Complex Motion Predicates in Hiaki by Alexandra Trueman

Complex Motion Predicates in Hiaki by Alexandra Trueman
Complex Motion Predicates in Hiaki
Alexandra Trueman
A Dissertation Submitted to the Faculty of the
In Partial Fulfillment of the Requirements
For the Degree of
In the Graduate College
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by Alexandra Trueman, titled Complex Motion Predicates in Hiaki and
recommend that it be accepted as fulfilling the dissertation requirement for the Degree of
Doctor of Philosophy.
Date: 06/24/2014
Heidi Harley
Date: 06/24/2014
Andrew Carnie
Date: 06/24/2014
Simin Karimi
Date: 06/24/2014
Amy Fountain
Final approval and acceptance of this dissertation is contingent upon the candidate’s
submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
________________________________________________ Date: 05/03/2015
Dissertation Director: Heidi Harley
This dissertation has been submitted in partial fulfillment of the requirements for
an advanced degree at the University of Arizona and is deposited in the University
Library to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission,
provided that an accurate acknowledgement of the source is made. Requests for
permission for extended quotation from or reproduction of this manuscript in whole or in
part may be granted by the head of the major department or the Dean of the Graduate
College when in his or her judgment the proposed use of the material is in the interests of
scholarship. In all other instances, however, permission must be obtained from the
SIGNED: Alexandra Trueman
No-one makes it through the dissertation process without a lengthy list of people to
thank, and the nagging, itching feeling that there are even more people whom one has
overlooked or failed to give their due. So let me just acknowledge, right from the outset,
that this page is barely the tip of the iceberg of all the help, support and encouragement I
received on my way to this point.
First and foremost, this dissertation could not exist without my extraordinary Hiaki
language consultants, Maria Florez Leyva and Santos Leyva, who have shown boundless
patience and generosity in sharing their wealth of invaluable knowledge, and didn’t even
laugh too hard at my continuing failure to become a minimally competent speaker.
Similarly, I could not have done any of this without the energy, insight and unending
forbearance of my advisor, Heidi Harley, who has been both inspiration and support
throughout this journey, and who first introduced me to Hiaki, and to Maria and Santos.
I’m not sure I can adequately express how grateful I am for the opportunity to work with
her, nor my awe at her intelligence, enthusiasm and her kindness. There is just no baked
good fancy or delicious enough.
I owe a huge debt of gratitude to all my committee members, not just for their insightful
commentary on my research, but also for their guidance and influence in how to approach
teaching, learning, and linguistics in general. Andrew Carnie’s clarity, his precision of
thought and expression is something I can only aspire to. Simin Karimi’s wealth of
knowledge and generosity of spirit is similarly inspiring. And Amy Fountain has been an
infallible source of solidarity and support, as well as peppermints and puppy pictures.
Again, I appreciate it more than I can say.
A thousand thank-yous to the other members of the UA Hiaki research group, including
Joe Sanchez and Mercedes Tubino Blanco, but especially to Hyun Kyoung Jung, with
whom I shared hours of elicitation time, ideas, equipment and funds. Chiokoe uttessia,
Acheka! Huge thanks to Jaehoon Choi, Hyun Kyoung Jung, Sylvia Reed, Deniz Tat,
Greg Key, Chen-chun Er, Megan Stone and Tatyana Slobodchikoff, for the sharing of
drafts, ideas, accomplishments, setbacks, and the Confessional. I am also indebted to
Tyler Peterson, Camilla Thurén, Lindsay Butler, Beth Stelle, Dave Medeiros, and Mary
Ann Willie, for their insights and influence on different aspects of my graduate work.
Nothing gets done without the hard work of administrative personnel, but Marian Wisely,
Kimberley Young, Shayna Walker and Jennifer Columbus have all been so much more
than that. Each one of them has gone well beyond expectations. More generally, I must
extend thanks to all the many UA linguists, faculty and students, past and present, who
have made my time here so valuable, and so joyful. In particular, I thank some of the core
regulars at Linguist Lunch – Shiloh Drake, Dane Bell, Mike Hammond and Adam
Ussishkin – for broadening my education immeasurably.
Closer to home, I extend my deepest thanks and appreciation to my wonderful, lovely,
brilliant housemates, Jelena Vukomanovic and Jessamyn Schertz, for everything (but
especially for wine and frozen yoghurt). Thanks again to Shiloh Drake, who took me in
during my last weeks in Tucson, to Kara Hawthorne, Chelsea Cady, Jessamyn Schertz
and Mary Dunggan, for tea, movies and dramatic readings, and to Maggie Camp, Amy
LaCross and Cat Bothelo for ‘bookclub’. And finally, of course, no thanks are sufficient
for my family, who have put up with me longer than anyone, and to whom I can offer no
compensation except to insist that you continue to put up with me indefinitely.
LIST OF ABBREVIATIONS ......................................................................................................................................................... 7 ABSTRACT ....................................................................................................................................................................................... 8 Chapter 1 Introduction ....................................................................................................... 9 1.1 Aims ............................................................................................................................................ 9 1.1.1 Roadmap ........................................................................................................................................... 10 1.2 Hiaki ......................................................................................................................................... 11 1.2.1 Background ..................................................................................................................................... 11 1.2.2 Structure and constituents ....................................................................................................... 13 1.2.3 (Morpho-­‐)Phonology .................................................................................................................. 14 1.2.4 Range of Hiaki verb/event combining strategies ........................................................... 17 1.2.5 Complex Motion in Hiaki ........................................................................................................... 21 1.3 Motion ...................................................................................................................................... 26 1.4 Complex Predication .......................................................................................................... 27 1.5 A note on theory and framework ................................................................................... 31 Chapter 2 Independent motion verbs in Hiaki ........................................................ 32 2.1 Basic Motion: siime ............................................................................................................. 32 2.1.1 Formal and featural properties of siime ............................................................................. 33 2.1.2 Interactions ..................................................................................................................................... 38 2.2 Other Verbs of Motion ........................................................................................................ 48 2.2.1 Path of Motion ................................................................................................................................ 49 2.2.2 Manner of Motion ......................................................................................................................... 49 2.3 Summary ................................................................................................................................. 51 Chapter 3 Complex Motion in Hiaki ............................................................................ 52 3.1 Overview of Hiaki Complex Verb Structures .............................................................. 53 3.1.1 Types of Complex Verbs in Hiaki ........................................................................................... 53 3.1.2 ECM V2s ............................................................................................................................................ 60 3.1.3 Raising V2s ...................................................................................................................................... 66 3.1.4 Control V2s ...................................................................................................................................... 72 3.2 Motion Compounds ............................................................................................................. 75 3.2.1 Clausal/Argument Structure ................................................................................................... 77 3.2.2 Effects of V1 ..................................................................................................................................... 83 3.3 Contrasting V2 -­‐sime with independent siime/saka ............................................... 90 3.3.1 Interactions with Morphological Processes ...................................................................... 92 3.4 Summary .............................................................................................................................. 102 Chapter 4 Complex Motion Crosslinguistically ..................................................... 105 4.1 Complex Predicates .......................................................................................................... 108 4.1.1 Definitions of CPr ........................................................................................................................ 109 4.1.2 Grammaticalization/Diachrony ............................................................................................ 113 4.2 Construction types and analyses ................................................................................. 117 4.2.1 Auxiliary Verbs ............................................................................................................................ 118 4.2.2 Light Verbs ..................................................................................................................................... 123 4.2.3 Restructuring Verbs .................................................................................................................. 132 4.2.4 Serial Verbs ................................................................................................................................... 139 4.3 Summary .............................................................................................................................. 153 Chapter 5 Struggles with Structure ........................................................................... 155 5.1 Questions regarding the structure of V-­‐sime .......................................................... 155 5.2 Close comparisons -­‐ Korean and Warlpiri ............................................................... 156 5.2.1 Korean ............................................................................................................................................. 156 5.2.2 Warlpiri ........................................................................................................................................... 163 5.2.3 Q1: V2 -­‐sime is lexical ................................................................................................................ 166 5.3 Modeling Structure ........................................................................................................... 167 5.3.1 Q2 -­‐ Size of V1 is vP, excludes VoiceP ................................................................................ 168 5.3.2 Q3 -­‐ VP1 is an adjunct ............................................................................................................... 172 5.3.3 Structural Alternatives ............................................................................................................. 177 5.4 Sorts and situations ......................................................................................................... 182 5.4.1 Situating V-­‐sime amongst Hiaki complex Vs ................................................................... 182 5.4.2 Situating Hiaki in the world of complex motion structures? ................................... 187 5.5 Summary .............................................................................................................................. 188 5.6 Other remaining problems, issues, directions for future work ........................ 188 References 193 6
first person
second person
third person
linking vowel
subject relative
This dissertation is an investigation into compound verbal structures in Hiaki in which a
verb of motion is modified by an adjoined lexical verb or verb phrase. It provides the first
in-depth documentation and analysis of this structure in Hiaki, an endangered language
indigenous to North America, and it explores the extent to which complex predicates of
motion may be said to form a discrete class crosslinguistically, either in structural or
semantic terms, by comparing Hiaki with genetically and typologically distinct languages
such as Korean and Warlpiri.
The study asks the following questions:
1) What is the underlying structure of a Hiaki compound verb? In particular,
what is the structure when the head verb is intransitive and thus cannot take
the second verb or verb phrase as its complement?
2) To what extent can complex motion predicates in different languages be said
to map to identical underlying syntactic structures? That is, if we compare
these constructions in Hiaki with those in languages with different surface
morphosyntactic realizations, how do the allowable surface forms constrain
the possible underlying structures?
3) Is there evidence to suggest a cline or typology of complex motion predicate
The overall goals of the dissertation project are the detailed documentation, description
and theoretical analysis of complex motion constructions in Hiaki, the crosslinguistic
comparison of these constructions, and the expansion of an existing database of
transcribed and interlinearized Hiaki texts.
Chapter 1 Introduction
Hitasa empo hiohtesisime? (What are you going around writing?)
1.1 Aims
This dissertation is an investigation into compound verbal structures in Hiaki in which a
verb of motion is modified by an adjoined lexical verb or verb phrase. It provides the first
in-depth documentation and analysis of this structure in Hiaki, an endangered language
indigenous to North America, and it explores the extent to which complex predicates of
motion may be said to form a discrete class crosslinguistically, either in structural or
semantic terms, by comparing Hiaki with genetically and typologically distinct languages
such as Korean and Warlpiri.
The study asks the following questions:
4) What is the underlying structure of a Hiaki compound verb? In particular,
what is the structure when the head verb is intransitive and thus cannot take
the second verb or verb phrase as its complement?
5) To what extent can complex motion predicates in different languages be said
to map to identical underlying syntactic structures? That is, if we compare
these constructions in Hiaki with those in languages with different surface
morphosyntactic realizations, how do the allowable surface forms constrain
the possible underlying structures?
6) Is there evidence to suggest a cline or typology of complex motion predicate
The overall goals of the dissertation project are the detailed documentation, description
and theoretical analysis of complex motion constructions in Hiaki, the crosslinguistic
comparison of these constructions, and the expansion of an existing database of
transcribed and interlinearized Hiaki texts.
1.1.1 Roadmap
The dissertation is organized thus:
Chapter 1 provides essential background information on the Hiaki language (§1.2), on
foundational thinking about the expression of motion in language (§1.3) and on how the
term ‘complex predicate’ is understood and defined (§1.4).
Chapter 2 provides a detailed description of (non-complex) verbs of motion in Hiaki,
with particular attention to the properties of the basic motion verb siime ‘go’, which is the
verb that appears in the complex motion compounds.
Chapter 3 describes the properties of Hiaki compound verbs in general, and examines the
differences and similarities between motion and non-motion compounds. In addition, it
examines the properties of the ‘go’ verb siime when it appears in a compound, in contrast
with its properties as an independent verb.
Chapter 4 surveys complex predicate types in languages from around the world, and the
analyses that have been proposed for them. It compares the properties of these
constructions with the Hiaki motion compounds, and considers whether a cline or
typology of complex motion predicate types may be posited, and where Hiaki might fit in
such a system. Ultimately, it is determined that the range of variation and overlapping
properties found in complex motion predicates makes such a proposition untenably
vague, and that Hiaki does not fit neatly into any of the previously described categories,
although some languages are identified which do bear important similarities.
In Chapter 5, drawing on similarities with constructions in Korean and Warlpiri, as well
as on language-specific properties, I propose a structural analysis of Hiaki complex
motion predicates. In this analysis, the first verb in a motion compound is a vP
constituent head-adjoined to the final motion verb, which is itself the head of a √P. I
discuss the limitations and problems with this analysis, as well as other potential
analyses, and the reasons for their rejection.
1.2 Hiaki
1.2.1 Background
Hiaki, more frequently named 'Yaqui' or ‘Yoeme’ in the literature, is a Uto-Aztecan
language spoken in Sonora, Mexico, and more recently in Arizona. The Arizona Hiaki
tribe currently exists in a state of increasingly precarious trilingualism between Hiaki,
Spanish and English. English is dominant amongst young people, and Hiaki is now
fluently spoken only by some 70 or 80 people, all of whom are over the age of fifty.
The word ‘Hiaki’, and its usual orthographic representation as ‘Yaqui’, is itself an
interesting example of the cultural and linguistic heritage of the Hiaki people, which
blends significant Spanish influence with a strong sense of core independent identity. The
spelling ‘Hiaki’ reflects the pronunciation, and conforms to the written conventions, used
by the Pascua Yaqui tribe today. ‘Yaqui’ was the orthographic representation of the name
given to them by the Spaniards when the two cultures first made contact in the mid
sixteenth century. In Mexico, the spelling ‘Jiaki’ is also used, in a Spanish-based
orthographic system for the language. ‘Yoeme’ is another alternative that is sometimes
used (Castile 2002). I use ‘Hiaki’ throughout, as it is the preferred spelling of my
consultants; it conforms to the English-based spelling system for the language adopted by
the Pascua Yaqui tribe, and is a more accurate representation of the pronunciation,
There is a relatively large amount of existing documentation and analysis of Hiaki,
including substantial work on the verbal morphosyntax carried out at the University of
Arizona by Heidi Harley’s research group. There are Hiaki-English and Hiaki-Spanish
dictionaries (Molina et al. 1999 and Fernandez et al. 2004, respectively). There is an
extensive compendium of Hiaki grammatical structures (Dedrick and Casad 1999), which
contains a wealth of empirical data. There have been four dissertations in English which
focus on the language: Lindenfield 1973, Escalante 1990a, Guerrero 2005, Martinez
Fabian 2006, as well as the dissertation of Hagberg 1993 on the phonology of the closely
related language Mayo, and a few theses in Spanish from the Universidad de Sonora, e.g.
Castro Llamas 1988. Finally, several papers on aspects of the language have appeared in
conference procedings, volumes and journals, including, e.g., Escalante 1990b, Jelinek
and Escalante 1988, Jelinek 1997, Demers, Escalante and Jelinek 1999, Guerrero and van
Valin 2004, Guerrero 2005, Felix Armendáriz 2005, Martinez Fabian and Langendoen
1996, as well as the work of the University of Arizona research group, Haugen, TubinoBlanco, Leyva, Sanchez, Trueman and Jung. However, the central focus of the present
proposal, motion constructions, addresses a significant gap in previous work on the
1.2.2 Structure and constituents
Hiaki has a basic SOV order; postpositional phrases are also preverbal. Guerrero &
Belloro (2010) describe the language as ‘syntactically rigid but pragmatically flexible’,
like English, in which deviation from canonical word order is relatively limited, and
information-structural status is prosodically indicated.
Broadly speaking, Hiaki clauses with lexical subjects conform to the following template:
Pre-subject (samarked wh-words,
certain adverbs;
discourse linkers)
Middle field
(objects, PPs,
Verb-word (including
particles, clitic object
elements: usu.
(Harley Trueman & Leyva 2012:12)
As well as lexical subjects, clitic pronominal arguments are also possible; they occur in
fixed positions with subject clitics generally following the first phrasal element, and
object clitics always immediately preceding the verb.
Hiaki has a straightforward Nominative-Accusative case system, and agglutinative
morphology, which is suffixal, except for reduplication. In the verbal domain in
particular, this morphology is extremely productive, and Hiaki verbal structures can be
quite complex. This rich system of verbal affixation includes derivational and argumentstructure-changing affixes (such as causative, applicative, desiderative and passive),
tense/aspect markers, and a complex set of reduplication patterns, as well as
compounding and incorporation.
The template in (2) shows the range of possible elements in the Hiaki verb structure.
Everything except the second V (underlined) is optional.
-(RED)-(V) -V-
(Harley 2013:12)
1.2.3 (Morpho-)Phonology
Adding to the complexity discussed above, approximately 15 verbs display agreementdriven suppletion. The motion verb siime is one of a class of suppletive verbs that
changes its form depending on tense/aspect, and on the plurality of its subject. The full
paradigm is shown in (3).
Present, singular subject:
Present, plural subject:
Past perfective, singular subject: siika
Past perfective, plural subject:
Stem forms
In addition to the suppletion, both the singular subject form siime and plural subject form
saka have bound forms, which are triggered by particular affixes.
Bound, singular subject: sim-
Bound, plural subject:
All Hiaki verbs have both a free stem and a bound stem form, illustrated for the verb ye’e
‘dance’ in (5)-(6) and there are several classes or sub-classes of verbs according to the
type of stem alternation they exhibit, including an invariable class. Examples of these are
displayed in the table in (7).
(5) Inepo
“I am dancing.”
(6) Inepo
“I will dance.”
(Escalante 1990:38)
Class 1: Truncation
a. poona
b. miika
c. bwase
Class 2: Echo-vowel
a. bwasa
bwasa'a- ‘cook’
b. kiima
c. yore
Class 3: invariable
a. kivacha
kivachab. hamta
hamtac. koko
(Harley 2013:7)
Some suffixes must attach to free stems – such as -k (past perfective), -n (past
imperfective), and -kan (past perfect) – but the majority attach to bound stems. Bound
stem suffixes include: -wa (impersonal/passive), -ne (future/irrealis), -na (future
impersonal/passive), -tua (direct causative), -tevo (indirect causative), -ria (applicative)
and many others. The affixal verbs that occur in complex verbs always attach to bound
stems. In the (frequent) event of affix-stacking, free-stem suffixes always follow suffixes
that take bound stem forms. (Harley 2013, Harley and Tubino Blanco 2012, Tubino
Blanco and Harley 2010)
Vowel shortening
Hiaki has long and short vowels. In both verbal and nominal stems that have a long
vowel, this may become shortened in particular morphological environments.
For example, long vowels in nouns shorten in the presence of the accusative suffix -ta.
è cat.NOM
Long vowels in verbs shorten in the presence of reduplication, the participial suffix -ka,
derivational suffixation, and a range of other morpho-phonological environments that
have not been exhaustively documented at this time.
1.2.4 Range of Hiaki verb/event combining strategies
Hiaki has four main strategies for combining verbs or events within an utterance. Most of
these are unambiguously multi-clausal structures in which each verb is inflected, either
for tense/aspect, or with a participial/subordinating affix.
Conjoined clauses
The first type of multi-clause utterance involves the conjunction (12) or disjunction (13)
of two clauses, each of which has a tense-inflected verb. Often, both verbs are inflected
with the same tense/aspect properties, but they are otherwise complete and distinct
Hoan tekipanoa-n, Anavela intok kari-ta tui-te-n.
Juan.NOM work-PST Anabel.NOM CONJ house-ACC good-make-PST
"Juan was working and Anabel was cleaning the house
empo ye'e-ka, taa aapo kaa ye'e-ka1
2s.NOM dance-PFV, but 3s.NOM NEG dance-PFV
“You danced but he didn't dance.”
Subordinate clauses
Subordinated clauses, particularly those describing temporally simultaneous or otherwise
connected actions performed by a single subject, are indicated by a participial suffix -ka
on the subordinated verb. In these examples, either or both verb may have an associated
argument or particle, which appears in its typical pre-verbal position and thus may
intervene between the two verbs. In addition, there is typically a significant intonation
break distinguishing the clauses.
Acheka yeu weye-ka(-su), kafe-ta woota-k
HK.NOM out walk-PCL(-SUB), coffee-ACC spill-PFV
"As HK was walking out, (she) spilled the coffee."
Acheka sime-ka-su, kafe-ta woota-k
HK.NOM go-PCL-SUB coffee-ACC spill-PFV
"As HK was leaving, (she) spilled the coffee."
Although I have here given it the gloss ‘SUB(ordinator)’, the suffix -su, which follows -ka
in these examples, is sometimes optional (14) and sometimes not (15), and it is not
-ka in this example is a allomorph of the usual perfective -k, occurs with a specific class of verbs, and
readily apparent what drives this optionality. Much more investigation into the precise
function of this affix is warranted, particularly since it is homophonous with a completive
Verbal modifier
The third, and least common form of multi-verb sentence involves two verbs that are
immediately adjacent to each other, with no intervening material or significant
intonational boundary. It can be clearly distinguished from the compound verbs in the
next section because the first verb does not occur in its bound stem form. The relevant
verbs in examples (16)-(17) are bolded, but it is that first verb (underlined) in each that is
the curious one. In this case, the verb siime occurs in what appears to be its perfective
form, albeit in (16) with a shortened vowel (siika -> sika), and in (17) with an added final
vowel (sahak -> sahaka).
Acheka sika we(y)e-ka-su,
kafe-ta woota-k
coffee-ACC spill-PFV
"As HK was leaving, (she) spilled the coffee." (Lit: As HK left walking, (she)
spilled the coffee”)
Ume veveme-m
sahaka kaate-ka-su, kafe-ta woota-k
DET.PL girl-PL
go.PL walk.PL-PCL-PRT
coffee-ACC spill-PFV
“As the young girls were leaving, they spilled the coffee”
In the closest English translation of this sentence, the verb ‘walk’ is a manner modifier of
‘go/leave’, however it is not clear that this is the appropriate analysis of the Hiaki
structure. It is also possible that the form sika ‘go/leave’ is functioning as a direction
modifier of weye ‘walk’, much as the particle yeu ‘out’ does in example (14). Directional
particles also occur in the immediately preverbal position, and within the same intonation
contour of the main verb in the clause.
Compound verb
The fourth and final kind of multi-verb strategy in Hiaki, and the focus of this
dissertation, is verb compounding. In this construction two verbs are combined in such a
way that nothing may intervene between them; not arguments, particles, nor any kind of
inflectional morphology. The first verb appears as a bound stem, and the second verb is
either bound or free, depending on whether it takes further suffixation of the relevant
type. Even in its free form, the second verb may show phonological reductions, such as
the shortening of long vowels. The properties of these compound verbs will be described
in further detail in §1.2.5 and in Chapter 3.
Chepa kari-po yeu bwan-sime
Chepa.NOM house-LOC out cry-go
"Chepa left the house crying." / "Chepa went crying out of the house."
1.2.5 Complex Motion in Hiaki
Hiaki has a rich and complex system of verbal affixation, which includes derivational and
argument-structure-changing affixes (such as causative, applicative, desiderative and
passive), tense/aspect markers, and a complex set of reduplication patterns, as well as
compounding and incorporation. In addition, many verbs display agreement-driven
suppletion. As noted above, the focus of this dissertation is compound structures in which
a tensed verb of motion is compounded with another lexical verb, as exemplified below:
Uu hamut
the woman
“That woman is pushing the child along.”
Haisa ne
eteho-u haisa ne
1sgNOM 2plACC together talk-to Q 1sgNOM
aa= hiohte-sim-ne
“When we are talking can I take notes?”
Constructions of this kind, although commonly occurring in Hiaki and other languages,
are poorly understood and present an interesting puzzle for our understanding of
argument structure. For instance, the head verb in the compound is an intransitive motion
verb ‘go’, and as such does not take a direct object; however accusative marked objects
may appear, as in (19), licensed by the non-head first verb. One question to be
investigated, then, is the degree to which objects are integrated into the argument
structure of the whole clause.
Hiaki has two major types of complex verbs, although the division between the two is a
little muddy. The first of these, which we will call ‘complex verbs’ following Escalante
(1990) involves a lexical verb with an affix (or affixes) which has some aspectual or
‘light’ verbal properties, and which for the most part does not have independent status.
The second is ‘compound verbs’, in which two (or potentially more) verbs are combined,
each of which has independent lexical verb status. Although complex verbal structures
are quite common in the language, verbal compounds appear to be a reasonably restricted
class, in the sense that there are relatively few verbs which commonly show up as the
final or head verb in such a structure.
One of the most common of the independent lexical verbs to occur as as the head verb, or
V2, of a compound verb is the intransitive motion verb siime ‘go/come’.
Vempo nee
3plNOM 1sgACC toward
“They came towards me.”
In a compound, siime usually indicates an interpretation like, roughly, ‘go along V1-ing’,
as examples (22)-(23) demonstrate.
DET.PL little boy-PL
“The little boys are going along kicking balls.”
Uu hamut
DET woman
little child-ACC
“The woman is pushing the little child along.”
Example (23) shows that the V1 (yu’a ‘push’ in this case) is occurring in its bound-stem
form, and in this form no independent tense marking is possible. Since V1 is a transitive
verb it has an object, which in this instance is clearly marked with accusative case,
however both Vs share the same nominative subject. In fact, because V2 is always an
intransitive verb of motion or stance, in these constructions both Vs always share a
subject, and this makes diagnosing clause structure a little difficult.
The reason that this is relevant is because in most of the better understood complex and
compound verb structures, such as causatives and desideratives, the V2 is typically
transitive, and V1 takes an independent subject. Escalante (1990) goes to some trouble,
using mostly binding facts, to show that sentences with compound verbs with a transitive
V2 are multiclausal, despite containing only a single Tense node. Harley (2011) treats
these as examples of “clause fusion by embedding a VP, not TP” which results in a single
case domain. Regardless, in such examples it is possible to have a distinct subject of the
embedded verb, which makes it possible to diagnose biclausality with binding.
In the structures under consideration here, however, the obligatory subject sharing
complicates matters. The nominative subject can bind the reflexive object of an
embedded transitive verb, as in (24) and (25), which would seem to be an argument for
Uu chuu’u
DET dog
always 3sgREFL scratch-RED-go
“The dog is always going around scratching itself.”
DEM.DISTAL little boy-PL
always 2plREFL
“Those little boys are always going around pushing each other.”
However, since there cannot be a subject of the V1 distinct from the nominative subject
of V2, in fact the binding facts can tell us very little here; a multiclausal structure
including a null PRO subject of V1 controlled by the matrix overt subject of V2 (or vice
versa) would generate the same binding patterns as a monoclausal analysis. A better
question is whether we can find evidence for a PRO subject of V1, controlled by the
nominative subject, which would indicate a biclausal structure. If no evidence for PRO
can be found, then we might conclude either that the structure is monoclausal, or that V2
-sime behaves as a raising predicate, and does not contribute an argument of its own. In
Chapter 3 I examine raising and control compounds in Hiaki in more detail, and show
that V-sime compounds do not fit either pattern. For example, a possible direction to look
for such evidence involves the interaction of V2 -sime with the impersonal passive -wa,
as in example (26) below.
Imi’i hiva
Here always
“There are always balls being kicked along here.”
In this example, the passive has been applied to the V2 -sime ‘go’, and the nominative
subject has disappeared. The morpheme -wa, as it happens, cannot be applied to
intransitive verbs that do not have an animate subject, and so this suggests that -sime may
have a thematic subject to contribute to the argument structure of the compound. If -sime
does not have a thematic subject, it could not be subject to deletion by -wa, and so this is
evidence that whatever is going on here, raising isn’t it. However, see Jelinek and Harley
(2014) and Harley (2014) for a contrary view.
Ultimately, I conclude that, unlike the majority of Hiaki verbal compounds, motion
compounds are in fact monoclausal, in the sense that they have only a single subject
position. However, this engenders further questions. Although it seems clear that in the
transitive V2 structures the embedded VP1 is occurring as the complement of V2, the
structural connection between VP1 and intransitive V2 is much more difficult to pin
Two broad hypotheses present themselves, and are taken up in following chapters. The
first, following Zubizarreta & Oh’s (2004, 2007) work on Korean serial verbs of motion,
which is one of the few scholarly works to touch on constructions of this nature, is that
VP1 is an adjunct to V2. This is a more feasible structure for a language like Hiaki than it
is for one like English, for reasons that will be discussed in detail in Chapter 5.
The other hypothesis is that verbs like -sime, the ‘verb-affix hybrids’ in Harley, TubinoBlanco and Haugen’s (2014) terminology, are undergoing grammaticalization into
something like an aspectual auxiliary verb. This would seem to be in keeping with the
presumed genesis of several of Hiaki’s obligatorily bound affixes, and with
grammaticalization trends in basic motion verbs crosslinguistically. This possibility is
investigated in more depth in Chapter 4 and Chapter 5.
I will argue that an adjunction analysis, in which VP1 is head-adjoined to a root node
instantiating V2 -sime, most closely fits with the semantic and morphosyntactic
properties of the V-sime construction in Hiaki, and explain in detail what this structure
would look like in Chapter 5.
1.3 Motion
In a series of influential works, Talmy (1975; 1985; 1991; 2000) decomposed motion
events into semantic sub-elements (Figure, Ground, Path and Manner) and proposed a
two-way language typology based on how different languages incorporate the Path
component into linguistic motion expressions (‘verb-framed’ vs ‘satellite-framed’). The
simplicity of Talmy’s typology has come under criticism in recent years, with the
recognition that many languages employ other strategies (such as symmetric, or doubleframing, where both Path and Manner are encoded in the verb) and that within any given
language, different strategies may be employed - “Talmy’s typological classification
applies to individual complex event types within a language, not to languages as a whole”
(Croft 2008:1). In particular, languages with serial verbs or other complex predicate
types are likely to exhibit either a variety of framing strategies, or strategies that do not fit
comfortably within Talmy's dichotomy (Slobin 2004; Son and Svenonious 2008).
Nevertheless, the majority of work on motion constructions in the literature remains
focused on more-studied languages like Germanic and Romance, and on the
lexicalization of either Path or Manner within a lexical verb (Beck & Snyder 2001; Folli
& Harley 2006; Jackendoff 1992, 1997; Levin & Rappaport Hovav 1991, 1995; Mateu
2001, 2002). Zubizaretta & Oh (2004, 2007) broaden this field somewhat by examining
not only Germanic and Romance, but also Korean, which is an interesting case because it
overtly decomposes manner and directed motion in serial verb constructions. The Korean
examples are particularly relevant because the language displays several morphosyntactic
similarities with Hiaki, as well as some interesting differences. The Korean facts, and
Zubizarreta &Oh's analysis will be reviewed in detail in Chapter 5.
1.4 Complex Predication
A wide array of phenomena, in various languages, has been lumped under the umbrella
term ‘complex predicate’ (henceforth CPr). These include, but are not limited to: light
verb constructions; coverb constructions; serial verb constructions; raising and
restructuring predicates; incorporation phenomena, including noun incorporation,
preposition incorporation, pseudo-incorporation and particle constructions; some types of
verbal classifier systems; resultatives; and even, contentiously, control constructions and
auxiliary verb constructions.
Further adding to the confusion, there is a great deal of mismatch in the interpretation of
many of these terms; ‘light verb’ is applied particularly freely, as for instance in Rosen
(1990) who argues that restructuring predicates are a type of light verb, thus collapsing
two of the categories above. Similarly, although Amberber et al. (2010) appear to
consider coverb constructions a distinct phenomenon, Bowern (2008) places them
decisively in the category of light verb constructions.
Complex predication involving motion is one of the most common types in the world,
particularly with respect to multi-verb structures. For example, Durie (1997:310) claims
to be unaware of any language with serial verb constructions that does not include a
category of motion serialization. Aikhenvald (2006:48) goes even further, stating
unequivocally that all serializing languages have, minimally, constructions involving
verbs of motion, posture, orientation and stance. (27) is an example that illustrates what
she labels an ‘asymmetric’ SVC, with the motion verb acting as a deictic or directional
lei lo di saam lai
you take PL clothing come
‘Bring some clothes’
Cantonese (Aikhenvald 2006:21)
Although complex motion constructions occur frequently, in many typologically distinct
languages, they may be realized quite differently, by a number of different structural
variants. Outside of the realm of serial verbs, Rice (2010) reports on a class of activity
incorporates in two Athabaskan languages, Ahtna and Koyukon. These differ from multiverb constructions because they involve a motion or stance verb in combination with a
nominal activity predicate, however they have very similar semantics to the Hiaki V-sime
shout-3pl.s-qual-qual-prog-l voice/
‘They are going along shouting.’
around-shout-qual-prog-l voice/
‘They (mosquitoes) are whining about.’
(Rice 2010:135-6)
Notably, in these constructions the motion/posture verb is the head of the construction,
and is modified by the activity noun. This contrasts with many of the motion SVCs
surveyed in Aikhenvald (2006) and Durie (1997), in which the motion/stance verb is
considered a ‘minor2’ verb and highly susceptible to grammaticalization of various kinds.
A 'minor verb' is one from a restricted class, such as motion or stance verbs, which serializes with an
unrestricted 'major verb' - the major verb controls argument structure and may be considered the semantic
and syntactic head of the construction. (Aikhenvald 2006:22)
In many languages of Central Australia, for example, ‘associated motion’ or ‘associated
path’ can be considered a grammatical category, usually indicated on an auxiliary verb,
and frequently forming a portmanteau with tense, aspect, and/or mood inflection (Wilkins
1991; Simpson 2002/4?; Nordlinger 2001, 2010, 2014).
return 1dl.incl.S-pst.twd
“We came back.”
return 1dl.incl.S-pst.awy
“We went back.”
return 1dl.incl.S-pst
“We returned.”
Wambaya (Nordlinger 2010:237)
So, not only are complex motion constructions similar to Hiaki motion compounds
common occurrences in a wide variety of languages, they also are instantiated by a range
of different structural types. The motion verb may represent the head of a complex verb,
a modifying element, or even grammatical category – each of these functions necessitates
a distinct structural representation, despite the apparent semantic similarities.
1.5 A note on theory and framework
My goal in undertaking this project is to provide an analysis that best models the data,
with no particular theoretical agenda, and which may work within a number of
frameworks. However, like anyone, I am subject to the biases of my own training and
preferences. Thus, my analysis inevitably betrays the influences of modern formal
approaches, in particular, Distributed Morphology, the Minimalist Program, and to some
extent, Bare Phrase Structure, without explicitly endorsing the tenets of any of these.
Chapter 2 Independent motion verbs in Hiaki
Haiseakai Alehandra si chuunti weye? (Why is Alex walking so fast?)
This chapter is primarily concerned with independent verbs of motion. §2.1 describes the
properties of the basic motion verb sime/saka ‘go’ when it functions as a main verb,
including its semantics, argument structure and interactions with various morphological
processes. §2.2 discusses the behavior of other motion verbs, including those that
lexicalize manner or path, and attempt to situate Hiaki motion constructions with respect
to Talmy’s (1975; 1985; 1991; 2000) typology of motion events.
2.1 Basic Motion: siime
Recall from §1.2.3 that many Hiaki verbs display suppletion, and that siime is one of this
class, exhibiting variation in form based on tense/aspect and on the plurality of its
Present, singular subject:
Present, plural subject:
Past perfective, singular
subject: siika
Past perfective, plural subject:
In addition, both the singular subject siime and plural subject saka have a bound form,
which occurs in particular contexts, such as with specific classes of affixes3.
Bound, singular subject:
Bound, plural subject:
More detail is provided in §1.2.3
The suppletive properties of siime, with respect to number features of the subject, have
important implications for argument structure, which will be discussed in §
Additionally, since the example sentences will necessarily exhibit a range of forms, some
familiarity with the possible forms of siime will be useful for the reader.
2.1.1 Formal and featural properties of siime
Semantics and deixis
Siime is usually translated as ‘to go’; it implies motion along a path, but neither
(necessarily) specifies the direction of that path nor the manner of motion. It is an
intransitive verb, requiring only a subject argument, although Hiaki seems to disprefer
simple two word sentences, so further information of some type is usually added.
Frequently, this comes in the form of a postpositional phrase indicating the path or
direction of motion.
When no path PP is present, the default interpretation of siime is ‘leave’ or ‘go from
here’, as in example (35). This is also true of the English verb go, however, unlike
English, Hiaki does not have a straightforward counterpart such as come. It is one of a
handful of typologically and genetically diverse languages that counter the prevailing
tendency of languages to possess a ‘class4’ of basic motion verbs, consisting minimally of
a pair of deicitic verbs encoding ‘motion-towards-speaker’ (come) and ‘motion-nottowards-speaker’ (go). However, even in languages that do have a clear oppositional pair
While motion verbs form a class according to notional definitions, cf Talmy (1975 et. seq.), there is some
dispute about whether they can be considered so on morphosyntactic grounds. (Wilkins & Hill 1995; Levin
& Rappoport 1992)
such as this, it is common for the ‘go’ verb to have a more generalized and non-deictic
use. Certainly in Hiaki, the apparent deictic interpretation of siime as ‘motion-nottowards-speaker’ can be subverted with the use of an appropriate Path PP, as in example
(36) where siime is translated as ‘come’. This indicates that the deictic interpretation of
siime in examples such as (35) is a pragmatic implication rather than a featural
specification. (Wilkins & Hill 1995)
Vempo nau
“They left together.”
“They came towards me.”
There is another verb, yaha/yepsa, which is sometimes translated as ‘come’, but more
usually ‘arrive’. It may not, for instance be used in the context of (36)—the sentence in
(37) is not acceptable.
“S/he came/arrived towards me.”
Siime is unaccusative
Distinguishing unaccusative intransitives from unergatives in Hiaki is not as
straightforward a proposition as one might wish.
Jelinek & Escalante (2000) argue that the passive/impersonal suffix -wa targets
Agent/Causer subjects, and thus can be used to diagnose unergative verbs. As in English,
in Hiaki passive sentences the semantic subject is absent, and the object of a transitive
verb gets promoted to the nominative subject position. However, in Hiaki you can also
passivize intransitive verbs, resulting in subjectless sentences – this is demonstrated by
the alternation in examples (38)-(39).
“Pete is singing.”
“Singing is happening.”
It is impossible to add agents in a ‘by’ phrase in Hiaki. Whenever -wa appears the
interpreted subject is generic, third person, plural and human – ‘people’. This also means
you can’t passivize sentences with a non-human subject (40)-(41).
Uu puato hamte-k
dish shatter-PFV
“The dish shattered.”
Intended: “Shattering happened.”
You can passivize siime (42), which lead Jelinek and Escalante to conclude that siime
must be agentive, and thus unergative.
Aman saka’a-wa
There go.PL-PASS
“(People) are going over there.”
However, -wa can also occur with the verb muuke/koko ‘to die’, which is interpreted with
an Experiencer subject.
Yellow(fever)-ACC-with die.PL-PASS-PST
“(People) were dying from yellow fever.”
Harley, Tubino Blanco and Haugen (2009) argue that muuke/koko is inherently
unaccusative, and disqualify -wa as a tool for diagnosing unergatives. They claim that the
applicative morpheme, -ria, may instead be used to identify unaccusative intransitives.
The applicative adds an accusative benefactive argument to the clause. It is productive
with all manner of agentive verbs, but doesn’t occur with non-agentive or suppletive
intransitive verbs, even when there’s no semantic/pragmatic conflict, as the examples in
(44)-(45) show. They take this to indicate that the suppletive intransitive verbs are
inherently non-agentive, as would be expected if they were unaccusative.
Santos Maria-ta vetchi’ivo San Xavierle-u weye
Santos Maria-ACC for
San Xavier-to walk
“Santos is going/walking to San Xavier for Maria.”
*Santos Maria-ta San Xavierle-u weye-ria
Santos Maria-ACC San Xanvier-to walk-APPL
“Santos is going/walking to San Xavier for Maria.”
(Harley, Tubino Blanco and Haugen 2009:48)
The ungrammaticality of the applicative with suppletive intransitive verbs (including
siime) taken together with the fact that suppletive transitive verbs' form is conditioned by
their internal, not their external argument, leads Harley, Tubino Blanco and Haugen to
argue that verb suppletion is always triggered by internal arguments. This is consistent
with Bobaljik’s (2012) claim that suppletion can only be conditioned in a strictly local
relationship, within the same maximal projection – ie, sisterhood. The consequence of
this locality constraint is that all suppletive intransitive verbs, including siime, must be
2.1.2 Interactions
Tense and aspect marking in Hiaki are often difficult to distinguish clearly. It is not clear
in all cases whether a given morpheme marks tense or aspect or a combination of both. I
will simply describe most of these morphemes in terms of both the tense and aspectual
information they typically indicate.
The future/irrealis suffix -ne is the only one of the tense/aspect suffixes that is
classed as ‘stem-changing’. (Dedrick & Casad 1999, Harley & Tubino Blanco
2010) That is, it requires the bound form of the stem shown back in (34). The
future inflected forms of siime are hence sim-ne (singular subject) and saka’a-ne
(plural subject).
Potam-meu sim-ne
tomorrow=1SG.NOM Potam-to
go.SG -FUT
“I am going to Potam tomorrow.”
tomorrow=1PL.NOM Potam-to
(Dedrick & Casad 1999:293)
go.PL -FUT
“We are going to Potam tomorrow.”
Present imperfective
Morphologically unmarked verbs indicate present tense, and imperfective (or
continuative) aspect. (Note that I gloss this form simply as ‘PRES’ for simplicity’s
vicha siime
3SG.NOM 1sg.ACC-to towards go.SG.PRES
“S/he is coming towards me.”
ne-u vicha
3PL.NOM 1sg.ACC-to towards go.PL.PRES
“They are coming towards me.”
(Past) perfective
The suffix -k is the usual marker of perfective aspect in Hiaki, but siime suppletes
to the forms siika and sahak instead. Perfective is translated as simple past in
English, and is the most common form used to describe past events, although
Dedrick and Casad (1999:318) argue that it does not indicate tense, per se, but
that perfective (completed) actions are considered ‘past’ by default.
vicha siika
3SG.NOM 1sg.ACC-to towards go.SG.PFV
“S/he came towards me.”
ne-u vicha
3PL.NOM 1sg.ACC-to towards go.PL.PFV
“They came towards me.”
Past (imperfective)
Dedrick and Casad consider that the suffix -n “may be the only genuine tense
marker in [Hiaki]” (1999:318). Affixed to the free, citation (present perfective)
form of a verb, it typically maintains imperfective aspect and adds past tense.
3SG.NOM 1sg.ACC-to towards
“S/he was (going to be) coming towards me…”
ne-u vicha
3PL.NOM 1sg.ACC-to towards go.PL-PST
“They were (going to be) coming towards me…”
This is not the whole of the story, however. In both (52) and (53), the implication
is that the intention to come was there, but that the event itself was prevented
from happening for some reason – these sentences are judged to be incomplete,
and speakers suggest adding an explanatory ‘but’ clause. This is consistent with
an aspectual contrast often seen in perfective/imperfective languges – perfective
forms are used for narrative advancement, while imperfectives are used for
backgrounding effects. Thus, the use of the imperfective in the examples above,
creates a sense of incompleteness – if an event is presented as background, there
must be another, foreground event to be described. (Sebastián & Slobin 1994:253254)
Past perfect5
The suffix -kan is, in the usual case, translated as something like past tense perfect
aspect, as in the example with hi’ibwa ‘eat’ below.
Alleh kaa hi’ibwa-k
Alex NEG
bwetuk ketwo
eat-PFV because morning
“Alex didn’t eat because (she) had eaten earlier.”
Because siime is a suppletive verb, we don’t get a form siime-kan or saka-kan.
What we do get is siika-n and sahaka-n – which looks like the perfective form
suffixed with past tense -n. This is interesting on the one hand because it suggests
that -kan may be historically derived from -k(a)6 + -n. More interesting, however
is that these forms do not receive the translation that we would expect to see if
these verbs fit with the usual V+kan pattern.
Although I am labeling the -kan suffix ‘past perfect’ for lack of a better alternative, that description is
almost certainly wrong, given that a perfective/imperfective language with a perfect category is highly
unlikely. More work certainly needs to be done to tease out the details of the aspectual system in Hiaki, but
such work is regrettably outside of the scope of the current project.
Note that this is not necessarily the past perfective –k(a). Hiaki also has a participial suffix -ka, which will
be discussed in §
3SG.NOM 1sg.ACC-to towards go.SG.PRF-PST
“S/he was coming towards me…”
Not “S/he had come towards me.”
nee-u vicha
3PL.NOM 1sg.ACC-to towards go.PL.PRF-PST
“They were coming towards me…”
Not “They had come towards me.”
In (55) and (56) the interpretation is past imperfective, which was expected in
(52) and (53). In those examples, the assumption is that the subjects never
embarked on their intended path of motion. In these cases, in contrast, the subjects
are assumed to have embarked, and the expected next clause would describe some
event that happened along the way.
Given these data, it is natural to assume that perhaps the analysis of these forms is
incorrect and that these are not examples of siime + -kan at all. If that is the case,
then there is no obvious way to inflect siime with –kan; the forms *siime-kan or
even *sim-kan are impossible.
Other aspectual categories
Hiaki has a number of other affixes that supply information of an aspectual
nature, such as -taite (inceptive), -yaate (cessative), -su (completive), -pea
(desiderative), and -vae (prospective). However these affixes behave differently
to those above. Dedrick & Casad (1999) consider them variably aspectual or
adverbial. Escalante (1990) classes them as bound verbs, elements in complex
verb constructions, as do Tubino-Blanco, Harley & Haugen (2009). These affixes,
unlike all other tense/aspect suffixes except –ne, suffix to bound forms of their
verbal stems. Further, this class of affixes co-occurs with the tense/aspect affixes
above, and always takes the inner position.
The most common interpretation of reduplication of Hiaki verbs is habitual aspect. Other
readings, such as plural subject, ongoing or progressive action, or intensification are also
possible in certain circumstances (Harley & Amarillas 2002). Siime is typical in this
regard; its reduplicated form is usually used as an indicator of habitual action (58).
Chepa kari-po
yeu siime
Chepa house-LOC
out go.SG.PRES
“Chepa is leaving the house.”
Chepa kari-po
yeu si-sime
Chepa house-LOC
out RED-go.SG.PRES
“Chepa leaves the house (regularly).”
However, in the appropriate context the reduplicated form can also be interpreted as
indicating an immediate and ongoing action (59).
3SG.NOM 1PL-to toward
“S/he is coming towards me (right now).”
Whether the reduplicated form is interpreted as habitual or ongoing, it is incompatible
with the perfective stem siika (61), as is typical of such reduplication in the language.
Chepa kari-po
yeu siika.
Chepa house-LOC
out go.SG.PFV
“Chepa left the house.”
**Chepa kari-po
Chepa house-LOC
out RED-go.SG.PFV
Intended: “Chepa left the house (regularly).”
It combines with the past imperfective to give a past habitual or ‘used to’ interpretation
Uu kamion wovusanim-po
The bus
“The bus used to leave at seven.”
As noted in §, when applied to transitive sentences, the suffix -wa behaves like a
typical passive morpheme – the expected subject argument is absent and the object (or
highest accusative argument) is obligatorily promoted to subject position and assigned
nominative case (63)-(64).
Vahi o’ow-im
Three man-PL
“Three men killed the deer.”
DET.NOM deer
“The deer was killed.”
(Escalante 1990b:290)
In addition -wa can be applied to intransitive verbs, and in these cases it results in a
subjectless clause, as in (38)-(39), reproduced here as (65)-(66). The unexpressed subject
is treated as an understood non-specific indefinite, glossable as ‘people’ or ‘they’; in this
regard it is an impersonal construction.
“Pete is singing.”
“(People/They) are singing.”
Like the future marker -ne, -wa is a stem-changing suffix. Furthermore, when it attaches
to a verb that suppletes for subject number, that verb will be in its plural subject form7.
This means that when siime is marked with the passive/impersonal, it is always in its
plural bound stem form saka’a-.
Aman saka’a-wa
There go.PL-PASS
“(People) are going over there.”
Regardless of whether -wa is affixed to a transitive or intransitive verb stem, it is
impossible to include the understood subject in the clause (such as in an oblique ‘by’
phrase, as in English passives8). However, the understood subject must be human, and
this is the only apparent restriction on the formation of -wa constructions; -wa can attach
Harley (2014) argues that this does not entail that the unexpressed subject must be interpreted as plural,
although it is in the default case. Even when the context indicates that the subject is likely singular, the verb
must still show plural agreement. I show this in examples (177)-(178) in Chapter 3.
Although Escalante (1990b) points out that oblique instrumentals may be included.
to any verb, so long as the subject that is suppressed is human (and not to any verb whose
suppressed subject is nonhuman). Thus, -wa can be shown to occur with unaccusative
and other non-agentive verbs. (Harley, Tubino Blanco and Haugen 2009; Harley 2014)
In subordinate adjunct clauses of the ‘while’ type, the semantics of siime are exactly what
we would expect from it in a main clause; in the absence of a specified path siime is
interpreted as ‘leave’, otherwise, as undirected motion. Subordinated siime appears in one
of two forms: sime-ka (68), which seems unambiguously to be an untensed participle, and
sika (69) which occurs in more limited contexts, and the analysis of which is somewhat
more mysterious.
Note that both of these forms have shortened vowels. In §, I demonstrated that
long vowels are often shortened in the environment of particular suffixes.
Acheka sime-ka(-su),
go-PCL-PRT coffee-ACC
“As she was leaving, HK spilled the coffee.”
Acheka sika
go.SG walk-PCL-PRT coffee-ACC
“As she was leaving, HK spilled the coffee.”
veveme sahaka, kaate-ka-su, kafe-ta
DET.PL girl.SREL go.PL walk-PCL-PRT coffee-ACC
“The young girls who were leaving, while walking, spilled the coffee”
In (69) sika appears in concert with the participial form of the verb weye ‘walk’, so this
sentence can perhaps be more fully translated: “As she left walking, HK spilled the
coffee”9. It is not entirely clear what the status of sika is in this construction - more data
and exploration is required to make a convincing analysis. It is possible that this is simply
an alternative participial form, and its range of use simply needs to be clarified with
further examples.
2.2 Other Verbs of Motion
In a series of influential works, Talmy (1975; 1985; 1991; 2000) decomposed motion
events into sub-elements (Figure, Ground, Path and Manner) and proposed a two-way
language typology based on how different languages incorporate these elements. Talmy
focused on how languages expressed Path in particular; hence in ‘verb-framed’ languages
path is expressed on the main verb of the clause, whilst in ‘satellite-framed’ languages, it
is expressed outside of the verb, by means of a particle or phrase (Slobin 2004).
The simplicity of Talmy’s typology has come under criticism with the recognition that
many languages employ other strategies (such as symmetric, or double-framing, where
the path is expressed both as part of the verb and in an accompanying satellite) and that
within any given language, different strategies may be employed. “Talmy’s typological
I note, however, that my consultants find this translation awkward and unnecessarily detailed.
classification applies to individual complex event types within a language, not to
languages as a whole” (Croft 2008:1).
2.2.1 Path of Motion
Hiaki has quite a small sample of verbs that lexicalize path, and there is a certain lack of
symmetry regarding which paths are encoded this way. For example, there is ‘fall’ but
not ‘rise’, ‘arrive’ and ‘enter’ but not ‘depart’ or ‘exit’. (As mentioned earlier, siime can
be used to mean ‘depart’ in some circumstances, but usually an additional particle – yeu
‘out’ – is required if one wishes to be explicit on that point.)
Intransitive verbs
Fall/drop (as of rain, fur)
Singular subject
Plural subject
Transitive verbs
Singular object
Plural object
(Molina, Valenzuela & Shaul, 1999)
2.2.2 Manner of Motion
Satellite-framed languages, which do not lexicalize Path of motion in the main verb,
commonly express Manner of motion in the verb instead (Slobin 2004). The verbs in
(73)-(74) do not need to form a compound with -sime in order to acquire a motion
interpretation. Although Hiaki does have more simple verbs that lexicalize manner than it
does verbs that lexicalize path, still these appear to be a far smaller percentage than in a
language such as English.
Intransitive verbs
Roll over
Singular subject
Plural subject
waka’ate / waka’aname / sunsunte
Transitive Verbs
Carry on head
Drive (herd)
Jump over
Shove (in)
Step over
himma, hissa
(Molina, Valenzuela & Shaul, 1999)
Most of the simple manner-of-motion verbs that I was able to find in the Hiaki dictionary
are relatively basic. Many more detailed or explicit manners are created by the use of
complex expressions. These include compound verbs, of the sort that are the focus of this
dissertation, as well as modification by non-verbal elements such as adjectives and nounincorporation.
It seems clear that, based on these data, Hiaki would be classified in Talmy’s typology as
a satellite-framing language, given the propensity for the Path component of the event to
be expressed in satellite expressions, such as particles, rather than by the main verb. This
classification becomes a little more complicated when we enter the territory of complex
and compound verbs, since it is no longer entirely clear exactly where the line between
‘verb’ and ‘satellite’ lies (Talmy 2000; Croft 2008). These issues will be explored
further in Chapter 3, which investigates Hiaki complex verbs in detail.
2.3 Summary
The motion verb siime is a complicated beast, with multiple suppletive forms. It may
imply, but does not stipulate, path or manner of motion – where these interpretations exist
they are presumed to be a product of pragmatics, rather than featurally specified. Siime is
unaccusative, and though it can occur with a Path or Goal PP, it need not. Other simple
motion verbs that do specify manner of motion are also capable of licensing the presence
of a Path or Goal phrase.
Chapter 3 Complex Motion in Hiaki
Acheka havivu atsisime (HK is always going about laughing.)
In Chapter 1 I gave a brief overview of some interesting features and puzzles in the study
of both motion expressions and complex predication, and in Chapter 2, I discussed Hiaki
simple motion constructions, focusing on the behavior of the basic motion verb
siime/saka ‘to go’. In Chapter 3, I delve into Hiaki complex verbal structures, and begin
to explore the convergence of motion and complex predication.
This chapter has four subsections. The first, §3.1, is an overview of the types and
properties of Hiaki complex verbal structures based on clausal and argument structure. I
claim that most Hiaki complex predicates are biclausal structures that fall into one of
three types: ECM, Raising or Subject control. In §3.2, I consider complex verbs of
motion, formed with -sime/-saka, and discuss how they compare to other complex verbs
in the language, again examining clause and argument structure as well as the role and
characteristics of the VP1. I argue that V-sime constructions do not clearly pattern with
any of the biclausal structures discussed in the previous section, but has properties that fit
with a monoclausal analysis. In §3.3 I compare the behavior of -sime/-saka in its role as a
participant in a complex structure with its behavior in a simple predicate clause, assessing
interactions with tense/aspect and other morphological operations. §3.4 summarizes the
main findings and conclusions thus far.
3.1 Overview of Hiaki Complex Verb Structures
This section is aimed at detailing as precisely as possible the range of structures involved
in Hiaki complex verbs, in order to then determine where the V-sime constructions fit
amongst them, or if they represent a different type altogether. Additionally, this laying
out of properties is a crucial step towards establishing how to fit Hiaki complex verbs in a
wider, crosslinguistic typology of complex verbal structures. This is not a trivial task,
since “One person’s complex predicate or compound verb is another person’s serial verb,
composite predicate, auxiliary construction, or even a control construction.” (Butt
To that end, §3.1.1 will include a brief sketch of previous descriptions and analyses of
Hiaki complex verb types, some of my assumptions regarding clausal and argument
structure, and some background information about verbal morphology and stem types.
3.1.1 Types of Complex Verbs in Hiaki
The division of complex verbs into ‘types’ or classes has different uses depending upon
the criteria used in the division. In §, for instance, I provide a brief rundown of the
division provided by Escalante (1990) which is based upon the ability of the V2 in the
compound to occur as a free lexical verb or not, briefly described in Chapter 1 above. I
will show, following Harley & Haugen (2010) that this division is not highly significant
to the question of clausal structure, since both bound and free V2s can demonstrate the
same range of clausal types.
In § I lay out some assumptions regarding clausal structure and its definition, and
in § provide some relevant background on the forms that Hiaki verb stems take in
particular morphosyntactic environments.
Free vs. Bound
Hiaki has been described as having two broad classes of complex verbs. This division is
based on the status of the V2 element as either bound or (potentially) free, rather than on
clausal or argument structure. The first of these, called ‘complex verbs’ by Escalante
(1990) involves a lexical V1 with a bound V2 affix (or affixes) which has some aspectual
or other verbal properties, and which cannot function as an independent verb. These
obligatorily bound items have functions such as: direct causative -tua, indirect causative tevo, applicative -ria, desiderative -pea, inceptive –taite, and more.
“I will make him/her sing.”
(Escalante 1990:40)
The second type described by Escalante is ‘compound verbs’, in which two (or
potentially more) verbs are combined, each of which has independent lexical verb status.
Although complex verbal structures are very common in the language, verbal compounds
appear to be a reasonably restrictive class, in the sense that there are relatively few verbs
which commonly show up as the final or head verb in such a structure. Independent verbs
that can occur in V2 position in compounds include: mahta ‘teach’, naate ‘start’, vicha
‘see’ and of course siime ‘go’.
3sgNOM 2sgACC
“He taught you to dance.”
Escalante (1990) points to the fact that several of the obligatorily bound V2 affixes may
have derived from historically independent verbs, so the line between these two classes is
inherently blurry. Harley and Haugen (2010:14) go so far as to refer to those independent
verbs that can occur in V2 as ‘verb-affix hybrids’, noting that when they occur in
compounds they behave identically to affixal verbs with respect to binding and the
assignment of case in embedded clauses – this is demonstrated in §3.1.2, §3.1.3 and
§3.1.4, in which both bound and free V2s are used to exemplify three distinct clausal
structures that can be identified amongst Hiaki complex verbs.
Defining clausality
One of the defining characteristics of a complex predicate as understood by scholars like
Butt (2003 et seq.) is monoclausality. In a monoclausal complex predicate, two elements
predicate as a single unit with a united argument structure. In a biclausal structure,
however, two predicates maintain individual syntactic domains with arguments shared
across them. (Butt 2003)
Biclausals may vary according to the size of the embedded structure. Larger embedded
structures such as CP or TP can be identified by the presence of an embedded topic or
embedded case/tense/agreement respectively. Embedded argument structure, defined by
the existence of binding domains, indicates a VP or vP sized constituent. (Butt 2003,
Harley 2008, Wurmbrand 2001)
Verbs that take a bare VP complement may be restructuring predicates, and result in
‘clause unification’. This terminology is indicative of the somewhat fuzzy position that
these structures inhabit between more clear-cut examples of monoclausal and biclausal
structures; the argument about whether a VP complement structure is considered monoor biclausal comes down to whether or not there is an embedded subject position
available. In a unified or monoclausal structure there is no embedded subject of the lower
VP, but semantic ‘sharing’ of the subject of the higher verb. (Wurmbrand 2001, Cable
2004) These distinctions are roughly sketched in (77)-(78), and will be examined in more
detail in Chapter 4 and Chapter 5.
Subject V1 DP
Object V1
Escalante (1990) goes to some trouble to show that sentences with both complex and
compound verbs (with a transitive V2) are biclausal with respect to binding domains.
They may not, however, embed Tense or even Negation, which indicates that the
embedded clause is limited to a VP or vP sized constituent. Tubino Blanco & Harley
(2012) argue that the relevant constituent is in fact VoiceP10.
Within a biclausal structure of this type, there are multiple methods of reconciling the
argument structure needs of both predicational elements. Arguments of the inner VP1
may be shared with11, or embedded under, the higher V2. Objects of V1s are always
assigned accusative case by their verb; it is the subjects of V1 that are treated differently
depending upon the properties of the V2. V2s can be grouped into one of these three
They argue for a 3-layered VP structure, which is illustrated in §5.3.1.
As in the case of subject raising.
1. ECM predicates, in which the subject of the lower VP is overt and accusative;
2. Raising predicates, which do not themselves contribute a thematic subject, and so
the subject of the lower clause raises to occupy the higher subject position;
3. (Subject) Control predicates, in which the subject of V1 is a phonologically null
element (PRO) that is bound by the coreferent subject of V2.
In § 3.1.2- § 3.1.4 I provide examples of affixal and lexical V2s that have each of the
properties described above. First, however, I take a small detour in the next section in
order to clarify how different classes of affixes affect the form of Hiaki verbal stems.
Verb Stems
Hiaki verbs have both a free stem and a bound stem form as illustrated in (79)-(80):
“I am dancing.”
“I will dance.”
(Escalante 1990:38)
There are several classes or sub-classes of verbs based on stem alternation, including an
invariable class, however, I will not discuss these classes in detail12. The more salient
point that I wish to make here is that some suffixes must attach to free stems – such as -k
(past perfective), -n (past imperfective), and -kan (past perfect) – but that the majority
attach to bound stems. Bound stem suffixes include: -wa (impersonal/passive), -ne
(future/irrealis), -na (future impersonal/passive), -tua (direct causative), -tevo (indirect
causative), -ria (applicative) and many others. The class of affixal verbs that occur in
complex verbs always attach to bound stems.
In verbal compounds, the first verb (V1) always appears in its bound form. The form of
the second verb (V2) varies depending upon the nature of any subsequent affixation. In
example (81) below, the inner VP ‘you dance’ is the complement of the V2 -mahta
‘teach’. The V1 ye’e ‘dance’ is in its bound form yi’i- (compare with (80) above) while
the form of V2 is free.
“He taught you to dance.”
(Escalante 1990:38)
As a loose generalization, we might say that free-stem affixes are inflectional, and boundstem affixes are typically derivational/compounding. Note, however, that this
generalization has clear prima facie exceptions: the future/irrealis -ne, and the
The full set of verb stem classes and their properties are outlined in Tubino Blanco & Harley (2010).
passive/impersonal -wa are bound-stem affixes. A more hard-and-fast rule is that, in the
event of affix-stacking, free-stem suffixes must always follow suffixes that take bound
stem forms. For example, the verb vuite ‘’ has a bound form vuiti-. The past
perfect* suffix -kan is a free stem suffix, in (82) we can see that the verb appears as vuite.
However, when combined with the completive -su, which is a bound stem suffix, -su
precedes –kan, and the verb stem is realized as vuiti- (83). (Escalante 1990, Harley 2011,
Harley and Tubino Blanco 2012, Tubino Blanco and Harley 2010)
“He/she had run.”
“He/she used to run.”
3.1.2 ECM V2s
What I am calling ‘ECM’ V2s are verbs that select for a subject and a propositional
complement (VP1). The subject of V2 receives the sole Nominative case marking
available; the second DP, which is always the semantic subject of V1, appears in the
accusative. The mechanics of how accusative case marking is received – by checking,
movement, feature valuation or some other technology – is not critical to my purpose
here. The important thing to note is that an ECM construction contains distinct – and
overt – subject arguments for each verb.
Affixal ECM: Direct Causative (-tua)
The direct causative suffix in Hiaki adds an external Causer argument, which takes
Nominative case; the embedded subject of the lexical verb is marked Accusative.
“He/she will sing.”
“I will make him/her sing.”
(Escalante 1990:40)
When the lexical verb is transitive, both of the embedded arguments are marked
Maria hitevi-ta
Maria doctor-ACC
“Maria made the doctor treat the child.”
(Harley 2011 (h5))
With -tua, the Causee argument – ‘the doctor’ in these examples – cannot be omitted, as
shown in (87) and this is also the argument that is promoted if passivization is applied to
the direct causative structure, as in (88).
Intended: “Maria made (someone) treat the child.”
Uu hitevi
“The doctor was made to treat the child.13” (Harley 2011 (h5))
When the Causer and the Causee are coreferent, the Causee is realized as a reflexive (89).
Inepo ino
bwiik-tua-vae-n, taa=ne
aa bwiika-k
lsgNOM lsgREFL sing-CAUS-PROPS-IMPF, but=1sgNOM
able sing-PFV
“I wanted to make myself sing, but I wasn't able to sing.”
(Escalante 1990:40)
If the Causer is co-referent with the embedded Theme, however, a pronoun must be used,
as in (90), and example (91) shows that an anaphoric object of V1 cannot be controlled
Passive/impersonal -wa does not permit the Causer argument to be added in a by phrase.
by the matrix subject. Since anaphors must be controlled within their binding domain14,
and pronouns must be free15, this is evidence that the VP embedded under the causative is
a distinct domain for binding. (Escalante 1990)
1sgNOM Art-ACC
1sgACC care.for-CAUS
“I make Art take care of me.”
1sgNOM Art-ACC
1sgREFL care.for-CAUS
Intended: “I make Art take care of myself.”
(Harley 2011)
This is further supported by the examples in (92)-(93), which also show that the Causee
argument can bind a reflexive object, but as in (91) the upstairs Causer cannot.
“Hei will make Petej shave himselfj/*i.”
“Hei will make Petej shave himi/k.”
Principle A
Principle B
Lexical ECM: ‘to teach’ (-mahta)
Example (94) shows the verb mahta as an independent verb; in (95) it is occurring in its
role as V2 of a compound.
3sgNOM 2sgACC
“He taught you the prayer.”
3sgNOM 2sgACC
“He taught you to dance.”
The compound structure in (95) has the inner VP ‘you dance’ serving as the complement
of mahta ‘teach’ – ye’e ‘dance’ occurs in its bound stem form (yi’i) , without any tense
marking of its own, making it clearly distinct from coordination structures such as (96)
and (97).
Empo bwiika-k,
into ye'e-ka
2sgNOM sing-pfv,
3sgNOM and dance-pfv
“You sang and he danced.”
Empo ye'e-ka,
taa aapo
kaa ye'e-ka
2sgNOM dance-pfv,
but 3sgNOM NEG dance-pfv
“You danced but he didn't dance.”
The embedded subject of ‘dance’ in (96) is accusative, and not nominative as we see in
the conjoined structures. Because the complement of mahta is a VP, not a TP, there is
only one case domain, and therefore only a single argument, the highest, can receive
nominative case.
The accusative phrase must nevertheless be a subject because it can control an anaphor,
which is a property of subjects (Escalante 1990). At the same time, the nominative
subject of V2 -mahta cannot bind a reflexive in the object position of the embedded VP.
Heidi Art-ta
Heidii Artj-ACC
“Heidi teaches Art to take care of himself/*her.”
Heidi Art-ta
Heidii Artj-ACC
“Heidi teaches Art to take care of her/*himself.”
The embedded VP thus behaves as a distinct domain for binding, indicating a biclausal
structure. In all these properties, lexical V2 -mahta is directly comparable with the
causative affixal V2 -tua.
3.1.3 Raising V2s
A raising construction contains only a single subject argument, and it is the thematic
argument of the V1. The V2 selects only for a propositional complement, thus the subject
of V1 is raised to the matrix clause and assigned nominative case marking.
The distinction between a biclausal raising construction and a monoclausal construction
lies in the availability of two distinct subject positions; crucially, the subject argument
must be selected for by the V1, and thus base-generated within the lower VP1 before
raising to the higher position.
Affixal Raising: Inception (-taite)
Although the inceptive suffix -taite is probably diachronically derived from a lexical
verb, synchronically it exists purely as a bound element.
(100) *Uu
Intended: “The car started.”
Unlike causative -tua, -taite does not add any arguments to the argument structure of the
(101) Ili
“The child is crying.”
(102) Ili
“The child started crying.”
It appears freely with intransitive (102) or transitive (103) main verbs, and has no impact
on case assignment.
(103) Hunume’e veha
“Those ones had already begun to teach me.”
The inceptive also has no restrictions with respect to subject properties such as animacy –
example (104) shows an inanimate subject – which would indicate a selectional
relationship between taite- and the subject argument.
(104) Uu karo
DET car.NOM break.down-INCEP-PFV
“The car is starting to break down.”
Example (105) demonstrates that -taite can also occur with a weather verb, which is
entirely subjectless. If -taite had an argument to contribute, it is in these conditions that
one would expect it to be overtly expressed; since no such argument appears, -taite is
classed as a raising predicate.
(105) Yuk-taite
“It is starting to rain”
Lexical Raising: ‘to seem’ –machi
Although it is relatively rare for an intransitive lexical verb, other than siime, to occupy
the V2 position in a compound, one that does is maachi ‘seem/look/appear, be light’, and
it has several interesting properties.
In its main verb use, it has two meanings or interpretations. Typically, it is used to talk
about outward appearance, as in (106).
(106) Uu mesa si
DET table very dirty looks
“The table looks very dirty.”
(107) Puatom si
Plates very
haiti ma-machi
dirty RED-looks
“The plates look very dirty.”
Its other use has to do with being or becoming light (as of the sky).
(108) Pa’akun haivu maachi
Outside already be(come).light
“Outside, it is already dawning/becoming light.”
Maachi also interacts atypically with reduplication – it reduplicates only in agreement
with plural subjects, as in (107) above; with a singular subject, as in (109) reduplication
is simply ungrammatical. The usual habitual reading of reduplication does not occur.
(109) *Uu mesa si
DET table very dirty
Intended: “The table regularly looks very dirty.”
When it appears in V2 position, -ma(a)chi is similarly complex. V2 -ma(a)chi also has
two different meanings or uses: the first is ‘seem/appear’, similar to its lexical use. This is
shown in examples (110), which also demonstrate that reduplication is restricted to plural
subjects, just as with main verb maachi.
(110) Vempo
“They seem really hateful.”
(111) Aapo
“S/he seems really hateful.”
(112) Vempo
“They all seem really hateful.”
(113) *Aapo
Intended: “S/he usually seems really hateful.”
In other compounds, however, -ma(a)chi creates a modal interpretation, translated as
(114) Ili uusi bwan-machi
Little child.NOM cry-appear
“That child should cry.”
(115) Merehilda Lioh-nok-machi
Merehilda god-talk-appear
“Merehilda should pray.”
It is interesting to note that in the ‘should’ examples (114) and (115), -machi is realized
with a short vowel; in the previous ‘seem’ examples, unreduplicated -maachi is realized
with a long vowel, like the independent verb.16 Since phonological reduction is a
common indicator of grammaticalization of a lexical item into a more functional role, this
distinction could potentially highlight either an in-progress shift, or two distinct paths of
grammaticalization from independent verb to raising verb in one case, and to a modal
auxiliary in the other.17
Like the inceptive affixal verb -taite, -ma(a)chi does not appear to add to the clausal
argument structure, and it can embed the subjectless weather verb yuuku, shown in (116).
It is perhaps significant that -ma(a)chi gets a modal interpretation in this construction;
any attempt to create a ‘seem’ reading requires a different construction altogether (117).
(116) Si
“It should really rain.”
(117) Yuke-m-ta
“It looks like rain.”
See § for discussion on vowel-shortening contexts in Hiaki.
Cf. Bowern’s (2008) discussion of distinct paths of grammaticalization for light verbs and auxiliaries – I
expand on this in §4.2.
From the data collected here, it appears that the ‘seem/appear’ interpretation is restricted
to complements that can be interpreted statively, similar to its main verb use with
adjectival predicates. With dynamic complements, the modal interpretation holds.
3.1.4 Control V2s
A (subject) control structure has two available subject positions, and both V2 and V1
select for subject DPs. However, only the higher DP may be overtly pronounced and
receive (nominative) case. The embedded subject is controlled by (coreferent with) the
matrix subject. Control constructions are distinguished from raising constructions by the
relationship between the subject argument and the V2. As we saw in the discussion of
raising predicates, raising V2s do not contribute a thematic subject – control V2s do.
Affixal Control: Inclination (-pea)18
The suffix -pea requires control of the embedded subject – it is not possible to have a
non-coreferential argument in this position, as shown in (119).
(118) Inepo
“I feel like singing.”
(119) *Inepo
“I’d like Maria to sing.”
This suffix is glossed ‘inclination’, in order to keep it distinct from the (ECM) desiderative -‘ii’aa.
It is in fact not possible to have any overt argument in the embedded subject position,
even one which is coreferential with the higher subject.
(120) *Inepo ino bwik-pea
Intended: “I want myself to sing.”
(121) *Inepo nee bwik-pea
Intended: “I want me to sing.”
Because -pea expresses inclination and attitude, its subjects are pragmatically restricted
to animate humans. For instance, the sentence in (122), while not rejected as strictly
ungrammatical, is nevertheless met with skepticism on the grounds that a chicken’s
feelings and desires are not particularly knowable.
(122) ??Uu toto’i
DET chicken.NOM
feathers drop-INCL
“The chicken feels like dancing” (Idiomatic)
“The chicken feels like dropping feathers” (Literal)
Unlike raising predicates, such as inceptive -taite, -pea is disallowed with subjectless
weather predicates such as yuuke ‘rain’, shown in (123). Here, the judgment is
unambiguously ungrammatical. It is only possible to use -pea with yuuke if it is
previously compounded with a predicate like causative -tua, which brings an appropriate
subject argument along with it (124).
(123) *Yuk-pea
Intended “It feels like raining”
(124) Si
“I wish it would rain!” (Lit: “I feel like making it rain!”)
Lexical Control: ?
Since there are relatively few intransitive lexical verbs that may take the V2 position in a
verbal compound, there is unfortunately no obvious example to fill this slot in the
paradigm I have sketched out here.
There is, a transitive verb - eiya/-eiya, glossed roughly as ‘feel’ – which intuitively seems
to be a candidate for subject control treatment, although further work would need to be
done to confirm this. Evidence of subject control properties, would be similar to that
shown for -pea; subject restrictions, inability to appear with weather predicates, etc.
(125) Nee
enchi eiya
1SG.NOM very 2SG.ACC feel
“I have feelings for you.”
(126) Nee
1SG.NOM sing-SUBJ.REL feel
“I like the singers.”
(127) Nee
1SGNOM DET.ACC snake-ACC fear-feel-PFV
“I was frightened of the snake.”
(128) Acheka taa temai
but ask
bwetuk aapo si’imeta hune’-eiya
because 3SGNOM all think-feel
“Ask HK, she always knows everything.”
3.2 Motion Compounds
Having looked in detail at the range of complex verb structures seen in Hiaki, the task in
this section is examine the properties of V-sime compounds in order to determine which
of the above three classes it most resembles, or alternatively, if V-sime compounds
comprise another type entirely.
Since -sime constructions always have only a single overt subject argument, they clearly
cannot be of the ECM type, which leaves the following possibilities:
1. Raising predicate
2. (Subject) Control predicate
3. Monoclausal predicate (of some kind)
Distinguishing monoclausals generally from the biclausal structures discussed above lies
in identifying whether VP1 contains an internal subject position. Recall that in § a
unified or monoclausal structure was defined as having no embedded subject of the lower
VP, but semantic ‘sharing’ of the subject of the higher verb. (Wurmbrand 2001, Cable
2004) Of course, monoclausal constructions themselves may fall into different categories
such as auxiliary constructions, light verb constructions, serial verb constructions or
restructuring constructions. This range of options will be explained and explored in detail
in Chapter 4.
More immediately, §3.2.1 examines the types of evidence for clausal/argument structure
that were used to elucidate the structure of the other complex verbs, in order to look for
commonalities between V-sime and these other structures, and shows that the evidence is
somewhat mixed in this regard. §3.2.2 examines the properties of the V1s that compound
with -sime, and shows both that -sime is quite promiscuous in the type of V1 it can
accommodate, and that it does not appear to impact the syntactic properties of that V1.
3.2.1 Clausal/Argument Structure
In motion compounds, the nominative subject can bind the reflexive object of an
embedded transitive verb, as in (129) and (130), which at first glance would seem to be a
solid argument for monoclausality.
Uu chuu’u
DET dog
always 3sgREFL scratch-RED-go
“The dog is always going along scratching itself.”
(130) -Hunume
DEM.DISTAL little boy-PL
always 2plREFL
“Those little boys are always going around pushing each other.”
However, since there cannot be a subject of the V1 distinct from the nominative subject
of V2, the binding facts can tell us very little here, because a null argument such as PRO,
controlled by the nominative subject, could occupy the subject of V1 position and license
the binding of reflexives within a lower clause. This would indicate a biclausal control
structure like that discussed in §3.1.4. If no evidence for PRO can be found, then we
might conclude either that V2 siime behaves as a raising predicate, and does not
contribute an argument of its own, as in§3.1.3, or that the structure is monoclausal and
contains no embedded subject.
One direction to look for evidence of an embedded subject position involves the
interaction of V2 siime with the impersonal/passive -wa, as in example (131) below.
(131) Imi’i
DEM.PROX always
“There are always balls being kicked along here.”
When -wa occurs with a V-sime compound, the nominative subject phrase is deleted and
the object of V1 - when V1 is transitive - is promoted to subject. Furthermore, as can be
seen in examples (132)-(133) below, the motion verb must now occur in its plural
form, -saka. Although in context the understood agent in (133)is uu hamut ‘the woman’
and singular, because -wa is an impersonal, the verb to which it is applied is typically
said to have an understood 3rd person plural subject ‘people/they’. The fact that the verb
here is showing plural number agreement suggests that there could be a syntactically
relevant null subject present that can trigger the suppletion of the verb.
(132) Uu hamut
DET woman
little child-ACC
“The woman is pushing the little child along.”
(133) Uu ili uusi
wam vicha
DET little child there toward
“The little child is being pushed along to there.”
In addition, consider the examples in (134)-(135) below. Example (134) shows that the
subject of a V-sime predicate may be inanimate and non-agentive. However we can also
add a Causer of the ball’s motion, as in (135) by using a transitive form of the V1
roakti/a ‘roll’
(134) Ume pelo’ota-m vo’o-t
straight roll.INTR-go
“The ball is rolling along the road”
(135) Rufiino pelo’ota-m vo’o-t
Rufinio ball-PL
road-LOC straight roll.TR-go
“Rufino is rolling the ball down the road “
(accompanied motion)
The V-sime construction necessitates an accompanied motion reading of this sentence.
That is, Rufino must also be in motion, as well as being the Causer of the ball’s motion.
In order to express caused motion without accompaniment, the V-sime construction must
be dropped, as in (136).
(136) Rufiino Simon-ta-u
bwe’u pelotam
Rufino Simon-ACC-to straight big
“Rufino rolled the big ball straight to Simon” (unaccompanied motion)
This requirement is evidence that -sime contributes a thematic subject to the construction.
Taken together, the passive and accompanied motion facts are suggestive of a subject
control structure. There is also, however, evidence against this conclusion. In §3.1, it was
argued that V2 -taite and -ma(a)chi do not contribute arguments, based on their ability to
compound with the subjectless weather verb yuuku ‘rain’. In fact, -sime is also able to
occur with yuuku. Example (138) shows yuk-sime operating, like yuuke, without an overt
subject. In examples (139) an overt subject is inserted, and the sentence is ungrammatical
(just as it would be with yuuke standing alone as the only verb in the sentence).
(137) (Si)
“It is (really) raining.”
(138) Avo vicha yuk-sime
This.way toward rain-go
“The rain is heading this way.”
(139) *Naamu si yuuke
Cloud very
Intended: “That cloud is really raining.”
(140) *Naamu yuk-sime
Cloud rain-go
Intended: “The cloud is raining as it moves.”
There are a couple of possible ways to interpret these facts.
1. Yuuke does not have a thematic subject, and -sime does not provide one, which
would mean that -sime is a raising predicate like -taite.
2. Yuuke does have a thematic subject, but of a very particular type. It must be
phonologically null, but with features such as 3sg, general and inanimate19.
Option 2 brings back the possibility that -sime could be a control verb. However, if it is a
control verb, it clearly differs from -pea in that it can accommodate this kind of peculiar
thematic subject20. In fact, as we have seen, -sime can co-exist with all manner of
inanimate and non-intentional subjects, while -pea cannot.
The other option to be entertained is that V-sime is a monoclausal structure. In this
scenario, there would be a single subject position available, and the subject would be the
thematic argument of the V2, -sime. The embedded VP1 would consist only of the V1
and any complements it might have – the nominative DP would be interpreted as the
semantic subject of both verbs, since no other options are available.
This follows Chomsky’s (1981:323-325) argument for weather ‘it’, that it is not a true expletive, but
more of a ‘quasi-argument’ that has some semantic and referential content. (Chomsky 1981, Svenonius
2002) It also matches the requirements on 3sg inanimate subject pronouns that were reported and
documented in Harley & Trueman (2012) – if yuuke had to have a weather-pronoun for a subject, this is the
properties we would expect it to have.
I think it is not entirely crazy to posit that ‘go’ might be able to select for the same kind of general subject
as a weather verb might. Consider, for example, the use of ‘it’ in expressions such as “How’s it going?”
where ‘it’ refers to some general, unspecified event or state.
As mentioned earlier, the binding facts in (129)-(130) work well with the idea of a
monoclausal structure, although they are far from definitive. The accompanied motion
facts, in (134)-(136), also align with a monoclausal analysis, since they show that the
subject DP is a thematic argument of -sime. The ability of -sime to appear in apparently
subjectless weather clauses (137)-(139) fits only if we assume that these constructions are
not truly subjectless, but that they have a particular type of null subject, with which both
yuuke and -sime are compatible.
Another possible challenge to a monoclausal analysis of V-sime presented here are the
facts in (131)-(133), which show that in a passive/impersonal construction -sime always
appears in its plural form, even when the deleted agent is interpreted contextually as
singular. I suggested that the plural agreement is suggestive of a syntactically relevant
null argument. However, the promotion of the embedded object to matrix subject should
not be possible if in fact an intervening embedded subject is available. This argument is
discussed in some detail in Chapter 4 and Chapter 5, so I will not elaborate on it here. It
does leave the question of obligatory plural agreement in the passive/impersonal open,
In sum, V-sime cannot, based on the mixed evidence in this section, be shown clearly to
be either a raising or a subject control biclausal construction, but appears to display
properties associated with both. Neither can the case for a monoclausal construction be
solidly made on this evidence alone, although I claim that this analysis is the best fit
overall. In Chapter 4 I examine analyses of monoclausal motion compounds in a range of
languages and compare how their properties line up with those of V-sime, in order to
make the argument that V-sime is a best understood as a monoclausal construction.
3.2.2 Effects of V1
Understanding the contributions of -sime to a compound requires some understanding of
what else is contributing arguments and predicational information. In this section I will
discuss the range and type of V1s that occur in -sime compounds, and show that the
appearance of -sime is not restricted to any particular V1, nor does -sime have any
obvious impact on the syntactic properpties of the V1, such as argument structure and
case assignment. Any restrictions on the V1 in a -sime construction appear to be based on
semantic/aspectual properties only.
Hiaki verbs are described in the existing literature as coming in three types: transitive,
intransitive, or ‘variable’ (meaning that some verbs can be either transitive or
intransitive). Many, though not all, of the verbs that are described as variable show
change in form depending on transitivity:
bwasa ‘be cooking it’
(142) chepta ‘jump/step over’
bwase ‘be cooking’
chepte ‘jump/step’
Both intransitive and transitive forms of verbs are compatible with –sime (assuming the
appropriate semantic context) with no change in argument structure, as shown in (143)(144) below.
(143) Ume pelo’ota-m vo’o-t
road-LOC straight roll.INTR-go
“The ball is rolling along the road”
(144) Rufiino pelo’ota-m vo’o-t
Rufino ball-PL
road-LOC straight roll.TR-go
“Rufino is rolling the ball down the road “
When the V1 is transitive, its object argument receives accusative case marking, as it
would in a simple-verb context.
(145) Uu hamut
DET woman
little child-ACC
“The woman is pushing the little child along.”
V1 can also be ditransitive, in which case both its direct and indirect objects are
accusative; this is consistent with their appearance in single-verb clauses.
(146) Aapo avai
“S/he is going along giving roasted ears of corn to them.”
‘Ball’ is one of a number of nouns in Hiaki that are obligatorily marked plural (‘pluralia tantum’), even
when a singular interpretation is intended – these are frequently, but not exclusively, loanwords.
Overt accusative case marking is usurped by the presence of the plural in determiners, adjectives and
The previous section showed that V1 may be transitive, and therefore have an agentive
external argument role to contribute. Unergative V1s are also possible, as shown in (147).
(147) Paola kia haisa
Paola just really
“Paola’s going around talking about you.”
The obvious next question, then, is whether unaccusatives may also take the V1 position,
and this question is a little more difficult to answer in Hiaki, because of controversies in
establishing a clear test for the unaccusative/unergative divide, as briefly discussed in
Chapter 2 above.
Jelinek & Escalante (2000) argue that the passive/impersonal suffix -wa targets
Agent/Causer subjects, and thus can be used to diagnose unergative verbs. However, this
position is not without controversy since -wa can be applied to the verb muuke/koko ‘to
die’, which seems to be interpreted with an Experiencer subject.
(148) Sawaria-ta-mak
Yellow(fever)-ACC-with die.PL-PASS-PST
“People were dying from yellow fever.”
In contrast, Harley, Tubino Blanco and Haugen (2009) argue that muuke/koko is
inherently unaccusative, which therefore disqualifies -wa as a tool for diagnosing
unergatives. They claim that the applicative morpheme -ria may instead be used to
identify unaccusative intransitives, arguing that these verbs cannot co-occur with the
(149) *Uu tasa Maria-ta hamte-ria-k
DET cup Maria-ACC break-APPL-PFV
“The cup broke for/on Maria.”
Harley, Tubino Blanco and Haugen (2009:44)
Neither of these claims is without its controversial examples. However, the examples in
(150)-(151) show that -sime can take a stative verb derived from an adjective, such as
alee ‘be happy’ and sioke ‘be sad’, as its V1, and (152) has the unaccusative verb hamte
‘break(intr)’ as its V1, all of which is good evidence that lacking an external agent role23
is not a barrier to the V1 position.
(150) Uu yoeme yu’in tomi yo’o-kai si
alee-sime-ka weye.
DET man a.lot money win-PCL very happy-go-PCL walk
“The man, having won a lot of money, is walking along very happily.”
I assume that predicates of this type have an Experiencer subject that is generated in an internal, rather
than an external argument position.
(151) Uu yoeme si’ime tomi-ta aman koove-k kialikun si-siok-sime-ka weye
DET man all money there lose-PFV RED-sad-go-PCL walk
“The man lost all the money there; that is why he is walking along very sadly.”
(152) Ume tasa-m mura karopo hamti-saka
DET.PL cup-PL mule kart-LOC break-go.PL
“The cups were breaking as they went along in the mule cart.”
Manner of motion verbs
Manner of motion verbs are the most likely to occur in a V-sime structure, and the
most flexible with regard to the appearance of reduplication. The manner verb vuite
‘run’ is the clearest example of this: vuiti- may occur with -sime with neither verb
reduplicated (153), with both reduplicated (154), with only vuiti- reduplicated24 (155)
or with only -sime reduplicated (156).
(153) Chepa aman vuiti-sime
Chepa there run-go
“Chepa is running along over there”
Note that V1-only reduplication is the least common and most context dependent option in V-sime
constructions. In (155) for instance, there is an implication that Chepa is crazy, or frenzied.
(154) Chepa vui-vuiti-si-sime
Chepa RED-run-RED-go
“Chepa is going running.”
(habitually, seriously, for fitness)
(155) ?Chepa
Chepa house-LOC out
“Chepa is repeatedly running out of the house.”
(156) Chepa (toto’i aso’ola-met cha’a-ka) vuiti-si-sime
Chepa chicken baby-after chase-PCL run-RED-go
“Chepa is running about (chasing after baby chicks).”
The semantic effects of the various reduplicated forms will be discussed in more detail in
§ below.
Stance/posture verbsV2
One of the unexpected and interesting classes of V1s that participate in this construction
are stance/posture verbs. The unreduplicated combination of a verb such as kikte/hapte
‘stand’ or yehte/hoote ‘sit’ retain the usual semantics of V-sime ‘go along V-ing’ with the
assumption of some kind of vehicle being involved, as in (157)-(158) below.
(157) Ume o’ow-im kamion-po hap-saka
man-PL bus-LOC stand.PL-go.PL
“The men are standing in the (moving) bus.”(Lit: going along standing)
(158) Heidi intok Aleh kavai-im-met hoo-saka
Heidi CONJ Alex horse-PL-on sit.PL-go.PL
“Heidi and Alex are riding on horses” (Lit: going along sitting)
Reduplication options are more restricted with stance V1s than with manner-of-motion
V1, as seen in the previous section. The only attested reduplication patterns for stanceVsime have -sime alone able to be reduplicated, as in (159)-(160).25
(159) Ume o’ow-im kamion-po hap-sa-saka
man-PL bus-LOC stand.PL-RED-go.PL
“The men are standing/milling about on the bus.”
(The bus is no longer necessarily moving)
(160) Hunum haamuch-im kia veha hunum hoo-sa-saka
DEM.PL woman-PL
just already there sit.PL-RED-go.PL
“Those women are just sitting around there.”
Interestingly, in these examples, the ‘motion’ semantics of the construction is no longer apparent. The
implications of this are discussed in detail in §
Any reduplication on the stance verb itself, regardless of whether -sime reduplicates, is
impossible (161)-(162) although, stance verbs are, on their own, usually quite amenable
to reduplication (163).
(161) *Ho-hoo-saka
(162) *Ho-hoo-sa-saka
(163) Santos intok inepo,
human ho-hooye
Santos CONJ 1SG.NOM, evening=1PL.NOM there
“Santos and I, we sit there in the evening (habitually)”
3.3 Contrasting V2 -sime with independent siime/saka
This section compares the properties of affixal -sime with those of independent siime that
we saw in the last chapter, in order to establish whether they should be considered
distinct lexical items with distinct functions and behaviors. In particular, I question
whether -sime retains the full lexical semantics of siime, or if it acts more like a
functional, grammatical element.
Before delving into an examination of how -sime interacts with particular classes of
morphology, the first step must be to look for examples of V-sime constructions in which
the lexical meaning of -sime as literal motion is not apparent, or ‘bleached’. To that end, I
have collected the sentences in (164)-(166).
(164) Itom
nau eteho-u, haisa=ne
tuisi (aa=)hiohte-sim-ne?
1PL.NOM together talk-while, Q=1SG good (3SG.ACC) write-go-FUT
“While we are talking, is it ok if I take notes?”
(165) Wari-ta-naat
“(He or she) is weaving a basket.”
(166) Inepo
“I’m beginning to understand a little bit.”
(Dedrick & Casad 1999:294(7))
This rather small sample consists of the only sentences I have been able to find in which
the expected ‘motion’ semantic contribution of -sime is obscured or not present.
Furthermore, whatever semantic contributions -sime is making in these cases are not
obviously consistent, so it would be difficult to make a case that affixal -sime is
grammaticalizing into a functional morpheme based on this evidence. These examples
appear to be simply idiomatic. It is not possible, for instance, to extend the interpretation
in (164) to a different V-sime combination; the example in (167) can only be interpretable
if the proposed singing is occurring along a path of literal motion.
(167) #Enchi hippona-u haisa=ne tuisi e-u bwik-sim-ne
2SG.NOM play-while Q=1SG good 2SG-to sing-go-FUT
Intended: “While you play, can I sing along with you?”
Achieved: “While you play, can I go along singing with you?”
Therefore, given these limitations, I turn to the interactions of V-sime with other
morphological processes.
3.3.1 Interactions with Morphological Processes
Affixal -sime interacts somewhat differently with certain tense/aspect categories than
does independent siime.
Although -sime is compatible with future -ne, speakers judge that the sentence in
(168) feels incomplete, and requires some further context, which is unnecessary when
siime is the only verb.
(168) Chepa kari-po yeu bwan-sim-ne…
Chepa house-LOC out cry-go-FUT
“Chepa will leave the house crying…”
Present Imperfective
Present imperfective interacts with -sime just as it does with siime. The sentence in
(169) is grammatical, complete, and has the expected interpretation.
(169) Chepa kari-po yeu bwan-sime.
Chepa house-LOC out cry-go
“Chepa is leaving the house crying.”
(Past) Perfective
Affixal -sime is entirely incompatible with perfective aspect. This is not entirely
shocking, since the usual interpretation of V-sime, ‘go along V-ing’ has a distinctly
progressive flavor, which is quite at odds with the perfective. However, this does
represent a clear difference between independent siime and affixal -sime.
(170) *Chepa karipo yeu bwan-sika
Chepa house-LOC out cry-go.PFV
Intended: “Chepa left the house crying.”
Past Imperfective
Past imperfective is perfectly compatible with -sime, just as we saw with the present
perfective, however in this case, as with the future, the sentence is incomplete and
requires more information to be added. In this case, there is a strong feeling that
something else must have happened while the events of (171) were ongoing.
Additionally, recall that in Chapter 2 we saw that the form siime-n had an unexpected
reading wherein the ‘going’ was prevented before it could begin. That stipulation
does not seem to feature here.
(171) Chepa karipo yeu bwan-sime-n…
Chepa house-LOC out cry-go-PST
“Chepa was leaving the house crying…”
Past Perfect26
The form -sikan – whose independent counterpart siikan was problematic for the
independent verb in that it obtained an ongoing or past imperfective reading, rather than
the expected past perfect – is completely impermissible in the compound construction.
Sentences with this form are judged to be extremely bad, regardless of the intended
aspectual interpretation.
(172) **Chepa karipo yeu bwan-sika-n
Chepa house-LOC out cry-go.PFV-PST
Intended: “Chepa had left the house crying…”
OR: “Chepa was leaving the house crying…”
As mentioned previously, this terminology is almost certainly incorrect, and is being used only for lack
of a more appropriate term.
Verb stacking
V-sime compounds can be further modified by adding subsequent verbs, both affixal, as
in (173)-(175), and independent (176). The properties of the final verb then control
interactions with tense/aspect, as evidenced by the appearance of perfective -k in example
(173) Rufiino pelo’ot-am vo’ot luula
Rufino ball-PL
road straight
“Rufino started to roll the ball down the road”
(174) Rufiino pelo’ot-am vo’ot luula
Rufino ball-PL
road straight
“Rufino is going to roll the ball down the road”
(175) Rufiino pelo’ot-am vo’ot luula
Rufino ball-PL
road straight
“Rufino feels like rolling the ball down the road”
(176) Rufino uka usi-ta
pelo’ot-am vo’ot luula
Rufino DET child-ACC ball-PL
road straight roll-go-teach
“Rufino is teaching the child to roll the ball down the road.”
The Impersonal Passive (-wa)
Hiaki impersonal/passive affix –wa, previously discussed in § and §3.2.1, has the
effect of removing the subject argument from a clause; if the clause is transitive, the
object will be promoted to subject. The demoted subject must be animate - -wa cannot
occur with an inanimate subject.
When -wa is applied to a V-sime compound, the process looks very similar. I showed in
§3.2.1 that the nominative subject phrase is deleted, and any accusative-marked object of
V1 will be promoted to subject. And, because -wa is an impersonal, the verb to which it
is applied has an understood 3rd person plural subject ‘people/they’. This process always
requires plural suppletion from -sime to –saka, possibly suggesting the presence of a
syntactically relevant null subject, of some kind, which can trigger the suppletion of the
(177) Uu hamut
DET woman
little child-ACC
“The woman is pushing the little child along.”
(178) Uu ili uusi
wam vicha
DET little child there toward
“The little child is being pushed along to there.”
Although, see discussion in §3.2.1 regarding difficulty for this idea, since the null subject doesn’t behave
as if it’s syntactically present in terms of occupying the subject position.
Verbal Reduplication
As discussed in Chapter 2, Hiaki has syllabic verbal reduplication, which manifests as a
verbal prefix, and may indicate habitual aspect, progressive aspect, or in some instances
emphasis, as exemplified in (179)-(181) below.
(179) Habitual
1plNOM there
‘We gather firewood there.’
(180) Progressive/continuative
totoi kava-m bwa-bwata
chicken egg-PL RED-stir
‘The woman is mixing the eggs.’
(181) Emphatic (often in Imperative examples)
hunum ma-mana
‘Don’t put that pot there.’
(Harley & Leyva 2009: 253)
When reduplication is applied to a complex verb, it is most commonly the outer element,
V2, which reduplicates (182) but in fact it can target either the first element in a
compound (183) or both (184).
(182) Paola kia haisa
Paola just really
“Paola’s going around talking about you.”
(183) Vempo si
3plNOM very
“They seem like really hateful people.”
(184) Vempo si
3plNOM very
“They really seem like really hateful people.”
Most of the bound V2 affixes reduplicate, and when they do, the reduplication scopes
over the whole complex verb. In these cases, if only the V1 reduplicates, then the
reduplicant does not scope over the V2. However certain of the bound suffixes do not
undergo reduplication (notably causative –tua and applicative –ria) - in these cases only
the V1 stem may reduplicate, and the interpretation is usually holistic. (Haugen and
Harley 2010)
In less common occasions when a complex verb has more than two parts, reduplication
can also target the central element.
(185) Yooko
Alehandra into Heidi kava’a-im-met
Tomorrow Alexandra and Heidi horse-PL-with
“Tomorrow Alex and Heidi are going to go around on horses.”
Although either or both verbs of a compound may in theory be reduplicated, not every
compound is compatible with all of the reduplication options. The vuiti-sime compound
is a significant example for this reason, because it can help to illustrate the effects of each
variable. Examples (153)-(156) from § are repeated here as (186)-(189) for this
(186) Chepa aman vuiti-sime
Chepa there run-go
“Chepa is running along over there”
(187) Chepa vui-vuiti-si-sime
Chepa RED-run-RED-go
“Chepa is going running.”
(purposefully, for fitness)
(188) ?Chepa
Chepa house-LOC out
“Chepa is repeatedly running out of the house.”
(189) Chepa (toto’i aso’ola-met cha’a-ka) vuiti-si-sime
Chepa chicken baby-after chase-PCL run-RED-go
“Chepa is running about (chasing after baby chicks).”
The examples above show that the different reduplication patterns available in V-sime
compounds create distinct patterns of interpretation. By changing the V1 from a mannerof-motion verb to a stance verb, we can shed further light on the function of -sime in
these constructions. Even though stance verbs don’t show the full range of reduplicative
patterns, there is an important difference in interpretation between an unreduplicated
Vstance-sime construction (190) and one in which -sime is reduplicated (191).
(190) Ume o’ow-im kamion-po hap-saka
man-PL bus-LOC stand-go.PL
“The men are standing in the (moving) bus.”
(191) Ume o’ow-im kamion-po hap-sa-saka
man-PL bus-LOC stand-RED-go.PL
“The men are standing/milling about on the bus.”
(bus no longer nec. moving)
In (191) we can see that the reduplication of -sime has the effect of negating the literal
interpretation of movement along a path, which is the defining semantics of almost every
V-sime construction. To emphasize this point, I include example (192) in which there is
no possibility of the men being carried along by a moving vehicle.
(192) Ume o’owim kari-po hap-sa-saka
man-PL house-LOC stand-RED-go.PL
“The men are milling about in the house.”
So what is happening here? There are a few significant points to notice. First, even
though all the verbs in these examples (stance and manner-of-motion verbs, as well as
siime) when used independently become habitual with reduplication, the reduplicated
compounds are not habitual, with the possible exception of (187). When reduplication
applies to the compounds, things get a little more complex. In these data, we can identify
the following patterns:
1. Unreduplicated compound: Action of V1 is occurring along a path of motion
(V1 along).
2. Compound with V2 reduplicated: Action of V1 is occurring either while
moving along multiple paths, or with several changes of direction. In the case
of stance verbs there is a sense of shifting positions multiple times (V1 about).
3. Compound with V1 and V2 reduplicated: multiple instances of V1 occurring
on multiple paths.
4. Compound with V1 reduplication: With a specified path, it may be
interpreted as multiple instances of V1 occurring along a single path of
motion. This configuration is the least common, and tends to be pragmatically
odd ((188) implies that Chepa is a crazy person).
From this I conclude that Hiaki verbal reduplication does not scope over both verbs in the
V-sime compound, and that the interpretation of the reduplicant is context dependent, but
can be broadly classed as ‘pluractional’. (Harley and Leyva 2009) More significantly for
the current purpose, I conclude that V2 –sime does not encode literal motion, but
something more like grammatical Path – in the sense of Talmy’s (1975; 1985; 1991;
2000) decomposition of motion events, described in the previous chapters – a subtle, but
important distinction.
3.4 Summary
In §3.1, I gave an overview of the basic properties Hiaki complex verbal structures,
which have previously been described in terms of two classes distinguished by whether
the V2 was an obligatorily bound element or was form-identical with an independent
verb. By comparing the clause and argument structure of a selection of both affixal and
independent V2s, we saw that this division is not significant at a structural level. Both
classes show evidence of being biclausal with respect to binding, though only a single
Tense node is available, which indicates that the embedded clause is VP or vP sized. (The
relevant boundary is VoiceP, according to Tubino Blanco & Harley (2012), although I do
not investigate this claim in detail here.) V2s in both groups can be classed as either
ECM, Raising or subject control predicates.
§3.2 examined complex verbs of motion formed with -sime/-saka, and compared them to
the complex verb types outlined in §3.1. Affixal -sime/-saka, however, resists
straightforward identification with any of the 3 previously identified V2 types. While
ECM can be easily ruled out, properties consistent with both and neither of the raising
and control predicates can be identified. It is possible, but not certain, that -sime/-saka is
of another type entirely, such as a restructuring predicate perhaps. There appears to be no
particular restriction upon the type of V1 that may compound with -sime/-saka, and it has
no obvious effect on the syntactic properties of its V1.
§3.3 contrasts the behavior of -sime/-saka in its role as a participant in a complex
structure with its behavior in a simple predicate clause, assessing interactions with
tense/aspect and other morphological operations. Affixal -sime/-saka has some
restrictions on aspectual categories with which it is compatible that its independent
counterpart does not; the affixal form is not compatible with perfective aspect, while
independent siime/saka is. However, the most significant issue in this section is
highlighted by the interaction of specific elements. These are the combination of affixal sime/-saka with a stance V1 and verbal reduplication, which show a consistent and
replicable non-literal interpretation of -sime/-saka. This effect suggests that V2 -sime has
functional, as well as lexical, properties and is perhaps indicative of incipient
grammaticalization. A crucial question in determining the structure of a construction like
V-sime is how to situate the V2 within the structure of the clause. Since functional items
occupy positions higher in the structure than lexical items, determining whether -sime
exhibits primarily lexical or functional properties is critical.
In Chapter 4, we will look at crosslinguistic examples of complex motion constructions
of other syntactic types –auxiliary verb constructions, light verb constructions, serial verb
constructions, and restructuring predicates – to ascertain if the V-sime/-saka
constructions align more closely with any of these. In some of these constructions the V2
is clearly a functional item (auxiliaries) and in some a lexical item (serial verbs).
Chapter 4 Complex Motion Crosslinguistically
Itepo ame venachi nonooka. (We talk like them.)
In the previous chapter I attempted to compare the characteristics of V-sime constructions
to siime’s behavior as an independent verb (Chapter 2) and to other V-V constructions in
Hiaki (Chapter 3). I showed that although affixal -sime retains the bulk of the semantic
weight of independent siime (literal motion), it does take on a more bleached or
functional character in a predictable and productive way in a specific context (i.e. when
reduplicated with a stance V1). I also showed that although some more well-understood
compound verb structures in Hiaki, such as causatives, have distinct binding domains and
a structural subject position available in the lower VP, the evidence regarding the
structural properties of V-sime is considerably less clear-cut.
Sentences with an intransitive V2 have a single overt subject DP and an interpretive
dependency between this DP and the unpronounced subject of V1. Since there can be no
overt lower subject in these constructions, we must look for other evidence to determine
whether both Vs are contributing distinct thematic subjects. We have seen, for instance,
that some V2 elements such as -pea ‘inclination’, impose strict selectional restrictions,
whilst others, such at -taite ‘inception’ have no such restrictions, so we can argue
that -pea does contribute a thematic subject which obligatorily binds the unpronounced
subject of the V1, whilst -taite does not contribute a thematic subject, and the subject of
V1 is raised to the nominative case position in the matrix clause. (Polinsky 2013)
Like -taite, V2 -sime does not impose any selectional restrictions on its subject – it even
occurs with weather predicates in clauses which entirely lack an overt subject – however
other evidence exists which suggest that –sime does contribute a thematic subject. When
V2 -sime is affixed by the passive morpheme -wa, which targets animate subjects, it
obligatorily suppletes for plural subject agreement, regardless of the number features of
the nominal argument. Since -sime fails to agree with the (singular) derived subject, this
suggests that it is not in a structural relationship with that subject. This raises the
possibility of a covert argument, with either plural or undetermined number features.
(Harley 2014)
(193) Uu hamut
DET woman
little child-ACC
“The woman is pushing the little child along.”
(194) Uu ili uusi
wam vicha
DET little child there toward
“The little child is being pushed along to there.”
Further, constructions such as (195)-(196) show that the presence of -sime requires an
accompanied motion interpretation, which also suggests that it does have a thematic
subject to contribute. If the nominative DP below was base generated as the external
argument of the V1 in (195) we would not expect it to also behave as the internal
argument of V2 (Haugen & Harley (2013) and Jung (2014) show independently that
siime is unaccusative.). That is, -sime introduces motion entailments for the external
argument of V1, which it should not be able to contribute if it was simply a propositional
operator like a normal raising verb. Example (196) shows that when the subject DP is not
in motion, the V-sime construction is not used.
(195) Rufiino pelo’ota-m vo’o-t
Rufinio ball-pl
road-LOC straight
“Rufino is rolling the ball down the road “
(196) Rufiino Simon-ta-u
Rufino Simon-ACC-to
(accompanied motion)
bwe’u pelotam
straight big
“Rufino rolled the big ball straight to Simon” (unaccompanied motion)
In this chapter I investigate the typology of V-V structures crosslinguistically, and
describe the analyses that have been put forward to account for them. In §4.1 I introduce
and define the notion of complex predication broadly, and identify the reasons for
including Hiaki compounds generally, and V-sime specifically, amongst this set. In §4.2 I
attempt to define and exemplify four recognized types of multi-verb complex predicates,
describe the kind of structural analyses that have been put forward for each, and show
which attributes of V-sime align with a given construction type.
Finally, in §4.3 I situate Hiaki motion V-Vs within this typology of complex predicates,
arguing that they best fit the description of verb serialization (SVCs) and summarize the
problems associated with identifying a structural model for V-sime constructions.
4.1 Complex Predicates
A wide array of phenomena, in various languages, has been lumped under the umbrella
term ‘complex predicate’ (henceforth CPr). These include, but are not limited to: light
verb constructions; coverb constructions; serial verb constructions; raising and
restructuring predicates; incorporation phenomena, including noun incorporation,
preposition incorporation, pseudo-incorporation and particle constructions; some types of
verbal classifier systems; resultatives; and even, contentiously, control constructions and
auxiliary verb constructions.
Further adding to the confusion, there is a great deal of mismatch in the interpretation of
many of these terms. ‘Light verb’ is applied particularly freely, as for instance in Rosen
(1990) who argues that restructuring predicates are a type of light verb, thus collapsing
two of the categories above. Similarly, although Amberber et al. (2010) appear to
consider coverb constructions a distinct phenomenon, Bowern (2008) places them
decisively in the category of light verb constructions.
Here I am concerned with multi-verb constructions only. In particular, I focus on: serial
verb constructions (SVCs); restructuring predicates (RPs); light verb constructions
(LVCs); and auxiliary verb constructions (AVCs). Each of these construction types
consists of two verbal elements, one of which is fully lexical; the other verbal element
varies between constructions, from a fully lexical item at one end of the spectrum (SVCs)
to an entirely, or almost entirely, grammatical item at the other (AVCs).
4.1.1 Definitions of CPr
Definitions and diagnostics for complex predication vary widely between scholars,
dependent largely upon the characteristics of the particular language/s with which they
are most familiar, as well as upon the theoretical model they advocate. Several attempts
have been made to survey the literature and tease apart common characteristics of the
range of structures which have been labeled ‘complex predicates’; these attempts have
lead to consensus on little more than that the world’s languages exhibit a dizzying array
of variable complex structures, and almost as extensive a range of descriptive
terminology. (Bowern 2008; Butt 2003, 2010; Seiss 2009)
Alsina, Bresnan and Sells, introducing their (1997) collection of papers describing a wide
range of complex predicate constructions in a variety of frameworks, provide one
frequently cited definition: “Complex predicates can be defined as predicates which are
multi-headed; they are composed of more than one grammatical element (either
morphemes or words), each of which contributes part of the information ordinarily
associated with a head” (Alsina et al 1997:1). Butt’s (2010) definition focuses less on the
complex nature of the predicate head, and more on the clausal and argument structure
implications: “…the term complex predicate designates a construction that involves two
or more predicational elements (e.g., nouns, verbs and adjectives) which predicate as a
single unit, i.e., their arguments map onto a monoclausal syntactic structure” (Butt
A less precise, but perhaps more accurate, summation of the current thinking around
complex predication is put forth by Sells:
“The usual understanding of the term ‘complex predicate’ is a semantic one: a
complex predicate consists in the argument structures of two separate predicates
being brought together somehow or other, and further, typically, the argument
structure of one of those predicates in isolation is taken to be incomplete, ‘light’,
or ‘bleached’. Within this basically semantic idea of a complex predicate, we can
find different structural manifestations” (Sells 1998:1).
Notably, Sells’ description avoids any mention of monoclausality, although it is a central
concept in many of the definitions surveyed here, in particular Butt’s. (2010:49)
At the other end of the conceptual spectrum, Hale & Keyser (1997) argue that simplex
verbs are themselves internally complex, so complex predication is in that sense the
normal state of affairs rather than a special circumstance and should be constrained by
normal processes of syntactic combination.
With all that vagueness and variation in mind, the properties of CPrs that are most
commonly agreed upon are: two predicators must be included under a single tense/aspect
node; they must share polarity (no negating of one element without the other); and they
should have a single intonation contour (ie, no intervening pause, as at a clause
boundary), all of which are criteria that Hiaki V-V compounds uncontestably meet.
However, this is a less strict definition of monoclausality than some (most notably Butt)
argue for. (Bowern 2008; Butt 2003, 2010; Seiss 2009)
Butt (2003 et seq.) argues that to qualify as a complex predicate, there must be no
evidence of distinct syntactic domains or unmerged argument structures. This criterion
would disallow control and raising constructions, which permit non-matching polarity,
and also constructions such as the Hiaki direct causative, which has distinct binding
domains. However, note that while Butt and others insist that argument sharing is a
requirement of CPrs, others include structures with no argument sharing – for example,
Aikhenvald describes a construction she labels ‘Event argument Serial Verb
Constructions’, in which “The event or state denoted by one component is predicated on
the entire situation referred to by an SVC. Event-argument SVCs provide the manner,
temporal order or locational specification for the other component” (2006:18). This
description also corresponds well with the semantics of V-sime, at least informally.
Finally, one of the stickiest areas of controversy in the complex predicate literature is that
of event structure. While some (Durie 1997; Aikhenvald 2006) claim that a CPr must
always have a single event structure, others (Baker & Harvey 2010) argue for at least a
limited range of combined or complex event structures. The disagreement largely seems
to be definitional. Durie and Aikhenvald use 'event' to describe a conceptual category that
is culturally specific; that is, an event may be complex, as long as it is conceived as a
singular act by the speakers of the language. Thus, the reasoning goes, in the following
examples from Alamblak, climbing a tree in search of insects is seen as a normal
complex event, but climbing a tree to look at the stars is not an inherently connected
sequence, because one can see the stars perfectly well from the ground28.
(197) mɨyt ritm muh-hambray-an-m
tree insects climb-search.for-1sg-3pl
"I climbed the tree looking for insects."
(198) *mɨyt guñm muh-hëti-an-m
tree stars climb-see-1sg-3pl
"I climbed the tree and saw the stars."
Alamblak (Durie 1997:329)
Baker and Harvey (2010) on the other hand, conceive of 'event' in terms of Jackendoff's
theory of Lexical Conceptual Structures (LCS) and use that to restrict what may be
conceived of as a single 'event'. From this they distinguish two distinct classes of
complex predicates. The first they call ‘merger’, assuming that two predicates with
similar LCSs may merge, resulting in a single, albeit complex event with shared
arguments. They associate this class with light verb constructions of the kind commonly
found in Australian languages. The second class they call coindexation structures, in
which the LCSs of the two predicates may be of different types and are related by
argument coindexation. The result is multiple events, albeit within a single clause; this
class is associated with serial verb constructions.
By this reasoning, the verb form in (198) should be logically possible, given the appropriate context in
which something could be seen only by means of climbing a tree.
Foley (2010), in the same volume as Baker and Harvey, shows that there is significant
crosslinguistic variation in how events are encoded; the same event that is expressed with
a monomorphemic root in one language may be necessarily expressed in structures of a
range of morphological and syntactic complexity in others. He asserts that complex
predicates express a diverse array of event structures, some simple and some extremely
In sum, while definitions of complex predicates abound, there is very little in the way of
consensus to be found, either for defining the class as a whole or for distinguishing
relevent sub-classes. We proceed, therefore, with the understanding that there is likely to
be no category in which to situate the Hiaki V-sime construction definitively.
4.1.2 Grammaticalization/Diachrony
Although this dissertation is concerned with developing a synchronic analysis of the
Hiaki language, and I do not have historical data readily available, diachronic forces
cannot be entirely ignored, particularly when dealing with constructions of this kind.
Grammaticalization processes and shifting categories add to the complicated landscape of
complex predicate types. For SVCs in particular “there is a very strong diachronic
tendency to lexicalization and grammaticalization of the meaning of serial complexes:
this can involve treating the whole serial complex as a single lexical(ized) item, or
‘demotion’ of the meaning and grammatical status of one of the verbs to that of a
modifier or case marker” (Durie 1997:291).
Bowern (2008) surveys the theories and data regarding the diachronic development of
complex predicates. She cites, among others, Givón’s (2008) symposium presentation, in
which he puts forward a grammaticalization cline, represented in (199) below.
(199) (a) parataxis > (b) hypotaxis > (c) serialisation > (d) light verb > (e) auxiliary > (f)
univerbated affix
This type of model assumes a clear, unidirectional path of grammaticalization from
distinct conjoined clauses, to hierarchically related clauses, through various types of
clause union, and finally, towards a single verb with bound grammatical affixes.
However, Butt (2003, 2010) disputes this model; in particular, she claims that light verbs
and auxiliary verbs do not belong on a single cline as above, but that they represent
distinct paths of grammaticalization, since the same verb form may occur as both an
auxiliary and a light verb in different constructions in the same language. She further
asserts that light verb constructions are historically highly stable and not subject to the
same forces of grammaticalization as other, more volatile structures. Bowern (2008),
while agreeing that Givón’s cline is too simplistic, provides evidence disputing the
purported stability of LVCs. Her conclusion is that the diachrony of complex predicates
is as tangled and diverse as their synchronic manifestations: “They appear to have no
single common historical source and few if any common paths of diachronic
development” (Bowern 2008:168).
Lord (1993) looks in depth at the changes that serial verbs may undergo, describing a
variety of patterns. For instance, transitive serial verbs, particularly those used to license
an additional argument, may eventually become adpositions, while intransitive serial
verbs are more likely to develop into adverbs or auxiliaries. This, too, disputes the
simplicity of Givón’s model, suggesting at least that a more nuanced model, allowing for
multiple directions of development, is required.
Hiaki has a pertinent example of how the same verb may take different paths of
grammaticalization within a language. There is a ‘purposive’ verbal affix, -se/-vo, which
is a particularly interesting case, because the singular form, -se appears to be historically
derived from the singular motion verb siime29. (The plural form, -vo, may be derived
from vo’o, ‘road’.) It is also the only verbal affix which suppletes for number.
(Dedrick&Casad 1990:295-6)
I have glossed -se/-vo as purposive, but more precisely it indicates motion towards some
purpose and is translated as “go (in order) to V”.
(200) Han=te, aman=te hi’ibwa-vo-k
let=1PL, there=1PL eat-PURP.PL-K
“Let’s go, we’re going over there to eat.”
Curiously, -se/-vo also displays an unexpected interaction with the suffix –k, which
hearkens back to a puzzle with the tense/aspect interactions of siime/saka seen in
Dedrick and Casad (1990) make this claim, presumably on the basis that –se is a plausible phonological
reduction of siime, and shares its suppletive character, as well as a ‘go’ meaning.
§ When -se occurs without a following suffix -k, it expresses intention. When -k
appears, the interpretation is that the movement towards a purpose is underway. This is
demonstrated in (201) below.
(201) Aman=ne aa=vit-se
There=1SG 3SG=see-PURP.SG
“I will be going there to see him/her.” (I intend to go, but haven’t yet left)
(202) Aman=ne aa=vit-se-k
There=1SG 3SG=see-PURP.SG-k
“I am on my way there to see him/her.”
Even more curiously, when the plural subject form -vo is used, it is considered more
natural to include -k in either of these contexts.
(203) Aman=tea a=vit-vo-k
There=1PL 3SG=see-PURP.PL-k
“We are going there to see him.”
(Either ‘We will be going OR ‘we are on our way’.)
In the imperative, however, neither the singular nor plural form may be inflected with -k.
Hiaki has more than one suffix with the form –k. The most common is the perfective aspect morpheme
discussed in Chapter 2, but there is also an adjectival –k and a possessive predicational –k. See Jelinek and
Escalante (1988) for discussion.
(204) Aleh, Heidi-ta aman vit-se(*-k)!
Alex, Heidi-ACC there see-PURP.SG-k
“Alex, go there to see Heidi!”
(205) Aleh intok Acheka, Heidi-ta aman vit-vo(*-k)!
Alex CONJ HK, Heidi-ACC there see-PURP.PL-k
“Alex and HK, go there to see Heidi!”
The presence of the prospective motion affix -se/-vo, in contrast with the associated
motion/path reading of V2 -sime/-saka, demonstrates that grammaticalization of a lexical
verb may occur along more than one semantic/functional path, even within the same
language. Although -sime/-saka retains a great many of its lexical properties, -se/-vo
demonstrates a form which is much more reduced and restricted, and exhibits many more
functional attributes, while still retaining some element of its motion semantics.
4.2 Construction types and analyses
Although I have described the broad categories of complex predicate above, my
discussion from this point on is restricted to multi-verb constructions; here, I discuss four
commonly described types. The first of these, the auxiliary verb construction (AVC), is
generally understood not to be an example of a complex predicate, however it does
involve more than one verb word, and since motion and posture verbs are a common
source of auxiliaries, it seems an important category to include (Seiss 2009). The other
constructions that I discuss here, which are more commonly understood as examples of
complex predication, are light verb constructions or LVCs, in §4.2.2, restructuring verb
constructions (RVCs) in §4.2.3, and finally serial verb constructions (SVCs) in §4.2.4.
As noted in §4.1.1, the label 'complex predicate' is applied to a wide range of
constructions with many different properties, and if it is difficult to define the class as a
whole, then clearly distinguishing sub-classes is a whole new can of worms. We will see
in what follows that there is some considerable overlap between the construction types
described here, and that V-sime has some property or properties in common with each of
4.2.1 Auxiliary Verbs
Auxiliary verbs, many of which have a lexical counterpart, have a functional role in a
clause. They are used to express grammatical categories, and must necessarily
accompany a lexical verb. However, Anderson (2006) has a somewhat broader view than
most. His definition allows for lexical contributions from auxiliaries – effectively, it is
expansive enough that for him ‘auxiliary’ covers territory normally attributed to light
verbs; these two categories are effectively collapsed in his treatment.
“‘Auxiliary verb’ is here considered to be an item on the lexical verb – functional
affix continuum, which tends to be at least somewhat semantically bleached, and
grammaticalized to express one or more of a range of salient verbal categories,
most typically aspectual and modal categories, but also not infrequently temporal,
negative polarity, or voice categories” (Anderson 2006:4-5).
This definition positions AVCs, for Anderson at least, within the category of complex
predicates. He breaks auxiliary constructions down along lines of headedness,
distinguishing semantic and syntactic construction heads. It is the examples wherein the
auxiliary verb functions as the phrasal or syntactic head of the construction − while the
lexical verb is the semantic head − that describes the class of more widely agreed upon
Although auxiliaries are most commonly associated with TAM categories, and perhaps
polarity, they may also be the locus for expressing direction and orientation, as in the
Australian associated motion examples discussed in §1.4. A similar function can be seen
in the Turkic language Tofa:
(206) onson vjertaljo:t-tar uhj-up kel-gen
then helicopter-PL fly-CV
‘then the helicoptors flew in’
(207) men ɲan-a
‘I set off for home’
Tofa (Anderson 2006:36)
The properties that V-sime has in common with these AVCs are the expression of
path/orientation. In constructions such as (208) below, the V1 could certainly be analyzed
as the semantic head − since the interpretation of the whole utterance has to do with
sitting and not with going − with the V2 acting as a functional modifier.
(208) Hunume hamuch-im kia hoo-sa-saka
woman-PL just sit-RED-go.PL
“Those women are just sitting about.”
Although auxiliaries have a primarily functional contribution to the clause, some
languages’ auxiliaries do show some verbal behavior; they may still carry some of their
original meaning in particular contexts, and this is particularly true of auxiliaries derived
from motion and posture verbs. The English 'going to' future construction is a good
example of this. In (20) the a) sentence is unambiguously a motion construction and c) is
clearly futurate, but the b) example is ambiguous between a grammatical (futurate) or a
verbal (motion) reading. (Seiss 2009; Heine 1993)
(209) a.
He is going to town
He is going to work
He is going to come
An even more ambiguous example is found in the Australian language Jingulu which has
highly suppletive 'auxiliaries' which typically co-occur with a non-inflecting predicative
coverb. In addition to hosting agreement prefixes and instantiating the usual TAM
categories, the Jingulu auxiliary also indicates a 3-way associated motion distinction:
motion towards the speaker, motion away from the speaker, and motion neutral.
However, unlike most auxiliaries, these are able to stand alone without a predicating
coverb, in which case they get the interpretations 'come', 'go' and 'do', respectively,
leading Pensalfini (1997) to argue that they are better understood as light verbs than
auxiliaries. The examples below show 'come' and 'go' both in complex constructions
(210)-(212) and in simple ones (211)-(213).
(210) Laja-ngardu kijurlurlu
Carry-1SG-go stone
“I’m carrying a stone.”
(211) Mindi-rruku jalyangku-ma Warranganku-ngka
1DL.INCL-went today-EMPH Beetaloo-ALL
“Today we went to Beetaloo.”
(212) Ngini-rni jundurru duwa-jiyimi
DEM(N)-FOC dust rise-come
“Dust is rising.”
(213) Wilinja ya-jiyimi jamaniki-rni
Countryman 3SG-come this(M)-FOC
“Our countryman is coming.”
Attempts to clearly distinguish auxiliaries from light verbs describe several defining
characteristics of auxiliaries. Although they may initially be form-identical with a main
verb, they need not retain this status (contrasting with Butt's (2003 et seq.) requirement
for light verbs) and are in fact highly prone to reduction. Auxiliaries may display
defective paradigms, unlike light verbs, and they may be restricted in their appearance
with particular tense and aspect forms. Light verbs may show combinatorial or
selectional restrictions, while auxiliaries do not, and light verbs can also affect case and
theta-role assignment but auxiliaries can't. (Butt, 2010; Butt and Lahiri, 2002; Seiss 2009)
Here we can identify more properties that might lead one towards an analysis of -sime/saka as an auxiliary. It appears to have no selectional restrictions, arguably displays a
defective paradigm and is restricted with respect to perfective aspect.
However, I have argued that -sime/-saka does affect theta-role assignment (in that it has a
thematic subject to contribute) and it still retains far too considerable a degree of lexical
weight for me to be comfortable labeling it a fully – or even primarily – functional
element in these constructions. This is particularly evident in the ability of affixal -sime
to reduplicate in these constructions31. It contrasts, for example, with more functional
verbal affixes such as the causative -tua – or indeed the fully grammaticalized
prospective motion affix -se/-vo – which may not be reduplicated.
Butt & Ramchand (2005) also points to reduplication as a feature of light verbs but not auxiliaries in
Urdu, although this is not a criterion that is necessarily applicable crosslinguistically.
Properties of V-sime that seem to correlate with those of AVCs, are as follows:
o No selectional restrictions
o Defective paradigm (arguable)
This is arguable – in V2 position -sime could be said to be lacking
a perfective form, but that feature is also covered by aspectual
restriction below.
o Aspectual restriction
V2 -sime is aspectually restricted relative to main verb siime, in the
sense that the may not occur with perfective aspect, whilst the
latter may do so.
Properties of V-sime that do not correlate with AVCs:
o Contributes a thematic argument
o Unbleached semantics
o Participates in reduplication, which is otherwise restricted to primarily
lexical categories
4.2.2 Light Verbs
LVCs are the most promiscuous of the complex predicate types listed here, at least in
terms of their analysis; the term ‘light verb’ was originally coined by Jesperson (1965) to
refer to constructions such as the English “take a bath”; ‘take’ in this example cannot be
said to be interpreted fully, or literally, and is therefore considered semantically 'light'.
Light verbs require another predicator in order to function. In 'take a bath' this is a DP,
however light verbs are attested in constructions with a wide variety of joint predicators
worldwide, including gerunds, nouns, adjectives, preposition phrases, coverbs32, and
other verbs.
Although usually described as semantically bleached (Jesperson 1965, Mohanen 2005,
Wittenberg 2014, and many others) light verbs are often argued to be, instead,
structurally impaired in some way, e.g having a deficiency of argument structure or theta
assignment (Grimshaw & Mester 1988, Rosen 1990, Sells 1998, Butt 2003, 2010,
Bowern 2008). Languages vary with respect to the size of their light verb inventory,
however they are typically high-frequency verbs with broad or general semantics (Butt
In languages with an inventory of more than half a dozen or so light verbs, they may be
used to categorize the predicate "for various aspect, event structure and trajectory
distinctions" (Bowern 2008:163). In the Uzbek examples (214)-(215) below, for example,
the light verb provides information about the structure of the event that is described by
the main predicator/s.
(214) Qush uchip ketib qoldi.
bird fly-IB come-IB remain-3.PST
“The bird flew away [unexpectedly].”
'Coverb' is the term used to describe a distinct uninflecting word class in many Australian languages that
cannot be used independently but must co-occur with an inflecting verb.
Bu kitobni o’qib borar ekanman, khayolim boshqa joyda edi
this book-ACC read-IB go-PART sow-ISG, mind-ISG.POSS’R other place-LOC be-3PST
“I was reading this book, but my mind was somewhere else.”
Uzbek (Bowern 2008:164)
Butt’s (2003) attempt to pin down a precise, measurable and crosslinguistically accurate
definition of light verbs, together with its revised and updated (2010) redux, is the
seminal work in this area. She tries, in particular, to draw a very clear distinction between
LVCs and AVCs, counter to Anderson’s (2006) assumptions.
Butt attributes the following properties to light verbs:
1. They are always part of a complex predicate (for which, as mentioned in §4.1.1,
she also has a strict definition),
2. They are always form identical with a main verb
3. They have a 'semi-lexical' status that distinguishes them syntactically from both
auxiliaries and main verbs
4. They structure or modulate the event described by main predicator in way that is
different from auxiliaries, modals or other main verbs
(Butt 2003, 2010)
Butt's insistence on form-identity between light verbs and some main verbs is linked to
her observations about the general semantics of light verbs; she considers light verbs as
not a distinct set of verbs, but the same items as their main verb counterpart, whose
interpretation may be either light or heavy depending upon the construction in which they
are used. This correlates somewhat with Cardinaletti & Giusti's (2001) proposal that the
'semi-lexical' properties of motion and other verbs in some constructions is the result of
merging a lexical item into a functional head, the specific resulting properties being the
result of the height of the functional node in question. That is, the same lexical element
will be interpreted with increasingly bleached semantics but with greater functional
properties at progressively higher positions above VP.
The syntactic and semantic differences Butt alludes to are left deliberately vague; she
describes them as variable between languages, leaving the identification of light verbs in
any given language to be diagnosed using language-specific tests (2003). In Butt and
Ramchand (2005) the syntactic distinctions between light verbs and auxiliaries in Urdu
are described, and I include them below as an example of the distinctions that motivate
Butt's claims.
Auxiliaries do not have an effect on the Case marking of the subject, light verbs
Light verbs may be reduplicated, just like main verbs; auxiliaries may not.
The main verb may be topicalized away from a light verb, but not from an
auxiliary verb.
(B&R 2005:8)
Even when clearly distinguished from auxiliaries, LVCs are still not a homogenous class
in Urdu. Butt & Ramchand (2005) distinguish three separate constructions involving light
verbs, distinguished by the form of the main lexical verb (V1), by the semantics of the
construction, and by their syntactic behavior. Of these, Type 1 is primarily of interest
here, since it most resembles, at least superficially, the kind of V-V structures found in
In Urdu Type 3 constructions, the V1 occurs in infinitive or gerund form and bears a
nominal case marker. These are, like Hiaki causatives, biclausal with regard to binding,
although dominated by a single Tense node, and are not complex predicates by Butt's
anjum=nee saddaf=koo
lik -nee]=koo
Anjum.F=ERG Saddaf.F=DAT letter.M=NOM write-INF.OBL=ACC say-PERF.M.SG
‘Anjum told Saddaf to write the letter’
In Type 2 constructions, V1 is in an oblique form of the infinitive. These constructions
have semantics of inception or permission, and are monoclausal by Butt's definition. In
Butt & Ramchand's analysis, the V2 in Type 2 constructions instantiates (little) 'v'.
(217) vo
cry-INF.OBL be.attached-PERF.F.SG
‘She began to cry’
Butt still refers to the V2 in these constructions as a light verb, despite her assertion that a light verb can
only occur in a complex predicate.
Type 1 constructions consist of the V1 in its stem form, along with the inflecting light
verb. The semantics of this type are more variable; they include notions such as
inception, completion, benefaction, suddenness and force, which Butt groups under a
general category of 'boundedness'.
naadyaa=nee xat
lıkh li-yaa
Nadya.F=ERG letter.M.NOM write take-PERF.M.SG
‘Nadya wrote a letter (completely).’
gir ga-yii
nadya.F=ERG fall go- PERF.F.SG
‘Nadya fell (down)’
Urdu (B&R 2005:23)
In the Type 1 examples, Butt argues, the light verb contributes to the Aktionsart of the
event described by the main predicate by rendering it telic, which motivates her claims
that “the function of light verbs is to modulate (sub)evental semantics” (2003:24). In both
her (2003) paper, and (2005) collaboration with Ramchand, Butt claims that this effect
can be mapped syntactically by decomposing the predicate structure into three layers
representing Cause/Initiation, Process/Change and Result/Telos. "Crucial to our proposal
is that idea that verbal predication decomposes (maximally) into these three distinct heads
with very specific semantic and argument structure connections" (2005:26) This analysis
is rooted in Ramchand’s (1997, 2006) theory of argument structure, which is an explicit
attempt to demonstrate how phrase structure reflects event structure and aspectual
v (cause/initiation)
Rv (telos)
Tree for example (218) (Butt & Ramchand 2005:26) Somewhat unintuitively, the light verb in this analysis does not itself occupy the telos
head, despite its proposed connection to boundedness, but instantiates process (or,
potentially, cause) instead - it is the V1 that represents the result state.
This analysis is interesting with respect to Hiaki V-sime constructions, which always
have an atelic or progressive Aktionsart (and are hence incompatible with perfective
aspect, as shown in § In Butt & Ramchand’s (2005) analysis, it is the V1 that
heads the Result phrase, while the light V2 instantiates either ‘cause/initiation’ or
‘process’; since V-sime is always atelic, the V1 in this construction could not occur in the
Result phrase, but presumably must occupy the higher Process phrase. This would
relegate V2 -sime to the remaining higher third of the structure, instantiating the
initiation/causation position head. This would be a very peculiar result, because it would
make V-sime structurally indistinguishable from a causative, which I am confident is
quite outside the intent of Ramchand's theory.
(221) Ramchand's (2006) structure for a simplex intransitive motion verb.
(222) Ramchand's structure for a transitive process verb.
a cake
So, despite initially promising similarities between V-sime and Urdu Type 3
constructions, the analysis that Butt and Ramchand developed for the latter is not a good
fit for Hiaki. Indeed, it may not be tenable for Urdu either: in her 2010 paper, Butt
apparently abandoned this approach in favor of returning to an analysis in LFG (which
was the framework she used in her 1995 dissertation). LFG utilizes linking theory, which
opposes UTAH (Baker 1998) in embracing a lack of one-one correspondences between
thematic roles and grammatical relations.
Properties of V-sime that seem to correlate with those of LVCs, are as follows:
o -sime has a basic or general semantics, which fits with the characterization
of verbs which may be utilized in LVCs
o It has a heavy or independent counterpart, siime
o V2 -sime impacts the aktionsart of the clause
Properties of V-sime that do not correlate with LVCs as they are described here:
o -sime does not appear to be semantically bleached to the degree that light
verbs are expected to be, and displays little to no metaphoric extension, for
o V-sime does not fit the structural model put forth by Butt and Ramchand,
4.2.3 Restructuring Verbs
'Restructuring' is the name given to a type of V-Vinf construction, found in Romance and
in Germanic, that does not exhibit clausal behavior − it does not show boundary effects in
the way that most infinitival clauses do, allowing effects such as clitic climbing and long
object preposing in Romance, and long passive, long distance scrambling, and verb
raising in Germanic. The label reflects the idea that a biclausal structure has been “restructured” to create a monoclausal one. Some analyses assume restructuring involves the
two verbs head-merging to form a complex verb (including Butt 1995 and Muller 2002
amongst others), however Rosen (1989; 1990) and Wurmbrand (2001 et seq) argue for
VP complementation for Romance and Germanic verbs of this type respectively, albeit
with somewhat different motivations. Restructuring constructions are explicitly
differentiated from serial verb constructions by the presence of overt infinitival marking
on the V2. (Rosen 1989, 1990; Wurmbrand 2001, 2003; Cable 2004; Seiss 2009;
Crowley 2002)
As with light verbs, the class of restructuring verbs are typically 'basic' verbs such as
want, begin, try, must, go, and come. (223) is an example of restructuring with a motion
verb, from an Italian dialect. (224) shows the structure commonly assumed for most
restructuring predicates (Rosen 1990; Wurmbrand 2001; Cable 2004).
(223) Vaju
a pigghiari u
go.1SG to fetch.INF
"I go to fetch the bread."
(Cardinaletti & Giusti 2001:3)
(224) [TP SUBJ [vP v [VP Vrestructuring-verb [VP Vhead-of-restructured complement ] ] ] ]
Although the V1 in Hiaki complex verbs does not display overt infinitival marking, Vsime does have some properties that are reminiscent of restructuring constructions.
One of the diagnostics of restructuring in Italian is clitic climbing, shown in (225), where
the accusative clitic lo, which is the object of leggere, 'climbs' to a position above the
restructuring verb vuole.
(225) Clitic climbing
Mario lo vuole leggere
Mario vuole leggerlo.
"Mario wants to read it."
(Rosen 1990:478)
We can see a similar effect in Hiaki, between a compound verb and a subordinate clause
(226) Jason ume koow-im am=su-sua-mahta"
Jason DET.PL pig-PL
"Jason is teaching them to butcher the pigs."
(227) Jason haisa ume koow-im su-sua-wa-u am=mahta
Jason what
pig.PL RED-kill-PASS-OBJ.REL 3PL.ACC=teach
"What Jason is teaching them is the killing of pigs."
As with the Italian example, (226) shows that a clitic argument of the second verb may
(in fact must) precede the first verb. In the Hiaki example this clitic then also intervenes
between the first verb sua ‘kill’ and its argument ume koowim ‘the pigs’. (The similarities
can only be pushed so far, however, because of the different ordering constraints in each
language - ie, because Hiaki is head final.)
Rosen (1990) claims that restructuring predicates in Romance are in fact light verbs. In
her analysis the restructuring verb lacks both arguments and an event specification,
leaving it entirely deficient. It therefore must merge somehow with another predicate in
order to acquire arguments. This happens at the level of Lexical Conceptual Structure. A
restructuring construction is realized syntactically as in (228) below, where the higher VP
represents the restructuring verb, which takes the lexical verb, and its arguments, as a
(228) Rosen 1990:488
Local Raising
Cinque (2001), also working with Romance, has a different approach. He assumes that
restructuring predicates are more like auxiliaries, in that they are generated in a functional
position above the verb phrase. He assumes a complex and universal hierarchy of
aspectual categories, each with its own position. The effects of the restructuring predicate
then are determined by the precise position in which it is generated.
Wurmbrand (2003) however disputes Cinque's claim that all restructuring is necessarily
functional in nature, and asserts that German restructuring facts can only be accounted for
by assuming two types of restructuring, functional and lexical. She, in effect, incorporates
both Cinque's and Rosen's structures, as shown in Error! Reference source not found.Error! Reference source not found. (although she differs from Rosen in that she
assumes that it is the lower verb that is deficient in the Error! Reference source not
found. structure, crucially lacking a subject).
(229) Functional restructuring
main verb
(230) Lexical restructuring
lexical RV
(Wurmbrand 2003:992)
Wurmbrand distinguishes functional restructuring from lexical restructuring by testing
for lexical or thematic properties on the matrix verb, which are held to be projected only
within the domain of VP, possibly as high as v'. Above this level is the non-thematic
domain - heads in this portion of the clause do not assign theta roles to arguments, nor do
they participate in creating special meaning (such as idiomaticity). Functional
restructuring predicates thus group with modal and raising verbs, while lexical
restructuring predicates group with ordinary lexical verbs with respect to a number of
tests, outlined by Wurmbrand in (231).
(231) Thematic properties
Inanimate subjects
Subject raising
Matrix passive
Raising predicates
restructuring verbs
(Wurmbrand 2003:997) As we saw in §3.2.1, V-sime is compatible with both weather predicates and inanimate
subjects, like the German functional restructuring verbs. The matrix passive test asks
whether the highest verb in the structure, V2 in Hiaki, can be passivized. That is, it asks if
the internal argument of V2 that is targeted for A-movement. This test can’t be applied to
intransitive -sime34, which lacks an internal DP argument, although it is possible with a
transitive V2 such as -mahta.
Matrix passive is to be distinguished from ‘long passive’, in which it is the argument of the embedded V1
that is targeted for A-movement, and which does apply to V-sime. See §5.3.
Cinque also claims that restructuring predicates cannot select for internal arguments.
Wurmbrand argues against this by showing that lexical restructuring predicates do select
for internal (dative) arguments; she proves that these are restructuring constructions
because they evidence long A-movement, which results in nominative case on the
embedded object, which triggers agreement with the matrix verb.
With respect to V-sime, example (194), reproduced here as (232), shows that when the
embedded object of V1 is A-moved in a (long) passive construction, it results in
nominative case on the DP, but does not result in agreement (for number) with the V2,
which is obligatorily plural in a passive construction.
(232) Uu
DET.NOM little child.NOM
wam vicha
there toward
“The little child is being pushed along to there.”
So, although V-sime has some apparent properties in common with restructuring
constructions, the picture is not particularly convincing. The discussion surrounding
restructuring predicate properties does, however, reinforce the conflict that emerges in
trying to draw a hard line between 'functional' and 'lexical' items in complex predication,
particularly in complex predicates where the notion of 'semi-lexical' items arises again
and again. However, attempts to define a discrete class of semi-lexical items or properties
have also met with limited success. (Cardinaletti & Giusti 2001)
Properties of V-sime that seem to correlate with those of RVCs, are as follows:
o Small (VP/vP) embedded structure)
o Some lack of clause boundary effects (eg 'long' passive, clitic placement)
Properties of V-sime that do not correlate with RVCs:
o Lack of agreement between long A-moved DP and V2
o Does not fit clearly into either lexical or functional restructuring categories
as defined by Wurmbrand's tests
4.2.4 Serial Verbs
The definition of serial verb constructions (SVCs) has much in common with the
definition of complex predicates generally. They are conceived of as verb sequences that
together behave as a single predicate and are monoclausal. They share a single value for
tense, aspect and polarity, and operate under a single intonation contour (ie, like a single
verb clause), but may have different transitivity values. The verbs should have no overt
markers of coordination, subordination or dependency, and they may share arguments some scholars argue that they must share at least a subject.35 Each of the verbs in the
SVC should be able to stand alone as full lexical verbs in an independent context.
(Aikhenvald 2006; Durie 1997; Bowern 2008; Johnson 2006; Seiss 2009; Lord 1993;
Baker 1989; Foley 2010)
Or an internal argument, for Baker (1989) - I discuss Baker's analysis as it pertains to motion verbs later
in this section.
Together, it has been argued, the verbs should describe what can be conceptualized as a
single event - although this criterion is a little more difficult to quantify, since
'conceptualized as a single event' seems to be understood in a culturally-specific sense,
and relies on fuzzy determinants such as "is best translated by a mono-verbal clause in
non-serializing languages" (Durie 1997:291). As discussed in §4.1.1 for complex
predicates generally, this criterion is subject to some dispute, and Foley (2010) explicitly
argues against it, on the grounds that it displays an inadequately precise description of
both events and structure. He argues that the class of SVCs may describe situations that
comprise either a single complex event or a group of loosely connected events.
SVCs are clearly not a uniform class, crosslinguistically, by almost any measure. They
may consist of two or more verbs, and these may be realized as independent phonological
words - (233), (234), (233) - or they may form a single word - (235), (236), (235). If they
consist of separate words, those words may be contiguous (236) or separable (233).
(233) ire reheson vakili reheha
(iire rehe-sooni vakilii rehe-haa)
1PL.INCL 1PL.INCL-distant-throw canoe 1PL.INCL-distant-go
"We will go, putting (throwing) our canoe to sea."
White Hmong:
(234) nws ntaus tus dev khiav kiag
(s)he hit CLF dog flee completely
"S/he beat the dog off."
(235) yo :h bəә:-ha m-yɔw
medicine spill-go-happen.straight.away
"The medicine spilt straight away."
(Durie 1997: 290)
(236) John-i
kel-e ka-ss-ta
walk-L go-PST-DECL
‘John went to the park walking.’/ ‘John walked to the park’
(Zubizarreta & Oh 2007:3)
The functions and semantics of SVCs also vary across languages. Most definitions
describe a prototypical SVC, although Aikhenvald (2006) points out that for any given
language the SVCs should have most, but not necessarily all of the prototypical
properties. She describes SVCs as inhabiting a continuum based on the expression of
properties such as structural symmetry, contiguity, and marking of grammatical
categories on serialized verbs; constructions may vary in their position on this continuum
both between and within individual languages.
Seiss (2009) considers this degree of diversity an indication that the range of
constructions labeled 'serial verbs' does not constitute a coherent, analyzable class. She
raises the possibility that putative SVCs may include auxiliary and/or light verbs. "As
serial verbs are a very diverse syntactic class, no claim can be made that all serial verbs
are light verbs or auxiliaries on the one hand, on the other hand it cannot be claimed that
no serial verb is a light verb or auxiliary either" (2009:510).
It will come as no surprise, therefore, that the range of structural analyses that have been
put forward for SVCs is similarly varied. Larson (1991) considers serialization to be
parallel to secondary predication. He considers a range of possible structural relations
between the two verb phrases - coordination, adjunction, or complementation - and settles
on the last of these. Johnson (2006) argues that the evidence for SVCs in Krio also
supports a complementation structure, and provides some tests to distinguish between the
three possibilities.
1. Coordination (Johnson 2006:42)
An advantage Johnson claims for coordination analyses is that coordinated events must
reflect the temporal order in which they occur, and this lines up with the temporal
iconicity which is often observed in SVCs. This is, however, actually a disadvantage for
languages like Hiaki, as well as Japanese, which are head final, as noted in Nishiyama
(1998) and Tomioka (2004).
More interestingly, Johnson observes that SVCs in many languages do not show island
effects as other coordinated structures do, and which are formalized in the Coordinate
Structure Constraint, attributed to Ross (1967). This constraint states that a conjunct may
not be moved out of a coordinated structure, as in: "Which book did you read Harry
Potter and?" (2006:43). Baker (1989) makes a similar arguments involving extraction of
an argument from a coordinated VP. "...if the NP argument of a verb in an SVC can be
extracted by WhMovement, it follows that the structure cannot be a coordination (by the
Coordinate Structure Constraint)" (1989:514).
2. (Phrasal) Adjunction (Johnson 2006:43)
Adjunction of V1 to V2
(239) Adjunction of V2 to V1
Adjunction is more difficult to dismiss out of hand than is coordination. Adjunction
structures are expected to show effects of asymmetrical c-command, but of course
complementation would show the same. Johnson argues specifically against the V'
adjunction proposed by Law & Veenstra (1992) and Veenstra (1993), which analysis he
claims to be insufficiently motivated by the evidence. In particular, he argues that Law
and Veenstra failed to demonstrate clearly that an adjunction analysis is preferable to a
complementation analysis36.
If we understand one of the VPs to be an adjunct then we might expect to see island
effects, such as restrictions on wh-extraction from the adjunct VP. We do not see this
effect in Hiaki V-sime constructions - objects of V1 are able to undergo wh-extraction, as
shown in (240).
(240) Hita-sa eme'e temu-sa-saka?
What-Q 2PL.NOM kick-RED-go.PL
"What are you going around kicking?"
Veenstra's (1993) argument is primarily for an asymmetrical phrasal relationship contra Baker (1989).
However, Truswell (2007) shows that A'-extraction is permitted from some untensed
verbal adjuncts in English, specifically from two kinds of secondary predicates, depictive
and causal37, so a lack of island effects here does not necessarily eliminate the possibility
that VP1 is an adjunct.
4. Complementation (Johnson 2006:44)
Subordination of V2 to V1
For Johnson, the distinguishing feature between an adjunction structure and a
complementation structure comes down to the issue of wh-extraction - however, he
concedes that this is only a viable diagnostic for languages which a) are not wh-in situ
and b) show ECP effects.
As (240) shows, wh-extraction is possible in V-sime constructions, however as
mentioned, this is not definitive evidence against an adjunction analysis. Furthermore, the
question of ECP effects in Hiaki has not been satisfactorily answered at this time (Harley
The examples that Truswell gives include: What did John arrive [whistling t]?
and What did John drive Mary crazy [trying to fix t]? (2007:1356)
5. Some other proposed structures
Baker's (1989) analysis cannot be omitted from a discussion of serial verbs; although
Baker's primary concern is in dealing with object sharing in trans-trans verb sequences,
his proposal has consequences which are relevant to motion constructions. For the
sentence in (242), Baker proposes the structure in (243), in which the verb phrase is
doubly-headed, with the second verb projecting only a V', and no empty categories are
(242) kofi naki amba kiri
Kofi hit Amba kill
"Kofi killed Amba."
(Baker 1989 from Sebba 1987)
(243) (Baker 1989:520)
naki Amba
(Ag. Th)
(Ag. Th)
In this structure, both the verb naki 'hit', which precedes the object, and kiri 'kill', which
follows it, assign theta roles to the object Amba. Since Amba is within a V' projection of
both verbs, it can only take the internal argument theta role of both. The first verb, which
is the structural sister of Amba, directly theta marks it, while the second verb indirectly
theta marks it, because Amba is the sister of one of its projections. However, in an SOV
language like Hiaki, Amba would not be the sister of the second verb's projection, so it is
uncertain how it would receive indirect theta marking in this situation.
Durie (1997) points out a specific problem with Baker's analysis related to motion
serializations, such as the one in (244)
(244) wan man go luka wan dansi
man go look a
"A man went to watch a dance."
(Durie 1997 from Baker 1989)
In this example, much like the Hiaki V-sime constructions, the internal argument of the
unaccusative motion verb is also the agent or external argument of the second verb. Durie
points out that if this construction is an example of serialization, then Baker's model is
not able to account for it. Indeed, Baker treats this as a form of complementation,
determining that in this case, 'go' must subcategorize for an infinitival clausal
complement with a controlled PRO subject, and thus places the construction outside of
the realm of serialization as he defines it.
Tomioka (2004; 2006) invokes a head-adjunction analysis for Japanese V-V compounds,
in order to account for transitive-transitive structures in a head-final language.
(245) Jiro-ga Ichiro-o shime-koroshi-ta
Jiro-NOM Ichiro-ACC strangle-kill-PST
"Jiro killed Ichiro by strangling (him)."
(Tomioka 2004:9)
(246) (Tomioka 2004:8)
In the structure she proposes, the first verb is head-adjoined to the second. It does not
participate in argument structure (evidenced by differences in the case pattern of selected
objects of the two verbs) and behaves as a manner modifier to the V2. Tomioka proposes
this structure contra Nishiyama's (1998) analysis, which analyzes the Japanese V-V
compounds as resultative serial constructions comparable to those found in Ewe, Yoruba
and Sranan. However, Tomioka shows that Japanese does not have resultative structures
in the usual sense, using evidence from directed motion constructions.
In Japanese, a manner-of-motion verb cannot take a Path PP, as shown in (247) - in order
to obtain the intended semantics, a V-V compound consisting of a manner verb and a
directed motion verb is employed, as in (248).
(247) *Taro-ga gakko-ni
Taro-NOM school-LOC walk-PST
"Taro walked to school."
(248) Risu-ga
squirrel-NOM tree-from
"A squirrel rolled down a tree."
Tomioka concludes that Path is then an argument of the directed motion V2, and the V1
is a manner adjunct, which does not contribute to argument structure. This pattern is very
similar to one proposed for Korean serial verbs of motion by Zubizarreta and Oh (2004;
2007), which I discuss in the next section.
Zubizarreta & Oh, Lim & Zubizarreta
In Korean, like Japanese, manner-of-motion verbs do not express directed motion, and so
the PP in (249)-(250) cannot be used to signify a goal:
(249) *John-I
“John ran to the park.”
(250) Cf. John-I
“John ran at the park”
(Zubizarreta & Oh 2007:84)
To unambiguously denote directed manner-of-motion, and license the presence of a goal
PP, a manner-of-motion verb must enter into a SVC with a basic motion verb such as ka‘go’.
(251) John-i kongwen-ey talli-e -ka-ss-ta
John-Nom park-Loc run-L-go-Past-Decl
“John ran to the park.”
(252) John-i kongwen-ey kel-e ka-ss-ta
John-Nom park-Loc walk-L go-Past-Decl
“John walked to the park.”
(Zubizarreta & Oh 2007:86)
Zubizarreta and Oh assume that the basic motion verb ka- ‘go’ and its counterpart o‘come’, are not really contentful lexical items in the usual sense, but a morphological
reflex used to spell out a V head in a particular configuration that signals directedmotion, that is, when “V takes a directional path complement and a specifier, of which
the path is predicated” (Zubizarreta & Oh 2007:78).
Importantly, for comparison with similar structures in Hiaki, Korean can also incorporate
transitive manner-of-motion verbs into a directed motion SVC, as in (253).
(253) John-i
chayksang-ul mil-e tul-i-e desk-ACC
push-L move.into-CAUS-L go-pst-decl
“John went into the living room, pushing the desk.”
(Zubizarreta & Oh 2007:106)
Like Tomioka, Zubizarreta and Oh (and subsequently Lim & Zubizarreta (2013)) argue
for an adjunction analysis, but the fact that the V1 in Korean can be transitive, and appear
with an attendant object argument, necessitates a slightly different formulation. In their
case, the entire VP1 adjoins to the V2 head. This structure is shown in (254).
(254) (Lim & Zubizarreta 2013:16)
Change of Location
Ploc Loc DP
Manner or means
The Japanese and Korean examples are very interesting with respect to Hiaki, which is
also head final and agglutinating. However, beyond the surface, there are some pertinent
differences to keep in mind. First, the Hiaki motion verb siime is not a directed motion
verb per se; although it does, when used in isolation, imply movement away from the
speaker, it is able to accommodate a PP directing movement towards the speaker as well,
as discussed in §2.1.1. Second, Hiaki manner of motion verbs are perfectly able to occur
with a goal or path PP without the necessity of adding a V2 such as –sime, as shown in
(255) Ume ili uusi o’o-im wahi-wa vicha tenne-k
DET.PL little child male-PL inside-to toward run.PL.PFV
“The little boys ran inside.”
Finally, in both Japanese and Korean the semantics of the adjoined phrase are identified
explicitly as manner modification, and it is not clear that this is the case in Hiaki.
Properties of V-sime that seem to correlate with those of SVCs, are as follows:
o Both V1 and V2 have independent counterparts
o Both verbs are (mostly) fully interpreted
o Both verbs share a single value for tense/aspect/polarity
o Both verbs come under a single intonation contour
o There are no overt markers of either subordination or coordination
Properties of V-sime that do not correlate with SVCs:
o No possibility of object sharing (if you're Baker)
o Lack of literal motion interpretation in the StanceV+RED-sime
4.3 Summary
In this chapter I showed that multi-verb motion constructions similar to V-sime exist in
many languages worldwide and in many forms, each of which requires its own structural
representation. I compared the properties of V-sime to motion in auxiliary verb
constructions, light verb constructions, restructuring constructions, and serial verb
constructions, and noted that the criteria for distinguishing these complex predicate subtypes is far from clear cut, and that there are plenty of grey areas and a considerable
degree of overlap.
For each of the construction types above, V-sime shares some number of the properties
associated with it, and fails to exhibit others. Of the four major sub types of complex
predicate discussed here, V-sime seems to have the most in common with SVCs,
although as noted, SVCs themselves constitute an extremely large and varied class.
I also discussed a number of structural analyses for the different types of complex
predicates, including a range of proposals for SVCs in various languages. While none of
these proposals seems to fit the Hiaki V-sime attributes perfectly, they represent a fairly
thorough range of possibilities and issues to be considered.
In Chapter 5, I will demonstrate that the properties of V-sime necessitate a structure
similar to that proposed by Zubizarreta & Oh (2004; 2007) and Lim & Zubizarreta (2013)
for analogous phenomena in Korean, although this structure is not without its challenges.
I will also consider the range of Hiaki strategies for combining verbs or events, as well as
evidence for a range of structural types within Hiaki V-V compounds. Although I have
shown in Chapter 3 that there is no major distinction between compounds with free or
bound V2s, I will argue that there is nevertheless a cline of complex V types, and
consider the structural ramifications for this.
Chapter 5 Struggles with Structure
Empo kia haana huni'i hita chatcha! (You just hang things any old way!)
5.1 Questions regarding the structure of V-sime
In Chapter 4, I surveyed a range of motion complex predicate types across a broad variety
of languages, as well as the kinds of analyses that they have engendered, and I compared
the properties of V-sime to the properties of these structures. Although V-sime has
commonalities with all of them, the greatest degree of similarity was shown to be with
(asymmetric) serial verb constructions. In particular, motion SVCs in Korean and
Warlpiri look a great deal like Hiaki V-sime constructions, both in terms of their
semantics and their morphosyntactic forms.
In Chapter 5, I address three questions with respect to the structure of V-sime:
1. What position does affixal -sime occupy in the phrase structure? Can it be best
classified as a functional or a lexical item?
2. What size and type of constituent is VP1?
3. What is the structural relationship between VP1 and V2-sime?
To answer question 1, in §5.2, I draw more explicit parallels between Korean, Warlpiri
and Hiaki and show that -sime must be regarded as a lexical item rather than a functional
one. In §5.3, based on the properties that have been laid out thus far, and drawing from
previously discussed analyses of related constructions, I answer questions 2 and 3,
showing that VP1 itself must be as large as, but no larger than vP, excluding VoiceP.
I also propose that the structure of V-sime constructions must be one of adjunction, with
VP1 head-adjoined to the root -sime.
In §5.4 I summarize the range of complex verbal types in Hiaki, and show how Hiaki
may be situated in the wider crosslinguistic typology of complex predicate types. In §5.5
I conclude with a number of outstanding questions and issues, with directions for future
5.2 Close comparisons - Korean and Warlpiri
In this section I show that affixal -sime is a lexical item by close comparison of V-sime
with similar constructions in Korean and Warlpiri.
5.2.1 Korean
In many languages’ motion expressions, the manner of motion and directed motion
components are decomposed such that each must be expressed by a distinct verb, and, as
shown in the previous chapter, Korean is one such language. In example (256), the verb
kel- ‘walk’ provides the manner of motion, while the directed motion verb
-ka ‘go’
allows the interpretation of motion along a path.
John-i kongwen-ey kel-e-ka-ss-ta
John-NOM park-LOC walk-L-go-PST-DECL
‘John went to the park walking.’/ ‘John walked to the park’
(Zubizarreta & Oh 2007:3)
The precise path of motion can be expressed by a PP, realized by the locative goal in
(257), or by a bound path verb as in example (258), which can occur with or without a PP
goal. The path verb cannot, however, occur without a tensed verb of motion as host.
John-NOM room-LOC go-PST-DECL
‘John went to the room’
(Zubizarreta & Oh 2007:81)
(pang-ey) tul-e
John- NOM room- LOC move.into- L-go- PST-DECL
‘John went in(to the room)
(Zubizarreta & Oh 2007:82)
As noted in Chapter 4, manner-of-motion verbs do not express directed motion in and of
themselves, and so the locative PP in (259)-(260) cannot be used to signify a goal.
kongwen-ey talli-ess-ta
John-NOM park-LOC run-PST-DECL
Intended: “John ran to the park.”
Cf. John-i kongwen-eyse38 talli-ess-ta
John-NOM park-LOC
“John ran at the park”
(Zubizarreta & Oh 2007:84)
To unambiguously denote directed manner-of-motion, and license the presence of a goal
PP, a manner-of-motion verb must enter into a SVC with a basic directed motion verb
such as -ka- ‘go’.
John-i kongwen-ey talli-e -ka-ss-ta
John-NOM park-LOC run-L-go-PAST-DECL
“John ran to the park.”
John-i kongwen-ey kel-e-ka-ss-ta
John-NOM park-LOC walk-L-go-PAST-DECL
“John walked to the park.”
(Zubizarreta & Oh 2007:86)
Recall Zubizarreta and Oh’s claim that the basic motion verb ka- ‘go’ and its counterpart
o- ‘come’, are not really contentful lexical items, but simply used to spell out a V head in
a particular configuration that signals directed-motion. That is, when “V takes a
directional path complement and a specifier, of which the path is predicated” (Zubizarreta
Although both -ey and -eyse are glossed 'locative' in Z&O, the former is goal-oriented ('to') while the
latter is not ('at'). In some of these constructions, native speaker consultants preferred -ulo ('towards').
& Oh 2007:78). Importantly, Korean can also incorporate transitive manner-of-motion
verbs into a directed motion SVC, as in (263) (previously shown as (253) in §4.2.4).
John-i kesil-ey
chayksang-ul mil-e
John-NOM desk-ACC push-L move.into-CAUS-L go-PST-DECL
“John went into the living room, pushing the desk.”
(Zubizarreta & Oh 2007:106)
Korean is an interesting case to compare the Hiaki constructions with because, on the
surface, the two languages share many similarities such as persistent head finality and
SOV word order, agglutinating morphology, and nominative-accusative case systems.
With respect to the complex motion constructions, both languages have a basic motion
verb in final position, inflecting for tense, and may have a transitive verb and its object
appearing in the VP1 position. Furthermore, in both languages, the complex motion
construction is required in order to denote an accompanied motion reading, in which the
subject of a transitive motion V1 moves, in addition to the object, as in example (263).
Removing the motion V2 in both cases results in an obligatory reading where the subject
remains stationary, and only the object of the V1 moves (264)-(265).
Rufiino Simon-ta-u lula bwe’u pelotam roakta-k.
Rufino Simon-ACC-to straight big ball-PL roll.TR-PFV
“Rufino rolled the big ball straight to Simon” (The ball moves; Rufino does not)
(265) Rufino-ka
Simon-eykey kong-ul sethwulukey kwu-li-ess-ta. (Korean)
Rufino-NOM Simon-DAT ball-ACC clumsily
“Rufino clumsily rolled the ball to Simon.” (The ball moves; Rufino in-situ)
There are some important points of distinction that must be acknowledged, however. The
first is that Korean basic motion verbs have a distal deictic component that Hiaki -sime
lacks; Korean -ka 'go' has a counterpart -o 'come'. Hiaki has no such distinction. The
second, related difference is that Hiaki intransitive manner of motion verbs are perfectly
capable of being independently combined with a goal PP; examples (266)-(267)
demonstrate that affixal -sime is not required to license the presence of a goal PP, as -ka
is in Korean.
(266) Marselo Suichi-u vicha si
chuumti weye
Marselo Swichi-TO toward very quickly walk.IMPFV
“Marselo is walking very quickly toward Swichi.”
(267) *John-i
kongwen-ulo kel-ess-ta
John-NOM park-TOWARD walk-PST-DECL
Intended: “John walked toward the park.”
Since Hiaki manner of motion verbs can license goal PPs then the V-sime structure
cannot be motivated in the same way as Zubizaretta &Oh propose for the Korean
constructions; that is, -sime cannot be analyzed as a reflexive spellout of a functional
head licensing a directed motion construction.
Finally, it is significant that although Korean –ka, like -sime, has a primary semantics of
spatial motion/path, it may also be extended into metaphorical or aspectual uses. In each
of the examples below, the 'path' that -ka contributes is metaphorical rather than literal,
and the aspectual properties of the V1 have been modified. In each case, the V1 is an
achievement verb; the use of -ka makes possible a durative interpretation, allowing for
modification by temporal adverbs such as 'gradually'.
(268) John-i (cemcem) cwuk-e ka-ss-ta.
John-NOM (gradually) die-L go-PST-DECL
‘John was on the path to death.’
(269) John-i (cemcem) salaci-e ka-ss-ta.
John- NOM (gradually) disappear-L go-PST-DECL
‘John was on the path to disappearance.’
(270) Yenkuk-i (cemcem) ttuthna ka-ss-ta.
performance-NOM (gradually) end go-PST-DECL
‘The performance approached the end.’
(Zubizarreta & Oh 2004:45-46)
As shown in §(271)-(272), the Hiaki V-sime construction resists metaphorical extensions
of this nature, even within the spatial realm.
(271) Hunume huyam vo’o-u tahti ha’abwek.
road-to all.the.way stand
“Those trees go all the way to the road.”
(272) *Hunume huya-m kora bwikola yo’outu-saka
Those tree-PL fence around grow-go.PL
Intended: “Those trees grow along the fence.39”
V-sime is incompatible with perfective aspect, which suggests that it may contribute a
durative semantics, but it never serves a purely aspectual function without also entailing
physical motion. This resistance to semantic bleaching of the lexical content of -sime
Admittedly, in the direct Korean equivalent to this example (below), -ka- ‘go’ does not
imply growth along a spatial path either – the ‘along the fence’ phrase can be omitted.
Rather, -ka- modifies grow, suggesting that the trees are growing over time – their
location is incidental. This is actually an even more abstract metaphor, but regardless,
neither interpretation is possible in Hiaki. (Choi & Jung, pers. comm.)
namwu-nun (tamcang-ul ttala)
that tree-TOP (fence-ACC along) grow-go-PRES-DECL
"{Those trees/that tree} grow(s) along the fence."
leads me to conclude that Hiaki -sime is less semantically 'light' than the Korean -ka and
has more characteristics of a lexical element than a functional one.
5.2.2 Warlpiri
The Australian language Warlpiri also has a type of serial construction with significant
similarities (and some differences) to Hiaki V-sime. The construction known as
‘associated motion’ or ‘associated path’ in Warlpiri consists of an infinitival verb40 –
which may be transitive or intransitive, simplex or complex, motion-denoting or not –
bound to a tensed verb of motion, usually the basic motion ya-ni ‘go’. Although Warlpiri
is a non-configurational language with extremely free word order and ergative-absolutive
case marking, it has agglutinating morphology and strict ordering of morphemes within
word boundaries. The associated motion verb is always the final, tense-bearing element,
like V-sime.
(273) pata-karri-nja-ya-ni
'fall while going along' / 'be falling'
This element is labeled VINF in these examples, in accordance with the glossing practices of my sources,
but should be considered basically equivalent to those items labeled V1 elsewhere.
(274) Nya-nja=rni
‘He would go along looking’
(Laughren 2010:187)
Like -sime, -ya-ni is also a non-deictic motion verb, with no 'come' counterpart. In
Warlpiri, direction of motion is typically indicated by the inclusion of one of a series of
directional enclitics (275), or by oblique case marking on a noun, or some other
periphrastic means.
(275) Winpirli-nja-ya-nu=rra
‘(He) went whistling all the way there.’
(Simpson 1991: 310)
Laughren argues that the tensed motion verbs in these constructions “do not express
argument-taking predicates; they serve to modify mainly spatio-temporal properties of
the situation or event denoted by the thematic core of the verbal constituent they are part
of” (2010:174). She considers them a type of aspectual auxiliary, along with the
inceptive. Regarding argument structure, certainly motion verbs are typically intransitive,
with an absolutive subject. In the example below, however, the VINF is transitive, and the
This example shows a directional enclitic and a 2nd position auxiliary clitic intervening between the
VINF and the tensed motion verb, which reflects the nonconfigurational properties of Warlpiri syntax,
however the generalizations about ordering and hierarchy within the verbal domain stands, scrambled
elements notwithstanding.
clausal arguments reflect that, with ergative case marking on the subject, showing that the
motion verb does not dictate the case frame of the whole clause.
(276) Kapi=li
FUT=3plS many-other-ERG
‘many others will come along spearing (them) the other way’
(Simpson 1991:112)
And, like the Korean constructions discussed in §5.2, the Warlpiri motion verb's uses
extend into the metaphorical and aspectual realm - the examples below show a similar
type of metaphorical path to a change of state, not dissimilar to the use of Korean -ka.
(277) Kurdu waku rdilyki-ya-nu.
'The child broke (his) arm.' = 'The child's arm broke.'
(278) Wati
=PRES good-become-INF-go-NP
'The man is getting better/becoming good.'
(279) Mangarri-rli
=PRES man
'Food is making the man better.'
(Laughren 2010:205)
In example (280) we can even see the 'go' verb occurring twice in the same complex - the
first instantiation expresses the literal motion, the second serves to affect the spatiotemporal aspect of the event, indicating durativity or continuation.
(280) Ya-ni.nja-ya-ni
go-INF -go-NP
'They are going along.'
Once again, there are some key similarities and differences between the Warlpiri and the
Hiaki constructions. Both have a non-deictic motion verb which lends a general 'path'
reading, but without the directional component of the motion verbs in a language like
Korean with its 'come/go' alternation. Like Hiaki, the Warlpiri -ya-ni construction is not
required to license a goal of motion, as was the case in Korean. However, both Warlpiri
and Korean show productive metaphorical extension into the realm of abstract paths,
while Hiaki resists non-literal readings in all but a very few specific cases, namely, those
which involve both reduplication of -sime and stance V1s, illustrated in examples (191)(192), in §3.2.2.
5.2.3 Q1: V2 -sime is lexical
By contrasting the finer details of the V-sime construction with some very similar
constructions in other languages, we are able to draw some important conclusions for
understanding the underlying structure, and to answer the first of the three structural
questions posed at the beginning of this chapter: What position does V2-sime occupy in
the phrase structure? Can it be best classified as a functional or a lexical item?
Because the V-sime construction is not required to license a goal PP with a manner of
motion verb, we cannot motivate the presence of V2-sime as a way to express directed
manner of motion, as Zubizarreta & Oh suggest for Korean. Additionally, despite similar
surface realizations and comparable semantics, Hiaki -sime does not exhibit productive
metaphorical or aspectual uses like those seen in Warlpiri and Korean. Although -sime
does impact the aspectual properties of the clause, being incompatible with perfective
aspect, it doesn't perform primarily aspectual functions. Therefore, although Korean -ka
and Warlpiri -ya-ni can be analyzed as functional elements - either as a light verb or an
auxiliary respectively - Hiaki -sime must be regarded as lexical rather than functional,
and therefore heads a root phrase (√P) rather than a higher category such as vP.
5.3 Modeling Structure
In §5.3.1, I address question 2 (‘What size and type of constituent is VP1?’) by drawing
on Wurmbrand’s (2001, 2004, 2007, 2013) analyses for German restructuring predicates
with similar properties, such as long passive, as well as properties of Hiaki complex
verbal morphology discussed in Chapter 3 and argue that VP1 must be a vP, crucially
excluding a VoiceP layer.
Subsequently, in §5.3.2, I address question 3 (‘What is the structural relationship between
VP1 and V2 -sime?’) by summarizing relevant argument structure properties of siime/
-sime as described in Chapter 2 and Chapter 3, and, following Zubizarreta and Oh’s
(2004; 2007) analysis of Korean motion SVCs, propose that VP1 is head-adjoined to V2.
I demonstrate how this structure might look, and discuss shortcomings of the analysis, as
well as alternative structures.
5.3.1 Q2 - Size of V1 is vP, excludes VoiceP
The second question regarding the structure of V-sime involves the size of the VP1
element. Following Jung (2014) and Harley (2013) I assume a three-layered structure for
Hiaki verb phrases42 consisting of a VoiceP, which introduces external arguments and
licenses accusative case, a vP which introduces verbalizing elements such as -tua 'cause'
and -te 'become', and a root phrase. The structure for a basic unaccusative verb like siime
is shown in (281).
I use V(P) as a shorthand reference for the layered verbal domain, since it is useful to occasionally refer
to the verbal domain without committing to its size in ever case (eg vP vs. VoiceP).
The example in (282) shows that the adjunct must be able to be at least as large as vP
because it is possible for a direct causative to serve as V1 - the causative morpheme -tua
can be shown to instantiate v0. (Tubino Blanco 2010; Harley 2013)
(282) Simon intok Hoan Rufino-ta bwik-tua-saka
Simon CONJ Juan Rufino-ACC sing-CAUS-go.PL
"Simon and Juan are going along making Rufino sing."
Note that in this example -sime takes the plural form -saka, in agreement with the
conjoined subject 'Simon and Juan'. This means that the DP Simon intok Hoan is basegenerated as the internal Theme argument of the motion verb43. Obligatorily, Simon and
Juan are also the joint Causers of the embedded VP 'make Rufino sing'.
There are a couple of possible explanations for this state of affairs. The first is that VP1
includes an external subject position (introduced by Voice), which is inhabited by PRO,
and controlled by the nominative subject. This option is represented in (283).
See § for justification of the analysis of siime as an unaccusative.
Simon intok Hoan
However this analysis runs into some problems with passivized examples such as (284).
(284) Imi’i hiva
Here always
“There are always balls being kicked along here.”
In this example the embedded object has been promoted to subject and received
nominative case, in a construction very like the 'long passive' that Wurmbrand (2013)
describes for German restructuring, as discussed in §4.2.3. Wurmbrand points out that
any analysis that involves an embedded PRO subject must, in order to account for the
long passive facts, assume first that the embedded accusative case feature can be
somehow eliminated or made inactive, and second that PRO can be circumvented in
some way in order to promote the embedded object over it to the matrix subject position
The other possibility, following Wurmbrand, is that VP1 includes vP but crucially
excludes VoiceP, and thus the external argument position44, which would host the
problematic PRO. The tree in (286) shows the proposed structure for a clause with an
embedded transitive vP1 such as (285).
(285) . Uu hamut ili usi-ta yu’u-sime
det woman little child-acc
“The woman is pushing the little child along.”
Uu hamut
ili usi-ta
Tubino Blanco and Harley (2011) also argue for the separability of v and Voice, using evidence from the
Hiaki indirect causative -tevo, bolstering the feasibility of proposing this structure for Hiaki.
The tree above is very similar to the previous tree (283), but it eliminates the VoiceP
layer of the embedded verb phrase. It maintains the v0 head, which is required for hosting
verbalizing affixes, but strips away Voice, which would introduce an external argument
position. This structure accounts for the long passive facts; without Voice to introduce it,
there is no longer a PRO argument to circumvent, and the embedded object is dependent
upon the functional domain of the matrix verb for case assignment (presumably the
matrix VoiceP). Furthermore, Wurmbrand (2002) makes the claim that an obligatory
control interpretation is a mark of semantic control (derived through strict locality and
other conditions) and that syntactic control (ie, PRO) gives a non-obligatory control
interpretation. Since V-sime constructions have an obligatory control interpretation45, this
suggests that semantic control, and not PRO, is responsible.
5.3.2 Q3 - VP1 is an adjunct
In Germanic languages such as Dutch (and English), manner-of-motion verbs can equally
well express an activity (atelic) or an accomplishment (telic); in the former case the verb
is syntactically unergative, in the latter, unaccusative. The difference in Dutch can be
seen by the use of different auxiliary verbs.
dat Jan naar Groningen twee uur lang heeft gewandeld
That Jan to Grongingen two hours long has walked
‘… Jan walked in the direction of Groningen for two hours.’
That is, the matrix subject is obligatorily interpreted as in motion – see §3.2.1 for the relevant examples.
dat Jan in twee uur naar Groningen is gewandeld
that Jan in two hours to Groningen is walked
‘… Jan walked to Groningen in two hours.’
(Zubizarreta & Oh 2007:2)
In Hiaki, although manner of motion verbs can take goal phrases, it is not clear that they
are variable between unaccusative and unergative in the same manner as in Germanic.
Some manner of motion verbs in Hiaki, aside from the basic 'go' verb siime/saka, show
suppletion conditioned by the number of the subject, notably vuiti/tenne ( and and weye/kaate ( and As discussed in §, Harley, Tubino
Blanco & Haugen (2009) show that in Hiaki, verbal suppletion is triggered by internal
arguments. They claim that factors affecting the spell-out of root nodes must be in a local
relationship with the root, as is also proposed by Bobaljik (2012) and bolstered further in
Bobaljik and Harley (2012)46. This entails that all suppletive intransitive verbs in Hiaki
are unaccusative.
In the previous section, I analyzed Hiaki VPs as three-layered constituents, and claimed
the structure in (281), repeated here as (289), as representative of a basic unaccusative
verb, using suppletive main verb siime as an example. The single argument of the verb is
base generated internal to the √P as the sister of the verb, and is thus in a sufficiently
local relationship for its number features to trigger suppletion of the verb.
The details of this analysis are presented within the framework of Distributed Morphology in which the
root node is an abstract bundle of features that are later realized (spelled out) by the insertion of
phonological content – it is this choice of phonological vocabulary item that must be determined by the
local environment.
The relationship between suppletion and unaccusativity is important for us, because
affixal V2 -sime/-saka also suppletes in agreement with subject number, and therefore
must also be unaccusative. This has two salient consequences. First, this supports the
earlier conclusion that -sime must be the head of a √P rather than instantiating v0 (which
position would be outside of a local relationship with a DP that could trigger suppletion).
Second, since an unaccusative structure allows only a single (internal) argument position,
it follows that the subject of siime/-sime must start out in that internal argument position.
In the case of the V-sime compound, this leaves no obvious structural position (such as
complement) for the VP1 to occupy.
Since no complement position is available, the only possible solution to the problem is
that VP1 is an adjunct47 of some sort. Furthermore in order to obtain the correct
linearization, VP1 must be adjoined such that it intervenes between √-sime and its
internal argument - but without disrupting the sisterhood relationship. This means that
Although it is possible to extract arguments from the VP1 (in the case of the ‘long passive’ constructions,
discussed in §5.3.1), and extraction is normally not considered a property of adjuncts, Truswell
(2005;2007; 2009) presents evidence that it is possible to extract arguments from adjuncts in some
VP1 can only be head-adjoined to √-sime, resulting in a complex head, which is
presumably not a barrier to the conditioning of the root by the number features of its
sister DP.
√ -sime
Head-adjunction is a famously awkward problem in Bare Phrase Structure (see, for
example, Matushansky 2006; Harley 2004 for details). Many of the problems noted in
those works apply to head-movement, rather than adjunction of a previously unmerged
element, and are not an issue here: for example, because this is external Merge, not
internal Merge, there is no copy and remerge operation, and there is no difficulty
associated with Chain Uniformity. Some issues persist, however.
If -sime is a head, which selects for a DP Theme complement, then the intervention of the
adjunct VP1 between the head and its complement must have been accomplished by one
of two derivational sequences, both of which are problematic in some way. Option 1 is
that -sime merges first with its complement DP, and the adjunction of VP1 subsequently
targets the √0 rather than the edge of the √P, which is a violation of the Extension
Condition (Chomsky 1995:190). Option 2 is that -sime merges first with the adjunct VP1
and only subsequently with its selected complement, which violates the principle of First
Merge (Adger 2003:105), in which complements are defined by their status as the
element which merges first with a head48.
In Bare Phrase Structure theory, a head is defined as a terminal or minimal projection (a
node that does not dominate a copy of itself) and a phrase is a maximal projection (is not
dominated by a copy of itself). Minimal and maximal projections are all that are
available to the computational system; intermediate (bar) levels are ‘invisible’ (Chomsky
1995:61). Although it is possible for an element to be both terminal and maximal (eg
clitics (Chomsky 1995:68-69)) a difficulty with the ‘head+adjoined phrase’ constituent
(in this case, VP-sime) is that it is neither terminal, nor maximal. Insofar as that
constituent behaves like a head in selecting for a DP sister, and exhibiting suppletion
conditioned by that sister, this could pose difficulties. These are not only difficulties for
the structure I propose – it is equally problematic for any adjunction structure, including
head-movement, and there have been a number of attempts to reconcile the conflict.
Carnie (1995; 2000), arguing for phrasal heads in Irish copula constructions, suggests that
the distinction between head and phrase is not a matter of primitives, but of behaviors,
and so phrases may have head-like properties and vice-versa.
Specifiers merge second, leaving adjuncts, in the usual case, to scramble for peripheral positions.
The structure in (290) is the only feasible structure, given the properties of Hiaki V-sime
constructions described here, but it is not entirely novel. It owes a great deal to the
structure that Zubizarreta and Oh provide for the Korean motion SVCs (described in
§4.2.4) although their proposal involves adjunction of a manner-modifying phrase to a
light verbal head, rather than to a lexical root.
5.3.3 Structural Alternatives
Given that I have acknowledged some problematic aspects of the structure I proposed, are
there no alternatives that might be advanced? In this section I discuss two alternative
means of achieving the correct linearization, and discuss why they are even less
satisfactory than the structure presented in (290) above.
Movement of the internal argument around the adjunct
The first and most obvious solution to the problem of an adjunct phrase apparently
intervening between a head and its sister is to propose that in fact the adjunct is adjoined
at the edge of the phrase like any normal adjunct, and that the DP argument of -sime then
moves around it to a higher position in the clause. Thus, a structure such as (291) might
be proposed, which allows for the correct linearization of elements, without the awkward
necessity of positing adjunct intervention.
(lands in Spec,TP)
-tua -
Simon intok Hoan
Rufiino-ta bwik
This structure has solved the difficulty associated with adjunct intervention, however it is
not without its own issues. The most significant problem with this structure is that it
reintroduces the problem of long passive, which was discussed in §5.3.1.
Recall that long passive involves the promotion of the embedded object of V149 to the
subject of the higher clause, and that this is only feasible when there is no intervening
argument position between that object and the matrix subject position. By requiring the
DP argument of -sime to move around VP1, we are positing that it moves past a
preceding argument position – that of the internal argument of V150. How then, can we
account for the fact that this preceding argument is not targeted for raising in this
Although I have used V1 and V2 throughout for simplicity’s sake, note that in example which contains
three verbal elements, the embedded accusative DP Rufinota, is both the semantic subject of the innermost
verb bwik- ‘sing’, but the ‘causee’ of -tua- ‘CAUS’, which is directly embedded under -sime/-saka.
If closeness is calculated in terms of m-command, then perhaps another solution is possible, since the
argument of -sime will m-command the argument of bwik-tua without being m-commanded by it.
However, m-command is not a simple relation to compute, and independent motivation for its exploitation
in grammar is fairly sparse.
instance? We know that the adjunct status of VP1 is not, in this case, a barrier to
extraction because embedded objects are subject to raising in the case of long passives.
Suppletion triggered within the phrase instead of (only) by
Bobaljik (2012) proposes a strict locality constraint on the conditioning of suppletive
arguments, such that suppletive vocabulary items can only be triggered by features within
the same maximal projection. Bobajik and Harley (2012) provide evidence to support this
claim, arguing that intransitive verbs in Hiaki whose suppletion is triggered by the
number features of the subject argument, such as siime, are unaccusative. This entails that
the subject argument is base generated as the internal argument of the verb, and thus in a
local sisterhood relation with the verb; hence, the suppletive verb form is licensed.
If ‘locality’ for the purpose of suppletion requires only that the triggering phrase be
generated within the same maximal projection of the verb, rather than specifically as the
sister of the verb, then it is feasible to suppose that the DP could be generated as a
specifier, rather than a complement, and still be capable of triggering suppletion.
Simon intok Hoan
In this structure, the subject DP is generated in Spec of √P, leaving the complement
position available for VP1, and suppletion is still triggered locally within the maximal
projection of the verb.
It must be noted at this point that the notion of ‘argument of √’ is not a universally
accepted one; in papers collected in a (2014) issue of Theoretical Linguistics devoted to
the topic of roots, Alexiadou, Borer, De Belder and van Craenenbroek all argue, for
various reasons, that a root is not a syntactic category, and therefore does not itself select
for arguments. Their position has been argued against, convincingly I believe, by Harley
(2014a) and Cuervo (2014), in the same volume, and by Bobaljik and Harley (2012); thus
I have assumed that position throughout. However, I have not thus far attempted to
distinguish whether a root may take more than one internal argument – that is, a specifier,
as well as a complement. Cuervo (2014) in reply to Harley (2014a) explicitly addresses
this question, pointing out that the arguments based on Hiaki suppletion data that Harley
(2014a) presents “provide evidence for the selection, combination and special relation of
roots with one argument: a complement/sister of the root” (Cuervo 2014:376).
Cuervo raises the following questions:
“Are there truly inherently dyadic roots? Are there roots that take a complement
but others take a specifier? How would a specifier of a root behave with respect
to locally conditioned phenomena, such as root suppletion?” (Cuervo 2014:377).
Cuervo’s own (2003, 2010, 2014) position is a nuanced one that distinguishes between
(at least two) types of unaccusative. Following Levin’s (1999) evidence regarding the
syntactic and semantic properties of ‘non-core’ transitives - that is, transitive verbs whose
objects have variable theta roles, not predictable from their status as direct objects.
Cuervo extends this discussion into unaccusative intransitives, and draws a similar
distinction between change-of-state intransitives, and predicates of happening or
movement (change-of-position in Harley’s (2014b) terms), whose arguments behave
similarly to the objects of Levin’s non-core transitives. Cuervo proposes that the
argument of a COS predicate is actually licensed as the specifier of stative vP. In
contrast, the argument of a COP predicate is the complement of the root. This distinction
fits well with the Hiaki data, since the set of suppletive intransitives are all verbs of
motion or stance – COS predicates do not supplete. (Harley et al. 2009; Bobaljik &
Harley 2012)
Incorporating the work of Levin, and others (Mateu & Acedo-Matellán 2012, e.g.)
Cuervo (2014:382) argues that a series of properties correlate with being the argument of
a root: being a complement, not being obligatory and not having a predictable, structural
meaning. She concludes, based on Harley’s evidence as well as the aforementioned
works, that roots can take only a complement, and not a specifier argument. What results
from this is the notion that the relation between a root and its complement, which is
maximally local, does not exist between a root and any other phrase in the structure.
Cuervo’s conclusions fit neatly with the Hiaki data, in providing an explanation for why
it is this particular set of verbs which show suppletion for number sensitivity, and not
other verbs regardless of their status as unaccusative or not.
The upshot of this is that if roots can only take a single argument - a complement - then
the tree in (292) cannot be correct. Both the Theme DP and VP1 cannot be internal
arguments of the root, and since the DP must be internal to √P in order to license
suppletion, then VP1 is necessarily left out in the cold. Thus, we are returned to our
original dilemma regarding linearization of VP1 in a position intervening between the
root and the DP which (must be) its sister.
5.4 Sorts and situations
5.4.1 Situating V-sime amongst Hiaki complex Vs
The analysis of the structure of V-sime that I have presented in §5.1 has some clear
differences from that of more well studied Hiaki complex Vs discussed in §3.1. Most
significantly, the V-sime construction - (285)-(286), reproduced here as (293)-(294) lacks a subject position in the embedded VP, whereas binding relations show that
complex Vs with transitive V2s – such as (98), reproduced as (295) and with the tree
structure provided in (296) - all have a lower subject position available, and therefore
must be VoiceP sized, rather than vP. In those structures, also, VP1 is the structural
complement of V2, rather than an adjoined element.
(293) Uu hamut ili usi-ta yu’u-sime
woman little child-ACC push-go.SG
“The woman is pushing the little child along.”
Uu hamut
ili usi-ta
(295) Heidi Art-ta
Heidii Artj-ACC
“Heidi teaches Art to take care of himself/*her.”
au j/*i
I do not claim, either, that the analysis presented here for V-sime constructions is
necessarily applicable to all constructions with an intransitive V2 - as shown in §3.1.1
there are at least two classes of intransitive V2s with different syntactic properties. I have,
rather, added to the range of understood structural configurations for Hiaki complex Vs.
It must be questioned, however, whether there is some principled way to understand
some of the differences in complex V construction types. It has been shown, in Chapter 3
that there is no syntactically useful line to be drawn simply between complex Vs with
bound V2s compared to those with free V2s. It is clear that a more nuanced approach is
One distinction that has been important for the current study is that between functional
and lexical items. Much of the literature on complex predication generally, and on motion
CPrs particularly have made mention of the 'semi-lexical' nature of the morphemes in
these kinds of constructions. For example, Cardinaletti and Guisti (2001) examined
motion in verb sequences in several language families, and concluded that the range of
properties are not consistent enough across languages to argue for a coherent class
existing between the lexical and functional poles. Instead, they argue that distinctions can
be understood by the insertion of lexical morphemes into functional nodes at different
levels in the structure (with lexical properties becoming progressively bleached the
higher the morpheme is inserted).
Butt (2010 etc) claims that light verbs, rather than being functional items distinct from
their lexical 'heavy verb' counterparts, are the same lexical item, with the difference in
interpretation being governed by the structural configuration - such as, for example
instantiating a functional category such as v0 - and this is the primary reason that she
contends that light verbs must, by definition, have a heavy counterpart.
In Hiaki, although it is tempting to assume that bound V2s are functional whilst free V2s
are lexical, there is clearly somewhat more to it than that. One determinant of a verbal
morpheme's lexical status within Hiaki is its ability to be reduplicated (Haugen & Harley
2013). In the discussion in §3.2.2 I showed that some bound V2s, such as causative -tua
are not reduplicable, whilst others, such as -pea ('inclination') are. Therefore we can
begin to establish a sort of feature matrix for V2s, as in (297) below.
+ bound
− bound
+ reduplication
− reduplication
Thus, while non-reduplicating bound V2s are functional, and must inhabit functional
nodes such as v0, lexical V2 may be either bound or free. This characterization runs
counter to Butt's claim about the obligatoriness of heavy counterparts. If we consider
inhabiting v0 to be the defining feature of 'light verbs' then by Butt's reasoning -tua
should have an independent verb counterpart, which it does not. However, if we consider
being a lexical item (a root) to be a necessary quality for a light verb, then -pea should
have an independent verb counterpart, which it does not. Either way, the stipulation that
having an independent counterpart is a defining feature of some syntactically significant
class is not upheld by the Hiaki data.
5.4.2 Situating Hiaki in the world of complex motion structures?
As was shown at length in Chapter 4 complex predicate structures inhabit a broad range
of structures across a wide variety of languages and language types. Efforts to distinguish
between CPr types have met with limited success - boundaries between construction
types are fuzzy at best, with several points of overlap. It seems that complex predicate
types may be better considered as a spectrum, rather than a clearly defined cline. As Seiss
(2009) points out, categories like SVCs may even subsume others, such as LVCs and
I have suggested that V-sime may be best labeled a Serial Verb Construction, largely
because of the lexical nature of both Vs in the compound, however it must be
acknowledged that this is a somewhat arbitrary distinction. For example, the directed
motion constructions discussed for Korean in §5.2.1 are typically also labeled SVCs,
although Zubizarreta and Oh (2004; 2007) analyze the V2 in these constructions as a
primarily functional element.
Another issue with labeling V-sime constructions SVCs (instead of, for example, LVCs)
particularly on the grounds that -sime is lexical, is that this means abandoning a single
classification for Hiaki complex Vs, since causative constructions, for example, do not
involve a lexical V2. Although it is, in principle, not inconceivable for a language to
contain more than one type of CPr, it does obscure the commonalities that V-V
compounds share that distinguish them from other kinds of multi-verb constructions in
the language, such as conjoined or subordinated clauses, and participial verbs used as
5.5 Summary
In response to the three questions investigated in this chapter, I have drawn the following
Question 1:
V2-sime displays attributes best associated with a lexical, rather than a
functional item, and occupies √0.
Question 2:
The VP1 constituent in a V-sime construction must be a vP, but crucially
cannot include a VoiceP layer (nor an external argument position)
Question 3:
The structure of V-sime compounds is involves adjunction of phrasal VP1
to the V2 head √-SIME.
The conclusion to this third question is, admittedly, a controversial claim, however I
believe that it is the best representation of the morphological and semantic characteristics
of the V-sime constructions.
5.6 Other remaining problems, issues, directions for future work
Several questions have arisen in the course of this study which I have been unable to
answer satisfactorily within the scope of the current work, but which are deserving of
further attention.
The first of these is the issue of reduplication and its position in the clausal structure.
Haugen & Harley (2013) analyze reduplication, in at least those cases where it results in a
habitual reading, as occupying the head of AspP, above external argument introducing
VoiceP. However, the evidence in §3.2.2 supports an analysis of Hiaki verbal
reduplication as broadly pluractional, with 'habitual aspect' merely one of a number of
pragmatic interpretations of this pluractional feature. Furthermore, there is the fact that
reduplication can apply to an embedded VP1 in a V-sime structure. Since these embedded
VPs can be shown to be smaller than VoiceP, it is not feasible to assume that they include
an AspP layer. Recall, also, that in §5.4.1 I pointed out that V2 elements such as
causative -tua, which are functional – that is, inhabit v0 – are not reduplicable.
Reduplication only applies to lexical items; ie, it is roots, specifically, which can host the
prefixal reduplicant.
Reduplication is a clearly productive inflectional process. In most cases, its interpretation
(habitual, in progress, intensification, etc) can be fairly clearly linked to the lexical
semantic properties of the verb root. The fact that a reduplicated verb, whose lexical
semantics is compatible with a habitual interpretation in isolation, cannot get that
interpretation when it appears in V1 position is a compelling puzzle. The first avenue of
investigation into this issue might be to re-examine V1 reduplication in other types of VV compounds that do not involve V2-sime. We know, for instance, that VP1 in other
compounds has a larger structure, including at least a VoiceP layer, in addition to a
different structural relationship with the V2 (complement instead of adjunct)51. If V1s in
these types of structures can be interpreted habitually, this might tell us either something
further about the domain of aspectual interpretation.
This structure is shown in (296).
Another issue of concern, particularly given the analysis presented here, is how to explain
the non-literal motion interpretation of -sime in the specific instance when a stance V1 is
compounded with a reduplicated -sime stem: VSTANCE-RED-sime. Compare the literal
motion reading of (298) with its reduplicated counterpart in (299):
(298) Ume o’ow-im kamion-po hap-saka
man-PL bus-LOC stand-go.PL
“The men are standing in the (moving) bus.”
(299) Ume o’ow-im kamion-po hap-sa-saka
man-PL bus-LOC stand-RED-go.PL
“The men are standing/milling about on the bus.” (bus no longer nec. moving)
Since -sime resists non-literal interpretations in almost every other instance, even in the
case of spatial paths or directions, it is curious to find this one very specific context for a
non-literal reading. My analysis, as it currently stands, does not account for this behavior.
It is, perhaps, conceivable that this behavior represents some early, nebulous step along
the path towards grammaticalization into a functional item, although considerably more
investigation into the stages of grammaticalization would be needed to determine if this is
a feasible proposition.
Another issue that deserves more thorough explication involves the aspectual semantics
of Hiaki and its interaction with the independent verb siime (and potentially other verbs
also). As I showed in §2.1.2 there are some interpretive idiosyncrasies which are
currently unexplained and which are deserving of more focused attention than was
possible within the confines of the current project. For example, the past imperfective
form gets an unexpected interpretation of ‘it was intended, but didn’t actually occur’, as
in (52), reproduced as (300).
(300) Aapo
3SG.NOM 1sg.ACC-to towards go.SG-PST
“S/he was (going to be) coming towards me…”
Not “S/he was (in the process of) coming towards me”
And, unlike most verb paradigms, siime does not obviously take the suffix -kan, which
usually gets something like a past perfect reading. Instead, we get an idiosyncratic form,
where the perfective stem takes a past tense suffix. It is this form that gets the
interpretation, expected in the previous example, of ‘(event) was in progress’.
(301) Aapo
3SG.NOM 1sg.ACC-to towards go.SG.PRF-PST
“S/he was (in the process of) coming towards me”
Not “S/he had been coming towards me”
All of these issues are at least potentially connected, dealing either with reduplication,
aspect, or both. For example, clarifying the role of reduplication in different contexts
could provide insight into the unexpected non-literal interpretations of reduplicated -sime
with stance V1s. The relationship between aspectual semantics and morphological
processes like reduplication (and perhaps suppletion) should also be investigated. Future
work, therefore, might focus on deeper understanding of Hiaki aspectual semantics in
general, in order to provide the necessary background information to properly illuminate
the questions raised here.
Alexiadou, Artemis. (2014). “Roots don’t take complements”. Theoretical Linguistics;
40(3/4): 287 – 297
Alsina, A., Bresnan, J. W., & Sells, P. (Eds.). (1997). Complex predicates. Center for the
Study of Language and Information.
Amberber, M., Baker, B. and Harvey, M. (Eds.) (2010) Complex Predicates: Cross-linguistic
Perspectives on Event Structure. Cambridge: CUP.
Anderson, G. D. (2006). Auxiliary verb constructions. Oxford University Press.
Baker, M. (1988). “Theta Theory and the Syntax of Applicative Constructions in
Chichewa,” Natural Language and Linguistic Theory 6, 353-389.
Berman, R. and Slobin, D. (1994) Relating Events in Narrative: A Crosslinguistic
Developmental Study. New Jersey: Lawrence Erlbaum Associates Inc.
Bobaljik, J. D. (2012). Universals in Comparative Morphology: Suppletion, superlatives, and
the structure of words (Vol. 50). MIT Press.
Bobaljik, J. D., & Harley, H. (2012). Suppletion is local: Evidence from Hiaki. Ms.,
University of Connecticut and University of Arizona.
Borer, Hagit. (2014). “Wherefore roots?” Theoretical Linguistics; 40(3/4): 343 – 359
Bowern, C. (2008). The diachrony of complex predicates. Diachronica, 25(2), 161-185.
Butt, M. (2003). The light verb jungle. In Workshop on Multi-Verb Constructions 1-28.
Butt, M. (2010). The light verb jungle: still hacking away. Complex predicates: crosslinguistic perspectives on event structure, 48-78.
Butt, M, and G. Ramchand. 2005. Complex Aspectual Structure in Hindi/Urdu. In The
Syntax of Aspect, Nomi Ertishik-Shir and Tova Rappaport (Eds). 117–153. Oxford:
Oxford University Press.
Carnie, Andrew. (1995). Head Movement and Non-Verbal Predication. Dissertation.
Massachussetts Institute of Technology.
Carnie, Andrew. (2000). On the definition of X0 and XP. Syntax 3 2: 59–106.
Castile, George. (2002) Yaquis, Edward H. Spicer, and Federal Indian Policy: From
Immigrants to Native Americans. Journal of the Southwest, 44(4), 383-435.
Chomsky, Noam. (1981). Lectures on Government and Binding. Foris, Dordrecht.
Chomsky, N. (1995). Bare Phrase Structure. In: Campos, H & P. Kempchinsky (eds).
Evolution and Revolution in Linguistic Theory, Georgetown University Press,
Washington DC, pp 51-109.
van Craenenbroeck, Jeroen. (2014). “On diagnosing complement-taking roots.” Theoretical
Linguistics; 40(3/4): 361 – 373
Croft, W., J. Barddal, W. Hollmann, V. Sotirova & Taoka, C. (2008). Revising Talmy's
typological classification of complex events. Unpublished ms.
Cuervo, Maria Cristina. (2014). “Arguments for a Root.” Theoretical Linguistics; 40(3/4):
DeBelder, Marijke. (2014). “Against Conflation.” Theoretical Linguistics; 40(3/4): 331 – 342
Dedrick, J. M. and Casad, E. H. (1999) Sonora Yaqui language structures. Tucson:
University of Arizona Press
Escalante, Fernando. (1990) Voice and Argument Structure in Yaqui. Ph.D. Dissertation. The
University of Arizona.
Fernández, Zarina Estrada et al. (2004) Diccionario yaqui-español y textos: Obra de
preservación lingüística. Hermosillo, Sonora, Mexico: Universidad de Sonora;
Mexico City: Plaza y Valdés Editores.
Folli, R., Harley, H., & S. Karimi. (2005). Determinants of event type in Persian complex
predicates. Lingua, 115(10), 1365-1401.
Folli, R and H. Harley. (2006). On the licensing of causatives of directed motion: Waltzing
Matilda all over. Studia Linguistica 60 2: 121–155.
Giusti, A. C. G. (2000). “Semi-lexical” motion verbs in Romance and Germanic. Working
Papers in Linguistics, 10(2).
Guerrero, Lilian. (2004). The syntax-semantic interface in Yaqui complex sentences: Arole
and reference grammar analysis. Doctoral dissertation, University of Buffalo.
Haugen, J. and H. Harley. (2013) On the implications of head-marking inflection for the
architecture of grammatical theory: Evidence from reduplication and compounding
in Hiaki (Yaqui). In A. Fountain, M. Miyashita, D. Cole and S. Bischoff, (eds), The
persistence of language: Constructing and confronting the past and present in the
voices of Jane H. Hill. pp 133-174 Amsterdam: John Benjamins.
Harley, H., Tubino Blanco, M., & J. Haugen. (2009). Applicative constructions and
suppletive verbs in Hiaki. Rice Working Papers in Linguistics.
Harley, H., M. Tubino Blanco, Jason Haugen. (2014) Affixal light verbs and complex
predicates in Hiaki. In Current issues in complex predicate research, edited by
Hideki Kishimoto and Yoko Yumuto. 257-290. Tokyo: Hituzi Syobo
Harley, H. and A. Trueman. (2012). "Hiaki pronominals and the typology of deficiency." In
Proceedings of WAIL 13, University of California Santa Barbara, 40-54. Available at
Harley, H. and M. Leyva. (2009). "Form and meaning in Hiaki (Yaqui) verbal reduplication."
International Journal of American Linguistics, 75.2, 233-72.
Harley, Heidi. (2004). Merge, Conflation and Head Movement: The First Sister Principle
revisited. In: Proceedings of NELS 34, U. Mass Amherst: GSLA
Harley, Heidi. (2008). On the causative construction. In: The handbook of Japanese
linguistics, Shigeru Miyagawa and Mamoru Saito (Eds); OUP.
Harley, Heidi. (2011) Morphosyntax of Hiaki. Class notes: Summer Institute of Linguistics
at the University of Colorado, Boulder.
Harley, Heidi. (2013) External arguments and the Mirror Principle: On the distinctness of
Voice and v. Lingua.
Harley, Heidi. (2014a). “On the Identity of Roots”. Theoretical Linguistics, 40(3-4), 225-276.
Harley, Heidi. (2014b). Aggressive Passives: Voice and Implicit Arguments in Hiaki. Paper
presented at the 45th annual meeting of NELS, MIT, October 31.
Jackendoff, Ray. (1992). Babe Ruth Homered his Way into the Hearts of America. In: T.
Stowell and E. Wehrli, (eds.), Syntax and Semantics, Vol. 26, Academic Press, 155178.
Jackendoff, Ray. (1997). Twistin' the Night Away. Language 73, 534-559.
Jelinek, Eloise Yaqui Language Workbook. Ms. The University of Arizona.
Jelinek, E. and H. Harley. (2014). Impersonal Agreement in a Non-Agreement Language:
The Hiaki Impersonal Construction. In Pronouns, Presuppositions, and Hierarchies:
The Work of Eloise Jelinek in Context. Carnie, A., & Harley, H. (Eds.). Routledge.
Johnson, Jean Bassett. 1962. El Idioma Yaqui. Mexico DF: Instituto Nacional de
Antropologia e Historia.
Jung, Hyun Kyoung. (2014). On the syntax of applicative and causative constructions.
(Doctoral dissertation, University of Arizona). (Accession order No. 3621440).
Laughren, Mary. (2010) Warlpiri verbs of change and causation: the thematic core. In
Amberber, Baker, Harvey (eds), Complex Predicates (CUP)
Levin, B., and Rappaport-Hovav, M. (1991). Wiping the slate clean: a lexical semantic
exploration. Cognition 41, 123–151.
Levin, B., and Rappaport-Hovav, M. (1995). Unaccusativity: At the Syntax-Lexical
Semantics Interface. Linguistic Inquiry Monograph 26. Cambridge: MIT Press.
Lindenfield, Jacqueline. (1979) Yaqui Syntax. University of California Press.
Mateu, Jaume: (2001). Small Clause Results Revisited. In N. Zhang (ed.), ZAS Papers in
Linguistics 26: Syntax of Predication, Zentrum fu¨ r Allgemeine Sprachwissenschaft,
Mateu, Jaume. (2002). Argument Structure: Relational construal at the syntax-semantics
interface. Unpublished Ph. D. dissertation, UAB, Bellaterra.
Matushansky, Ora. (2006). Head-movement in linguistic theory. Linguistic Inquiry 37.1.
Mohanan, Tara. "Grammatical Verbs (with Special Reference to Light Verbs)." The
Blackwell Companion to Syntax. Everaert, Martin and Henk van Riemsdijk (eds).
Blackwell Publishing.
Molina, F., Shaul, D. & Valenzuala, H. (1999). Yoeme-English English-Yoeme
Standard Dictionary. New York: Hippocrene Books.
Polinsky, M. (2013). Raising and control. The Cambridge Handbook of Generative Syntax,
Marcel den Dikken (Ed).
Ramchand, G., & Svenonius, P. (2002). The lexical syntax and lexical semantics of the verbparticle construction. In Proceedings of WCCFL (Vol. 21, pp. 387-400).
Sebastián, E., & Slobin, D. (1994). Development of linguistic forms: Spanish. Relating
events in narrative: A crosslinguistic developmental study, 239-284.
Sells, P. (1998). Structural relationships within complex predicates. In Proceedings of ICKL
11. Seoul, International Circle of Korean Linguists.
Seiss, M., Butt, M., & King, T. H. (2009). On the difference between auxiliaries, serial verbs
and light verbs. In Proceedings of the LFG09 Conference (pp. 501-519).
Slobin, Dan. (2004) The many ways to search for a frog: Linguistic typology and the
expression of motion events. In S. Strömqvist & L. Verhoeven (Eds.) (2004),
Relating events in narrative: Vol. 2. Typological and contextual perspectives (pp.
219-257). Mahwah, NJ: Lawrence Erlbaum Associates.
Simpson, Jane. (1991). Warlpiri morpho-syntax: A lexicalist approach. Kluwer academic
Simpson, Jane. (2001) Preferred Word Order and the Grammaticization of Associated Path.
In Butt, M. & King, T. (Eds). Time Over Matter: Diachronic Perspectives on
Morphosyntax. Stanford; CSLI Publications.
Snyder, William. (2001) On the Nature of Syntactic Variation: Evidence from Complex
Predicates and Complex Word-Formation. Language 77:324-342.
Son, M., & Svenonius, P. (2008). Microparameters of cross-linguistic variation: Directed
motion and resultatives. In Proceedings of the 27th west coast conference on formal
linguistics (pp. 388-396).
Svenonius, P. (Ed.). (2002). Subjects, Expletives, and the EPP. Oxford University Press.
Svenonius, P. (2008). Complex predicates and the functional sequence. Nordlyd, 35(1).
Talmy, L. (1975) Semantics and syntax of motion. In Kimball, J.P. (ed.) Syntax and
Semantics (Vol. 4). New York: Academic Press. 181-238.
Talmy, L. (1985) Lexicalization patterns: semantic structure in lexical forms. In Shopen, T.
(ed.) Language typology and syntactic description III: grammatical categories and
the lexicon. Cambridge: Cambridge University Press. 57-149.
Talmy, L. (1991) Path to realisation: a typology of event integration. In Buffalo Papers in
Linguistics. 91. 147-187.
Talmy, L. (2000) Toward a Cognitive Semantics. Cambridge, MA: MIT Press.
Tomioka, Naoko. (2004). The lexicalization patterns of verbs and V-V compounds. In D.
Beermann and L. Hellan (eds.), Proceedings of the 26th Annual Boston University
Conference on Language Development. Somerville, MA: Cascadilla Press.
Tomioka, Naoko. (2006). The morphology of transitivity. Toronto Working Papers in
Linguistics 26: 115-122.
Tubino-Blanco, Mercedes and H. Harley. (2010) "Dos tipos de base verbal en Hiaki (yaqui)".
In Análisi lingüístico: enfoques sincrónico, diacrónico e interdisciplinario, ed. by
Rosa María Ortiz Ciscomani, p. 97-128. Editorial Unison, Hermosillo, Son., Mexico
Wittenberg, E., Jackendoff, R., Kuperberg, G., Paczynski, M., Snedeker, J., & Wiese, H
(2014). The processing and representation of light verb constructions. Structuring the
Argument: Multidisciplinary research on verb argument structure, 10, 61.
Wurmbrand, S. (1998). Restructuring control. U. Penn Working Papers in Linguistics, 5,
Wurmbrand, S. (2001). Infinitives: Restructuring and clause structure. Mouton de Gruyter.
Wurmbrand, S. (2002). Syntactic versus semantic control. In Studies in Comparative
Germanic Syntax: Proceedings from the 15th Workshop on Comparative Germanic
Syntax (Vol. 53, p. 93). John Benjamins Publishing.
Wurmbrand, S. (2004). Two types of restructuring—Lexical vs. functional. Lingua, 114(8),
Wurmbrand, S. (2007). How complex are complex predicates? Syntax, 10(3), 243-288.
Wurmbrand, S. (2013). Complex predicate formation via voice incorporation. Ms., University
of Connecticut, Storrs.
Zubizarreta, Maria Luisa, and Oh, Eunjeong. (2004). The lexicon-syntax interface: The case
of motion verbs. Ms., University of Southern California, Los Angeles.
Zubizarreta, Maria Luisa, and Oh, Eunjeong. (2007). On the Syntactic Composition of
Manner and Motion. Linguistic Inquiry Monograph 48. Cambridge: MIT Press.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF