense and the properties attributed to the Lexicon
are assigned to three distinct lists (p. ):


For some previous studies on light verbs from a formal perspective see Grimshaw
and Mester ( ), Miyamoto (), Saito and Hoshi () for Japanese, and Jeong-
Ryeol () for Korean. See also Jackendoff () and Culicover and Jackendoff
( : –).
 The structure of the light verb construction in Amharic 307

() a. Syntactic terminals: The list containing the roots and the abstract
morphemes.
b. The vocabulary: The list of vocabulary items, rules that provide
phonological content to abstract morphemes.
c. The encyclopaedia: The list of semantic information that must be
listed as either as a property of a root, or of a syntactically con-
structed object (idioms like kick the bucket.)
The syntactic terminals – roots and abstract morphemes – are defined as
follows:
() a. Abstract morphemes: composed exclusively of non-phonetic fea-
tures, such as [past] or [pl], or features that make up the deter-
miner node D of the English definite article eventuating as the.
b. Roots: these include items such as √CAT, √OX, or √SIT, which are
sequences of complexes of phonological features, along with, in
some cases, non-phonological diacritic features.
While the class of abstract morphemes is closed and universal, the class of
roots is open-ended and language-specific. A further crucial assumption in
DM is that roots are category-neutral and are categorised only when they
combine with category-defining functional heads (such as v, n, a) in a local
environment. The abstract morphemes acquire their phonological fea-
tures through a mechanism called Vocabulary Insertion. For example, the
phonological exponent of the abstract morpheme [pl] (plural) in English is
/-z/: [pl]
z.

.. The morphosyntactic derivation of light verbs

We assume that minimally a verb is an amalgam of an idiosyncratic root and
an event-type functor – essentially specifying the event-type encoded by the
verb. For the present purposes, we assume that there are four main types of
event-type functors: CAUS, BECOME, BE, ACT which roughly correspond to the
Vendlerian verb classes of accomplishment, achievement, state, and activity
(see also Baker and Harvey’s Jackendovian analysis in this volume).
Let us take a specific example to see how the various pieces of the verb are
assembled. Consider the example below:
() k’ɨbe – w k’əllət’-ə
butter – DEF melt.PF-M
‘The butter melted.’
The event expressed by the construction in () is a simple change of state.
Let us assume that most change of state events are headed by the event-type
functor BECOME (see also Harley ; Amberber b). Following some
308 Amberber

standard assumptions about the syntactic representation of predicate struc-

ture (Hale and Keyser , ; Harley, ; Kratzer ), we assume a
layered VP structure with a functional higher head. The higher projection
is headed by Voice (following Kratzer ), which is also referred to as lit-
tle v in much recent work (see, for example, Chomsky (, ), Marantz
(), among others). Thus, consider the configuration in ():
()
VoiceP

VP

BECOME

DP V

k’ bew √rootP V
CONJ.CL

√klt’

The little v head can be regarded as a kind of voice head and determines
whether or not the structure projects an external argument. In () the
VoiceP is headed by a change of state event, BECOME, which by hypothesis
does not project an external argument. Following Arad’s () analysis of
verb formation in Hebrew, we assume that the consonants that form the root
need to be syllabified in order to be pronounceable. The V head provides a
range of CV- patterns, in effect conjugation classes (the traditional Semitic
binyan). The root moves to V to get syllabified. The choice of a particu-
lar CV- pattern is sensitive to the type of root selected. For example, the
root √klt’ ‘melt’, maps onto the conjugation class Type A with the pattern
[CəC CəC]. As pointed out by Arad (: ) ‘the combination of roots
with patterns serves a double purpose: it makes the segmental root into a
pronounceable string and turns the (category-neutral) root into a noun, a
verb or an adjective’.
Once the root has moved into the CV- pattern, it needs to spell out its
event type (represented by the event-type functors CAUSE, BECOME, BE, and
ACT respectively). Again, the vocabulary item that is inserted as an exponent
of the event-type functor depends on the type of √root. For example, the
root √klt’ ‘melt’ can occur either as a change of state or causative predicate.
Thus, it can be simply phonologically zero, [ø], or may be realised by one
 The structure of the light verb construction in Amharic 309

of the two overt affixes [a-] or [t(ə)-] (see Yabe (a, b) for a simi-
lar and insightful analysis of causatives in Amharic within a Distributed
Morphology framework).
The derivation for k’ɨbew k’əllət’-ə ‘the butter melted’ is roughly repre-
sented in ().
()
VoiceP

VP
v
BECOME ↔ [Ø] + [k’ ll t ’e]
e e

DP V

k’ bew √rootP
V
↔ [C1 C2C2 C3] ↔ [k’ ll t ’e]
e e
[CONJ CL.TYPA]
e e
√t
√t

Notice that the combination [√root + conjugation class] merges with the
voice head BECOME whose phonological exponence is [ø] for this particular
root.
() BECOME
[ø]
One important assumption in DM (see Bobaljik ), and shared by a
number of other approaches to morphology, is that the insertion of phono-
logical material (or vocabulary insertion in DM terms) proceeds cyclically:
() CYCLICITY: The interpretive procedure (vocabulary insertion) pro-
ceeds root-outwards (Bobaljik ).
Thus, given the string such as this [[[√root] conj cl -] voice], the phono-
logical exponence of the conjugation class is inserted in the first cycle. The
phonological exponence of the voice feature is inserted in the second cycle.
In this way, the insertion of a vocabulary item at each cycle will ‘see’ the
morphophonological features of the previous cycle.
Now let us consider the derivation of the light verb construction. Consider
() and the structure in ():
() k’ɨbe – w k’ɨll ɨt’t’ al -ə
butter – DEF melt.CV.INT say.PF-M
‘The butter melted.’
310 Amberber

()
VoiceP

VP
v
BECOME ↔ [al
e
‘say’]

DP V

k’ bew √rootP V
[CONJ CL.CV] ↔ [C1 C2C2 C3C3][k’ ll t’t’]

√t

There are two crucial differences in the derivation of the light verb con-
struction: (a) the phonological exponence of the conjugation class is the
pattern [CɨC C ɨCC]; (b) the phonological exponence of the event type
functor is the light verb alə ‘say’. We assume that when the event type func-
tor is spelled out by an independent form, the [√root + conj cl] form does not
move into it, by hypothesis.
() BECOME
[alə ‘say’]
The question of which vocabulary item is inserted to spell out the event
type functor is determined by contextual allomorphy. The identity of the
conjugational class determines the choice of exponent for BECOME.
() BECOME
[ø] {√root + [CəC C əC ], …}
BECOME
[alə ‘say’] {√root + [CɨC C ɨCC], …}
Thus, BECOME is spelled out as [ø] in the context of √root + [CəC CəC].
It is spelled out as alə ‘say’ in the context of √root + [CɨC C ɨCC]. The
above derivation shows how the monadic predicate ‘melt (INTR)’ is repre-
sented in the syntax. The transitive predicate ‘melt (TR)’ is derived in essen-
tially the same way. Consider the following transitive forms:
() a. aster k’ɨbe – w- ɨn a -k’əllət’-əčč
A. butter – DEF-ACC CAUS-melt.PF-F
‘Aster melted the butter.’
b. aster k’ɨbe – w- ɨn k’ɨllɨt’t’ adərrəgə-əčč
A. butter – DEF-ACC melt.CV.INT do/make.PF-F
‘Aster melted the butter.’
We assume that in both (a) and (b) there is a CAUS head which can be
spelled out either by a causative prefix or by an independent light verb.

A similar analysis is proposed by Megerdoomian () to account for complex predicates
in Persian. However, the Persian complex predicate is different from the construction we
 The structure of the light verb construction in Amharic 311

() CAUS
[a-] {√root + [CəC CəC ], …}—

CAUS
[adərrəgə ‘do/make’] {√root + [CɨC C  ɨCC], …}—
The predicate ‘melt’ in (a) takes two arguments – the Agent (‘Aster’) and
the Patient (‘the butter’). The verb itself is marked by the causative prefix a-,
which is the phonological exponence of the CAUS head. The verb root √k’lt’
merges with little v and the complex so formed moves to the CAUS head:
()
VoiceP

V

DP

VP
aster

CAUS ↔ [a− ]

DP V

k’ bew √rootP V
[CONJ CL.TYPA] ↔ [C1 C2C2 C3] ↔ [k’ ll t ’ ]
e e e e e

On the other hand, in the light verb construction (b), the phonological
exponence of the CAUS head is the light verb adərrəgə ‘do/make’. The root √klt’
merges with v but the complex so formed does not move to the CAUS head:
()
Voice P

V’
DP

VP
aster 
CAUS ↔[ad rr g-]
e e

DP V

k’ibew √rootP V
[CONJ CL.CV] ↔ [C1 C2C2 C3] ↔ [k’ ll t ’ ’]

√t

are interested in here. Crucially, in Persian the preverbal element can be a noun, an adjec-
tive, an adverb, or a prepositional phrase. See also Folli et al. ().

The light verb adərrəgə is itself morphologically complex – consisting of the causative pre-
fix a- and the form dərrəgə (which does not occur by itself). Given this fact, it is probably
312 Amberber

The phonological exponents of the verbalising head are partially deter-

mined by some diacritic features of the root, for example, the specification of
conjugation classes Type A, Type B, Type C, among others.
Notice that our analysis has important consequences for the categorisa-
tion of the coverb element in the light verb construction. Recall that the
coverb by itself does not belong to any of the lexical categories: verb, adjec-
tive, noun, adverb, etc. Therefore, one may be tempted to stipulate that
the coverb is a new part of speech in its own right. If the present analysis
is on the right track, no such stipulation is necessary: the coverb is simply
another verb pattern formally equivalent to other verb patterns in the lan-
guage. Nevertheless, while this analysis is plausible for the productive coverb
exemplified in the derivations above, it is problematic for coverbs that are
clearly related to interjections and onomatopoeic forms. Thus, consider the
coverbs in the following examples:
() a. aster zɨ mm al- əčč
A. silent.CV say.PF-F
‘Aster became silent.’
b. aster k’uč’č’ al- əčč
A. sit.CV say.PF-F
‘Aster sat down.’
c. zɨnab – u t’əbb t’əbb al -ə
rain-DEF drip.drip.CV say.PF-M
‘The rain dripped.’ [lit. ‘The rain said: “t’əbb t’əbb” ’]
We suggest that these coverbs have to merge with v whose exponence is
[ø], as shown below for the coverb root √zɨmm ‘silent’.
(59) VP

YP v

v
v BECOME ↔ [al ‘say’]

e

R v
zimm [ø]

more appropriate to analyse this light verb as composed of a root √drg plus a verbalizing
head v that merges with the causative head a-.
 The structure of the light verb construction in Amharic 313

The analysis of such roots as zɨmm as a category-less root also makes sense
as they can be nominalised by merging with an n whose phonological expo-
nence is ɨta, as in √zɨmm + -ɨta = zɨmm ɨta ‘silence’:
()
n’

R n ↔ [- t a ]
√z mm

To summarise, we have shown that the light verb construction is derived

in essentially the same manner as its morphologically simple counterpart. In
both constructions a category neutral root merges with v – a verbalising head.
In the unmarked case the (partial) phonological exponence of the verbalising
head is the verbal pattern (the traditional Semitic binyan). Event-type func-
tors, such as CAUS and BECOME, specify the event type of each construction.
The phonological exponence of these event-type functors can be ø, a- (for
CAUS), or the light verbs alə ‘say’, adərrəgə ‘do/make’, among others. The
√root + v complex forms a single morphophonological unit with the event-
type functor at PF only when the functor is spelled out as ø or as an affix.
One consequence of this analysis is that while some light verbs such as alə
‘say’ are exponents of a single terminal head – an abstract morpheme – light
verbs such as adərrəgə ‘do/make’ are internally more complex and have a
hybrid property involving the merger of two types of terminal heads – roots
and abstract morphemes.
Incidentally, recall that light verb adərrəgə ‘do/make’ occurs not only
with accomplishments but also with activities such as ‘dance’ and ‘sing’.
These are unergative intransitives and do not permit the light verb alə as can
be seen in the following examples:
() a. č’əffərə ‘dance’
*č’ɨff ɨrr alə č’ɨff ɨrr adərrəgə
b. zəffənə ‘sing’
*zɨff ɨnn alə zɨff ɨnn adərrəgə
There are a number of formally compatible ways of capturing this phe-
nomenon. One possibility is to argue that unergatives are underlyingly
diadic, as they can admit some sort of cognate object as an internal argu-
ment, for example: she danced a dance; she sang a song (see Hale and Keyser
). Thus, the fact that adərrəgə and not alə can be used in this context is
consistent with the otherwise robust generalisation that the light verb con-
struction that involves adərrəgə is transitive.
314 Amberber

. Remarks on the typology of light verbs

It is important to emphasise that, cross-linguistically, verbs which function
as light verbs are similar: they include verbs such as ‘be’, ‘become’, ‘come’,
‘do’, ‘get’, ‘give’, ‘go’, ‘have’, ‘make’, ‘put’, ‘say’, ‘take’. Of course, there are
some light verbs that are found only in some languages due to the cultural
saliency of the event encoded by the verbs in question, for example, ‘spear’
in Australian languages. In general, it is true that semantically similar verbs
function as light verbs cross-linguistically.
Many linguists assume that at least some of these verbs behave like a
closed-class category in that they are semantically ‘near-empty’ or impover-
ished. Some claim that such verbs have general and abstract meanings when
employed as light verbs or ‘auxiliary’ verbs (see Heine () and references
therein). This is usually what is assumed for English verbs like be, become,
make, do.
It is assumed that most light verbs are not born as light verbs but rather
are historically derived from lexical sources through a process of both mor-
phophonological and lexical semantic ‘depletion’ or ‘bleaching’ (see also
Bowern, forthcoming). For instance, regarding the verb be in English,
Jespersen (: ), who is believed to have coined the term ‘light verb’ (see
Butt ), claims:
The verb be has become what it is through a long process of wearing down
more concrete signification (‘grow’); it took a predicative in exactly the same
way as many other verbs with a fuller signification still do: he grows old |
goes mad | the dream will come true | my blood runs cold | he fell silent |
he looks healthy | it looms large | it seems important | she blushed red | it
tastes delicious | this sounds correct, etc.
The basic idea is that when lexical verbs undergo the process of bleaching
their meaning becomes more constant and less idiosyncratic – facilitating their
transition from a lexical item into a grammatical one. Likewise, Weinreich
(: ) has the following to say about the nature of verbs such as take.
When we contemplate the varieties of ‘meanings’ which a word like take
has in English (take offence, take charge, take medicine, take notice, take
effect, etc.), we come to the conclusion that this is a case not of abnor-
mally overdeveloped polysemy of a word, but rather of its semantic near-
emptiness.
What is not usually acknowledged in the literature on light verbs is the
fact that in most cases verbs that are used as ‘light’ have a rather ‘schematic’
meaning, even in their ‘heavy’ or ‘lexical’ usage. For example, consider the
English verb take. The basic meaning of take can be paraphrased as ‘cause a
thing to undergo a change in location’ ( Jackendoff ). There are a num-
ber of verbs that incorporate this basic meaning but are also more specific.
 The structure of the light verb construction in Amharic 315

For example, English has a large class of specialised take verbs, such as the
steal class of verbs (Levin : ff). This class includes such verbs as
abduct, capture, confiscate, cop, emancipate, embezzle, thieve, wangle, winkle.
Each verb encodes a kind of ‘taking’ but ostensibly adds something more
having to do with the manner of the ‘taking’ and other parameters including
agency, volition, intent, and outcome.
The same is true with the verb say. English has a number of so-called
‘manner of speech verbs’ (Zwicky , Levin ), such as shout, scream,
yell, holler, whisper, shriek, lisp, growl, mumble, mutter, etc. A manner of
speech verb incorporates the basic meaning of ‘say’ (whatever that may be)
but is also more specific: Mary whispered something to him means ‘Mary said
something to him in a whispering manner’. Similar remarks can be made
with regard to other common ‘light’ verbs. Thus, the lexical semantics of
‘give’ in English forms the subset of the lexical semantics of verbs such as
present, award, confer, etc.
This raises the possibility for a ‘monosemic’ analysis of light verbs, i.e.
at least some light verbs should be analysed as inherently ‘light’ and that
their apparent ‘heavy’ interpretation should be derived constructionally
(similar to Borer (a, b)). For example, consider the English verb
take:
() a. She took a book. (Heavy)
b. She took a shower. (Light)
Suppose that in both (a) and (b) the verb ‘take’ is the spellout of a
functional (event) head – CAUS/ACT. The ‘heavy’ interpretation arises when
the inner vP involves a phonologically null root ÿ with a meaning some-
thing like ‘move’ that takes ‘a book’ as its internal argument. The ‘light’
interpretation obtains when the inner vP involves the root √shower which
merges with n. If something along these lines is on the right track, it will
obviate the need for two ‘take’ verbs and derives the difference between the
‘heavy’ and ‘light’ usages from the construction as a whole.

. Conclusion
In this chapter we investigated the lexical semantics and syntactic represen-
tation of the light verb construction in Amharic. The language has two pro-
ductive light verbs that combine with a range of preverbal elements (referred
to as coverbs) to form a complex predicate.
Overall, the transitivity frame of the light verbs is predictable: ‘say’ occurs
with intransitives and ‘do/make’ tends to occur with transitives. In terms
of event structure, ‘say’ light verbs are typically used to encode achieve-
ments whereas ‘do/make’ light verbs often occur with accomplishments and
activities.
316 Amberber

It is argued that both the light verb construction and its simple verb vari-
ant are derived in the same way – utilising the same structural operations
and represented in the same configuration. In both constructions the verb is
assumed to involve the merger of a category-neutral root with a verbalising
head.
This analysis provides a natural account for the otherwise descriptively
problematic aspect of the construction: the fact that the coverb does not
belong to any lexical category. In the proposed analysis the productive co-
verb is formed through the merger of a root and a verbalising head in exactly
the same way as the formation of a morphologically simple verb.

References
Amberber, M. . Functional Verbs and Predicate Formation: Event-type
Hierarchy and Grammaticization. In Ahlers, J. and Bilmes, L. (eds) The
Proceedings of the Twenty-First Annual Meeting of the Berkeley Linguistics
Society, –. Berkeley Linguistic Society.
a. The Transitivity of Verbs of ‘saying’ Revisited. In Camacho, J., Choueiri,
L., and Watanabe, M. (eds) The Proceedings of the Fourteenth West Coast
Conference in Formal Linguistics, – Stanford: CSLI Publications.
b. Transitivity Alternations, Event-types and Light Verbs. PhD dissertation,
McGill University.
. Quirky Alternations of Transitivity: The Case of Ingestive Predicates. In
Amberber, M. and Collins, P. (eds) Language Universals and Variation, –.
Westport, Connecticut and London: Praeger.
Appleyard, D. . The Verb ‘to say’ as a Verb ‘Recycling Device’ in Ethiopian
Languages. In A. Zaborski (ed.) New Data and New Methods in Afroasiatic
Linguistics:Robert Hetzron in Memoriam, –. Wiesbaden: Harrassowitz.
Arad, M. . Roots and Patterns: Hebrew Morphosyntax. Dordrecht: Springer.
Armbuster, C. H. . Dongolese Nubian. Cambridge: Cambridge University
Press.
Baker, B. and Harvey, M., this volume. Complex Predicate Formation.
Bender, M. L., Bowen, J. D., Cooper, R. L., and Ferguson, C. A. . Language in
Ethiopia. Oxford: Oxford University Press.
Beyene, T. . Aspects of The Verb in Amharic. PhD dissertation, Georgetown
University.
Bobaljik, J. . The Ins and Outs of Contextual Allomorphy. In University of
Maryland Working Papers in Linguistics : –.
Borer, H. a. In Name Only: Structuring Sense, vol I. Oxford: Oxford University
Press.
b. The Normal Course of Events, vol II. Oxford: Oxford University Press.
Bowern, C., forthcoming. The Historical Linguistics of Complex Predication.
Butt, M. . The Light Verb Jungle. Harvard Working Papers in Linguistics
: –.
Chomsky, N. . The Minimalist Program. Cambridge: MIT Press.
 The structure of the light verb construction in Amharic 317

. Derivation by Phase. In Kenstowicz, M. (ed.) Ken Hale. A Life in Language.

Cambridge: MIT Press.
Culicover, P. and Jackendoff, R. . Simpler Syntax. Oxford: Oxford University
Press.
Embick, D. and Noyer, R. . Distributed Morphology and the Syntax/
Morphology Interface. In Ramchand, G. and Reiss, C. (eds) The Oxford
Handbook of Linguistic Interfaces. Oxford: Oxford University Press.
Ferguson, C. A. . The Ethiopian Language Area. In Bender, M. L., Bowen,
J. D., Cooper, R. L., and Ferguson, C. A. (eds) Language in Ethiopia, –.
Oxford: Oxford University Press.
Folli, R., Harley, H., and Karimi, S. . Determinants of Event Type in Persian
Complex Predicates. In Astruc, L. and Richards, M. (eds) Cambridge
Occasional Papers in Linguistics : –.
Grimshaw, J. and Mester, A. . Light Verbs and Theta-Marking. Linguistic
Inquiry : –.
Gutt, E. . The Silte Group (East Gurage). In Hetzron, R. (ed.) The Semitic
Languages, –. London: Routledge.
Haile, G. . The Suffix Pronouns in Amharic. In Kim, C-W. and Stahlke, H.
(eds) Papers in African Linguistics, –. Edmonton: Linguistic Research,
Inc.
Hale, K. L. and Keyser S. J. . On Argument Structure and the Lexical
Expression of Syntactic Relations. In Hale, K. and Keyser, S. (eds) The View
From Building , –. Cambridge, MA: MIT Press.
. Prolegomenon to a Theory of Argument Structure. Cambridge, MA: MIT
Press.
Halle, M. and Marantz, A. . Distributed Morphology and the Pieces of
Inflection. In Hale, K. and Keyser, S. (eds) The View from Building 20, –
Cambridge: MIT Press.
Harley, H. . Subjects, Events, and Licensing. PhD dissertation, MIT.
Harley, H. and Noyer, R. . Distributed Morphology. Glot International , :
–.
Haspelmath, M. . More on the Typology of Inchoative/Causative Verb
Alternations. In Comrie, B. and Polinsky, M. (eds) Causatives and Transitivity,
–. Amsterdam and New York: John Benjamins.
Heine, B. . Auxiliaries: Cognitive Forces and Grammaticalization.
Oxford: Oxford University Press.
Hetzron, R. . The Verbal System of Southern Agaw. Berkeley and Los
Angeles: University of California Press.
. Ethiopian Semitic: Studies in Classification. Manchester: Manchester
University Press.
. Semitic Languages. In Comrie, B. (ed.) The World’s Major Languages,
–. Oxford: Oxford University Press.
. The Semitic Languages. Oxford: Routledge.
Huddleston, R. and Pullum, G. K. . The Cambridge Grammar of the English
Language. Cambridge: Cambridge University Press.
Jackendoff, R. . A Deep Structure Projection Rule. Linguistic Inquiry, ,
: –.
318 Amberber

. Semantic Structures. Cambridge: MIT Press.

Jeong-Ryeol, K. . A Lexical-Functional Grammar Account Of Light Verbs. PhD
dissertation, University of Hawai’i.
Jespersen, O. . The Philosophy of Grammar. London: Allen & Unwin.
Kane, T. L. . Amharic-English Dictionary. Wiesbaden: Harrassowitz.
Kogan, L. E. . Tigrinya. In Hetzron, R. (ed.) The Semitic Languages, –.
Oxford: Routledge.
Kratzer, A. . Severing the External Argument From its Verb. In Rooryck, J. and
Zaring, L. (eds) Phrase Structure and the Lexicon. Dordrecht: Kluwer, –.
Leslau, W. . Reference Grammar of Amharic. Wiesbaden: Harrassowitz.
Levin, B. . English Verb Classes and Alternations. Chicago: Chicago University
Press.
Marantz, A. . No Escape From Syntax: Don’t Try Morphological Analysis in the
Privacy of Your Own Lexicon. In Dimitriadis, A., Siegel, L., Surek-Clark, C.,
and Williams, A. (eds) Penn Working Papers in Linguistics 4:2: Proceedings of
the 21st Annual Penn Linguistics Colloquium, –. Philadelphia: University
of Pennsylvania.
Megerdoomian, K. . Event Structure and Complex Predicates in Persian.
Canadian Journal of Linguistics , /: –.
Miyamoto, T. . The Light Verb Construction in Japanese. The Role of the Verbal
Noun. Amsterdam: John Benjamins.
Raz, S. . Tigré. In Hetzron, R. (ed.) The Semitic Languages, –.
London: Routledge.
Reid, N. . Complex Verb Collocations in Ngan’gityemerri: A Non-derivational
Strategy for Encoding Valency Alternations. In Dixon R. M. W. and
Aikhenvald, A. Y. (eds) Changing Valency: Case Studies In Transitivity, –.
Cambridge: Cambridge University Press.
Saito, M. and Hoshi, H. . The Japanese Light Verb Construction and the
Minimalist Program. In Martin, R., Michaels, D., and Uriagereka, J. (eds) Step
by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge:
MIT Press.
Tosco, M. . Is There an ‘Ethiopian Language Area’? Anthropological Linguistics
, :  –.
Weinreich, U. . On the Semantic Structure of Language. In Greenberg, J. (ed)
Universal of Language. Cambridge: MIT Press.
Yabe, T. a. Morphological Transitivization and Causativization in Two
Ethiopian Semitic Languages, Amharic and Ge’ez. Handout from the 
(S/C-U-)NY(-U) Mini Conference,  April .
b. Segmenting Amharic Causative as- into Two Verbal Morphemes: Semitic
a- and Cushitic –s. MS, The Graduate Center of the City University of
New York.
Zwicky, A. . In a Manner of Speaking. Linguistic Inquiry : –.
 INDEX

Alpher, B.  Bohnemeyer, J. , , 

Abeillé, A.  Borer, H. , , 
Ackerman, F.  Boroditsky, L. , , 
Adams, B. , , , , , , , Bowern, C. , , , –, 
–, , , , ,  Bowers, J. 
Adjer, D.  Bradshaw, J. 
Agaw  Bresnan, J. , , , , , , , ,
Ahtna - , 
Aikhenvald, A. , , , , , , , , Briem, D. 
, , –,  Bril, I. , 
Aissen, J.  Brinton, L. J. 
Akimoto, M.  Broadwell, G. A. , , , , , 
Alamblak  Bruce, L. , , 
Alsina, A. , , , , , , , , Bugenhagen, R. –
, ,  Butt, M. , –, , , , , ,
Ambae  , , , , , , 
Amberber, M. , , , , , ,  Bybee, J. 
Amha, A. , , , , , , ,
–, , , ,  Campbell, L. 
Amharic , , , , , , – Cantonese 
Andrews, A. , , , ,  Carter, R. , , 
Appleyard, D. , , ,  Cattell, R. 
Arad, M. ,  Chadwick, N. , 
Arce-Arenales, M.  Chatterji, S. K. 
Armbuster, C. H.  Choctaw 
Austin, P.  Choi, S. , 
Awngi ,  Chomsky, N. , , 
Axelrod, M. , , ,  Christaller, J. G. 
Cohen, D. 
Baker, B. , , , , , , ,  Comrie, B. 
Baker, M. ,  Crowley, T. –, 
Bamgbose, A.  Culicover, P. 
Barai , , 
Bardi  Dagaare 
Bench  Darlymple, M. , 
Bender, M. L. et al  Dehé, N. 
Bengali  Denison, D. 
Berlin, B.  Deo, A. 
Beyene, T.  Dhoore, C. S. , 
Bilin  Diller, A. 
Bitima, T.  Dillmann, C. F. A. 
Bobaljik, J.  Dimmendaal, G. , , , , ,
Bodomo, A. B.  , , –, –, 

319
320 Index

Dixon, R. M. W.  Hook, P. E. , –, 

Dowty, D. , ,  Hopper, P. J. , , , 
Durie, M. , , , ,  Hoshi, H. 
Dutch  Huang, J. 
Huddleston, R. –, 
Embick, D. –
Enfield, N. ,  Ikegami, Y. 
English , , , , , , , , , Isoda, M. 
, 
Evans, N.  Jackendoff, R. , , , –, , , ,
, , , , , 
Ferguson, C. A.  Jaminjung –, –, –, , 
Fitzpatrick-Cole, J.  Japanese , , , , 
Fleming, H.  Jarkey, N. , , –
Fodor, J.  Jelinek, E. , 
Foley, W. , , ,  Jeong-Ryeol, K. 
Folli, R. , , , , , ,  Jespersen, O. , , 
Fox, B.  Jetté, J. –
French , , , – Jingulu 
Johns, B. 
Gentner, D. , ,  Jones, E. –
Geuder, W. ,  Joshi, S. 
Ghomeshi, J. 
Giiz  Kamabata 
Givon, T.  Kamu 
Glass, A. ,  Kane, T. L. 
Godard, D.  Kaplan, R. 
Goldberg, A.  Kari, J. –
Green, R. ,  Karimi, S. , , , , 
Griefenow-Mewis, C.  Keyser, S. J. , , , , , , , ,
Grimshaw, J. , , ,  
Gurr-goni  King, T. H. , 
Gutt, E.  Kogan, L. E. 
Korean , 
Hacker, P.  Koyukon , –
Hackett, D. ,  Kratzer, A. , 
Hale, K. , , , , , , , , , Kriol 
, , , , , , , 
Halle, M. ,  La Polla, R. , , , , 
Halpern, A.  Lahiri, A. –
Hamar ,  Lakoff, G. 
Hansen, K.  Lamberti, M. , , 
Harley, H. , , , , , ,  Lane, J. 
Haro  Lapointe, S. 
Harris, A.  Laughren, M. , , , , , ,
Harvey, M. , , , , , , , , , , , , , , , 
,  Legate, J. , 
Haspelmath, M.  Leslau, W. , , , 
Hayward, R. J. ,  Levin, B. , , ,  –, , , , , ,
Heath, J. – , , , , , , , , 
Hebrew  Levinson, S. 
Heine, B.  Li, C. , 
Hendriksen, H.  Lieber, R. 
Hetzron, R. , , ,  Lightfoot, D. 
Hindi , , ,  Lord, C. –
Hinrichs, E.  Lydall, J. , 
Hirut, W. 
Hoekstra, T. ,  Malagasy , 
 Index 321
Malak-Malak  Reesink, G. , 
Mandarin , –,  Reid, N. 
Mangap Mbula – Reimer, N. 
Manning, C. , , , ,  Rice, K. –
Marantz, A. , , ,  Ritter, E. 
Marathi  Roberts, I. 
Marra – Rosen, S. , , 
Masica, C.  Roussou, A. 
Massam, D. 
Matisoff, J. A.  Sag, I. 
Matngele  Saito, M. 
Matthews, P.  Sanskrit , 
Matthews, S. – Saxon, L. 
McConvell, P.  Scalise, S. 
McGregor, W.  Schultze-Berndt, E. , , –, –,
Mchombo, S.  –, , 
Megerdoomian, K. , , , , ,  Scott, B. , 
Mester, A. , , ,  Sebba, M. , 
Miyamoto, T.  Sells, P. , , 
Mohanan, T. , , ,  Silte 
Müller, S.  Simeone-Senelle, M-C. 
Simpson, J. , , , , , , ,
Nash, D. , , , , , , , , , 
, , , , ,  Smith, C. S. , , , 
Nelemwa  Somali 
Ngaliwurru ,  Sottile, R. , , 
Ngan’gityemerri – Speas, M. 
Nordlinger, R. , , , , , , Strecker, D. 

Noyer, R. – Tagalog , 
Numbami ,  Talmy, L. , , 
Tantos, A. 
Olson, M. , ,  Tenny, C. , 
Oromo , ,  Thai 
Thompson, C. , , 
Paamese – Thompson, S. , , 
Pagliuca, W.  Tigre , 
Pardeshi, P.  Tigrinya , , , , 
Pawley, A. , , ,  Tikkanen, B. 
Pensalfini, R  Tosco, M. , , 
Pentland, C.  Trask, L. 
Perkins, R.  Traugott, E. C. , , 
Perlmutter, D.  Travis, L. , , , , , 
Persian , , , , ,  Turkish 
Pesetsky, D.  Tversky, B. 
Plank, F.  Twi –
Pullum, G. K. –, 
Pustejovsky, J. , , ,  Urdu –, , 
Pylkkanen, L. 
Van Staden, M. 
Qafar , , ,  Van Valin, R. , , , , , , 
Vanhove, M. 
Ramchand, G. , , –,  Vendler, Z. , 
Rapold, C. , 
Rappaport Hovav, M. , , ,  –, , , Wagiman –
, , , , , , , , ,  Wambaya , –
Ratliff, M. ,  Warlmanpa , 
Raz, S.  Warlpiri , –
322 Index

Warner, A.  Xamtanga 

Watam , –
Webelhuth, G.  Yabe, T. 
Weinreich, U.  Yimas , –
White Hmong , – Yip, V. –, –
Whitney, W. D. ,  Yoruba 
Wilhelm, A. 
Willie, M.  Zachs, J. 
Wilson, S. , , –, , , , ,  Zargulla , , , 
Wolaitta , – Zayse 
Wurmbrand, S. ,  Zwicky, A. 