Professional Documents
Culture Documents
Complex Predicates - Cross-Linguistic Perspectives On Event Structure
Complex Predicates - Cross-Linguistic Perspectives On Event Structure
Edited by
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/
A catalogue record for this publication is available from the British Library
Index
v
List of figures and tables
vi
Contributors
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
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
(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
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
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
. 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
() 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.
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
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:
BE (
Thing , place )
State
Event MOVE ( Thing )
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
(
(
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
[ 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
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])]
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
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
() 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])])]
() 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)])])]
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)])])]
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
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
[ ]Word Syntax
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, : )
We omit from consideration here reflexive and reciprocal constructions.
38 Baker and Harvey
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.’)
() 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, : )
() 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
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
Intransitive
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
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
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
MIRIAM BUTT
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
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
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.
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.
PRED ‘Nadya’
SUBJ
CASE ERG
PRED ‘Yassin’
OBJgo
CASE DAT
PRED ‘plant’
OBJ
CASE NOM
TENSE PAST
56 Butt
XCOMP SUBJ
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
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.
.. 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)
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.
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
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
.. 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.
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
. 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:
(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
‘You two pick up the canoe and put it into the water!’
Division of Dominance
← Cognitive dominance
linguistic dominance →
← open class
closed class →
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
I'
S I
S S. . . . S TNS
MOOD
IF
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
(21)
(a) [Eventi ] [Eventj ]... [Eventn ] Lexical Conceptual Structure
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:
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
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]
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:
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]
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
[sub-event] [sub-event]
[sub-event] [sub-event]
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
. 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
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
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 ..
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’.
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 (..).
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.
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’
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
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.’
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.
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 …’
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.
‘And Nzeu leapt up and stood in the middle, crying [and] scolding [and] crying
[and] scolding, saying …’
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
. 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
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 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
... 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
.. 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.
() 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.
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
different patternings likely form part of the justification for different types
of listings.
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
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
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.
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.
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
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
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 ..
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
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
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.
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
Directional deictic enclitics are =rra ‘thither’, =rni ‘hither’ and =mpa ‘across, by’.
Warlpiri verbs of change and causation: the thematic core 177
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
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
(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
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
MoodP
Inflectional structure
TenseP
IncepP (-nji)
PurpDisP (-parnka-)
Thematic core
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
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).
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
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
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
() 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
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
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
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
See Appendix , Table C for other examples of monadic stance PV-mani verbs.
192 Laughren
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
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
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
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
AKTP5 (ACCOMPLISHMENT)
Yj
CAUSEP
AKT
+/– dyn
AKTP4 (EVENT/SITUATION)
Yj PREDP
AKT
+/– dyn CAUSE
+/– dur (Negative) AKTP3 (ACHIEVEMENT)
Xi CHANGEP
AKT
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.
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’
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
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
forcing the external causer role and the changer of stance role to be linked
to the same referent via syntactic binding.
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
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
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
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.
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
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.
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
‘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]
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.)
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.
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 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
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
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
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)
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
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
“?” 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
Parri-rni is a dependent verb with very restricted distribution as a causative ‘removal’
verb.
236 Laughren
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
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 ():
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
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
()
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.
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
b. ‘-any’:
{ Event GO ([ ] ,
Thing Path TO [ Place THERE ]
)} ET
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.
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
{
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.
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
(
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).
{
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)
CAUSE
]) ,
([ ] D
, BE ([ ] , AT HIP (D ) ))
Eventx GO ([Thing D ] , [ Path
Eventy AFF [ ( ]
D
, )
. 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
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
AZEB AMHA
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
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
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 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
section .. for discussion). Accordingly, the discussion in section .. will
only address the same-subject anterior converb.
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
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
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 Vs 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
Table .. Asymmetrical V verbs in Wolaitta (Amha and Dimmendaal a: ).
Table .. Symmetrical set in Wolaitta (Amha and Dimmendaal a: ).
[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
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.
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.
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 Vs 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.
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
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.
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, Vs 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
. 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
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 CəC 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 -CCəC.
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
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
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
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’ *
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
Intransitive Transitive
[COVERB.INT. + alə] [COVERB.INT. + adərrəgə]
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
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.
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.
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
[CəC CəC]. 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 [CɨC C ɨCC]; (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 + [CəC C əC ], …}
BECOME [alə ‘say’] {√root + [CɨC C ɨCC], …}
Thus, BECOME is spelled out as [ø] in the context of √root + [CəC CəC].
It is spelled out as alə ‘say’ in the context of √root + [CɨC C ɨCC]. 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
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
YP v
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
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
319
320 Index