Ergativity: Argument Structure and Grammatical

Home , Labile verb, Syntactic pivot

ERGATIVIT Y ARGUMENT STRUCTURE

AND GRAMMATICAL RELATIONS

a dissertation

submitted to the department of linguistics

and the committee on graduate studies

of stanford university

in partial fulfillment of the requirements

for the degree of

doctor of philosophy

Christopher D Manning December

I certify that I have read this dissertation and that in my

opinion it is fully adequate in scop e and in quality as a

dissertation for the degree of Do ctor of Philosophy

Joan Bresnan

Principal Adviser

I certify that I have read this dissertation and that in my

opinion it is fully adequate in scop e and in quality as a

dissertation for the degree of Do ctor of Philosophy

Ivan A Sag

I certify that I have read this dissertation and that in my

opinion it is fully adequate in scop e and in quality as a

dissertation for the degree of Do ctor of Philosophy

Peter Sells

I certify that I have read this dissertation and that in my

opinion it is fully adequate in scop e and in quality as a

dissertation for the degree of Do ctor of Philosophy

Mary Dalrymple

Approved for the University Committee on Graduate

Studies iii

Abstract

This dissertation considers the prop er treatment of syntactic ergativity arguing for

a framework that decouples prominence at the levels of grammatical relations and

argument structure The result is two notions of sub ject grammatical sub ject and

argument structure sub ject as in Schachter and Guilfoyle Hung and Travis

and a uniform analysis of syntactically ergative and Philippine languages

Both these language groups allow an inverse mapping in the prominence of the two

highest terms b etween argument structure and grammatical relations A level of argu

ment structure although app earing in much recent work is shown to b e particularly

well motivated by the examination of ergative languages A study of Inuit Tagalog

Dyirbal and other languages shows that constraints on imp erative addressee and

controllee selection antecedent of anaphors and the controller of certain adverbial

clauses are universally sensitive to argument structure Thus these phenomena are

always accusative or neutral and we can explain why passive agents and causees

can generally bind reexives However constraints on relativization topicalization

fo cussing or questioning sp ecicity or wide scop e coreferential omission in co ordina

tion etc are shown to b e universally sensitive to grammatical relations Examining

just these phenomena which are sensitive to grammatical relations we see that many

languages are indeed syntactically ergative and so this option must b e countenanced

by linguistic theory iv

Acknowledgments

My choice for a dissertation topic can b e traced to an interest in ergativity that was

sparked during my time as an undergraduate at the Australian National University

More recently Bob Dixon suggested to me that I should write a thesis on how well

or badly mo dern syntactic theories fare in treating ergative languages However

I havent altogether done that although much of this thesis b ears on that issue

mainly b ecause it would have b een dicult to start writing on such a topic without

having rst sorted out my own views on the typology of ergative languages Having

started with Australia let me complete my thanks for that part of the world Ive

appreciated having Avery Andrews as an email corresp ondent and o ccasional coau

thor Ive also b een lucky enough to meet various other Australian linguists while at

Stanford mainly ANU graduates of an earlier time and so I am in the unexp ected

p osition of knowing more Australian linguists now than when I b egan at Stanford

If the choice of topic still harks back to my Australian education the treatment

of it b ears much to my Stanford education I had a wonderful four years at Stanford

I also remember the climate fondly as winter b egins to descend in Pittsburgh The

p eople on my dissertation committee were among those that contributed most to

my time at Stanford Joan Bresnan was welcoming right from when we rst met

provided brilliant and inspiring lectures organised workshops and discussion groups

at Stanford and did everything she could to assist me in writing this dissertation

in less than ideal circumstances I think p erhaps my biggest debt to Ivan Sag is for

helping me to b ecome part of the larger linguistics community but I learned a lot else

from Ivan over the years including much ab out how to do and present research Peter

Sells is legendary among Stanford students for the quantity and quality of help he

gives to students b oth individually and in lectures and I appreciate what I received v

even more now that I can see how busy life is on the other side of the fence Thanks

also to the many others at Stanford from whom I to ok courses and seminars

The nal member of my committee was Mary Dalrymple who I thank not only

for her many comments on the content and organization of this thesis but also for

her advice help and friendship throughout my time at Stanford Most of my contact

with Mary was not actually at Stanford but at Xerox PARC I was very fortunate

to receive a Xerox internship for the summer of and then to have a continuing

asso ciation with Xerox PARC Not all that much of what I learned at PARC app ears

in this thesis but it has b een exceedingly useful in other places not least in my new

job I learned ab out many things from Ron Kaplan including LFG theoretical com

putational linguistics and the history of computing He also always dealt eciently

with nding ways to pay me despite my irregular schedule Others that contributed

to the quality of my time at PARC include Jeanette Figueroa Marti Hearst Julian

Kupiec John Maxwell Hinrich Schutzeand Hadar ShemTov

Nearly all the examples in this dissertation come from published sources and thus

this dissertation would not have b een p ossible at all without the dedicated eldwork

of others A dozen or more examples are taken from the work of each of Maria

Bittner Bob Dixon Michael Fortescue Martin Hasp elmath Paul Schachter and

Jerry Sado ck and so they deserve sp ecial thanks although I am also grateful to the

many other p eople on whose work I have drawn

Thanks also to all the other p eople who help ed me to write this dissertation

Edna Paneatak MacLean willingly discussed I nupiaq with me despite the fact that

we had to go slowly b ecause of my p o or knowledge of the language Brett Kessler

corrected and help ed me gloss the Sanskrit examples G uven G uzeldereokayed some

Turkish examples MaraEugeniaNi nowas on hand when needed to provide Spanish

judgments and Tagalog references Maria Bittner Miriam Butt Martin Hasp elmath

and Jerry Sado ck discussed various asp ects of their and my work with me by electronic

and regular mail

Not everyone help ed sp ecically with the thesis More general thanks to the others

in my year Lynn Cherny Hyewon Choi Yookyung Kim Hinrich Schutzeand Hadar

ShemTov May they have the b est of luck with their dissertations Particular thanks vi

to Hyewon for b eing my rst friend at Stanford and for submitting this dissertation

for me Even more particular thanks to Jane for her love and help This thesis is

longer than Janes so I must have disturb ed her more than she disturb ed me Thanks

also to our friends in particular Jill Jennifer and Pollo In Gina and Michelle I was

fortunate to have some of the nicest administrative sta around And nally to my

family not all of whom will b e able to read this vii

Transcriptions Abbreviations and Conventions

This section describ es conventions and abbreviations that I have used I have tried to

make this dissertation useful as a reference do cument There is an index of languages

and topics although I should stress that it is incomplete The bibliography also

acts as an index of citations at the end of each entry is a list that gives the page

numbers where the work is cited At the rst mention of each language strictly the

rst mention from Section on outside of fo otnotes I list in brackets its family

aliation and where it is sp oken

Almost all the examples in this thesis are drawn from previously published sources

The source of each example is given in the App endix Sources of Examples Refer

ences are to the example number of the cited work where available in the form

chex when examples are numbered separately within each chapter otherwise to

the page number Most examples app ear using the transcription conventions of my

source Some attempts have b een made to make transcriptions and the glossing of

grammatical formatives more consistent as outlined b elow

All Dyirbal examples are transcrib ed using a form of the practical orthography

now in widespread use by Australianists including Dixon The corre

sp ondences with the system employed by Dixon are ny N j d r

and rr r However the name Dyirbal is not written in the practical orthography

where it would b ecome Jirrbal

The transcription of Inuit examples is not completely consistent Some West

Greenlandic examples are in the ocial orthography while others are in a pure phone

mic variant thereof the new orthography continues to distinguish ie uo and vf

although these dierences are not phonemic Examples from other varieties of Inuit

are transcrib ed as in their source viii

Most Lezgian examples use the transliteration system of Hasp elmath but

some examples from Mel cuk follow his transcription conventions

Mayan examples app ear in the practical orthography of Terence Kaufmann the

most unusual feature of which is that is used for glottal stops

Tagalog examples are shown using the style and glossing of Kro eger in

particular case markers are attached to the following word with an equals sign

which indicates cliticization

The following abbreviations are employed in the glosses

rst second third excl Exclusive

p erson

equ Equalis case

fourth p erson Inuit

erg Ergative case

anaphoric form

freq Frequentative

i i i i i i iv noun class markers

fut Future

A Absolutive agreement

gen Genitive case

abl Ablative case

ger Gerundive

abs Absolutive case

immed Immediately following

acc Accusative case

event

antip Antipassive

impv Imp erative

asp Asp ect

incept Inceptive

aux Auxiliary

ind Indicative mo o d

av Active Voice

inf Innitive

caus Causative

instr Instrumental case

cmpltv Completive asp ect

interrog Interrogative

cond Conditional

intr Intransitive ending

cop Copula

iv Instrumental voice

dat Dative case

def Denite lnk Linker

dep Dep endent asp ect

loc Lo cative

dir Directional

masc Masculine gender

dm Determinate

mod Mo dalis case

dirs Directional Sux

msd Masdar event

dv Dative Voice

nominalization verb

form

E Ergative agreement

emph Emphatic marker neg Negative ix

psub Past sub ordinate mo o d nfut Nonfuture

nom Nominative case

ptcl Particle

nomlz Nominalizer

ques Question marker

O Ob ject agreement

rec Recent past tense

ogen Oblique Genitive

rel Relativizer Relative

extension

Case

ov ob jective voice

S Sub ject agreement

part Participle

sg Singular number

pass Passive

subj Sub ject agreement

part Participle

with

pat Patient

term Terminalis case

perf Perfective

th There series

pl Plural number

demonstratives

pm Prop er name marker

tm determinate Topic

purp Purp osive

Marker

pres Present tense

tns Tense

pret Preterite

tr Transitive ending past Past tense

In glossing the agreement of Eskimo transitive verbs agreement with the A argu

ment is always listed rst in the gloss and agreement with the O argument is listed

second even when this do esnt corresp ond with the historical order of morphemes

for which see Section

The following abbreviations are used for grammatical frameworks

ALS Autolexical Syntax Sado ck

GB GovernmentBinding Theory Chomsky Chomsky

GPSG Generalized Phrase Structure Grammar Gazdar et al

HPSG Headdriven Phrase Structure Grammar Pollard and Sag

LFG Lexical Functional Grammar Bresnan a

RG Relational Grammar Perlmutter x

Contents

Abstract iv

Acknowledgments v

Transcriptions Abbreviations and Conventions viii

Cutting the ergativity pie

What is ergativity

Anderson

Dixon

The scarcity of syntactically ergative languages

Towards a new typology of mixedpivot languages

Tagalog

Inuit

Notions of Sub ject

Historical Origins

Word order a necessary digression

Other languages for which an inverse analysis seems correct

Mayan languages

Chukchi

Toba Batak

Tsimshian languages

Nadeb

Summary xi

Theoretical Foundations

Grammatical relations and argument structure

Approaches to ergativity

Argument pro jection linking and valence changing op erations

HPSG

The four underlying primitives mo del

An analysis of binding

Binding in Tagalog

Binding in Inuit

Echoes in Accusative Languages

Conclusion

Inuit West Greenlandic

Basic background on Inuit

Genetic Aliation

Case Marking and Case Markedness

Word Order

Termhoo d Passive and Antipassive

Arguments for the absolutive b eing a surface pivot

Relative clauses

Semantic scop eSp ecicity

Nominalizations

Co ordination

Order of agreement suxes on verbs

Conclusion

Phenomena sensitive to a level of argument structure

Derivational morphology

The Inuit Innitive

Binding Phenomena

Imp eratives

Expressive adjectival p ostbases xii

Summary

Approaches to Inuit

Johnson on Central Arctic Eskimo

Marantz and Levin

Three GB analyses

Bobaljik and his precursors

Sado ck forthcoming

Lezgian

Basic Background on Lezgian

Genetic Aliation

Case Marking and Case Markedness

Agreement

Word Order

Termhoo d

Valence changing morphology

Phenomena exp ected to b e sensitive to argument structure

Binding Theory

Imp eratives

Controlled adverbial clauses

Control of innitival complement clauses

Idioms

Is there a surface pivot

The case for an SO pivot Mel cuk

The case against an SO pivot

Sometimes pivotsensitive phenomena that are neutral

The evidence for a weak SA pivot

Concluding discussion

Dyirbal

Evidence compatible with an SO pivot

An apparent problem xiii

The Oblique Analysis of Dyirbal

Observations p erhaps not requiring structural explanations

Ergative case marking in syntactically accusative languages

Absence of syntactically ergative morphologically accusative

languages

Absence of antipassive in syntactically accusative languages

Conclusions

Sources of Examples

Bibliography

Index xiv

CHAPTER

Cutting the ergativity pie

This dissertation considers the prop er treatment of some of the phenomena that

have led languages to b e considered ergative Dixon Comrie At present

most such languages have b een only sparsely studied and many fundamental ques

tions in their analysis seem at b est incompletely answered Thus the dissertation

tries to concentrate on such basic issues as when ergativity should b e analyzed as

syntactic or morphological whether there is a division into two classes of syntacti

cally and morphologically ergative languages and if so where should it b e drawn

whether ergative arguments are always core roles or not and so on

The plan of the dissertation is as follows The rst half of this chapter introduces

the lines of the analysis that I will prop ose within a broad typological sweep This sec

tion includes some review of standard approaches to ergativity from the structuralist

functionalist literature The second half discusses some p ossible mo dels of ergativity

within the tradition of generative grammar and briey introduces the framework that

I will b e assuming

The next chapter examines Inuit Eskimo It shows how the main language

phenomena supp ort a division b etween grammatical relations and argument structure

with Inuit showing an inverse mapping b etween the two The chapter also serves a

second function since so much of the recent generative literature on ergativity deals

in whole or in part with Inuit I will continue to survey the literature from the vantage

p oint of Inuit

Chapter Cutting the ergativity pie

Chapter Three then turns to the analysis of Lezgian While Inuit is a dicult

case b ecause it is what Dixon describ es as a mixed pivot language Lezgian

is problematic b ecause there is little evidence for surface grammatical relations at all

Chapter Four briey considers the placement of Dyirbal within the typology that

I have b een developing and summarizes the conclusions of my study

The dissertation views language from a basically lexicalist p ersp ective It isnt

a lengthy defense of lexicalism for which see among others Bresnan and Mchombo

and Sells but adopting a lexicalist p ersp ective is an imp ortant part

of the argument and is discussed esp ecially in Section The dissertation also

attempts to describ e grammatical phenomena by factoring constraints onto dierent

levels in the tradition of LFG rather than trying to treat everything in terms of

the congurational relationships dened on a single tree structure as in much GB

work The bulk of the dissertation ends up b eing ab out the interplay b etween a

level of grammatical relations and a level of argument structure Questions of phrase

structure and word order have seemed less cogent to the issues that I concern myself

with or at least I have not seen how to make sucient use of them

What is ergativity

It is a longstanding observation that many languages exhibit not the pattern of case

marking known from the ma jor Europ ean languages but rather an ergative pattern of

case marking as shown for Burushaski isolate Pakistan in and West Greenlandic

EskimoAleut Greenland in

a ne hre phalo boki

themasc manerg seedplabs sowpretsgmascsubj

The man planted the seeds

b ne hir yalti

themasc manabs yawnpretsgmascsubj

The man yawned

a Olip neqi nerivaa

Olierg meatabs eatindtrsgsg Oli eats meat

Chapter Cutting the ergativity pie

b Oli sinipp o q

Oliabs sleepindintrsg

Oli sleeps

In such languages the more patientlike argument of a transitive verb app ears in

the same absolutive case as the single argument of an intransitive verb while the more

agentlike argument of a transitive verb is marked dierently in what is known as the

ergative case This contrasts with the accusative system of case marking illustrated

by Latin ancient IndoEurop ean Italy in

a puella venit

girlnom comepresindsg

The girl comes

b puer puellam audit

b oynom girlacc hearpresindsg

The b oy hears the girl

To avoid constant use of clumsy lo cutions like the more patientlike argument of

a transitive verb I will use for descriptive purp oses the abbreviations introduced by

Dixon whereby A represents the agentlike argument of a transitive verb

O the patientlike argument of a transitive verb and S the single actant of intransitive

verbs The accusative and ergative systems of case marking then pick out the classes

shown in

A Ergative

Nominative

Absolutive

Accusative

I will always refer to the two direct cases in an ergative language pretheoretically as

Ergative and Absolutive a term ultimately from the Eskimo literature Thalbitzer

Many prefer to call the Absolutive case Nominative in part b ecause b oth are

usually unmarked as in the examples ab ove but that is a theoretical decision whose validity I will discuss

Chapter Cutting the ergativity pie

From a basis in case marking the term ergative has b een extended to other sub

systems of language verbal agreement p otentially word order syntactic and even

discourse prop erties that treat S and O in one way and A in another The term

accusative is similarly used when a subsystem treats S and A in one way and O in

another These are not the only two groupings that app ear in language but they are

the two most common ones Others will b e introduced later

Before pro ceeding let me introduce a little more basic terminology that I will

use throughout Most grammatical theories distinguish b etween the basic means by

which the actants of verbs are expressed and more sp ecialized means which have

denite semantic restrictions I will refer to the former as core roles or terms and

the latter as oblique roles In casemarking languages terms are most commonly

marked by the basic cases of the language variously called direct or structural cases

whereas obliques and sometimes some terms are marked by oblique cases which

have clear semantic content I will sometimes use the expression displaced term for

an argument which in some verbal diatheses app ears as a term but which in a certain

conguration has b een demoted or has b een unable to b e linked to a term such as

the agent of passive constructions and the dative or oblique causee in the Romance

languages

Returning to our initial examples the question is what signicance this dierent

pattern of morphological marking has for the analysis of ergative languages Unlike

in Latin the sub ject of traditional grammar the class of S and A NPs is no longer

consistently marked by a single case Do these dierent morphological markings

show that the languages concerned have a fundamentally dierent syntactic charac

ter from familiar accusative languages This was widely assumed by grammarians

in the premo dern era Various analyses could b e viewed as denying there are any

grammatical relations in an ergative language or suggesting that there is only one

general grammatical relation of mo dier but the most common view was that the

ergative is something like a passive This analysis holds that the O b ears the same

It is hard to precisely characterize this distinction Many have noted that ergative cases have

semantic correlates but then some p eople such as Wierzbicka have suggested that all

cases including nominative and accusative have meanings Nevertheless I feel that there is a basic distinction that has to b e made

Chapter Cutting the ergativity pie

grammatical relation as the S of an intransitive as in the English passive while the

A has a dierent grammatical relation This allows one to maintain a simple link

b etween morphological form and syntactic function which seems desirable ceteris

paribus and indeed most traditional work was averse to recognizing syntactic re

lations that were not overtly marked However it complicates the statement of the

mapping b etween semantics and syntax where it is often assumed that universally

an agent argument should b e mapp ed onto the sub ject p osition in the basic verbal

voice

Anderson

This traditional analysis is challenged by Anderson and Dixon Ander

son argues that for most morphologically ergative languages there is a clear notion of

grammatical sub ject which picks out the same roles A and S that the term grammat

ical sub ject picks out in accusative languages despite the dierences in case marking

He suggests that the b est understo o d tests for sub jectho o d are the ma jor cyclic

syntactic rules of transformational grammar equiNP deletion raising reexive

binding and conjunction formation Anderson shows that many ergative languages

have analogs of these op erations and that they app ear to b ehave in an accusative

fashion He examines equiNP deletion in Basque isolate France and Spain sub

ject raising in Tongan Austronesian Tonga conjunction reduction in KateHuon

Papua New Guinea and reexivization in Abkhazian AbkhazoAdyghean Georgia

and suggests that they all pick out a sub ject relation comprising the S and A NPs To

present just a couple of Basque examples at this stage shows how an absolutive

recipro cal can b e b ound by an ergative NP a while the reverse is imp ossible b

Recipro cals never app ear in the A or S sub ject p ositions but an S can also bind an

oblique recipro cal c

a Gudariek elkar hiltzen zuten

soldierserg recipabs kill aux

The soldiers killed each other

i i

b Gudariak elkarrek hiltzen zituenzituzten

soldiersabs reciperg kill aux

Chapter Cutting the ergativity pie

c Lagunak elkarrekin joan dira

friendabs recipwith go aux

The friends have gone with each other ie together

i i

Anderson presents the data in suggesting that they show that equiNP

deletion followed an accusative pattern in Basque

a nahi dut joan

desire havesgsg goinf

I want to go

b nahi dut egin

desire havesgsg doinf

I want to do it

Thus Andersons argument is that the morphology of ergative languages is misleading

In general syntactic op erations show these languages have a grammatical relation of

sub ject grouping A and S just like accusative languages

However Anderson notes that this is not true for all languages In particular he

suggested that Dyirbal PamaNyungan Queensland Australia has pro cesses of b oth

equiNP deletion and relativization that work on an ergative basis by applying only

to S and O NPs For a language such as this Anderson concludes that something

like the underlying passive theory app ears to b e correct p

Dixon

This line of approach led to the current ortho doxy as presented in Dixon the

standard survey of ergativity Dixon analyzes direct arguments in terms of what

However Ortiz de Urbina argues that these examples do not consist of a biclausal

equi control construction with an embedded PRO sub ject but rather a mono clausal complex

predicate construction Thus they should b e handled at argument structure as I suggest for similar

constructions in Inuit in Section

It is unclear whether Anderson was actually adopting the analysis that in these languages surface

ergative structures arise by passivization or whether he was simply referring to a certain conguration

of grammatical relations The distinction is discussed further in Section

Dixon up dates his ideas in the monograph Dixon The summary in this section is of

Dixon but it is generally compatible with the p osition of Dixon Some ideas and

references from Dixon are used further on

Chapter Cutting the ergativity pie

he regards as universal semanticsyntactic primitives A S and O dened as in Sec

tion Dixon argues that there are two types of ergative languages a large class

of morphologically ergative languages and a small class of syntactically ergative lan

guages In the former tests that are sensitive to surface grammatical relations these

include some of the tests Anderson uses in particular tests involving clausal

co ordination and sub ordination show that the underlying grammatical relations do

not mirror the surface morphology Rather all of these pro cesses clearly pick out a

notion of grammatical sub ject or pivot as Dixon terms it grouping A and S For

syntactically ergative languages like Dyirbal similar pro cesses pick out a grouping

of S and O Dixon is careful to distinguish the presence of a pivot which not all

languages need have or they may use a mixture of pivots from what he regards as

the true universal concept of deep sub ject which I will refer to here as the logical

subject following Jesp ersen The logical sub ject is the highest argument at ar

gument structure of the basic form of a predicate normally the agent or exp eriencer of

transitive verbs Dixon suggests that all languages have a category of deep sub ject

formed by grouping A and S For semantic reasons certain grammatical pro cesses

will universally pick out this notion of deep sub ject regardless of the surface pivot of

the language This is b ecause it is the deep sub ject that can control events These

pro cesses include deciding the addressee of imp eratives control relations with jussive

complements roughly complements to verbs b elonging to Sag and Pollards

inuence sort and the similar control relations with asp ectualmo dallike verbs such

as can must try b egin nish when they o ccur as lexical verbs in a language

Thus for example Andersons evidence for the syntactic accusativity of Tongan is

that there is optional sub ject raising a pro cess aecting A and S b b but not

Note that for Dixon valence changing op erations change the assignments of A S and O so

that passive for example puts an underlying O NP into derived S function Dixon This

is compatible with a framework like mine see Section Grimshaw Bresnan and Zaenen

etc where passive manipulates argument structure but not with an approach like in RG

or Bresnan b where passive manipulates grammatical relations

Dixon f deletes discussion of jussive complements and adds discussion of reexives

Note also that on Dixons theory op erations like passive must change the conguration of A S

and O in the case of passive the former O b ecomes what is called a derived S This is need to explain how the theme can control events in the grammatical English gloss of for example

Chapter Cutting the ergativity pie

O with the verb lava b e p ossible can

a oku lava ke h u a mele ki hono fale

pres p ossible tns enter abs Mary to his house

It is p ossible for Mary to go into his house

b oku lava a mele o h u ki hono fale

pres p ossible abs Mary tns enter to his house

Mary can go into his house

a oku lava ke taai e siale a e fene

pres p ossible tns hit erg Charlie abs def woman

It is p ossible for Charlie to hit the woman

b oku lava e siale o taai a e fene

pres p ossible erg Charlie tns hit abs def woman

Charlie can hit the woman

oku lava a e fene o taai e siale

pres p ossible abs def woman tns hit erg Charlie

The woman can b e hit by Charlie

But Dixon do es not accept this data apparently originally from Chung as

evidence of syntactic accusativity in Tongan suggesting that it follows from universal

prop erties of deep sub jects Dixon fn Indeed as he suggests all

Chungs evidence for syntactic accusativity is of this sort involving the b ehavior of

asp ectualmo dal verbs like can b egin must and therefore unconvincing This

idea that pro cesses of control and imp erative addressee have a partly semantic basis

and op erate indep endently of the language particular pivot is imp ortant and it is one

that I will try to build on

This universal notion of deep sub ject is indep endent of whether a language is

ergative or accusative at either or b oth of the levels of morphology or syntax Dixon

Cf also argumentation by Mosel and Hovdhaugen concerning the Samoan Austrone

sian Samoa construction that corresp onds to the Tongan examples cited ab ove they suggest that

the restriction on raising follows simply from the corresp onding Samoan verb meaning not just can but can do something see Co ok for a dissenting view on the status of sub ject in Samoan

Chapter Cutting the ergativity pie

sees syntactic and morphological ergativity as two separate parameters although there

is a one way implication in which syntactic ergativity implies some morphological

ergativity Note also that morphological ergativity at least is not an all or nothing

thing Dixon suggests that no language he surveys is entirely ergative at the surface

morphological level

Whether a language has an ergative or accusative syntactic pivot is to b e deter

mined according to Dixon by the b ehavior of language particular syntactic op er

ations such as co ordination sub ordination and related phenomena that he sees as

applying at a level of shallow structure Dixon suggests that most languages use

an SA pivot for these op erations but a few use an SO pivot He gives Walmatjari

PamaNyungan Western Australia as an example of a morphologically ergative lan

guage with an accusative SA pivot In three clause linking constructions including

the one shown in clauses can only b e linked when they share the SA NP

tikiryanula manalunja manawanti patjani

returnafter indexclSOplSplO treeplabs chopped

Having returned we chopped trees

Another example is Basque In a sentence like the A of the rst clause must

b e interpreted as the S of the second regardless of whether the rst clause is in a

reduced form without the auxiliary

Semea eskolan utzi zuen eta klasera joan zen

sonabs schoolat leave aux and classto go aux

She left herhis son at school and went to class

This contrasts with similar data from Dyirbal where clause linking requires the clauses

to share an SO NP

bayi yara bangul gubingu mundan bangun jugumbiru

iabsth manabs iergth gubierg bringnfut i iergth womanerg

balgan

hitnfut

The gubi shaman brought the man here and the woman hit him

I take the ab ove to b e the essence of Dixons analysis But the pap er naturally

discusses much more the functional bases of NP grouping the dierent types of

Chapter Cutting the ergativity pie

split ergativity that o ccur and the pro cess of diachronic change There is also much

other relevant work in the structuralistfunctionalist literature esp ecially the survey

article of Comrie and the work of Silverstein on the hierarchies condi

tioning ergativity splits but since Dixon ably summarizes much of this work I will

move on

The scarcity of syntactically ergative languages

Examples showed a clear dierence in syntactic b ehavior b etween two lan

guages with ergative morphology However there are some reasons to doubt whether

the ab ove division into syntactically ergative and syntactically accusative languages

correctly or optimally sub divides the class of languages that have partially erga

tive morphologies or otherwise exhibit patient prominence One problem is that

attempts to add other languages to the class of syntactically ergative languages thus

dened have tended to prove unsustainable Anderson claims without sub

stantiation that Hurrian ancient isolate Turkey is also syntactically ergative But

this is questioned by Pullum in large part b ecause the textual record is

so scant that the prop osition is untestable and the manuscripts in preparation in

which Anderson was supp osed to supp ort this claim have not app eared Marantz

attempts to argue that Central Arctic Eskimo EskimoAleut Canada is syn

tactically ergative as opp osed to certain other forms of Eskimo in particular West

Greenlandic which he analyzes as morphologically ergative However this claim for

a fundamental typological dierence b etween Central Arctic Eskimo and other vari

eties of Eskimo simply cannot b e substantiated see Johns and Section

b elow Andersons tests of equiNP deletion and reexive binding would identify

Central Arctic Eskimo as syntactically accusative just like all other forms of Eskimo

0 y

Levin suggests Yup ik EskimoAleut Alaska and Yidin PamaNyungan

Australia But the only evidence she provides for the syntactic ergativity of Yup ik is

awed see Chapter fn and Dixon describ es Yidin as less ergative

than Dyirbal as a mixed pivot language

Thus Dyirbal has tended to stand alone as the one true syntactically ergative

language Perhaps b ecause of this some researchers have maintained that there are

Chapter Cutting the ergativity pie

no syntactically ergative languages such is the p osition of Johnson Pullum

and apparently the one of Perlmutter and Postal at the Linguistic In

stitute lectures on Relational Grammar Now this apparent rarity of syntactically

ergative languages may just b e an historical accident certain basic word orders are

apparently very rare but nonetheless exist but I would like to suggest that this result

has instead come ab out from cutting the pie in the wrong places Given the cuts I

prop ose there will turn out to b e a large supply of syntactically ergative languages

Another problem is that many languages do not app ear to t into the current

typology Chapters and of this dissertation will b e sp ent concentrating on lan

guages that do not t in in one of two ways Chapter lo oks at Lezgian Nakho

Daghestanian Russia and Azerbaijan which like some other NakhoDaghestanian

languages is problematic b ecause there is little or no evidence for a pivot of any

sort However here and in Chapter I will concentrate on the opp osite problem

Many languages have a variety of syntactic phenomena that one might take as ev

idence in determining the sub ject or pivot but the phenomena are not consistent

and some groupings work along ergative lines and others along accusative lines this

is what Dixon termed the problem of languages with mixed pivots

A classically problematic case is the Philippine languages and in particular Tagalog

Austronesian Philippines Schachter Schachter My aim is to show

that the socalled mixed pivot b ehavior of these language is not random and is in

fact to b e exp ected under an appropriate denition of syntactic ergativity

Towards a new typology of mixedpivot languages

Tagalog

Tagalog has a rich system of socalled voices up to seven dep ending on the verb

some of which I will term following Kro eger active voice av ob jective

voice ov dativelocative voice dv and instrumental voice iv which determine

the semantic role which is b orn by the ang marked NP of which there is normally

Chapter Cutting the ergativity pie

one in every sentence The rst three of these voices are illustrated in The

ang marked NP is in turn the agent the theme and the lo cation

a Bumili anglalake ngisda satindahan

perfavbuy nomman gensh datstore

The man b ought sh at the store

b Binili nglalake angisda satindahan

perfbuyov genman nomsh datstore

The man b ought the sh at the store

c Binilhan nglalake ngisda angtindahan

perfbuydv genman gensh nomstore

The man b ought sh at the store

The p erfective asp ect marker is an inx the stem ab ove is bili and there app ears

to b e a sp ecial fusional inx for the p erfect in the active voice Here and generally in

the nonfuture asp ects Ob jective Voice is morphologically zero that is the unmarked

form is when the O is the ang marked NP but this is not always the case

Schachter p oints out that Tagalog had a split in apparent sub ject

prop erties in roughly the sense of Keenan b b etween those b orne by the ang

marked NP and those b orne by the Actor Some of these prop erties can b e lined up

as follows

Ang marked NP Topic Actor

Obligatory element of clause Possible antecedent of reexives

Launches oating quantiers Equi target

Relativization Imp erative addressee

Sp ecic

Because Keenans criteria do not consistently pick out a notion of sub ject in Tagalog

Schachter concludes that the ang marked NP is the Topic but that various other

Here and elsewhere in this chapter certain complexities in the data have b een suppressed for

the purp oses of a clean presentation of the basic idea Some of these lacunae are addressed in later

chapters The omission at hand here is that certain Tagalog constructions do not exhibit voice

morphology or an ang marked NP but the description is accurate for what Schachter called Tagalog simple narrative sentences

Chapter Cutting the ergativity pie

prop erties key o the macrorole of Actor and that Sub ject isnt a useful notion in

the description of Tagalog However he is careful to p oint out that these Philippinist

conceptions of Topic and Actor are somewhat at variance with normal usage the

Topic can b e what would normally b e called a fo cus for example and the Actor can

have various thematic roles in the context of Tagalog verbs meaning roughly receive

and endure But the Topic has referencerelated prominence while the Actor has

rolerelated prominence serving as the protagonist

I do not wish to illustrate all these prop erties nor the other tests discussed by

Kro eger but let me just work through a few which have further application

in my crosslinguistic study Schachter describ es the topic as regularly denite

or more carefully as a term whose reference is presupp osed This was indicated very

approximately in the dierence in the translation b etween a and b

Another key prop erty of the Topic is that it is the only p osition that can b e

relativized on This is illustrated in where active voice and ob jective voice

are used when relativizing the actor and patient resp ectively It is not p ossible to

form relative clauses unless the gap representing the relativized NP is in the Topic

slot

a Iyon angbabaeng bumili ngbaro

that nomwomanlnk perfavbuy gendress

Thats the woman who b ought a dress

b Iyon angbarong binili ngbabae

that nomdresslnk perfbuyov genwoman

Thats the dress that athe woman b ought

a Iyon angbarong bumili angbabae

that nomdresslnk perfavbuy nomwoman

b Iyon angbabaeng binili angbaro

that nomwomanlnk perfbuyov nomdress

On the other hand Schachter shows that the Actor can always control a reexive

regardless of whether it is the Topic ab while it cannot itself b e a reex ive c

Chapter Cutting the ergativity pie

a Nagaalala anglolo sakaniyang sarili

avworry nomgrandfather dathis self

Grandfather worries ab out himself

b Inaalala nglolo angkaniyang sarili

ovworry gengrandfather nomhis self

Grandfather worries ab out himself

c Inaalala anglolo ngkaniyang sarili

ovworry nomgrandfather genhis self

Also in the basic pattern of control it is always the actor that is the gapp ed

controllee regardless of the verbal voice of the complement For example a

shows a topic actor controllee while b shows a nontopic actor controllee see

Kro eger for more extensive paradigms

a Iniwasan kong tumingin kayLorna

perfavoiddv Igencomp avlo okat datLorna

I avoided lo oking at Lorna

b Binawalan ko siMariang awitin ang Dahil sa

perfforbiddv Igen nomMariacomp singov nom b ecause dat

iyo

yousg

I forbade Maria to sing Because of you

Thus the kind of sub ject prop erties that Keenan b identies are split

b etween two NPs in Tagalog sentences except when the Actor and the Topic coincide

and it is not immediately obvious to which of these we should apply the term sub ject

Inuit

But this problem where sub ject prop erties are split b etween two NPs in a sentence

is not conned to the Philippine languages Indeed a very similar sort of split

The controller is determined on semantic grounds and can b e either a nontopic a or a topic

b This follows from the sort of semantic theory of controller selection given in Sag and Pollard

and Pollard and Sag where in a the controller is an experiencer and in b

an influenced participant

Chapter Cutting the ergativity pie

app ears to o ccur in various ergative languages For example the Eskimo languages

are widely regarded as syntactically accusative having an SA pivot and hence only

morphologically ergative According to the criteria of Anderson it seems that

all Eskimo languages would b e regarded as syntactically accusative evidence from

ma jor cyclic transformations like control and binding suggest syntactic accusativity

This is the p osition argued for in Johnson Marantz also concludes

that West Greenlandic Eskimo is syntactically accusative on rather dierent grounds

Givon endorses the analysis of Kalmar that the language is not

deep ergative but rather surface ergative

However Woo dbury a notices that there is a mixture of evidence for b oth

SO and SA b eing pivots in Greenlandic and this analysis is mentioned by Dixon

who suggests Eskimo as another mixed pivot language family We can

b egin to pro duce a split of prop erties quite similar to that shown for Tagalog in Inuit

Absolutive marked NP Actor

Sub categorized element of every clause Antecedent of reexives

Relativization Equi target

Sp ecicWide Scop e Imp erative addressee

Some of these same prop erties are briey illustrated b elow they are exemplied

in greater detail in Chapter Examples are from West Greenlandic except where

noted All verbs sub categorize for an absolutive argument although it may not

app ear overtly b ecause of free prodrop Relative clauses are restricted so that the

relativized role must b e the absolutive within the relative clause So ab show

relativization of O and S NPs while c shows that relativization of an A NP is

imp ossible

a nanuq Piitap tuqutaa

p olarb ear Piitaerg killtrpartsg

a p olar b ear killed by Piita

Relative clauses in Inuit are actually participial nominalizations but I am essentially accepting a functional denition of what a relative clause is following Comrie

Chapter Cutting the ergativity pie

b miiraq kamattuq

childabs angryrelintrsg

the child that is angry

c angut aallaat tigusimasaa

manabs gunabs takeprfreltrsgsg

the man who to ok the gun

Thirdly the absolutive NP has certain sp ecial interpretive prop erties which much

of the traditional literature has interpreted as deniteness but p erhaps sp ecicity

is nearer to correct and which Bittner accounts for in terms of scop e

The avor of the distinction b etween the transitive Central Arctic Eskimo sentence in

a and the intransitivized variant in b comes from the sp ecicity referentiality

or givenness asso ciated with the absolutive NP in a unsatisfactorily captured by

translating it with the

a Jaaniup tuktu takuvaa

Jaanierg carib ouabs seeindtrsgsg

Jaani sees the carib ou

b Jaani tuktumik takuvuq

Jaaniabs carib oumod seeindintrsg

Jaani sees a carib ou

On the other hand some prop erties seem to b e sensitive to a notion of sub ject

linking A and S This is the kind of evidence that following Anderson is often

taken to suggest that most ergative languages are syntactically accusative Thus a

p ossessive reexive can b e b ound by an A or by an S ab but not by an O

a ataatani Juunap tatigivaa

fathersgabs Juunaerg trustindtrsgsg

Juuna trusts his father

i i

In Inuit p ossessor agreement is suxed to nouns A separate set of reexive agreement axes are customarily referred to as the th p erson

Chapter Cutting the ergativity pie

b Aani illuminut ingerlavoq

Anneabs housesgterm goindintrsg

Anne is going to her house

i i

c Anaanami Piita nagligijana Inuktitut

mothersgerg Piitaabs lovesgsg

His mother loves Piita

i i

and in cases of control the controllee NP is again the A or S and not the O

a Miiqqat Juuna ikiussallugu niriursuipput

childrenabs Juunaabs helpfutinfsg promiseindintrpl erg

The children promised to help Juuna

b Miiqqat qitissallutik niriursuipput

children abs dancefutinfpl promiseindintrpl

The children promised to dance

The correlation of prop erties b etween Philippine languages and certain other erga

tive languages compare again and is actually quite impressive but do es

not seem to have gained wide currency However after I b egan investigating the

correlation b etween prop erties of Kro egers analysis of Tagalog and prop erties

found in the Inuit literature I b ecame aware of a number of prior observations of this

correlation Johnson refers to the remarkable coincidence of results b etween

her work on Central Arctic Eskimo and Schachters discussion of Tagalog

and Payne makes a similar argument with resp ect to Yup ik EskimoAleut

Alaska Similarly Blake compares Tagalog to Kalkatungu PamaNyungan

Australia At the end of his pap er Andrews also tentatively considers col

lapsing the syntactically ergative and Philippine language types although the parallel

with the class of syntactically ergative languages as traditionally dened is less con

vincing than the one that I have b een considering I thus conclude that within a

syntactic typology the basic voices of Tagalog and Inuit should b e analyzed the same

something that some previous researchers have b een reluctant to do b ecause of the dierentness of Philippine voice systems

Chapter Cutting the ergativity pie

Notions of Sub ject

What then is the sub ject of an Inuit sentence We might conclude that Inuit should

also b e put in the to o hard basket and we could start talking ab out the Actor and

Topic of Inuit sentences following Schachter However this seems undesirable If we

have any conception whatso ever of universals of human language then we should seek

syntactic categories that can b e applied across languages of dierent typological sorts

This is the condition of generality of Chomsky It would b e irresp onsible

to exclude Philippine Eskimo and many other ergative languages from consideration

when determining the meaning of the notion Sub ject There is absolutely no reason

to suggest that basic grammatical primitives should not also b e found in Inuit since

it has syntactic pro cesses that are in general familiar from languages around the

world Rather it suggests only that our traditional grammatical categories may need

renement But this is the whole p oint of doing crosslinguistic work

If we accept that linguistic notions should b e applicable to Philippine and Eskimo

languages the question then b ecomes which NP is the sub ject There are clearly

two choices b efore us Either the Philippine Topic and the Eskimo absolutive is the

sub ject or the Philippine Actor and the Eskimo SA NPs are the sub ject It is

logically p ossible that the answer to this question is quite arbitrary Supp ose that

these groupings in Philippine and Eskimo languages pick out two notions A and B

and that it happ ens that the notion of sub ject from common Europ ean languages is

precisely the conjunction of the notions A and B Then it would b e quite arbitrary

which of A and B we call sub ject and which we nd a new term for In practice

however I do not b elieve that this is the case

Nevertheless part of the problem is that sub ject is not used uniformly Some

frameworks and analyses dene sub ject in terms of a system of surface grammatical

relations others such as Dixon consider sub ject as a basically semantic

notion while yet others seem to confuse these dierent criteria such as the list of

sub ject prop erties in Keenan b In this sense Schachter is right in suggesting

Given the convergence of prop erties b etween the two language families that I have outlined it

would app ear typologically irresp onsible to consider the other two p ossibilities declaring the Topic the sub ject in Philippine languages and the SA NP the sub ject in Eskimo languages or vice versa

Chapter Cutting the ergativity pie

that we need two notions of sub jectho o d similar in content to what he terms Topic

and Actor although one could disagree with the choice of names

The rst notion of sub jectho o d that I want to establish is sub ject as the privi

leged term in a system of surface grammatical relations what I sometimes refer to as

the grammatical sub ject to avoid ambiguity This notion is similar to the notion of

sub ject in traditional grammar but it is more clearly seen in Dixons usage of pivot

the notion of nal in RG or sub ject in LFG Kro eger argues extensively that

the correct analysis of Tagalog is that the socalled Topic is actually the grammatical

sub ject Since the Ob jective Voice is basic in Tagalog Cena the arguments

are summarized in Foley and Van Valin this means that Tagalog has an

essentially ergative character Kro eger argues that grammatical sub ject should b e

decided on syntactic rather than semantic or discoursepragmatic criteria and that

a range of these criteria quantier oat relativization number agreement raising

control of secondary predicates sub ject obviation p ossessor ascension and conjunc

tion reduction all pick out the ang marked NP as the syntactically privileged NP

Schachter was not the rst to distinguish these two notions of sub jectho o d although pre

generative work did not generally distinguish surface grammatical relations indep endent of mor

phological form Allen wrote Discussions of this type often distinguish b etween a

grammatical and a semantic sub ject The former is formally denable and usually refers to the

noun or pronoun which is in a particular eg nominative case andor in concord with the verb

or with a particular verbal element The latter also variously describ ed as real or psychological

generally remains undened and intuitive since situational or logical correlates such as actor or

topic break down in a number of instances

I take Kro eger as my main reference on Tagalog However it should b e noted that a

morphologically ergative analysis of Philippine languages was prop osed earlier by Gerdts

and De Guzman and in various other works Blake Payne Mithun

Gibson and Starosta and by various other p eople mentioned in this last reference The

Relational Grammar analysis of Gerdts and De Guzman diers considerably from Kro egers in two

ways it simultaneously maintains notions of nal and yielding accusative sub jects as well as

the relations ergative and absolutive and it generates Active Voice by antipassivization whereas

Kro eger argues that the genitive patient in the Active Voice is still a term hence using the term

voice is somewhat misleading but I have retained it for want of a b etter term The binding

facts discussed b elow supp ort Kro eger against almost all other analyses in suggesting that none

of the voices of Tagalog result from pro cesses of passivization or antipassivization that demote

terms The claim that the Topic is the sub ject in Tagalog was also put forward without argument

in various early treatments of Tagalog such as Blo omelds and was explicitly argued for earlier

by H McKaughan see Kro eger for references Cena gives the same analysis for

ob jective voice sentences as Kro eger he do esnt make clear how he intends the other voices to b e generated

Chapter Cutting the ergativity pie

which is termed the sub ject Conversely the prop erties app earing to favor the Ac

tor p ossible antecedent of reexives imp erative addressee and equi target have a

decidedly semantic avor p erhaps meaning that they should b e treated at the level

of argument structure Two of the three are listed among Dixons prop erties that

always work accusatively due to the nature of Deep Sub jects and Andrews

suggests that binding p ossibilities in Tagalog should in general b e dened in terms of

the thematic hierarchy

But an adequate linguistic theory also needs another notion of sub jectho o d

the more semantic one This one is related to Dixons use of the term sub ject and

Schachters notion of Actor which is in turn equivalent to Jesp ersens notion of logical

sub ject I want to prop ose that their category is not quite what we need to capture

the other notion of sub jectho o d My analysis crucially involves the p ostulation of a

level of syntactic argument structure astructure with its own distinguished term

which I call the asub ject While all logical sub jects are asub jects I prop ose b elow

that the comp ound argument structures that result from derivational op erations like

passive causative and antipassive yield additional asub jects As a result the other

half of sub jectho o d in Inuit and other languages is not prop erly describ ed by the

grouping A and S but rather by the larger set of all asub jects For example in a

passive b oth the logical sub ject and the surface sub ject count as asub jects

Purely syntactic pro cesses like relativization and topicalization should b e sensitive

to the hierarchy of grammatical relations while the more semantic prop erties of

binding control and imp erative addressee should b e mainly sensitive to the level of

syntacticized argument structure Tagalog and Inuit are b oth syntactically ergative

languages which implies that the obliqueness ordering of grammatical relations in the

basic verbal voice do es not match the obliqueness ordering at argument structure

Rather there is an inverse relationship as shown in

GRs astructure

subj agent asub ject

obj patient

The prop osal in might b e familiar as the Inverse Analysis of previous work which is discussed b elow in Section but my prop osal diers in not exp ecting the surface

Chapter Cutting the ergativity pie

sub ject or pivot I will use the two terms interchangeably to b e the basis of syntactic

organization throughout the grammar of the language Dowty Rather

I exp ect to nd in any language a principled division b etween phenomena that are

sensitive to the level of grammatical relations and phenomena that are sensitive to

argument structure

Historical Origins

It is p ossible that the parallels in the synchronic situation b etween Philippine and

Eskimo languages result in part from the diachronic path that these language families

have followed An historically inuential approach to Eskimo has b een the nominalist

tradition In this tradition Eskimo sentences are analyzed as b eing an equation b e

tween a sub ject and a nominalized complement so that a sentence like a actually

has the structure suggested by the calque in b

a angutiup nanuq kapijaa

manerggen p olarb earabs stabpasspartsgsg

The man stabb ed the p olar b ear Qairnirmiut

b The p olar b ear is the mans stabb ed one

This analysis seeks to explain the parallels in case marking b etween p ossessors and

ergative NPs and b etween the agreement on p ossessed NPs and verbs I take it

that the analysis has little synchronic app eal among other reasons b ecause of the

unmarked word order shown in a but that it app ears to have a certain histor

ical truth Eskimo transitive verb endings do app ear to have evolved from passive

participle morphemes

It is thus very interesting that something of the same explanation has b een pro

p osed for Philippine languages Starosta et al argue that the fo cus morphology

The idea that Eskimo clauses are like nominalizations was prop osed by Thalbitzer and Ham

merich supp orted by SchultzLorentzen Lowe and Dorais carefully discussed but not altogether

endorsed by Woo dbury b and adopted as the basis of a GB analysis by Johns

See Johns f and Johns fn for references

16 Eskimo participles are nominal the language lacks a category of adjectives

Chapter Cutting the ergativity pie

of Philippine languages can clearly b e traced back to nominalizing axes in proto

Austronesian They argue that the origin of a Tagalog sentence like a is again

to b e found in the calque given in b although again the current freer word order

suggests that reanalysis has taken place

a Kakanin ngmaestro angpapaya

eatov genteacher nompapaya

Athe teacher will eat the papaya

b The papaya will b e athe teachers eatee

This explanation might also extend to the Mayan languages that I consider b elow

They also use the same agreement for ergatives and p ossessors but I have not further

investigated their historical reconstruction

These languages thus stand in contrast to cases where ergativity arises from a

p erfective or maybe from a passive where the ergative is frequently homophonous

to the instrumental or a case like ablative that expresses passive agents such as in

Hittite ancient Anatolian IndoEurop ean Garrett and the Indic branch of

IndoEurop ean Anderson I b elieve that historical origin could b e a go o d

guide in sub dividing the types of ergative languages although the matter would re

quire much further investigation Making an initial cut b etween ergativity arising

from nominalization versus ergativity arising from a p erfective or passive origin rein

terpreting an oblique instrumental or agent as the ergative NP seems promising

Note however that I am making quite dierent predictions to those given in Trask

Trask prop oses two types of ergative languages Type A resulting from a

passive made obligatory and Type B from a nominalized verb form often a stative

participle incorp orated into sentences by means of a p ossessive construction On

Trasks typology Type A languages should at least briey go through a syntacti

cally ergative stage while Type B ergativity should always b e relatively sup ercial

and not extend b eyond casemarking In contrast I am suggesting that many lan

guages where ergativity arises from nominalization are syntactically ergative whereas

the ergativity in the Indic IndoEurop ean languages for example seems sup ercial

from the p oint of view of syntactic b ehavior

Chapter Cutting the ergativity pie

Word order a necessary digression

In the typological tradition the basic word order of a language is generally denoted

in symbols as SOV OVS or whatever Where there is an unmarked p osition for

oblique NPs this is typologically imp ortant additional information and we can write

SOXV or SXOV The question is what notion of sub ject are these lab els referring

to the pivot or the logical sub ject Past work in word order has suered by not

adequately distinguishing these two notions of sub jectho o d when stating word order

generalizations as noted by Pullum

Dixon fn faults the SOV notation for not distinguishing his primitive

S and A and suggests that for a universal framework one should describ e basic word

order in terms of A O and S as some languages present congurations like SVOVA

or VSAVO I susp ect that this is the wrong direction to head Given the distinction

b etween the two notions of sub ject prop osed ab ove I b elieve that word order should

b e describ ed in terms of the surface grammatical relations pivot P other core C

oblique O and verb V rather than in terms of A S and O which refer to terms

at a level roughly akin to my level of argument structure

For example I would like to argue that Tagalog should b e analyzed as a VCP

ie VOS language like other Austronesian languages such as Malagasy Keenan

a and Toba Batak Schachter whereas it has sometimes b een wrongly

analyzed as VSO Once this analysis is adopted the unmarked order of full NPs

in Dyirbal is seen to fall into the common PCV ie SOV type rather than it

b eing an example of what has conventionally b een the least common sort OSV

Somewhat unsettlingly this argument would move Eskimo languages into the least

attested type CPV ie OSV However to really evaluate this result one would

need a new wide coverage survey of word order working consistently with the notion

For instance VSO was one of two options Pullum oers for Tagalog while arguing

against the idea that Tagalog is a VOS language as had b een argued earlier by others I must add

however that word order in Philippine languages is complicated and facts on Actor placement in

Cebuano Kalagan and Pangasinan noted noted by Schachter would need to b e integrated

into a complete word order typology

This p ossibility was noticed but not endorsed earlier by Pullum a work that should now

b e approached with caution b ecause of the many typological discoveries and developments since its publication

Chapter Cutting the ergativity pie

of surface pivot For example given the analysis of Tagalog and other Austronesian

languages suggested ab ove note that VCP ie VOS languages b ecome much more

common than has typically b een thought to b e the case

Further evidence for whether word order should b e describ ed in terms of surface

grammatical relations should come from syntactically ergative languages We should

b e able to tell whether word order groups O and S b ecause it is working from the

category of surface pivot or whether word order groups A and S using Dixons notion

of deep sub ject In practice evidence has b een thin on the ground b ecause most

ergative languages are verb initial or verb nal However where there is word

order evidence it generally app ears to supp ort a statement of word order in terms

of surface grammatical relations although further syntactic information would b e

needed to conrm the identication of the pivot in the languages in question Below

are several examples all of which I draw from Dixon

Sanuma Yanomami Brazil and Venezuela has ergative noun morphology and

Dixon describ es the basic word orders XSV and AXOV Huastec Mayan Mexico

has ergative verbal crossreferencing as in other Mayan languages discussed b elow and

the most frequent constituent orders are what Dixon calls VSAVO Pari Western

Nilotic Sudan has ergative case marking and according to Dixon the basic word

orders SVOVA

a ubur atuuk

Uburabs cmpltvplay

Ubur played

b joob akeel uburr

bualoabs cmpltvsho ot Uburerg

Ubur shot the bualo

I susp ect that all the ab ove languages could b e analyzed in terms of a surface ergative

pivot so that Sanuma would b e viewed as a COPV language Huastec as a CVP

language and Parias a PVCO language There isnt conclusive evidence in Andersen

Indeed the fact that ergative languages are usually verb initial or nal is an imp ortant part of certain theories such as those of Anderson and Trask

Chapter Cutting the ergativity pie

to supp ort an absolutive pivot in Pari but there is suggestive evidence the

language has antipassive constructions generally only found in languages with SO

pivots and Andersen suggests that the diachronic source of ergativity

in Pari is a topicalization of O and S that evolved to b ecome the unmarked form

protoNilotic app ears to have b een an accusative VPC language

Other languages for which an inverse analysis seems

correct

Let us then examine the applicability of this new typology wherein syntactically

ergative languages are exp ected to show the kind of split of prop erties that we have

seen for Philippine languages and Inuit An Inverse Grammatical Relations analysis

of the type I have prop osed for Tagalog and Inuit actually seems to b e appropriate

for a considerable number of ergative languages although still p erhaps a minority

Below I suggest other languages that seem to t into the same class as Tagalog

and Inuit The size of this class is consonant with the observation of Wierzbicka

that deep ergativity in the sense of patientorientedness or no overall agent

orientedness is much more widespread than is often assumed

Mayan languages

A breakdown of prop erties fairly similar to that found for Tagalog and Inuit app ears

to b e widespread in the Mayan languages England b a and Campana

present evidence from Mam Mayan Guatemala and Mexico Mam has a

consistently ergativeabsolutive system of verbal crossreferencing The absolutive

app ears to b e the grammatical sub ject b ecause it is privileged to p erform various

surface syntactic functions S and O NPs may b e fo cussed by fronting ab while

an A NP may not c Antipassivization is required d

Abbreviations p eculiar to Mayan are dir directional dirs directional sux rec recent

past tense rn relational noun sp ecial always p ossessed structural nouns used to indicate oblique case relationships

Chapter Cutting the ergativity pie

a xiinaq suul

man depaspsgAarrivehere

The man arrived here

ttzyun xiinaq b qacheej xhi kub

dir sgEgrabdirs man plhorse depaspplA

The man grabb ed the horses

ttzyun qacheej c xiinaq chi kub

dir sgEgrabdirs plhorse man plA

The man grabb ed the horses

d xiinaq xkub tzyuun te qacheej

man depaspsgAdir grabantip sgrnpat horse

The man grabb ed the horses

Similar evidence of syntactic ergativity app ears in question formation The S and

O can b e simply questioned b b while questioning an A is imp ossible c

without a change to antipassive verbal morphology d

xiinaq a maa chi tza j ttzyun Cheep kab

two man recemph pA dir sEgrabds Jose

Josegrabb ed the men

b alkyeeqa xhi tza j ttzyun Cheep

whopl recdeppA dir sEgrabds Jose

Whom did Jose grab

c alkyee sa j ttzyun kab xiinaq

who recdepsAdir sEgrabds two man

Who grabb ed the men

xiinaq d alkyee sa j tzyuun kye kab

two man who recdepsAdir grabantip plrn

Who grabb ed the men

a ma chi b eet xiinaq

rec pA walk man The men walked

Chapter Cutting the ergativity pie

b alkyee xhi b eet

who pAdep walk

Who walked

However control picks out the Actor for the controllee

0 0 0

tx eemal sii ma tz ok nq ona

cutinf woo d asp sAdir sEgivedirssgsg

I made you cut woo d

Larsen and Norman suggest that a pattern of syntactic ergativity of this

sort is widespread throughout the Kanjobalan Mamean and Quichean subgroups of

Mayan Larsen suggests that the pattern of data in Quiche Mayan Guate

mala is not quite as neat as that given ab ove but nevertheless ends up concluding

that there is syntactic ergativity

Van Valin describ es a similar split of prop erties in Jacaltec Mayan Guate

mala based on data from Craig Relativization Whquestion formation and

clefting are again sensitive to an SO pivot which I am terming the grammatical

sub ject Control is more complicated as it app ears that only intransitive verbs may

b e controlled A prop osed explanation for this BokBennema Murasugi

is that Mayan innitive complements are unable to assign the unmarked here ab

solutive case in parallel with the way that Europ ean innitives except Portuguese

do not assign the unmarked nominative case of accusative languages However I will

not investigate this hypothesis nor any alternative explanation here Van Valin also

lists preferred crossclause coreference as a feature that groups A and S but it is

not altogether clear how this sort of notion of discourse coherence ts in with the

syntactic systems I have b een considering It seems more likely to b e a fact ab out

the world that narratives most often continue to refer for a p erio d to the actions of

the same Actor

The bare ob ject noun in adjacent to the verb is p ossibly incorp orated as suggested in

Campana I am aware of no p ersuasive evidence either way The b ehavior here may b e

similar to the required intransitivization of controlled sub ordinate clauses in Jacaltec noted b elow

For instance Co oreman showed that even in Dyirbal the canonical syntactically ergative

language ASAS crossclausal referential continuity is more common than OSOS crossclausal referential continuity

Chapter Cutting the ergativity pie

Chukchi

Comrie provides some evidence of a similar split of prop erties in Chukchi

ChukotkoKamchatkan Russia Comrie notes that the negative participle can b e

used to relativize on an S or O but never an A Antipassivization is required

a etipeynekelin newecqet ragtege

negsingnegpartabssg womanabssg gohomesg

The woman who was not singing went home

b iger ayokelete enmete menelqenmek

now negreachnegpartto hillto plgopl

Now let us go to the hill which someone didnt reach

enagtatkela qaak aaceka

antipchasenegpartinstr reindeerloc youthinstr

winreterkeninet newecqetti

helppressgEplA womanabspl

The youth who do es not chase the reindeer is helping the women

On the other hand the controlled argument of innitivals is the A or S as shown in

and so Chukchi ts in with the generalizations for control constructions that we

have b een observing in other languages

a gemnan get tite mewinretget ermetwik

Ierg yousgabs sometime sghelpsg abs growstronginf

Let me help you some time to grow strong

b morgenam get metrewinretget riwlek emelo

weerg yousgabs plfuthelpsg erg moveinf allabs

geceyot

collectpasspartabspl

We will help you to move all the collected items

For p ositive participles it seems that participial relativization is restricted to intransitive sub

jects alone S A and O are relativized via active antipassive and passive participles resp ectively

Some other constructions like coreferential omission in co ordination show no syntactic constraints in Chukchi

Chapter Cutting the ergativity pie

Toba Batak

Most of the observations made ab ove ab out Tagalog are in fact true for all Philip

pine languages and many extend to western Austronesian languages more generally

However many of these other languages have more rigid congurationality in their

phrase structure than Tagalog and hence indicate more clearly the indep endence of

binding from surface structure command relationships One such language is Toba

Batak Austronesian Sumatra Indonesia describ ed in Schachter

Toba Batak also has a distinction b etween active voice mang and ob jective

voice di

a Mangida si Ria si Torus

avsee pm Ria pm Torus

Torus seessaw Ria

b Diida si Torus si Ria

ovsee pm Torus pm Ria

Torus seessaw Ria

However in Toba Batak there is strong evidence that a verb and the following NP

of a transitive clause form a constituent that I will call a VP regardless of the verbal

voice chosen Emmorey shows that the pitch accent of a sentence denoted

b elow o ccurs on the last stressed syllable of the predicate where the rst following

NP of a transitive clause counts as part of the predicate regardless of the verbal voice

chosen

a Mu uli anggna

marry brotherhis

His brother gets married

b Mangalean eme halak an tu malim an

avgive rice man to priest

The man gives rice to the priest

Schachter provides evidence that b oth arguments in b oth voices in are terms as Kro eger argued for Tagalog

Chapter Cutting the ergativity pie

c Diboto malim na man uhor eme pangula

ovknow priest buy rice farmer

The priest knows that the farmer buys rice

An adverb cannot app ear in the middle of the VP b etween the verb and the NP

though adverbs can generally o ccur b etween other ma jor constituents VPs can b e

co ordinated regardless of the voice chosen

a Manuhor baoang jala mangolompa mangga halak an

avbuy onions and avco ok mango es man

The man buys onions and co oks mango es

b Dituhor si Ore jala dilompa si Ruli mangga

ovbuy pm Ore and ovbuy pm Ruli mango es

Ore buys and Ruli co oks mango es

In contrast the VPexternal pivot b ehaves similarly to the ang marked NP in

Tagalog This NP may optionally b e fronted b efore the verb in questions or as a

topic while the VPinternal NP may not b e As in Tagalog one must relativize on

the pivot NP Pretheoretically the ab ove evidence motivates the phrase structure

shown in

Sp ec S

VP NP XP

V NP

However despite this clear evidence for phrase structure and a pivot NP re

exivization shows that an asub ject can bind a nonasub ject and not vice versa

regardless of the verbal voice Sugamoto

a Mangida dirina si John

avsaw selfhis pm John

John saw himself

i i

Chapter Cutting the ergativity pie

b Mangida si John dirina

avsaw pm John selfhis

Himself saw John

i i

a Diida dirina si John

ovsaw selfhis pm John

Himself saw John

i i

b Diida si John dirina

ovsaw pm John selfhis

John saw himself

i i

To account for these reexivization patterns using a surface structure based notion of

command would mean suggesting that the phrase structure of the sentences in

and were radically dierent despite all the evidence I outlined ab ove indicating

that the phrase structure is the same despite the changing verbal voice

Schachter and Sugamoto suggest that binding p ossibilities are de

ned by the thematic hierarchy However whereas most thematic hierarchies place

recipient ab ove theme Kiparsky Bresnan and Kanerva they note that

a patient argument can bind an oblique recipient in Toba Batak They take this

as evidence that Toba Batak has the languageparticular thematic hierarchy shown

Agent Patient Dative

However this seems most unlikely Many theories would rule out languageparticular

thematic hierarchies in principle and at any rate evidence from various other Aus

tronesian languages suggests that in this language family to o goalsrecipients out

rank themes Rather I take this as strong evidence that binding is dened not on the

thematic hierarchy but rather on the indep endent syntactic level of argument struc

ture These facts follow a common pattern namely that at the level of astructure

For instance Bell for Cebuano Forsberg f for Tb oli Some other Austrone

sian sources give thematic hierarchies like the one in the text for Toba Batak but all cases I know of result from what I regard as misinterpretation of evidence from binding

Chapter Cutting the ergativity pie

terms can bind obliques b ecause they are less oblique at astructure regardless of

their semantic role This is further discussed in Section

One wrinkle in the Toba Batak data is that the controllee is the pivot regardless

of the verbal voice

a Mangelek si Bill si John manuhor biang

avp ersuade pm Bill pm John avbuy dog

John is p ersuading Bill to buy a dog

b Mangelek si Bill si John dipareso doktor

avp ersuade pm Bill pm John ovexamine do ctor

John is p ersuading Bill to b e examined by a do ctor

This contrasts with the Tagalog facts cited ab ove where the asub ject is the con

trollee and is unexp ected on my account However control in Tagalog is more com

plicated than indicated ab ove and in certain circumstances the controllee in Tagalog is

the pivot as well see the discussion in Kro eger Here the pivot always

seems to b e the controllee while the controller is determined on semantic grounds

as discussed b efore I will leave this as an outstanding problem but p erhaps one

has to say that determination of the controllee sometimes dep ends on grammatical

relations as well as astructure prominence

Tsimshian languages

Tarpent argues for an ergative analysis of Nisgha Tsimshian British Colum

bia Canada Where there are suppletive verb stems dierentiating number they

mark the number of the SO argument Tarpent argues that co ordination is along

ergative lines suggesting that has the translation shown while noting that it

One should b e able to dierentiate b etween these two hypotheses in Toba Batak by lo oking at

binding in the Dative Shift Construction Schachter which places recipients into a core

role But the appropriate data is not available to me

This is probably not evidence either way since Durie notes that where verbs have sup

pletive forms for number it is invariably for the number of the SO NP this p erhaps rather shows

how an SO NP is more involved in delimiting an event roughly along the lines of Tenny or Dowtys incremental theme

Chapter Cutting the ergativity pie

could mean and he kissed Mary but only if main stress fell on Mary rather than

humtsaxs and the context made the intended meaning obvious

tsint Fred iit humtsaxs Mary

comeintm Fred anderg kissdm Mary

Fred came in and Mary kissed him

Any of S A and O can b e fo cussed or relativized but nevertheless Tarpent nds

certain evidence grouping S and O in these constructions Nisgha has an antipassive

which app ears to make the demoted O more indenite but Tarpents discussion is

to o imprecise to say much more Nisgha also has a passive but one that do es not

allow expression of the agent These are all prop erties consistent with an SO pivot

in the language

Tarpent do es not cover binding and there is no evidence from control since

Nisgha uses mo dal pro clitics and morphological causatives rather than verbal com

plement structures However the A or S is the addressee of an imp erative while

incorp oration is of internal arguments including the demoted theme of antipassives

suggesting that Nisgha argument structure is again the same as that of other lan

guages that I have discussed

The situation in Gitksan Tsimshian British Columbia Canada app ears similar

Rigsby There is less evidence in supp ort of an inverse analysis in Smalgyax

Tsimshian British Columbia and Alaska although Mulder a b nds

some supp ort for an ergative syntax in the fo cussing and relativization constructions

Smalgyax also app ears to lack passive and antipassive op erations It may b e changing

into a syntactically accusative language

Nadeb

From the limited information available in Dixon Nadeb Maku Brazil also

app ears to have an SO surface pivot The pretheoretical phrase structure seems to

b e as in

Chapter Cutting the ergativity pie

S Sp ec

NP VP

abs

NP V

erg

The unmarked basic word order is PCV with an SO pivot but the pivot can also

b e p ostp osed to the indicated Sp ec p osition The situation is thus a mirror image of

Toba Batak Dixon notes that in the third p erson an SO pivot is used

for co ordination when two co ordinated main clauses show coreference b etween their

SO NPs then the appropriate NP may b e omitted from the second clause

On the other hand there is some evidence that Nadebhas the usual sort of argu

ment structure Weir states that only O and S arguments incorp orate and

all the examples of incorp orated Ss app ear to b e with unaccusative verbs and some

t work do make incorp orated constituents have b ecome idioms such as mooh wu

t b e in movement from mooh hand arm and wu

Summary

This section has shown that there are a considerable number of languages which

seem as if they should b e describ ed in the same manner as Tagalog and Inuit The

ab ove data demonstrates that mixed pivot languages really have a quite consistent

breakdown of prop erties The apparently accusative ones are prop erties like binding

imp erative addressee and control which are sensitive to argument structure These

phenomena are a sup erset of the ones that Dixon identies as always sensitive

to the notion of deep sub ject Other phenomena such as co ordination sp ecicity

relativization and topicalization are sensitive to a level of grammatical relations and

at this level the absolutive is the pivot in all these languages b ecause they are syn

tactically ergative One other group of languages Dyirbal and related Queensland

Chapter Cutting the ergativity pie

languages which I think should also b e analyzed similarly are discussed in Chap

ter

However it is not my claim that all languages with ergative morphology should b e

analyzed in this way As is discussed later other languages have an ergative case for

historical or semantic reasons but the absolutive NPs are not surface pivots Note also

that not all the phenomena listed ab ove will pick out a notion of asub ject or pivot

in all languages In many languages various phenomena are neutral particularly

phenomena that are sensitive to surface grammatical relations The prediction is

rather that if a phenomenon is restricted then it should b e sensitive to relations at

the appropriate level

Kazenin which I got to read only after writing this chapter contains a study of mixed

pivot languages from a functionalist p ersp ective and covers many of the same languages Kazenin

seeks to establish an implicational hierarchy where

conjunction reduction is ergative ! coreferential deletion in purp osive constructions

is ergative ! relativization is ergative

I disagree with certain of Kazenins classications of the data I regard coreferential omission in

co ordination as patterning ergatively in Tagalog following Kro eger and not accusatively

as Kazenin suggests The argument for accusativity of coreferential omission in co ordination in

Asiatic Eskimo Central Sib erian Yupik is actually based on sub ordinate innitive forms akin to

the Inuit innitive discussed in Section In fact Kazenin notes work by Vaxtin that

co ordination of nite forms seems to follow the ergative strategy ::: however ::: this problem

is not quite clear My suspicion would b e that coreferential omission in co ordination in Central

Sib erian Yupik is in fact neutral as in other forms of Eskimo Section although p erhaps

the preference for SOSO coreference is nevertheless consonant with the SO NP b eing a surface

pivot Finally it is unclear that what Kazenin rep orts on as conjunction reduction in Aguacatec

in fact involves conjunction reduction and at any rate his description inappropriately simplies the

facts presented in Larsen In contrast to Kazenin my prediction would b e that b oth

coreferential omission in co ordination and relativization are ergative or neutral in al l languages with

an SO pivot Purp osive constructions should also b e ergative in such languages if they are prop erly

analyzed as clause chaining constructions as in certain Australian languages see Section but

should dep end on argument structure and hence app ear roughly accusative if they are actually

complement or controlled adverbial clauses With the dierent interpretations of the data mentioned ab ove taken into account I nd no data in Kazenin that conicts with my hypotheses

Chapter Cutting the ergativity pie

Theoretical Foundations

Grammatical relations and argument structure

This dissertation centrally involves the interplay b etween two levels of syntactic rep

resentation which I call grammatical relations structure grstructure and argument

structure astructure There is nothing that I present that could not b e imple

mented in various theories of grammar and so I have chosen an at least somewhat

neutral format of presentation although it is closest to LFG The level of surface

grammatical relations is familiar from a number of frameworks It is the level of nal

grammatical relations in RG and the level of fstructure in LFG In GB the con

gurational p ositions of NPs at sstructure are basically equivalent to grammatical

relations

Use of a level of argument structure is also familiar from various places However

great caution should b e used in p ostulating new levels of structure esp ecially when

they app ear to overlap the functionality of other levels of representation and some

may feel that previous uses of argument structure have not b een suciently justied

I would like to argue that consideration of syntactic ergativity provides strong mo

tivation for a separate level of argument structure b ecause it demonstrates a clear

disso ciation b etween grammatical relations and argument structure prominence The

goal of this thesis is to show that problems of syntactic ergativity can b e insightfully

handled by use of a separate level of syntactic argument structure whereas solu

tions that attempt to use only grammatical relations or structural p ositions andor

thematic roles either do not work or are unrevealing

Not all conceptions of argument structure are quite the same The notation I

use for astructure is similar to that of recent work in LFG Mohanan Alsina

Butt but I should emphasize that I am thinking of argument structure

as a syntactic representation as in Bresnan and Zaenen while some work in

LFG such as Alsina has suggested that astructure is a purely semantic

representation Notions of astructure are also present in the GB literature eg

Grimshaw However my notion of syntactic astructure is actually more similar

to VPinternal relationships in many recent versions of GB and related Minimalist

Chapter Cutting the ergativity pie

work Note particularly that I am assuming that larger units like clauses and sentences

have comp ound argument structures akin to the content corresp onding to these

units in HPSG or their astructure corresp ondent in LFG argument structure is not

simply a feature of lexical items

While grstructure and astructure are syntactic representations I see them as

ultimately resulting from the grammaticization of two dierent sets of relationships

in the world Grstructure results from the grammaticization of discourse roles In

particular there is a distinguished entity at grstructure which is often called the

sub ject but which I often call the pivot following Dixon and I take it that

this pivot is derived by grammaticization from the notion of the topic or fo cus of

an event Because of this there is still an asso ciation b etween sub jectho o d and topi

cality Keenan b although since we are now dealing with a syntactic

notion not all pivots are topical and languages often have additional pro cesses of

topicalization

Argumentstructure in turn results from the grammaticization of notions of seman

tic prominence roughly along the lines of the protoagent and protopatient prop erties

of Dowty The distinguished entity at astructure I will call the asub ject

Typically asub jects are the argument that has the most protoagent prop erties but

astructure is also a syntactic level and so there can b e atypical asub jects like the

asub jects of suer undergo and receive

Given the two levels of astructure and grstructure we need to determine two

mappings The rst would determine the argument structure of a verb or other

predicator based on its meaning Let me call this a theory of argument pro jection

Secondly we need a mapping from astructure to grstructure which I will refer to

as a linking theory In this work my main aim is to motivate the two levels and

I only briey discuss the mappings Ab out the theory of argument pro jection it

See among others Bresnan and Mchombo for the evolution of grammaticized topics

Note however that the term linking theory is sometimes used to refer to something more like

a theory of argument pro jection or a theory that maps directly from the semantics to surface

grammatical relations for instance in Alsina I b elieve that the ergative syntactic phenom

ena that I study here show quite clearly that it is wrong to try to do linking directly b etween the

semantics and surface grammatical relations Rather an indep endent level of argument structure is strongly motivated and the the factorization of the mapping into two parts is to b e greatly preferred

Chapter Cutting the ergativity pie

will b e sucient to know that agents and exp eriencers normally b ecome asub jects

but see Dowty for development of a linking theory which I regard as generally

compatible with what I present here The p ossibilities for linking are discussed briey

b elow

Approaches to ergativity

If one is working in a categorical or discrete framework as opp osed to a fuzzy or

squishy one then regardless of whether the terms of reference are congurational

p ositions grammatical relation lab els or p ositions in sub categorization lists there are

only so many fundamentally dierent approaches for capturing ergative phenomena

As far as I am aware a clear classication of the p ossible approaches was rst provided

in the RG work of Johnson and Postal While analyses have many

variants in terms of other elements of the framework they assume or the level at

which various groupings o ccur it is notable that most subsequent analyses of ergative

languages can b e classied into one of the categories that Johnson and Postal outlined

Here I will present a slightly augmented and reworded classication into six classes

of treatments of ergativity At the end of each section I will mention analyses that

could b e regarded as exemplars of the approach

Syntactic Accusativity At the level of grammatical relations the A and S argu

ments are in the sub ject p osition while O arguments are in the ob ject p osition

In this analysis grammatical relations are the same as in accusative languages

and the language is only morphologically ergative This is the analysis of Ander

son for all ergative languages except Dyirbal and Hurrian the analysis

prop osed by Chung for certain Polynesian languages including Samoan

and the analysis prop osed for morphologically ergative languages by Marantz

Johnson argued that this was the correct analysis for Dyirbal

Postal states his classes in terms of initial grammatical relations but this was b efore

the integration of the Unaccusative Hyp othesis Perlmutter into RG To maintain Postals

denition of the classes one has to examine the level after unaccusative Ss have advanced to b e s

Many of these analyses are describ ed in more detail later others are exemplars listed by Postal to which the reader is referred for references

Chapter Cutting the ergativity pie

and most likely for all languages

ErgativeasPassive An ergative clause is the result of obligatory passivization of

transitive verbs The details dep end on the treatment of passivization in a

framework but b efore passivization applies the p osition of A S and O NPs is

the same as in accusative languages while after passivization of transitive verbs

S and O NPs are sub jects and A NPs are agentive obliques This analysis is

sometimes asso ciated with Schuchardt but by referring to transitive sentences

in ergative languages as having a passive character he seems rather to have

b een adopting the Oblique Analysis discussed b elow see Plank fn

This analysis is considered for some Australian languages esp ecially Warlpiri

in Hale although in his concluding remarks he doubts that such an

analysis will prove to b e correct It app ears to b e the analysis Anderson

adopts for Dyirbal though the reference is brief and it o ccasionally surfaces

in other stillb orn analyses in the generative literature such as George

and Jake

AbsolutiveSasOb ject At a certain level of structure A NPs are treated as

sub jects while b oth O and S NPs are treated as direct ob jects Perhaps the

rst prop osal of this form app ears in Tragers analysis of Taos Tanoan

New Mexico USA Postal sees the analysis of Mohawk in his own dissertation

and an analysis of Chino ok by Silverstein in this class A clear recent exemplar

of this approach is Bobaljik Larsen is also an analysis of this

type as is Legendre et al

The Oblique Analysis The language has no transitive clauses The O and S ar

guments are treated as sub jects while A arguments are obliques functioning

something like instrumental NPs Postal saw the elements of this analysis in

prop osals by Martinet for Basque and by Jacobsen Clear recent advocates

of this analysis are Mel cuk for Lezgian and BokBennema

and Kiparsky for Dyirbal The surface nature of this analysis is like

the ErgativeasPassive account but it do es not p ostulate passivization as the

means by which this conguration is derived

Chapter Cutting the ergativity pie

The Inverse Analysis In this analysis S and O NPs are sub jects and A NPs are

direct ob jects This conguration is often but not necessarily seen as b eing

set up b efore valency changing rules apply so that antipassive in syntactically

ergative languages can b e said to b e the same rule as passive in accusative

languages This approach is advocated by Dixon for Dyirbal a line

of research in Montaguecategorial grammar that has examined Dyirbal and

QuicheSchmerling Trechsel Dowty Dowty apparently

by Mel cuk for Dyirbal and Marantz in his treatment of languages

that he classes as syntactically ergative Dyirbal and Central Arctic Eskimo

The Four Relations Analysis The level of grammatical relations must recognize

all of the grammatical relations sub ject grouping A and S ob ject O NPs

absolutive grouping O and S and ergative A NPs Dierent languages will

have dierent rules sensitive to various of these grammatical relations This is

the analysis of Postal although he regards the grammatical relations

sub ject and ob ject as primitives while ergative and absolutive are dened rela

tions and it is also the analysis of Woo dbury a

It is quite p ossible that the same analysis is not correct for all languages that have

ergative forms in their morphology Splitting languages into dierent groups with re

sp ect to this classication has already b een adopted by Anderson and Marantz

I will advocate a split as well

Certain analyses of ergativity cannot b e simply placed into one of the ab ove cat

egories b ecause they are attempting to capture a combination of ergative and ac

cusative phenomena not by the the metho d of the Four Relations Analysis but by

having dierent levels organized accusatively and ergatively However generally the

dierent elements of the analysis can individually b e analyzed in terms of the cate

gories listed ab ove This applies to the analysis of syntactically ergative languages

that I present I adopt the inverse analysis at the level of surface grammatical rela tions but assume an accusative organization for all languages at the level of argument

Chapter Cutting the ergativity pie

structure

It would b e premature to evaluate these ve alternatives here but let me just

mention a few p oints that can b e made briey The p erceived absence of passives

antipassives in ergative languages has sometimes b een taken as an argument in favor

of the ErgativeAsPassive analysis as passive has already applied we should not

exp ect it to apply again However while the languages of the Caucasus tend to

lack passivesantipassives it is now wellknown that many ergative languages have

these op erations It has b een suggested that an advantage of the Inverse Analysis

is that antipassive in a syntactically ergative language b ecomes the same op eration

as passive in a syntactically accusative language Marantz But this analysis

also has only limited app eal once it is noticed that many languages have b oth passive

and antipassive for example the Mayan and Eskimo languages Both the Inverse

Analysis and the Oblique Analysis are antithetical to RG In RG the Universal

Alignment Hyp othesis requires that an A starts as an initial and an O as an initial

The relationships of the Oblique Analysis could thus only b e generated at a later

stage and this would mean that an ErgativeAsPassive analysis results Short of

unmotivated use of s as a holding space there is no way to generate the Inverse

Analysis at all as whichever of or rst advanced or retreated should send the

other relation en chomage

What I wish to prop ose in this thesis is that there are two types of ergative

languages syntactically ergative languages and morphologically ergative languages

I am prop osing a much enlarged class of syntactically ergative languages as outlined

ab ove and for all these languages I prop ose that the Inverse Analysis is correct

For the remaining morphologically ergative languages I naturally adopt Syntactic

Accusativity Related to the enlargement of the class of syntactically ergative lan

guages is the prop osal of an enlarged role for argument structure which always has

an essentially accusative organization I prop ose that binding theory control and

the determination of imp erative addressees is in general sensitive to argument struc

ture and hence roughly accusative I will call my analysis of syntactic ergativity the

I think roughly the same statement can b e made ab out the treatment of Inuit by Bittner though the classication of this analysis is admittedly less clear

Chapter Cutting the ergativity pie

Inverse Grammatical Relations analysis in an attempt to emphasize that while the

grammatical relations are inverted in such languages argument structure remains the

same as in all other languages

Argument pro jection linking and valence changing op erations

While many theories of linking are designed on the assumption that agents should

always b ecome sub jects adoption of an Inverse Grammatical Relations analysis means

allowing much greater exibility in the linking b etween astructure p ositions and the

level of grammatical relations so that agents can b ecome ob jects for instance As

Kro eger notes this exibility do es not fall out of most current theories

of linking but as I suggested ab ove I think that progress can b e made when we

clearly separate the two mappings of argument pro jection and linking with argument

structure b eing the level at which they meet

At least three linking p ossibilities seem to b e observed crosslinguistically West

ern Austronesian languages allow great exibility in which argument at astructure

b ecomes the pivot Dep ending on the voice marker chosen a variety of mappings are

p ossible without demotion of higher arguments to oblique roles Secondly there are

many languages which always use a straightthrough mapping in which the oblique

ness ordering of terms is the same at astructure and at grstructure This gives

syntactic accusativity

grstructure astructure

subj asub ject agent

obj patient

Finally there are languages that always use an inverse mapping for transitive verbs

like this

grstructure astructure

subj asub ject agent

obj patient

Such languages are syntactically ergative In b oth these last two language types

I am unaware of any syntactically ergative languages that have ditransitive verbs where all

Chapter Cutting the ergativity pie

the mapping is invariant and dierent realizations of arguments at grstructure can

o ccur only as a result of derivational morphology like passive that leads to comp ound

argument structures

Let me lay out more explicitly my conception of the level of syntacticized ar

gument structure In my framework argument structure is a syntactic level and

valence changing op erations are op erations on argument structure The basic argu

ment structure for a verb is an ordered list of the verbs arguments for example

a yawnh i John yawned

i Sarah nished her book b nishh

c presenth i Judith presented an award to Cynthia

There are two principles governing the obliqueness ordering of arguments within a

single level argument structure Firstly direct arguments precede obliques If desired

this division can by indicated explicitly by setting o oblique arguments with a bar j

Thus the two argument structures of give are

a giveh j i Stacey gave the chocolates to Simon

i Stacey gave Simon the chocolates b giveh

At any rate we need an oblique argumentdirect argument distinction since various

subsystems of language such as case marking are sensitive to it Within each of the

direct and oblique arguments arguments are ordered according to thematic oblique

ness The thematic hierarchy of Bresnan and Kanerva will b e sucient

Ag Ben RecipGoExp Inst ThPt Lo c

This separate ordering of direct and oblique arguments has b een previously moti

vated by Hellan and adopted by Dalrymple Note that it also

three arguments are terms and so the interesting question of how the mapping would lo ok in such

cases app ears not to arise

I call it syntacticized since working out the consequences of these prop osals for languages with

expletives would app ear to require representing the expletives at argument structure which is pre

cisely what some advocates of more semantic argument structures eg Alsina do not

wish to do See Bresnan and Zaenen for indep endent evidence from resultatives that nonthematic arguments interact with the rest of argument structure

Chapter Cutting the ergativity pie

explains the Toba Batak binding facts mentioned in Section without claiming

that thematic obliqueness is dierently dened in Toba Batak

In my framework valence changing op erations result in complex nested argument

structures T Mohanan suggested a representation for apparently mono clausal

or clause union causatives where they are biclausal at argument structure but mon

o clausal at the level of grammatical relations and this has b een widely adopted by

others Alsina and Joshi Alsina Butt The argument structure of

causatives is then said to b e either a or b where the rst argument of cause

is the causer and the second is the causee The causee role is fused with one of the

arguments of the base predicate and the alternation in which role the causee is iden

tied with corresp onds to the two types of causatives familiar from many languages

roughly those with an indirect ob ject versus an oblique expression of the causee

when a transitive verb is causativized

a causeh lo okafterh ii

b causeh lo okafterh ii

This analysis of causatives is similar at a high level to a GB analysis such as

the one in Baker the causative verb is a higher predicate and there is some

form of predicate comp osition This is in contrast to theories which have b een based

around the demotion or promotion of certain arguments such as Williams and

Zubizarreta The analysis of causatives contrasts with Bakers analysis

of applicatives where applicatives are introduced as a sub ordinate predicate namely

a prep osition However Austin has argued from Australian data that the

correct analysis of applicativescomitatives is also as a higher predicate roughly along

the lines of where the applicative morpheme introduces the higher predicate

affect

affecth separateh ii

36 See also Alsina f for a related but slightly dierent prop osal

Chapter Cutting the ergativity pie

One argument for this approach is that Austin notes that the same morphemes are

used for applicatives and causatives in many Australian languages and so a similar

representation is to b e preferred

Now for passives many p eople have assumed that the argument structure of the

verb is roughly as in such is roughly the account of Grimshaw Bresnan

and Zaenen and Kiparsky The argument structure is the same as for

the active verb except that the logical sub ject or external argument has b een

suppressed This is marked by the empty set symbol and indicates that such a role

cannot b e expressed by a direct argument

i lo okafterh

In contrast I would like to suggest that the argument structure of passives should

rather b e as in Again passive is now represented as a higher predicate that

mo dies the argument structure of the basic ro ot The single nominal argument

of the passive axs argument structure is identied with the patient of the stems

argument structure

passh lo okafterh ii

In Korean it is almost always the same morpheme that is used to form passives

and lexical causatives This provides an argument in favor of this representation

of passives which is parallel to Austins argument for choosing a representation for

applicatives

What then of antipassives Antipassives are regarded as an abstract noun by

Baker so again an antipassive is a sub ordinate predicate However I will

analyze antipassives in parallel with passives as shown in

antiph lo okafterhj ii

Essentially the same representation for the passive is prop osed by Pinker Although

the many dierences in framework make precise comparison dicult there is also a clear conceptual relationship with the transformational analysis of Hasegawa

Chapter Cutting the ergativity pie

The lower patient b ecomes an oblique At present I will just stipulate this

Conceptually I b elieve that there is a nice parallelism in these prop osals All

argument structure changing op erations now introduce higher predicates and could

b e viewed as combinators in the sense of Hindley and Seldin As we will

see so on the binding facts of Inuit and other languages provide imp ortant evidence

that this approach is correct Below I tend to write astructures as attributevalue

matrices but this is just a notational dierence of no theoretical imp ortance The

following should b e taken as two dierent depictions of the same argument structure

a askhshe him lo okafterh childrenii

ask

asb she

arg him

lookafter

xarg asb

arg children

Note that under my analysis antipassive in a syntactically ergative language is not

the same op eration as passive in a syntactically accusative language b ecause valence

changing op erations apply to argument structure and not to surface grammatical re

lations Rather passive in a syntactically ergative language is the same op eration as

passive in a syntactically accusative language This is a welcome result terminolog

ically it means that my technical notions corresp ond with the established terms in

the typological literature whereas the prop osal of Marantz that antipassives

in syntactically ergative languages were really passives created obvious terminological

chaos Indep endent of this I b elieve it is also the right result evidence presented

in Chapter shows quite clearly that passive and antipassive are argument structure

op erations

The remainder of this section explores a couple of connections with other prop os

als Firstly I briey address how what I have just prop osed ts certain predictions of

HPSG but also causes some problems The nal subsection then deals with reconcil ing the three primitives of Dixon A S and O with the more binary nature of most

Chapter Cutting the ergativity pie

generative theories of grammar I examine prop osals for four primitives and suggest

that they can b e more smo othly integrated

HPSG

That at least some parts of binding may b e sensitive to argument structure or simply

thematic relations has b een suggested in many places including work by Jackendo

Wilkins and Williams on English Sells on Albanian

Hellan on Norwegian Grimshaw on Japanese and Dalrymple

and Joshi on Marathi Nevertheless most work on binding in GB assumes that

binding is dened over congurations at sstructure following Chomsky and

most work on binding in LFG such as Dalrymple argues that the syntactic

domains that constrain binding p ossibilities are dened on fstructure the level of

surface grammatical relations

In contrast it seems to me that HPSG the framework used in Pollard and Sag

Chapter and within other more recent work is actually predicting a split of

prop erties with al l binding sensitive to argument structure even if this prediction is

p erhaps somewhat an historical accident Let us briey review recent developments

in HPSG In HPSG the rest of Pollard and Sag there was a unied subcat

list which listed all the sub categorized arguments of a verb or other predicator For

example the transitive verb likes had the lexical entry in many details omitted

likes

head n

verb

subcat h NPnom NPacci

In HPSG following work by Borsley the unied subcat list is divided

into three lists for complements comps sub jects subj and sp eciers spr The

Sub categorization Principle is reformulated as the Valence Principle which ensures

that arguments on all three lists are correctly disp ensed However these three new

valence lists have not replaced the old subcat list as Borsley p erhaps originally

intended The subcat list has b een retained in the sign of lexical entries as the

lo cus of binding theory among other things Canonically the values of a words

Chapter Cutting the ergativity pie

valence features add up via the app end relation to the words subcat value but

other noncanonical relationships are allowed for example in prodrop languages

This gradually led to the realization that the old subcat list had b ecome simi

lar to notions of argument structure in other frameworks esp ecially the somewhat

syntacticized argument structure suggested by Kiparsky lectures and

in recent work the old subcat list has b een renamed args to reect this Sag and

Go dard Iida et al The result is that argument structure is distinguished

from a words valence as illustrated for the English word likes in

likes

head n

verb

NPnom i subj h

comps h NPacci

args h i

The imp ortant observation here is that if a language diered from English in its

assignment of verbal arguments to the valence lists for example a syntactically

ergative language as I am dening it would have the mapping in then HPSG

is predicting that everything that dep ends on subcatargs in particular binding

theory will b e unchanged

likes

head n

verb

NPabs i subj h

NPergi comps h

args h i

That is in an argument structure based theory of binding just falls out of the

current HPSG formalism see for instance Iida et al

What ab out control If I suggest that the least oblique element on the args

list is the controllee of complement or adverbial clauses it might seem as if control

theory also falls out of the current framework However this do esnt actually work

Indeed as a technical p oint for those wellversed in HPSG control theories this prop osal allows

an argumentstructurebased theory of control that can handle Philippine languages like Tagalog

Chapter Cutting the ergativity pie

in the current formalism since the args is not carried up to phrasal signs only the

unsatised members of the valence lists Suggesting that the args of the parent of a

headed phrase is the same as that of the child would destroy a ma jor lo cality principle

of HPSG whereby currently a no de can only see unsatised valence list entries of a

child or sibling

Nevertheless for my theory to work I need a parent or sibling to b e able to see the

asub ject of a clause rather than or in addition to the pivot in control relationships

Two choices are available to me Either I could just ignore lo cality and pass up the

args to parent no des anyway by making args a head feature or a part of content

that is always inherited from the head child of a headed phrase Some evidence to

supp ort this approach could p erhaps b e gained from phenomena that have b een said

to b e sensitive to logical ob jectho o d Alternatively I could introduce an attribute

maxasubj which is inherited b etween the parent and head of a headed phrase and

which is token identied with the highest asub ject of the heads argument structure

Under this prop osal the lo cality that results from HPSGs current feature geometry

would b e undermined less The prop osal p erhaps also derives a little supp ort from

the fact that while binding is commonly sensitive to all asub jects control and

imp erative addressee is only ever sensitive to the highest asub ject in an argument

structure

Also the subj and comps lists as currently congured in HPSG do not correctly

capture what I take to b e the most imp ortant divisions among grammatical relations

a division rst b etween terms otherwise known as core roles and nonterms followed

by a second cut that distinguishes a certain term as in see Alsina

among others for motivation of this classication I will therefore make use of

even on the Pollard and Sag Chapter treatment of control it is not necessary to revert to

Sag and Pollards version which worked by adding an external argument extarg attribute

to content

See T Mohanan for motivation of the notion logical ob ject Within my prop osal the

logical ob ject can b e dened as the most oblique term As an example of a phenomenon to study

Joshi f argues that the gap in prenominal past participial phrases which is identied

with the nominal head is always the logical ob ject but it may b e p ossible to capture this notion purely through semantic entailments

Chapter Cutting the ergativity pie

keyword grammatical relations in what follows

terms nonterms

pivot nonpivot

ie sub ject ie obj core

While I mean by pivot neither more nor less than is understo o d by the term gram

matical sub ject I will henceforth use the term pivot since as Dixon has observed

it can b e very confusing to refer to a grouping of O and S as the sub ject I will then

lab el other core roles simply as core this grammatical relation is fully equivalent

to object in standard systems of grammatical relations

Except for the qualications expressed ab ove I b elieve that my theory can b e

easily implemented within HPSG It also ts naturally into an LFG framework With

lesser or greater work dep ending on the case I b elieve the main ideas could in fact

nd their way into any syntactic framework

The four underlying primitives mo del

Dixon argues for the primacy of A S and O as syntacticsemantic primi

tives but this tripartite division is somewhat at o dds with the binary divisions most

prevalent in generative grammar In particular most frameworks would want to

map these three classes into four categories given by the crosscutting distinctions b e

tween transitive and intransitive verbs and b etween internal and external arguments

Dixon notes that some languages mark some Ss like As and some like Os

and introduced the notations S and S for such cases Such languages can b e referred

a o

to as ActiveStative or by just one of these dep ending on which is the marked term

Dixon feels that any such division of intransitives is secondary to the fundamental

division of actants into A O and S but his arguments for the primacy of S seem less

That is grammatical relations have names rather than b eing dened by a notion such as oblique

ness in a SUBCAT list

More precisely it is fully equivalent in systems that allow multiple ob jects undierentiated by grammatical relation for instance the system of Alsina

Chapter Cutting the ergativity pie

than convincing More recent work suggests that this language type deserves a fuller

treatment many of the languages traditionally describ ed as ergative such as Basque

Levin f and Georgian see Harris but also the critique in Hewitt

actually have an at least partly activestative character and other languages

such as Eastern Pomo Hokan California McClendon GuaranTupGuaran

Paraguay VelaquezCastillo and Acehnese Austronesian Sumatra Indonesia

Durie app ear to b e organized on activestative lines

In contrast to Dixon some others in the typological literature such as Kibrik

have worked with a basic four term system of syntacticsemantic primitives

distinguishing one and two argument verbs and agent and patientlike arguments

Argument type

Agentlike Patientlike

One argument A S O S

i a i o

Two argument A O

t t

This kind of system where the basic division is into agentlike and patientlike terms

compare also Foley and Van Valins central use of the notions of Actor and Un

dergo er is much more compatible with what is normally assumed in generative gram

mar following the adoption of the Unaccusative Hyp othesis of Perlmutter

where we can map A and O onto the externalinternal argument distinction at D

structure in GB or initial s and s in RG This organization seems to give a more

natural representation of activestative languages while not precluding the treatment

of ergative and accusative languages Thus it do es not app ear that Dixons argu

ments for the use of A S and O necessitate revision of the standard conceptions of

Dstructure or argument structure in generative grammar

I b elieve it may turn out most insightful to analyze b oth activestative languages

and the other common system contrastive marking where all of A S and O are

Dixon presents cases where ActiveStative languages treat all Ss the same or all SA NPs

or all SO NPs the same But it is not clear that such evidence is inconsistent with a mo del that

provides a four way classication at argument structure

Although it must b e remarked that dividing verbs into unergative and unaccusative classes is not

without problems Levin Zaenen Ultimately verbs may need to b e crossclassied

by semantic features rather than there b eing a binary division of intransitives into unaccusative and unergative classes

Chapter Cutting the ergativity pie

marked dierently in terms of either an extended ergative case or an extended ac

cusative case as suggested in Bittner and Hale forthcoming b However I do not

examine activestative languages further in this dissertation

An analysis of binding

Binding is generally taken to require some kind of command relationship dened on

some form of hierarchical structure I will argue here for binding relationships dened

on a form of syntacticized argument structure the command relationship will thus

b e called acommand I dene it formally in Chapter

If the binder of reexives or the controller of the gap in certain adverbial clauses

is restricted my claim is that it will generally b e restricted to arguments that are

the most prominent at some level of argument structure For an active verb this will

b e the logical sub ject for a passive verb either the logical sub ject or the promoted

patient for a causative verb either the causer or the causee but for antipassives just

the logical sub ject since it is the highest argument at b oth levels of the argument

structure I will call arguments that are most prominent rst on some level of

astructure asub jects I take this notion as distinct from logical sub ject b ecause

Jesp ersens use of logical sub ject was clearly meant to indicate the agent of a passive

to the exclusion of the patient

My rst approximation to a universal binding theory is then in

a Command is dened on astructure

b If antecedence is restricted it is restricted to asub jects

c If antecedence dep ends on obliqueness it is obliqueness at the level of a

structure

Note that in general I am suggesting that phenomena that have b een analyzed as

sometimes sensitive to logical sub jects are in fact sensitive to asub jects and should

in general pattern the same for all asub jects Later I will discuss the extent to which

this prediction is conrmed

44 For the impatient the answer is nearly always but not exceptionlessly

Chapter Cutting the ergativity pie

Binding in Tagalog

Tagalog provides some initial plausibility for this approach It app ears that regardless

of which NP is the sub ject the ang marked NP an NP can bind thematically more

oblique NPs ab but not the other way around c where thematic obliqueness

is dened via some version of the conventional thematic hierarchy An ob ject binding

a grammatical sub ject is shown in a while b shows a recipient able to bind a

grammatical sub ject theme

a Iniisip nila angkanilang sarili

dvthinkab out genthey nomtheir self

They think ab out themselves

b Nagiisip sila sakanilang sarili

avthinkab out nomthey dattheir self

They think ab out themselves

c Iniisip sila ng kanilang sarili

dvthinkab out nomthey gentheir self

a Sinaktan ngbabae angkaniyang sarili

dvhurt genwoman nomher self

Athe woman hurt herself

b Sinabi niJuan kayMaria angkatotohanan tungkol

perftellov genJuan datMaria nomtruth ab out

sasarili niya

datself sg

John told Mary the truth ab out self

i j ij

That binding in Tagalog is sensitive to nothing other than argument structure

obliqueness follows I would argue from the fact that Tagalog has no argument struc

ture aecting op erations such as passive and antipassive Under my analysis if these

op erations existed they should aect binding p ossibilities To the extent that the

However there are causatives These do aect the argument structure and b oth the causer

and the causee can bind a reexive See for instance Miller for Tagalog Bell for Cebuano

Chapter Cutting the ergativity pie

framework in general seems reasonable this could b e taken as evidence for Kro egers

analysis of Tagalog over others where various of the voices are claimed to b e passives

or antipassives

Binding in Inuit

Kro eger provides considerable evidence that the choice of which Tagalog NP

to call sub ject the Actor or the Topic is principled and not arbitrary Returning to

Inuit I b elieve I can make the same sort of arguments A closer examination of the

Inuit evidence also suggests that prop erties of the absolutive should b e regarded as

grammatical sub ject prop erties while apparent prop erties of A and S are correctly

treated in argument structure terms

Earlier it was suggested that binding in Inuit was accusative which could b e

captured by prop osing a sub ject grammatical relation grouping S and A However

more careful analysis due mainly to Bittner but see also Sado ck forthcoming

and Woo dbury a shows clearly that neither control in adverbial clauses nor

binding is controlled just by A and S NPs but is also controlled by other asub jects

such as passive agents and causees I will very briey examine binding here and put

aside the largely similar control facts A full analysis of b oth app ears in Chapter

Inuit has b oth lexical pronouns and reexives and agreement suxes that mark the

same pronominalanaphoric distinction the socalled rd p erson axes are obviative

like pronouns while the socalled th p erson markers require to b e b ound in the

same manner as reexives This rdth p erson alternation app ears with b oth the

agreement suxes of sub ordinate verbs and suxes to a head noun that agree with

the p ossessor

The ma jority of cases of binding have the anaphoric pronoun or sux b ound by

an A or S NP In addition to the examples in ab ove a shows an overt

reexive b ound by an S NP and b shows a reexive p ossessive ending b ound by

an A NP

I do not illustrate here a direct equivalent to a plain transitive such as John cut himself b ecause

a languageparticular condition on binding of coterms in Inuit means that such a sentence has no

syntactic form preserving translation See Chapter for much more on this and other asp ects of

Chapter Cutting the ergativity pie

a imminut uqarvigaanga

selfterm sp eaktoindintrsg

I sp oke to myself

b palasip nulii tuquppaa

priesterg wifesgabs killindtrsgsg

The priest killed his wife

Inuit allows long distance antecedents so the binder need not b e in the same clause

but can b e in a higher clause

But these are not the only p ossible binders Other asub jects can also bind re

exives There are two classes of cases to consider One is derived verbs containing

double transitive suxes Kleinschmidt that is causative and similar suxes

including in Inuit suxes with the basic meanings of think and ask The other

case is passives

Given the argument structure representations and binding theory that I have

developed ab ove the prediction is that b oth the agent and the surface sub ject of

passives should b e able to bind NPs that they acommand but the agent cannot bind

the surface sub ject for example b ecause it do es not acommand it since the higher

lo cation of the patient is not acommanded by the agent consider again the prop osed

argument structure shown in

passh b eath ii

Indeed the patient can b e seen to acommand the agent in the argument structure

of the passive This correctly predicts the licensing of the reexive p ossessor shown

Hansi nuliaminit unatarniqarpuq

Hansiabs wifesgabl b eatpassindintrsg

Hansi was b eaten by his wife

i i

An example illustrating binding by either asub ject of a passive is shown in binding in Inuit

Chapter Cutting the ergativity pie

Na ja Tobiasimit uqaluttuunniqarpuq taassuma

Na jaabs Tobiasabl tellpassindintrsg demsgerg

itigartissimagaani

turndownprfprttrsgsg

Na ja was told by Tobias that he had turned self down

j i k ij

The grammatical relations and astructure for this sentence are shown in

pred b e told

h i

pivot pred Na ja

h i

obl pred Tobias

pred turndown

h i

pivot pred pro

comp

h i

core pred pro

grs

pass

asb Na ja

tell

asb Tobias

arg

xarg

turndown

arg asb pro

arg re

Here the reexive can b e b ound by either asub ject asb or asb

Similarly for derived forms with double transitive axes the complex argument

structure of anginirusinnaannginnirarpaa shown in allows an oblique reexive

to b e b ound by either asub ject a whereas a reexive asso ciated with a simplex

verb with the same surface arguments can only b e b ound by the ergative b

sayh couldnotb ebiggerh ii

It actually cannot b e b ound by asb b ecause of the coterm disjointness condition mentioned

3 in the previous fo otnote

Chapter Cutting the ergativity pie

a Kaalip Pavia imminit

Kaalierg Paviaabs self abl

anginir u sinnaa nnginnirar p a a

big cmp becan neg say indtr sgsg

Kaali said that Pavia couldnt b e taller than self

i j ij

b Juunap Kaali imminik uqaluttuuppaa

Juunaerg Kaaliabs selfinstr tellindtrsgsg

Juuna told Kaali ab out self

i j ij

While I have shown that a reexive in a passive can b e b ound either by the agent

logical sub ject or by the surface absolutive sub ject the ab ove prop osal may still seem

undermotivated it might just b e the case that there is a disjunctive sp ecication of

p otential binders which allows b oth the ab ove p ossibilities Bittner provides

crucial evidence that this is not the case and I will present this evidence within my

framework in Chapter

Echoes in Accusative Languages

While the disso ciation b etween surface grammatical relations and argument structure

is shown most vividly in ergative languages of the type for which the Inverse Analysis

is appropriate it cannot b e stressed to o strongly that the argument structure based

account of binding and control in adverbial clauses that I have outlined here and

which I present in more detail in Chapter is not a prop osal designed only to account

for the strange binding b ehavior of ergative languages Rather I would submit that

an argument structure based approach to constraints on binding is necessary in any

monostratal theory of syntax that recognizes complex predicates formed from embed

ded argument structures such as much recent work in LFG including Alsina

Andrews and Manning and Butt and also certain work in HPSG such

as Iida et al Indeed the need for ma jor revisions in the binding theory

that have b een caused by the introduction of complex predicates has so far not b een

suciently appreciated

Thus in accusative languages as well we nd construal phenomena of binding

and control that are sensitive to all asub jects One language in which this can

Chapter Cutting the ergativity pie

b e seen clearly is Japanese these data are drawn from Iida et al Japanese

adverbial clauses in the nagara while form are generally describ ed as having their

sub ject controlled by the sub ject of the higher clause but actually their sub ject can

b e controlled by any asub ject including the causee of a causative a and the

logical sub ject of a passive b

a Tanakawa kodomotatini utainagara tegamio kakaseta

Tanakatop childrenpldat singwhile letteracc writecauspast

Tanaka made the children write a letter while he they sang

i j i j

b Hanakoga Tarooni arukinagara aisatu sareta

Hanakonom Tarooby walkwhile greet dopasspast

Hanako was greeted by Taroo while she he was walking

i j i j

Similarly with resp ect to binding theory Kitagawa suggested that b oth

the overt pronoun kare he and the zero pronoun little pro should b e regarded

as pronominal elements and sub ject to typical Principle B eects as shown in

Taroowa Hanakoni kareo sarakedasita

Tarootop Hanakodat heaccpro revealpast

Taroo revealed him to Hanako

However in the morphological causative construction as shown in either kare

or the zero pronoun in the lower ob ject p osition may b e b ound by the grammatical

sub ject The obviation condition is correctly stated in terms of argument structure

it is necessary and sucient that a pronominal in the lower ob ject p osition b e disjoint

in reference with the nearest asub ject which is the dative causee Kitagawa

a Taroowa Ziro oni kareo b engosaseta

Tarootop Ziro odat heacc defendcauspast

Taroo made Ziro o defend him

i j

ijk

b Taroowa Ziro oni b engosaseta

Tarootop Ziro odat pro defendcauspast

Taroo made Ziro o defend him

i j

ijk

Indeed much of the evidence in the literature for doing morphology in the syntax

Baker and b oth prior and subsequent literature in which all asub jects are

Chapter Cutting the ergativity pie

treated as sub jects at some stage in the syntactic derivation can b e reinterpreted as

evidence that the current approach pro ceeds along the right lines given a lexicalist

framework In other words for the case of passives I am accepting the argument

of Perlmutter from Russian Slavic Russia reexives that the passive must

have a complex representation of some sort Perlmutter argues that b oth the logical

sub ject and surface sub ject of a passive must b e a at some level b ecause either can

b e the antecedent of a reexive

a Boris mne rasskazal anekdot o seb e

Borisnom medat told joke ab out self

Boris told me a joke ab out himself

i i

b Eta kniga byla kuplena Borisom dlja seb ja

this b o oknom was b ought Borisinstr for self

This b o ok was b ought by Boris for himself

i i

I would merely argue that the correct representation is embedded argument structures

rather than multiple strata of grammatical relations the b ehavior of syntactically

ergative languages seems go o d evidence of this

In Sanskrit ancient IndoEurop ean India as well either a logical sub ject a

or a grammatical sub ject b can bind a reexive but binding by the logical sub ject

is preferred

a sarpas tenatmana svalayam ntah

snakenom heinstr selfinstr selfhouseacc broughtpasspartnom

The snake was brought by him himself to self s house

i i

b anrtam tu vadan dan dyah svavittasyamsam

untruthnom but tellingnom negernom selfprop ertygen partacc

But a p erjurer is to b e ned one eighth lit part of self s prop erty

i i

There are numerous cases from other languages that could b e added reex

ives in Cebuano Austronesian Philippines Bell Turkish Turkic Turkey

His other argument from Acehnese is awed see Durie

For a similar argument cf Grimshaw

This contrasts with Russian and Inuit where the grammatical sub ject is the preferred antecedent I have no explanation for this it may just b e conventional

Chapter Cutting the ergativity pie

Kiparsky Lithuanian Slavic Lithuania Kiparsky ChiMwini Bantu

Somalia Marantz Hindi IndoEurop ean India T Mohanan

Chamorro Austronesian Guam Baker one reexive swataah in

Marathi IndoEurop ean India Joshi control of the unexpressed

argument of gerund phrases and sense without phrases in Catalan Alsina

It should b e p ointed out that there are however some languages where a causee

or the logical sub ject of passives do es not have all the binding capabilities of other

sub jects For example K P Mohanan see also Marantz describ es

Malayalam as such a language

a amma kuttiyekkonte aanaye swan tam wittil wecce

mothernom childinstr elephantacc self s house at

nulliccu

pinchcauspast

Mother made the child pinch the elephant at self s house

i j k

ijk

b jo on iyaal meeri swantam wiittil wecce nullapp et tu

Johninstr Maryabs self s houseloc at pinchpasspast

Mary was pinched by John at self s house

i j

In cases like this I would like to suggest that binding is most likely restricted to a

sub jects that are also terms Alsina describ es a similar conjunction

of conditions as determining the binding of the Catalan reexive phrase per si sol on

ones own This reexive can b e b ound by either a grammatical sub ject which is

also an asub ject or a causee expressed as a term but it cannot b e b ound by either

the logical sub ject of a passive or an oblique causee

I know of only one rep orted case of an anaphor that can b e b ound by a logical

sub ject but not by a grammatical sub ject and that is the Marathi anaphor aapan

Dalrymple Joshi If such is indeed the case this binding

While either the logical sub ject or the grammatical sub ject of a passive may bind the reexive

apnaa the causee of morphological causatives may not Mohanan analyzes it as not b eing a logical

sub ject Its not clear to me how I wish to incorp orate this case into my theory

Some earlier work regarded only grammatical sub jects as p ossible antecedents of swataah but logical sub jects are now recognized as p ossible antecedents by b oth Joshi and Rosen and Wali

Chapter Cutting the ergativity pie

p ossibility cannot b e excluded entirely but it nevertheless seems very marked

I see no particular reason why this argument structure based account of binding

should not extend gracefully to other languages I will leave this as basically a promis

sory note but there are interesting comparisons The ab ove binding theory is largely

identical to the one prop osed by Pollard and Sag once the subcat list is

reinterpreted as argument structure args along the lines suggested ab ove HPSGs

lo cal ocommand corresp onds to the prop erty describ ed in c and ocommand is

derived from lo cal ocommand by incorp orating a notion of command as in a

Williams argues that binding holds at the level of theta roles although the

details of the approach are dierent Dalrymple collects a variety of

evidence for binding of coarguments largely dep ending on thematic sup eriority

It is standardly accepted that binding theory cannot b e dened solely in terms

of an ordering of roles on a thematic hierarchy For if this were the case we would

exp ect to b e go o d since agent theme

Himself was washed by Sam

But in my theory since passive applies at the level of argument structure giving

the argument structure for the sentence in it is predicted that this binding

relationship is imp ossible In the agent do es not acommand the highest p osition

of the theme shown in italics

passhhimself washhSam ii

So standard arguments for dening binding at the level of grammatical relations

or via sstructure command are not arguments against the current theory On the

other hand my prop osal can hop e to explain not only data from other languages

such as Tagalog but also apparent thematic conditions on binding in English such

as Jackendo suggested to explain the following data from Postal

a John pleases himself

b John likes himself

In English it could b e argued that it is a lexical gap that there is no nominative case reexive but the same argument can b e made more convincingly in languages that have nominative reexives

Chapter Cutting the ergativity pie

a Bill talked to Mary ab out herself

b Bill talked ab out Mary to herself

a Bill showed Mary herself in the mirror

b Bill showed Mary to herself in the mirror

See Dalrymple and Hellan for some suggestions along these

lines Both of these authors also argue for the p osition that terms acommand obliques

see also Sells As well as giving an account of the Toba Batak facts Sec

tion in English to o this is required to allow binding patterns like where

an oblique recipient is b ound by a direct argument theme whereas most thematic

hierarchies place recipient ab ove theme

a Mary explained John to himself

i i

b John introduced Bob to himself

i i

Conclusion

My theory is thus that in all languages construal prop erties such as binding im

p erative addressee and the controllee of adverbial and complement clauses should

b e determined at the level of argument structure that I have outlined This level

of argument structure has an essentially accusative nature and multiple asub jects

result at this level when valence changing verbal derivations o ccur such as passive

antipassive and causative

Indep endently of this in many languages one can motivate a surface structure

relation of grammatical sub ject or pivot It is language sp ecic whether certain op

erations will treat the pivot sp ecially or whether they apply to all NPs or to all

terms But the type of surface structure op erations that sometimes distinguish piv

ots include relativization role in sub ordinate clause restrictions on topicalization

Although sometimes there will b e a secondary syntactic condition such as requiring that the antecedent is also a term

Chapter Cutting the ergativity pie

fo cussing or cleft formation necessarily wide scop e or sp ecic element question for

mation restrictions questioned element launcher of quantier oat coreferential

omission in co ordination and raising

The obliqueness ordering at the level of grammatical relations can b e dierent to

the obliqueness ordering at the level of argument structure Because of this there are

two sorts of ergative languages In syntactically accusative morphologically ergative

languages the A and S NPs are the pivot while in syntactically ergative languages

the S and O NPs are the pivot Under my more restricted conception of what phe

nomena are p otentially sensitive to the surface pivot I have shown that there are

many syntactically ergative languages

We cannot exp ect to o much uniformity There is some irreducible variation For

example some languages with ergative case marking such as Inuit also have an

ergative pattern of agreement suxes it is hard to tell this from the third p erson

forms that have app eared ab ove but it is more obvious with other forms However

other languages with ergative case marking have an accusative pattern of agreement

suxes for example Warlpiri PamaNyungan Australia or Burushaski cf

Nevertheless I have tried to motivate two universally valid notions grammatical

sub ject pivot and asub ject I b elieve that there is a core intuition and a core set

of prop erties related with each and b oth can b e applied crosslinguistically

CHAPTER

Inuit West Greenlandic

Inuit is an appropriate language to fo cus on b ecause not only have the Eskimo

languages b een studied longer and in greater detail than most other ergative lan

guages notably including the pioneering work of Kleinschmidt but many

recent theoretical prop osals on ergativity are based largely or solely on data from

Eskimo languages The rst section will briey review some of the basic facts of Inuit

in particular West Greenlandic if no other variety is mentioned The next section

b oth presents much more data and shows how Inuit supp orts the distinction b etween

grammatical relations structure and argument structure argued for in the previous

chapter The nal part of the chapter then compares what I have presented to the

other main generative analyses of Inuit ergativity from the eighties and nineties

Basic background on Inuit

Genetic Aliation

Inuit b elongs to the EskimoAleut family The Eskimo languages in turn break up

into the Yupik languages sp oken in Alaska and the eastern tip of Russia and Inuit

the distribution of which includes northern parts of Alaska Canadian Inuktitut and

all Greenlandic languages including West Greenlandic Kalaallit o qaasii While the

emphasis b elow is on West Greenlandic most of what follows is applicable to all

Eskimo languages

Chapter Inuit West Greenlandic

Case Marking and Case Markedness

All nounspronounsdemonstratives etc are consistent with ergativeabsolutive case

marking Actually many forms such as st and nd p erson pronouns and plural

and where they exist dual nouns have noncontrastive forms with all of S A

and O marked the same but singular nouns distinguish absolutive and ergative The

absolutive is unmarked while the ergative has a sux The ergative is also used as

the p ossessive case The ablative mo dalis and terminalis are the three cases used for

displaced terms Further details ab out the names form and function of Inuit noun

cases are found in There is little consistency in the use of case names in the

Eskimo literature I will use the names in the left column except that the relative

case will usually b e glossed erg or gen dep ending on its function in a sentence

Alternative Case names

Absolutive Nominative

Relative Ergative Genitive

Mo dalis Instrumentalis Comitative Accusative Secondary

Terminalis Dative Allative

Ablative Distantialis

Lo cative Lo calis

Vialis Translative Prosecutive Perlative

Equalis Equative Aequalis Similaris Simulative Comparative

Conformative

Eastern Inuit Greenlandic

Case sg pl

Absolutive it

Relative up it

Mo dalis mik nik

Terminalis mut nut

Ablative mit nit

Lo cative mi ni

Vialis kkut tigut

Equalis tut tut

Woo dbury a suggests that an exception to this is plural demonstratives where a sux

ku is used only to mark O but this do es not corresp ond to the description of Fortescue

where ku is regarded simply as a demonstrative plurality marker to which optionally a is added in the relative case and collo quially also in the absolutive

Chapter Inuit West Greenlandic

Absolutive marks S and O

Relative marks A and p ossessors In niq nominalizations the relative can

mark S or O with participles it can mark A

Mo dalis means with in various contexts and is used for instrumentals It is

used for the demoted theme in the antipassive for the theme in unmarked

ditransitives see the discussion around for an NP representing

concealed questions or prop ositions or acting as a secondary predicator

They call him Ajuqimod and as the case that mo dies incorp orated

nouns

Terminalis marks directional towards and goals and b enefactives Stan

dardly used to co de the reexive theme argument in the intransitivized

reexive of semantically transitive verbs Expresses denite agent in im

p ersonal agent construction Used for the demoted causee sub ject when

a double transitive ax is added to a verb Section

Ablative Means from and is used for demoted agents in the passive

Lo cative marks lo cation or temp oral inaton

Vialis marks overthrough or a frequentative temp oral meaning

Equalis means aslike and can b e added after other case endings

Word Order

Though it has sometimes b een claimed that word or rather constituent order in

Inuit is free Johnson Bittner most authors accept that there is a neutral

ordering of constituents Fortescue Woo dbury Bittner This is as

in where Adv stands for sentence adverbials and Adv for predicate adverbials

An example sentence is shown in

Adv erg abs obl Adv Verb

However the theme of an antipassive sometimes also app ears in the Terminalis

This is now pretty much obligatory in standard West Greenlandic Other dialects use the

Mo dalis Woo dbury a suggests the choice is dialectal and insensitive to the semantic

value of the cases

In Yupik where the ablative is lacking the terminalis is used in passives and the mo dalis for

from Even just within West Greenlandic there is variation in oblique agent marking and any

of ablative mo dalis or terminalis is p ossible Woo dbury a Johnson fn

suggests that in Central Arctic the terminalis is used in certain semantically predictable cases as when stateofmind predicates love worry ab out are involved

Chapter Inuit West Greenlandic

Kaalip arnaq aalisakkanik tunivaa

Kaalierg womanabs shplmod giveindtrsgsg

Kaali gave the woman some sh

This unmarked order of NPs applies even in semantically biclausal constructions

with double transitive suxes of the sort to b e considered in Section Fortescue

Hansip quaq uatsinnut niriqquaa

Hansierg frozenmeatabs weterm eataskindsgsg

Hansi told us to eat the frozen meat

Inuit lacks sub ordinating conjunctions like that if There is no overt Wh

movement and there are no relative pronouns Word order is quite variable reecting

emphasis newness and heaviness Informally it seems that noun phrases are fronted

when they are a topic or fo cus or can b e placed after the verb when they are a fo cus

For simple sentences any order of the noun phrases and the verb is an acceptable

sentence

Complement clauses generally o ccur to the right of the verb Light complement

clauses can also precede the verb It app ears likely that the base p osition of comple

ment clauses is preverbal but that most clauses b eing relatively heavy app ear in a

right extrap osed p osition In general heavy constituents can b e extrap osed after the

verb

Given the typological patterns in observed word orders word order can b e taken

as very weak evidence for sub jectho o d in a language To the extent that PCV ie

SOV languages are much more common than CPV ie OSV languages this could

b e taken as evidence that the ergative NP is the sub ject Others such as Bittner

see Section have sought to downplay the imp ortance of word order

suggesting that the ergativerst order results from topicalization and thus that the

language is underlyingly PCV with an absolutive pivot

Termhoo d Passive and Antipassive

Both the Absolutive and the Ergative should b e regarded as terms or core arguments With the exception of the innitive mo o d they are b oth always crossreferenced on the

Chapter Inuit West Greenlandic

verb The unmarked word order app ears to distinguish b etween terms and obliques

as do es the binding theory as will b e discussed b elow Both the ergative and the

absolutive can also b e demoted to oblique roles the absolutive of transitive verbs

can b ecome a mo dalis by antipassivization as shown in b while the ergative

can b ecome an ablative or other cases by passivization bc There is further

discussion of the antipassive in Section

a Hansip inuit tuquppaa

Hansierg p eoplenom killindsgsg

Hansi killed the p eople

b Hansi inunnik tuqutsivuq

Hansiabs p eoplemod killantipindsg

Hansi killed p eople

a Juunap miiqqat paarivai

Juunaerg childrenplabs lo okafterindtrsgpl

Juuna is lo oking after the children

b miiqqat Juunamit paari ni qar p u t

childrenplabs Juunaabl lo okaftergerhaveindintr pl

The children are lo oked after by Juuna

c miiqqat Juunamit paari sa u pp u t

childrenplabs Juunaabl lo okafterppartcop indintrpl

The children are lo oked after by Juuna

There are two passives in Greenlandic the dynamic passive shown in b and the

stative passive shown in c Fortescue describ es the rst as unrestrict

edly pro ductive while the latter is somewhat less so A third tit pseudopassive

o ccurring mainly in the sp oken language is formed from the intransitive reexive

form of the causativepermissive double transitive ax John let himself get bitten

by a dog Diachronically the dynamic passive is formed from the niq nominalized

form of the verb followed by the p ostbase for have while the stative passive derives

from the passive participle verb form followed by the copular p ostbase and then

intransitive inection as shown in bc However it is unclear to what extent

these derivations are still part of the synchronic grammar of Inuit

Chapter Inuit West Greenlandic

Arguments for the absolutive b eing a surface pivot

In this section I will examine a number of phenomena that treat the S and O NPs in

Inuit as a class Taken individually dierent explanations could p erhaps b e found for

some of these phenomena but I wish to argue that the only way to give a coherent

explanation for this whole class of phenomena in Inuit and to explain the parallels

with other languages discussed in Chapter is to recognize an SO pivot in Inuit

Relative clauses

As noted in the last chapter the only relative clauses in Inuit are actually participial

nominalizations which are strictly clauseb ound Greenlandic has a series of par

ticipial mo o ds used b oth in sub ordinate clauses and in relative clauses note that

a dierent participial mo o d is used for intransitive and transitive relative clauses

In some other forms of Inuit the participial mo o d is also used in main clauses see

Section It is a somewhat surprising fact that participial relative clauses can

b e formed by relativizing on either the S a or O b of the embedded clause

but not by relativizing on an A argument c

a miiraq kamattuq

childabs angryrelintrsg

the child that is angry

b nanuq Piitap tuqutaa

p olarb earabs Piitaerg killtrpartsg

a p olar b ear killed by Piita

c angut aallaat tigusimasaa

manabs gunabs takeperfreltrsgsg

intended the man who to ok the gun

This immediately recalls the KeenanComrie Accessibility Hierarchy Keenan and

Comrie according to which if the gap in a relative clause can o ccur with

There are also some marginal examples of relativizing the p ossessor of S or O Bittner although I will not further discuss them

Chapter Inuit West Greenlandic

only one grammatical relation it is the sub ject relation This suggests that Inuit

absolutive NPs serve as the surface sub ject or pivot Inability to relativize on A

arguments cannot b e explained as merely a morphological gap there is a transitive

particple form but it cannot b e used to form relative clauses

Semantic scop eSp ecicity

It was mentioned in the last chapter that the pivot NP in Tagalog is normally denite

it is an NP whose reference is presupp osed Thus dierent voices cause contrasts

in presupp osed reference such as the following

a Kinain ko angisda

perfeatov Igen nomsh

I ate the sh

b Kumain ako ngisda

avperfeat Inom gensh

I ate some sh

This follows from the nature of pivothoo d Being an asub ject is a relation in a syn

tactic system but the diachronic ro ots of asub jectho o d are thematic prominence

Similarly although pivot is also a syntactic notion the origin of sub jectho o d is dis

course prominence This connection b etween sub jectho o d and presupp osed reference

has b een noted b efore in the literature A number of the sub ject prop erties of Keenan

b refer to this prominence He notes that sub jects tend to have absolute ref

erence A student owes John a report implies the existence of a student but not of

a rep ort p it is harder to susp end the presupp osed reference of sub jects than

of other NPs p sub jects are normally the topic of the sentence p p osi

tions that must b e lled by denite or highly referential NPs are likely to b e sub ject

p ositions p Keenan b also notes that sub jects preferentially have

wide scop e This claim is consonant with the results of Ioup Ioup rep orts

the preference hierarchy for certain grammatical relations to have wide scop e shown

underived sub ject derived sub jectlogical sub ject indirect ob ject

prep ositional phrase direct ob ject

Chapter Inuit West Greenlandic

See also Givon for further exploration of the idea that the sub ject although a

grammaticized syntactic category exists to co de referentialitytopicality at the level

of pragmaticsdiscourse

Of course as Kro eger notes it is not easy to sp ecify exactly

what the discourse prominence of the ang marked NP consists of It is not exactly

that the pivot has to b e denite Examples like with an indenite but apparently

sp ecic pivot are p ossible or the pivot can b e a generic

Kinuha niya ang isang aklat

perftakeov hegen nom onelnk b o ok

He to ok a certain b o ok

And it is not exactly that the sub ject must t in with typical discourse notions of

topic It can actually b e a newly introduced fo cus in answer to a question as in

Ano ba ang binili mo sapamilihan

what ques nom perfbuyov yougen datmarket

Binili ko itong damit

perfbuyov Igen thisnomlnk dress

What did you buy at the market

I b ought this dress

Nevertheless the sub ject NP has a strongly felt discourse prominence that has b een

remarked on by nearly all observers of the language

Exactly the same can b e said ab out the absolutive NP in Inuit The traditional

wisdom of Eskimologists is that the absolutive NP of a transitive verb must b e denite

This thesis originated with Kleinschmidt and has b een widely adopted

by later authors see for instance the discussion in Fortescue which

is roughly along traditional lines though with a careful discussion of the various

With some trepidation let me oer the following working denitions An NP is specic within

a certain context of use when it has unique individuation for the sp eaker I take this term as

synonymous with presupposed reference but sp ecicity do es not imply existence since the unique

individuation referred to ab ove is within the mentally pro jected world of the sp eaker An NP is

denite within a discourse context if it is given information if the sp eaker can assume it to b e

within the consciousness of the hearer This is hard to dene exactly in part b ecause of the use of

accommo dation as discussed by Lewis A generic expression uses an NP attributively to refer to a type

Chapter Inuit West Greenlandic

nuances and p ossibilities that o ccur The contrast is generally drawn b etween the

interpretation of the absolutive NP in transitive sentences versus the interpretation

of the mo dalis NP in the corresp onding intransitivized antipassive form Compare

for example the pairs in where the translations indicate how the contrast has

generally b een p erceived

a atuakkat atuarpai

b o okplabs readindsgpl

He read the b o oks

b atuakkanik atuarpuq

b o okplmod readindsg

He read some b o oks

a Si gmiam katinnia gaa a gnaq Canadami n I nupiaq

Si gmiaqerg marryfutsgsg womanabs Canadaabl

Si gmiaqis going to marry the a sp ecic Canadian woman

b Si gmiaq katinniaqtuq a gnamik Canadami n

Si gmiaqabs marryfutsg womanmod Canadaabl

Si gmiaqis going to marry a Canadian woman

However a statement simply in terms of deniteness has proven unsatisfactory

for Inuit as well as for Tagalog Bittner suggests that the dierence is

actually one of scop e Thus she shows that the absolutive NP of b oth an intransitive

and a transitive sentence must take wide scop e with resp ect to sentential

op erators such as negation whereas other NPs such as the ergative NP can take wide

or narrow scop e

atuagaq ataasiq tikissimanngilaq

b o okabs oneabs comeperfnegindsg

i One b o ok hasnt come yet

xb o okx comex

Kleinschmidt refers to these semantically transitive but syntactically intransitive forms as halftransitive

Chapter Inuit West Greenlandic

ii No b o oks have come yet

Juunap atuagaq ataasiq tigusimanngilaa

Juunaerg b o okabs oneabs getperfnegindsgsg

i There is a b o ok which Juuna hasnt got yet

xb o okx getJ x

ii Juuna hasnt got any b o oks yet

atuartup ataatsip Juuna uqaluqatigisimanngilaa

studenterg oneerg Juunaabs talktoperfnegindsgsg

i No student has talked to Juuna yet

xstudentx talktox J

ii One student hasnt talked to Juuna yet

xstudentx talktox J

The remainder of this subsection will contain further inconclusive discussion

of the virtues and pitfalls of various attempts to characterize these contrasts but

in terms of the main theme of this section the parallel with Philippine and other

languages should b e clear The absolutive NP in Inuit p ossesses the same prop erties

of discourse prominence or presupp osed reference as the pivot NP in Tagalog and

this is thus further strong evidence for recognizing a pivot or grammatical sub ject

in Inuit that groups S and O

Further discussion

As stated ab ove traditional wisdom held that in pairs such as those given ab ove in

that the absolutive NP in transitive sentence must b e denite while the mo dalis

NP of the intransitivized antipassive variant must b e indenite However it has

b een noticed that a simple division into denite and indenite NPs is not satisfactory

Johnson Bittner denite NPs such as prop er names can

app ear in the mo dalis case with antipassives and the translation equivalents of indef

inite ob ject NPs from Europ ean languages o ccur as absolutives in transitive clauses

This has led to various attempts to rene this distinction Kalmar attempts to

Chapter Inuit West Greenlandic

work in terms of given and new arguments of discourse but crucially used a culture

sp ecic denition of what is given in Inuit discourse Johnson describ es the

data in terms of pragmatic backgrounding and foregrounding We might also wish

to invoke notions such as sp ecicity or referentiality or the strongweak distinction

in NP readings argued for by De Ho op who includes some discussion of the

data from West Greenlandic

The most radical reformulation of this distinction is presented by Bittner

who attempts to explain this phenomenon in terms of semantic scop e in marked

contrast to Johnson which argues that this was a pragmatic rather than a

semantic issue Bittner observed certain interactions b etween NP scop e and senten

tial op erators such as negation expressed in Inuit as verbal suxes The following

table summarizes Bittners data on whether dierent items can have wide andor

narrow scop e with resp ect to negation and other similar sentential op erators

Scop e wrt op erator

Wide Narrow

Ergative agent of transitive X X

Ablative agent of passive X X

Nonovert agent of passive X

Mo dalis Instrumental of antipassive X X

Incorp orated theme of active intransitive X

Absolutive of intransitive including passive X

Absolutive argument of transitive X

The entry marked with an asterisk deserves futher comment Bittner stated

that the mo dalis NP of an antipassive must take narrow scop e but Bittner

suggests that either scop e is in fact p ossible It is merely that narrow scop e is preferred

for essentially Gricean reasons given the availability of the necessarily wide scop e

active transitive alternative Thus Bittner suggests the readings shown for

the active and antipassive pair in but the account of Bittner would hold

that the reading of a is also available to b but pragmatically dispreferred

a atuartut ilaat ikiurtariaqarpara

ofstudents oneofthemabs helpmustindtrsgsg

xx is one of the students it is necessary that I help x

Chapter Inuit West Greenlandic

b atuartut ilaannik ikiuisariaqarpunga

ofstudents oneofthemmod helpantipmustindintrsg

It is necessary that xx is one of the students I help x

Bittner suggests that these scop e facts all hold of a wide range of axal

op erators i negation nngit ii tense and asp ect ssa fut sima perf tar

freq lir incept and qqi again iii mo dals of necessity and p ossibility sinnaa

can gunar probably and tariaqar must and iv other mo o d op erators such as

the conditional interrogative imp erative and contingent eg ku cond and pi

interrog see Bittner for examples Further she suggests that

the same scop e facts hold with resp ect to adverbial phrase sentential op erators and

worldcreating predicates such as b elieve say that and lo ok for

While Bittner has provided a wealth of new data there are still some problems

i with her theory for explaining this data ii with certain data and iii with

resp ect to her arguments against other accounts

To explain the data Bittner suggests that the sstructure as she denes

it see Section acts as the default logical form for a sentence and that other

NP scop es are made p ossible by quantier raising QR which can only increase the

scop e of an NP with resp ect to sstructure Thus on her account the absolutive

NP is in Sp ec IP at sstructure ab ove sentential op erators like negation and so

must scop e outside such op erators whereas the ergative NP is a VPinternal sub ject

the distinguished adjunct of VP and so the ergative NP can scop e within sentential

op erators like negation or it can raise by QR at LF to scop e over sentential op erators

Bittner makes the following comparison b etween the Inuit data pre

sented in and the situation holding in English

In English sub jects are normally restricted to take wide scop e relative

to negation and other VPlevel op erators while ob jects can take either

narrow or wide scop e

a One b o ok hasnt come yet

b One student do esnt know John yet

c John hasnt received one b o ok yet

Chapter Inuit West Greenlandic

In English the scop e patterns exemplied in can b e plausibly

attributed to the ccommand relations which hold at sstructure that

is at the syntactic level which determines structural Case assignment

agreement and syntactic binding relations For English this correctly

predicts that the nominative sub ject will b e outside the scop e of VPlevel

op erators such as the negation since it is outside the VP at the

sstructure level Accusative ob jects on the other hand can take

narrow scop e by virtue of their VPinternal sstructure p osition Their

alternative wide scop e readings are due to the option of argument raising

at LF

However notwithstanding the inclusion of the hedge normally this scarcely seems

satisfactory as a statement of the situation in English and totally ignores Car

den a b and subsequent work There app ear to b e many examples in

English where many sp eakers allow the sub ject AS NP to scop e inside a senten

tialVP op erator

a All that glitters is not gold

b Someone always interrupts me when I start talking

Carden a b studies the interpretation of structurally identical sentences

such as those in discussing whether sp eakers assigned them a negq interpre

tation the quantied NP has scop e inside the negation op erator a or a negv

interpretation the quantied NP has scop e outside the negation op erator b

a All the b oys didnt leave

b Everyone didnt go to the store

a bRb

b bRb

Carden nds that whether sp eakers get either or b oth the negq or the negv inter

pretation varies apparently randomly among sp eakers in a way not correlated with

standard so ciolinguistic variables or even family membership Gil refers to

This example and some of the references on this topic were provided to me by Ivan Sag Others I owe to Paul Kiparsky

Chapter Inuit West Greenlandic

work in his MA Thesis suggesting apparent geographical variation in interpre

tation Midwest American English sp eakers generally obtain only the negq interpre

tation while sp eakers from the NorthEast may get either or b oth readings consistent

with Cardens data

If we accept for the moment Gils claim and the form of explanation for it oered

ab ove by Bittner the results are prep osterous I would suggest We would have to

say that Midwest sp eakers not only have the sub ject in a VPinternal p osition at

sstructure but that for some reason it is unable to raise at LF Among North

Eastern sp eakers the grammars of some are the same while others have the sub ject

in a VPinternal p osition at sstructure but allow raising at LF while yet others have

the sub ject in Sp ec IP at sstructure so that only the negv reading obtains But

there do es not app ear to b e any indep endent evidence that would supp ort or even

p ermit such a claim For example the sub ject precedes negation and what have

b een analyzed as VPinitial adverbs in all forms of American English The sub ject

also app ears preceding auxiliaries rather than near the verb of which it is a thematic

dep endent as would b e exp ected if it were in a VPinternal sub ject p osition for

more detailed substantiation of the p osition that the sub ject is in Sp ec IP at s

structure in English see Ko opman and Sp ortiche and references cited therein

also the argument from VP Ellipsis in McCloskey Incidentally Gil

also rep orts work from his MA thesis that only the negv interpretation is obtained

when the same sort of sentences are tested in Mo dern Hebrew Hebrew would then

b e an accusative language with apparently the same scop e prop erties as for Inuit the

pivot must have wide scop e with resp ect to op erators like negation

In fact it app ears unlikely that the data of Carden and Gil indicate dialect or even

idiolect variation Lab ov shows how the scop e judgements sp eakers

rep ort dep end greatly on context and suggests that the eliciting context can b e

controlled to pro duce negq and negv dialects at will This suggests that in

general we should aim for a p ermissive theory of scop e that can explain the readings

that emerge in any context

It is notable that all the data in Bittner came from lo oking at what are

broadly existential NPs mostly of the form n Xs It might thus b e thought that the

Chapter Inuit West Greenlandic

contrast in available interpretations comes from existential NPs receiving a sp ecic

or strong reading when they are in the absolutive case It seems that existential NPs

usually do take wide scop e when in the pivot p osition in b oth English and Inuit in

accord with Keenans sub ject prop erties Thus there is a clear contrast b etween

English and Inuit transitive sentences that are translation equivalents in English

there is a strong preference for the A NP to have wide scop e while in Inuit there is

a strong preference for the O NP to have wide scop e

In fact Bittner app ears to weaken her case for a scopal account by

showing that if one lo oks at what are broadly universal quantiers the same facts

no longer app ear to hold Either the sub ject or the ob ject of a transitive verb can

take wide scop e with resp ect to VP op erators

a suli atuartut tamarmik Juuna uqaluqatigisimanngilaat

yet studentpl allpl Juunaabs talkwithperfnegindplsg

None of the students have talked to Juuna yet or

Not all of the students have talked to Juuna yet

b Aanip miiqqat tamaisa takunngilai

Aanierg childpl allpl seenegindsgpl

Aani saw none of the children or

Aani didnt see all of the children

Bittner attempts to account for these observations by noting that universal quan

tiers in Inuit seem to b e ultimately derived from adverbs They still take prox

imativeobviative endings thirdfourth p erson see Section when mo difying

terms rather than nominal cases which they take when mo difying oblique roles

and they can oat She thus suggests that these universal quantiers are adjectives

and this allows them to have sp ecial scopal prop erties I omit the details here But

this suggestion is not totally convincing It is unclear how much weight to put on the

morphological evidence the Inuit p ersonal pronoun endings also ultimately derive

from adverbial endings not nominal endings Michael Fortescue pc March

But note b where an English A NP do es take narrow scop e

Interestingly oating seems to b e restricted crosslinguistically to broadly universal quantiers many of which have a somewhat adverbial nature

Chapter Inuit West Greenlandic

And when all is said and done in examples like there are still regular nouns

app earing as the head of absolutive noun phrases which are now exhibiting dierent

scopal prop erties

The contrasting b ehavior of universal quantiers might make one think that the

scop e data that Bittner observes should actually b e explained as a prop erty

of sp ecicity or referentiality that absolutive NPs acquire As well as the denite

indenite distinction it seems that natural language additionally requires a sp ecic

nonsp ecic distinction

Fodor and Sag suggest that the readings of indenites can b e divided into

two b etween quanticational readings where English aan b ehaves like an extensional

quantier and cases where the indenite refers to a sp ecic item but not one that

has b een denitely established in the context This leads to the ambiguity in the

sentence b etween the contexts delineated in

Id like to buy a b o ok

a This is such a lovely store that I think Id like to buy something Id like to

buy a b o ok But I have no idea what I want

b Excuse me Id like to buy a b o ok Im lo oking for the edition of Le rouge

et le noir published by Garnier freres

The question arises as to whether in Inuit we are really dealing with some kind of

presupp osition of sp ecicity that is attached to the absolutive case rather than a

demand for wide scop e One reason that such an approach seems app ealing is that

a scopal account apparently cannot explain all the data Consider again a pair such

a ujarak tiguvaa

stoneabs takeindsgsg

He to ok the stone

b ujarqamik tigusivoq

stonemod takeantipindsg

He to ok a stone

Chapter Inuit West Greenlandic

If we assume that the pronominal ending on the verb is nonquanticational then on

a purely scopal account there is no semantic or pragmatic contrast b etween these

sentences a result strongly at o dds with what is rep orted by native sp eakers

This p oint draws from the corresp onding observation made for English by Fodor

and Sag Fodor and Sag argue that sp ecicity cannot b e reduced to the scop e

of indenites since a sentence like

A cousin of mine is pregnant

is ambiguous b etween a sp ecic and a nonsp ecic reading although this cannot

p ossibly b e for scopal reasons They argue that rather there is a lexical ambiguity

in indenite NPs b etween a quanticational existential reading and a referential

reading On a referential reading indenites app ear to have widest scop e in the same

way that denite NPs always app ear to take widest scop e They present a number

of further arguments attempting to show how sp ecic referential indenites dont

b ehave like quantiers they can escap e socalled scop e islands just as referential

denites can they do not prevent VPdeletion as quantiers with wide scop e do and

they can only escap e to have maximally wide scop e intermediate scop e p ositions

are not op en to them These last two claims have b een disputed in the literature and

it seems indeed that there are valid counterexamples but nevertheless the basic

observation on the ambiguity of remains as do other contrasts like the ability

of indenites to take nonrestrictive relative clauses whereas other quantied NPs do

not allow them Indeed it is commonly accepted in the literature now that sp ecicity

cannot b e reduced to scop e Fodor and Sag Enc De Ho op

What then is the evidence in Bittner against a nonscopal account Bittner

argues only against a treatment in terms of deniteness suggesting that the alterna

tive accounts of Johnson and Kalmarmentioned ab ove were to o vague to b e testable

Bittners ob jection to the traditional deniteness account is that one can nd exam

ples of b oth the other p ossible combinations denite NPs like prop er names in the

Hirschbuhler provides counterexamples to the VPdeletion argument Farkas Lud

low and Neale and Abusch all argue that indenites can escap e to the supp osedly

unavailable intermediate scop e p ositions De Ho op fn notes but do esnt endorse another challenge by Ed Ruys

Chapter Inuit West Greenlandic

mo dalis case of halftransitive sentences and indenites in the absolutive p osition of

transitives Thus she concludes that the deniteness analysis cannot b e right

Not all of Bittners examples are fully convincing She suggests that names

which are necessarily denite can b e in the mo dalis

Jesusimik takusivuq

Jesusmod seeantipindintrsg

He saw Jesus

However is the only example she gives and the choice of name is p erhaps some

what unfortunate Michael Fortescue pc Mar questions how go o d sentences

with prop er names in the mo dalis case are in contexts where the intransitivization

is not required for grammatical reasons Fortescue suggests that Greenlanders nd

this sentence fairly strange It cannot refer to a straightforward act of seeing a p erson

but must have some meaning to do with seeing the concept of Jesus

It actually app ears as if there might b e some variation in the acceptibility of prop er

names in the mo dalis in antipassives They seem to app ear quite freely in Canadian

Inuktitut sources Kalmar Johnson although this could p ossibly b e an

artefact of elicitation In contrast Edna MacLean an I nupiaq sp eaker did not

accept sentences with names in the mo dalis judging bad

John tautuktuq Marymik I nupiaq

Johnabs seesg Marymod

John sees Mary

Several of the counterexamples Bittner presents of absolutive indenites

are actually partitive readings

a arlaat tiguniaruk

oneofthem takeimpersgsg

Take one of them

Fortescues p osition is that the absolutive can b e denite or indenite but that the mo dalis of

the antipassive can only b e indenite except in contexts like relative clauses where an expression

can only b e formed by antipassivization and there is no contrast

Johnson p oints out that sp eakers sometimes nd such sentences o dd but nevertheless

grammatical

Even when as here the names are linguists names not the names of real p eople in contrast to what Kalmar rep orts

Chapter Inuit West Greenlandic

b illut taakkua ilaat nuannarigaluarpakka

houses these someofthem likeactuallyindtrsgpl

I actually like some of these houses

Although these phrases are indenite they refer to part of a set that has b een previ

ously established De Ho op develops a theory dividing NP interpretation into

strong and weak readings where the dierence in readings is related to the Case of

the NP In De Ho ops system the strong readings are sp ecicreferential A friend

of mine is a paleontologist partitive Two of the sh are black generic Fish are

vertebrates and generic collective Three fossils are more expensive than two Thus

on an analysis via strongweak readings the examples in would b e covered

as would many of the other examples that Bittner discusses For example

Bittner suggests that the contrast in rep eated b elow as cannot b e

explained by a deniteness account but it app ears to follow on an account where the

absolutive must b e sp ecic while the mo dalis must b e nonsp ecic

a atuartut ilaat ikiurtariaqarpara

ofstudents oneofthemabs helpmustindtrsgsg

xx is one of the students it is necessary that I help x

b atuartut ilaannik ikiuisariaqarpunga

ofstudents oneofthemmod helpantipmustindintrsg

It is necessary that xx is one of the students I help x

Indeed De Ho op attempts to explain the distinction b etween use of the

mo dalis in the halftransitive and use of the absolutive in the transitive as a Case

based distinction b etween a weak and strong NP Case resp ectively Thus the

distinction would b e similar to the alternations b etween partitive and accusative case

in Finnish or the presence or absence of accusative case in Turkish Enc

Another counterexample that Bittner gives to the claim that mo dalis NPs

must b e indenite is

This involves her adopting BokBennemas analysis that the mo dalis case is actually a

direct case namely accusative rather than halftransitive forms resulting from antipassivization although I nd that analysis unlikely Section

Chapter Inuit West Greenlandic

Jaaku ilinnik suqutiginnippuq

Jaakuabs youmod b einterestedinantipindintrsg

Jaaku is interested in you

This example cannot b e analyzed as nonsp ecic but having noted that the Finnish

accusativepartitive distinction can indicate either a deniteindenite distinction

or irresultativity vs resultativity De Ho op suggests that this example is

p ossible b ecause the verb is irresultative But this suggestion cannot deal with all

the remaining counterexamples as we will see b elow

Bittner presents data that argue against the adequacy of a Casebased

approach She notes that for many sp eakers an antipassive morpheme can app ear

outside a double transitive sux as well as inside it see Section for discussion

of these suxes However when the antipassive morpheme is outside the double

transitive ax the mo dalis case NP must take wide scop e with resp ect to other

sentential op erators as absolutive NPs normally do whereas when it is inside them

the preference is for narrow scop e

a Aanip Juuna atuakkamik ataatsimik

Aanierg Juunaabs b o okmod onemod

tigusi simannginnirarpaa

getantipperfnegsayindtrsgsg

Aani said that Juuna hadnt got any b o oks yet

sayA xb o ok x getJ x

b Aanip Juunamut atuakkamik ataatsimik

Aanierg Juunaterm b o okmod onemod

What I have b een glossing as antipassive morphemes do have a clear asp ectual meaning see

Bittner for some discussion The semantics of the antipassive morphemes are still

p o orly understo o d but several of them seem to have an imp erfective or p erhaps irresultative

meaning It is p ossible that ultimately antipassivization as reexivization should b e seen mainly

as a consequence of attaching intransitive verbal endings to a transitive stem with the socalled

antipassive morpheme b eing basically asp ectual Indeed sometimes there is no overt antipassive

morpheme eg This is an area where further research is needed

Chapter Inuit West Greenlandic

tigusimannginnirai vuq

getperfnegsayantipindintrsg

i xb o okx sayA getJ x

ii sayA xb o okx getJ x

iii sayA xb o okx getJ x

Such data cannot b e accounted for by any account along the lines of De Ho op

where strong readings are asso ciated with absolutive case and weak readings with

mo dalis case whereas Bittner is able to give a successful account following from

scop e interactions of the antipassive and negative morphemes

Finally there are some clear cases where a sp ecic NP app ears in the mo dalis

with a resultative verb Fortescue notes that this is p ossible in relative clauses

where agent relatives are necessarily formed by antipassivization

qimmit marluk nannumik saassussisut

dogplabs twopl b earmod attackantipipartpl

two dogs attacking athe b ear

But Bittner mentions other corpus examples of sp ecic mo dalis NPs In

one a b ear is introduced Ab out a fortnight after that funeral a full grown b ear abs

came to our village but then later it is nevertheless referred to by a mo dalis NP in a

sentence like

Miki nannumik saassussivuq

Mikiabs b earmod attackantipindintrsg

Miki attacked the b ear

Thus it seems that an account like De Ho ops cannot b e maintained in the form in

which she presented it

So it seems that neither account is fully successful The scopal account cannot

explain the dierences in interpretation that o ccur even when no scopal contrast is

p ossible and the prop osed theoretical explanation do es not extend well to other lan

guages The weakstrong readings account fails to explain the full range of data on

17 And see Bittner for another such argument

Chapter Inuit West Greenlandic

scopal interactions and sp ecicity that Bittner provides indeed it seems broken

b eyond repair It is p ossible that a scopal account could b e made to work by aug

menting it with a pragmatic account of how sp ecicity presupp ositions are generated

this would b e consonant with the arguments of Ludlow and Neale At the

other extreme recent work from UCLA Beghelli et al Beghelli et al

has argued against classic theories of scop e where scop e is uniform and blind to the

semantics of individual quantied noun phrases arguing rather that the traditional

concept of scop e is a result of a conspiracy of factors like sp ecicity and distributivity

Their work would instead suggest that a solution is to b e found by dismantling the

current dichotomy b etween scop e and sp ecicity

This is where I will leave this topic It would have b een nice to b e able to say

something more denitive to make some clear progress but I susp ect that that job

must b e left to someone with a greater understanding of the Inuit language than

myself But none of the ab ove undermines the basic observation that the absolutive

NP has a sp ecial discoursescopal status

Nominalizations

Another construction that indicates an SO pivot in Inuit is the formation of niq

nominalizations In this construction the SO argument of the corresp onding verb

can b e expressed in the genitive case mo difying the nominalization The

ergative argument of the corresp onding transitive verb can only b e expressed as an

ablative oblique like the agent of a passive Sado ck forthcoming Bittner

a umiarsuup qassinut tikinnissaa nalunngilara

shipgen howmanyterm arrivenomlzfutsg knowindsgsg

I know when the ship will arrive

b atuakkat atuarnirat nuannirpuq

b o okplgen readnomlzplsg funindintrsg

Reading b o oks is fun

::: somewhat surprisingly given that genitive case is the same as ergative

Chapter Inuit West Greenlandic

Hansip Aanimit ilinniartinnera

Hansgen Anneabl teachnomlzindsgsg

the teaching of Hans by Anne

Co ordination a nonargument

Co ordination in Inuit is by means of a clitic that attaches to the right of the rst

word of the second conjunct In his discussion of Yup ik T Payne suggests

as a further argument for an SO pivot in the language that all else b eing equal the

unrealized S of an intransitive verb following a transitive verb will b e interpreted as

coreferent with the O of the previous verb He cites examples such as

Tomam Dorisaq cingallrua tuallu quyillruuq

Tomerg Dorisabs kisspasttrsgsg thenand coughpastintrsg

Tom kissed Doris and then she coughed

I think one should b e suspicious whether such a discourse preference should actu

ally b e seen as evidence for a syntactic constraint This view is reconrmed by the fact

that Fortescue app ears to argue the opp osite for West Greenlandic

He suggests that omission in co ordination is generally p ossible but that when the O

of the rst clause is coreferent with the S of the second deletion may b e dicult but

is p ossible when the ambiguity is claried by context

I would like to suggest that the gapping of NPs in co ordination is not syntactically

restricted in Inuit Any NP core or oblique can have wide scop e into the second

conjunct of a co ordinate sentence where felicitous Indeed I do not b elieve that there

is any sp ecial syntactic pro cess that is gapping NPs in Inuit co ordination Rather I

think it is just the normal omission of arguments in discourse prodrop Fortescue

suggests that the only case where it is not p ossible simply to delete an

actant in the second clause which is coreferent with an actant in the rst clause is

when an O of the rst clause is coreferent with the A of the second Here he suggests

a demonstrative pronoun is required

Chapter Inuit West Greenlandic

Hansip Kaalat takuaa taassumali

Hansierg Kaalatabs seeindsgsg sheergbut

takunngilaa

seenegindsgsg

Hansi saw Kaalat but she didnt see him

But even here I think the infelicity is likely to b e for reasons of discourse coherency

rather than it b eing a syntactic restriction given the other options for coreferential

deletion that are available

Thus I conclude that co ordination is a neutral construction in Inuit and neither

supp orts nor undermines the establishment of an SO pivot Note also that in practice

clause chains are usually expressed not by co ordination of nite verbs but by using

innitive verbs for all forms but one and then the coreferential asub ject condition

discussed in Section applies

Order of agreement suxes on verbs

Some further evidence for a surface syntactic relation grouping S and O can b e had

from Inuit agreement morphology although the evidence is more equivocal than is

sometimes suggested

Most transitive verb forms register b oth core participants but in the innitive

mo o d only a single agreement sux o ccurs and it crossreferences the S and the O ar

guments of intransitive and transitive verbs resp ectively This part of the agreement

system thus has a clear and synchronic ergativeabsolutive basis See Section

for further discussion

For other verbal mo o ds many recent authors Bittner Bittner and Hale

forthcoming a Johns suggest that the order of agreement suxes for

transitive forms is always StemErgativeAbsolutive where determinable and set up

The existence of ergativeabsolutive agreement systems go es against Jelineks pronominal

argument hypothesis which seems to predict that agreement systems should always b e nominative

accusative with only lexical NPs displaying ergativity However indisputably ergativeabsolutive

agreement systems are also found in other language families such as Mayan

In nite clauses in Inuit the outermost agreement marker is always construed with the nomina tive argument while the innermost marker in transitives is construed with the ergative argument

Chapter Inuit West Greenlandic

their functional pro jections on this basis However historical reconstruction and

the surviving surface evidence shows quite clearly that the present West Greenlandic

indicative and that of other varieties is a mixed system The historical source for

the third p erson ob ject forms is a marker of O followed by a marker for A ie Stem

abserg while other forms are Stemergabs see Fortescue BokBennema

and particularly Fortescue Since the third p erson forms are the least

morphologically transparent one could argue that their construction is opaque to a

mo dern sp eaker while the StemAO order of other forms is apparent but I know

of no indep endent evidence to supp ort this

In sub ordinate mo o ds such as the conditional from whence the ma jority of the

mo dern indicative suxes were apparently b orrowed Fortescue the ordering

is consistently Stemergabs for transitive verbs However the morphology there is

clearly formed on historically nominativeaccusative lines Examination of even the

partial paradigms shown in should b e sucient to convince the reader that the

transitive form has the same morphology marking A as is used for S in the intransitive

forms although sometimes it is reduced to an assimilated geminate consonsonant

while there is a dierent following marker for O sg gu pl git

West Greenlandic Conditional partial paradigm

Intransitive Transitive

Sub ject sg Ob ject pl Ob ject

sg guma gukku gukkit

sg guit gukku gukkit

sg ppat ppagu ppagit

sg guni guniuk gunigit

pl gutta gutsigu gutsigik

pl gussi gussiuk gussigik

pl ppata ppassuk ppatigik

pl gunik gunikku gunikkik

In contrast Woo dbury a presents the following negative conclusion Most often the

agreement suxes are phonologically fused to such a degree that it is imp ossible to tell whether they

work on a nominativeaccusative basis or on an absolutiveergative basis This is to o conservative

but is p erhaps not an unreasonable view of a mo dern sp eakers knowledge

22 0

See also Reed et al for the same generalization with resp ect to the Central Yup ik indicative

Chapter Inuit West Greenlandic

The morphological evidence from agreement is thus much more complicated than

it is sometimes p ortrayed revealing a complex mixture of diachronic and synchronic

levels However the overall ordering of axes ie excluding the indicative third

p erson ob ject forms can b e argued to exhibit an ergative organization Among lan

guages with sub ject and ob ject agreement nominative agreement is generally outside

accusative agreement ie further from the verb in cases when b oth are marked on

the same side of the verb while absolutive agreement is generally outside ergative

agreement With the exception of the indicative third p erson ob ject forms the or

der would thus b e a typical ergativeabsolutive agreement system but an unusual

nominativeaccusative agreement system

Conclusion

My conclusion is that data from relative clauses scop esp ecicity and niq nominal

izations with some supp ort from the order of agreement suxes show clearly that

the SO NPs the absolutive NPs are the pivot in Inuit In the next section I will

show that phenomena that have b een taken as indicative of an SA pivot in Inuit are

actually phenomena that are sensitive to argument structure Thus the surface gram

matical relations of the Inuit sentences in are as in with an absolutive

pivot

a Ani atuarp o q

Aniabs readindintrsg

Ani reads

b qimmit aalisakkat nerivaat

dogplerg shplabs eatindtrplpl

The dogs eat the sh

pred Ani

pivot num sg

case abs

pred read grs

Chapter Inuit West Greenlandic

pred sh

num pl

pivot

case abs

pred eat

pred dog

core num pl

case erg

grs

Phenomena sensitive to a level of argument structure

In contrast to the previous section this section lo oks at phenomena that have tradi

tionally b een taken to motivate an AS pivot in Inuit I will argue that most of these

phenomena must and the rest can b e describ ed instead in terms of the argument

structure notion of asub ject rather than in terms of surface grammatical relations

Derivational morphology

The Phenomenon

Inuit has a wide range of derivational suxes often called postbases in the Eskimo

literature some of which cover the semantic eld of control verbs in other less p olysyn

thetic languages These p ostbases come in two types One type we might call equi

sub ject suxes where this name is to b e taken simply as a convenient lab el rather

than as an analysis With these verbs the actant of the sux ie the wanter tryer

etc is necessarily identied with the surface ergative of a transitive stem a and

the surface absolutive of an intransitive stem b If the stem is transitive the

verb agrees with b oth actants

a Aanip miiqqat ikiurumavai

Aanierg childrenabs helpwantindtrsgpl

Aani wants to help children

b Hansi sinikkumavuq

Hansinom sleepwantindintrsg

Hansi wants to sleep

Chapter Inuit West Greenlandic

The other class is termed by Kleinschmidt double transitive suxes b e

cause the sux seems to add a new actant to the event expressed by the stem This

new actant is expressed in the ergative If the base verb is transitive its logical

sub ject is marked terminalis and its logical ob ject absolutive and the latter triggers

agreement ab if intransitive the stems actant is marked absolutive c

a miiqqat uannut paaritippai

childrenplabs meterm lo okaftercausindtrsgpl

He had me lo ok after the children

b Aanip miiqqat Juunamut paasisurivai

Aanierg childrenabs Juunaterm understandthinkindtrsgpl

Aani thinks that Juuna understands children

c Aanip miiqqat qasunirarpai

Aanierg childrenplabs b etiredsayindtrsgpl

Aani said that the children were tired

Examples like a are parallel to the morphological causatives of many languages

However Inuit uses syntactically identical p ostbases to also express other meanings

like think say and ask as shown in bc

There has b een a tradition in Eskimo studies which argues that such verbal forms

and other structures involving noun incorp oration need to b e analyzed syntactically

Rischel Woo dbury a Sado ck Smith The details dep end on

the framework assumed but for instance these verbal forms might b e derived by the

same cyclic syntactic transformations that were p osited to derive equi constructions

in English in the framework of Chomsky Such an approach is consonant with

Baker and related work that do es derivational morphology in the syntax If

such an approach were correct then these axes would provide evidence for sub ject

ho o d in Inuit since equi is generally analyzed in terms of a controlled sub ject It

would b e argued that the equi sub ject suxes are evidence of syntactic accusativity

since the SA of the stem is the controlled argument However recall that this is just

the kind of evidence that Dixon argues should b e discounted suggesting that

such predicates always function accusatively The double transitive suxes if any

thing represent an ergative syntactic pattern but this is not strong evidence either

Chapter Inuit West Greenlandic

b ecause the pattern resembles a common causative pattern from canonical accusative

languages such as Turkish or the Romance languages Comrie

Some putative evidence in favor of a syntactic analysis is that these axes inter

act with pro cesses of passivization and antipassivization The stem can b e passivized

a or antipassivized b b efore adding an equi sub ject sux resulting in con

trol of the derivedS of the stem The derived form can also undergo further valence

changing op erations such as passivization c and marginally antipassivization

a Miqqat Aanimit ikiurniqarumapput

childrenabs Aaniabl helppasswantindintrpl

The children want to b e help ed by Aani

b Kaalat Suulumik ikiuiniarpuq

Kaalatabs Suulumod helpantipintendindintrsg

Kaalat intends to help Suulut

c Ilisimatuuq inuaaqqanit ikiurumaniqarpuq

scientistabs pygmyplabl helpwantpassindintrsg

lit The scientist was wanted by the pygmies pro to help t

j i i j

ie The pygmies wanted to help the scientist

d Kaalat Suulumik ikiurniaivuq

Kaalatabs Suulutmod helpintendantipindintrsg

Kaalat intends to help Suulut

The situation with double transitive axes is similar Passivization a and

antipassivization b can o ccur b efore a double transitive ax is attached if the

stem is transitive Passivization c and for most sp eakers antipassivization

d can also o ccur after the double transitive ax is attached

This description of the data follows Bittner Woo dbury and Sado ck and

Sado ck draw a contrast b etween double transitive suxes and equi sub ject suxes

suggesting that external passivization or antipassivization as in cd is basically imp ossible

with equi sub ject axes although they recognize that there are a few exceptions However Sado ck

concludes that this could well b e for semantic rather than syntactic reasons a p osition

supp orted by the existence of examples such as those shown Woo dbury and Sado ck also suggest

that all sp eakers accept examples like d

Bittner notes that some sp eakers reject antipassivization outside either a double tran

sitive ax or an equi sub ject ax This is denoted with a p ercent sign

Chapter Inuit West Greenlandic

a Jaakup ammit qimminit niriniqatsaalivai

Jaakuerg skinplabs dogplabl eatpasspreventindtrsgpl

Jaaku prevented the skins from getting eaten by the dogs

b Aanip Juuna miiqqanik

Aanierg Juunaabs childplmod

paarsisurivaa

lo okafterantipthinkindtrsgsg

Aani thinks that Juuna is lo oking after the children

c ammit Jaakumit qimminut niritsaaliniqarput

skinplabs Jaakuabl dogplterm eatpreventpassindintrpl

lit The skins were prevented by Jaaku from the dogs eating t

i i

d Aani miiqqanik Juunamut

Aaniabs childplmod Juunaterm

paarisurinnippuq

lo okafterthinkantipindintrsg

Aani thinks that Juuna is lo oking after the children

Indeed suxation of p ostbases can pro ceed through several cycles Smith

provides the following example

taku jau tit tau gasugi jau juk Labrador Inuttut

see passcauspassb elieve passsg

She was b elieved to have b een made to b e seen

An argument structure based account

However as Smith app ears to observe such interactions in no way prove

that a syntactic analysis of these constructions is correct Rather these data only

sharp en a disjunction b etween p ossible analyses The data show that either all deriva

tional pro cesses including passivization antipassivization and the suxation of equi

sub ject and double transitive axes are lexical or else that all these same pro cesses

are syntactic

Smith ultimately argues for a syntactic approach However much of the argument seems

more rhetorical than substantial He argues that Duplicating syntactic rules in the morphological

Chapter Inuit West Greenlandic

There is a p erfectly adequate account of this data in terms of a lexical analy

sis whereby complex lexical forms are built up by argument comp osition Mohanan

Alsina Such an account i explains the lexicality of these morphologi

cally complex verb forms ii correctly predicts that these op erations are sensitive to

argument structure rather than surface grammatical function and iii explains why

these forms b ehave as complex predicates in terms of agreement as noted ab ove

on a syntactic account it would b e quite mysterious why there was agreement with

an argument of the lower verb word order as discussed ab ove surface case

marking patterns and relativization relativization is clauseb ound but the absolutive

of these complex verb forms can b e relativized on Johnson Double transi

tive suxes are quite parallel to the causatives of other languages and equi sub ject

suxes are quite parallel to the asp ectual and mo dal suxes or light verbs of other

languages Recently there has b een much literature on how to treat these forms as

complex predicates in frameworks broadly compatible with the one I am using and

I refer the reader to these works for fuller exemplication of what I present briey

b elow Rosen Manning Alsina Andrews and Manning

Butt and Iida et al

The actant of equi sub ject suxes always fuses with the asub ject of the stem

and so they should have argument structures of the form shown in

intendh Ph ii

The other cases were already discussed in Chapter Double transitive suxes

have the argument structure suggested for causatives by T Mohanan In Inuit

when a double transitive sux is added to a transitive verb the embedded logical

sub ject is always an oblique and so I will prop ose that in Inuit the second argument

of the double transitive sux always fuses with the logical ob ject as shown in

comp onent of the grammar would increase the expressive p ower of a grammatical theory far b eyond

what is necessary p as if the solution he prop oses word formation by cyclic transformational

rules were somehow more constrained

A lexicalist account of Inuit derivational morphology is also provided by Jensen and Johns

although it is unclear how their approach could deal with the binding facts discussed in Section

Chapter Inuit West Greenlandic

For intranstives the second argument of the causative fuses with the single direct role

of intransitive verb stems

preventh P h ii

Passives are also treated via a higher predicate at astructure

passh P h ii

Asub jects that are not the highest asub ject of a clause such as b oth the the em

b edded asub ject in the argument structure of passives the logical sub ject and the

one that results when double transitives axes are added to transitive verbs the

causee will app ear on the surface as an oblique case This is broadly compatible

with the linking theory prop osed by Alsina

Antipassives are analyzed in parallel with passives as shown in

antiph P h ii

The embedded logical ob ject of the stem will again surface as an oblique

These op erators will then comp ose in the correct way to predict all the forms

shown ab ove The argument structures of the verbs in are shown b elow in

a wanthchildren passh helphAani iii

abs abl

b intendhKaalat antiph helph Suulut iii

abs mod

c passhscientist wanthpygmy helph iii

abs abl

d antiphKaalat intendh helph Suulut iii

abs mod

a preventhJaaku skins passh eath dog iii

erg abs abl

Chapter Inuit West Greenlandic

b thinkhAani Juuna antiph lo okafterh children iii

erg abs mod

c passhskins preventhJaaku h eath dogs iii

abs abl term

d antiphAani thinkh children lo okafterh Juuna iii

abs mod term

passh she b elieveh passh caush passh seeh iiiiii

abs

In the remainder of this section I would like to discuss further the distinction

b etween supp osedly lexical and supp osedly syntactic accounts of derivational mor

phology in Inuit I think much of this debate has b een somewhat misguided as I

hop e to demonstrate by a comparison of LFG with Autolexical Syntax

On some corresp ondences b etween Autolexical Syntax and other

constraintbased theories

This section will discuss some parallels b etween Sado cks mo del of Autolexical Syntax

ALS and work in constraintbased frameworks such as LFG or HPSG The theme

of the rst section will b e to try to develop corresp ondences b etween the dierent

frameworks and then I will use these parallels to suggest that an earlier debate over

whether a lexical or syntactic approach to Eskimo word formation should b e adopted

Grimshaw and Mester Woo dbury and Sado ck is no longer a debate at

all the supp osed dierence in approach largely dissolves

The conclusion will b e that at one time the claims of lexicalist and nonlexicalist

analyses seemed clearly distinct but the dividing line has largely evaporated There

are ways in which words containing complex derivational morphology b ehave dier

ently from morphologically simple words For example they tend to have dierent

Below I continue to use pivot and core rather than subj and obj to lab el grammatical

relations as elsewhere in this thesis but it should b e clear that nothing in the following section dep ends on that it is an indep endent assumption

Chapter Inuit West Greenlandic

bindingtheoretic prop erties in Inuit Japanese and other languages This contra

dicts Grimshaw and Mesters claim that complex verbs are in fact syntac

tically indistinguishable from simplex verbs Thus lexicalist analyses have had to

provide more abstract levels b eyond the surface syntax such as some form of argu

ment structure where such dierences in b ehavior can b e represented In contrast

nonlexical analyses have to provide mechanisms to deal with the surface facts of mor

phology see for example Bonet i Alsina who develops a theory of the surface

syntax constraints on clitic templates in Catalan following the line of work initiated

by Perlmutter In the end there is convergence in the mo dules provided by

various theories and Woo dbury and Sado cks claim that Eskimo morphology requires

a syntactic treatment turns out to b e quite compatible with lexical theories like

LFG after all

Equating levels Sado ck forthcoming assumes a mo del of grammar with

three indep endent constraining levels morphology syntax and semantics each gen

erated by a contextfree grammar In this section I will examine a matchup of these

levels with the levels of LFG attempting to equate morphology with cstructure

syntax with fstructure and semantics with astructure argument structure and a

bit of what one would include in a real semantics as is provided by Dalrymple et al

Trying to equate LFGs cstructure with the morphology of ALS may initially

seem quite o dd and indeed it is an imp erfect match but key in that it allows the

imp ortant alignment b etween ALSs syntax and fstructure which is at the heart of my

discussion A p ossible LFG cstructure for a simple West Greenlandic sentence such

as a is b and this might seem much more like the syntax tree of ALS c

from Sado ck forthcoming than representations from the level of ALS morphology

It seems most convenient to compare ALS with LFG b oth b ecause LFG is an overtly multilevel

framework like ALS and b ecause the prop osal by Grimshaw and Mester that I will discuss used

LFG But I b elieve that a very similar corresp ondence can b e made with HPSG This would follow

from transitivity if one accepts some of the corresp ondences b etween HPSG and LFG that I have

observed elsewhere Manning At any rate I will include a few remarks on similarities with HPSG as we pro ceed

Chapter Inuit West Greenlandic

which deal only with the makeup of individual words as indicated in d

a Niisip ernini asavaa

Niisierg sonsg loveindsgsg

Niisi loves his son

i i

b S

NP NP V

N N asavaa

Niisip ernini

c S

NPerg NPabs VP

N NPerg N V

Niisip ni erni asavaa

d N N V

N INFL N INFL V INFL

Niisi p erni ni asa vaa

The LFG cstructure and ALSs syntax tree initially lo ok quite similar and the

similarity could b e increased by adding a VP no de to the LFG cstructure if one

wished In contrast the cstructure seems to have little to do with ALSs morphology

which indicates no relationships ab ove the word level But the crucial thing to fo cus

on is how the word ernini has b een treated in ALSs syntax tree the p ossessive

ending ni of ernini has b een separated o and placed b efore the noun stem in the

p osition where lexical p ossessors app ear This sort of decomp osition do esnt o ccur

in the LFG cstructure The imp ortant parallel b etween ALSs morphology and

LFGs cstructure is that in b oth frameworks this is where the notion of wordhood

29 0

Note that Sado ck uses negative numbers for the X levels in the morphology

Chapter Inuit West Greenlandic

or lexicality is captured cf Bresnan and Mchombo T Mohanan f

for discussion of this p oint in LFG Both theories need a notion of wordhoo d to

explain phonological and other pro cesses see Sado ck for a thorough

discussion of wordhoo d tests in Greenlandic

This is illustrated more dramatically if we lo ok at a complex derivational form For

the sentence a the LFG cstructure remains the same b while a syntactic

analysis of the sentence would b e as in c

a Niisip ernini itersarpaa

Niisierg sonsg wakeuptryindtrsgsg

Niisi tried to waken his son

i i

b S

NP NP V

N N itersarpaa

Niisip ernini

c S

NPerg S VP

N NP NPabs VP V

Niisip PRO NPerg N V sarpaa

ni erni iter

Nevertheless this do esnt erase the dierence that the cstructure in b lo oks

only ab ove the word level while the ALS morphology in d lo oks only b elow the

word level The degree of corresp ondence can b e increased by noting that LFGs c

structure has often b een extended b elow the word level to provide an X like treatment

This kind of multiclausal syntactic analysis is what Smith used and is presumably what

Woo dbury and Sado ck were assuming I b elieve Sado ck no longer holds this view Sado ck

forthcoming says of the double transitive axes that the syntax of verbs formed with such suxes do es not dier from that of underived transitive verbs but rather that their semantics is dierent

Chapter Inuit West Greenlandic

of morphology just as Sado ck do es with ALSs morphology For example Simpson

provides the cstructure b for the Warlpiri word in a

a pirli ngka ku

ro ckloc dat

on the ro ck the second case marker showing a higher function

b N

Af N

Af ku N

pirli ngka

Thus LFG is clearly prepared to extend the cstructure to represent sublexical

constituency Indeed this seems to b e the preferred approach of LFG practictioners

eg Ishikawa Cho and Sells Such a solution is also prop osed for West

Greenlandic nouns and verbs and Japanese causatives in Bresnan Similar tree

structured representations of sublexical structure have also b een explored recently in

HPSG see for instance Krieger and Nerb onne Riehemann and Iida et

al The ab ove and b elow word level trees can b e shown simultaneously and

so we could replace the cstructure in b with that in

V NP NP

A V N N

in stem

paa A V A N A N

drv stem in stem in stem

sar iter ni erni p Niisi

This structure now lo oks much more similar to the ALS morphology However

there is still a remaining dierence that cant just b e swept under the rug the

LFG cstructure is representing structure ab ove the word while the ALS morphology

isnt Given the corresp ondences that I am attempting to establish we might wonder

Chapter Inuit West Greenlandic

whether this ab ove word cstructure information is needed in LFG if it apparently

isnt needed in the corresp onding ALS structure This is a question that can only

b e answered in the context of the whole theory and lo oking closely at what other

structures there are to provide this information And so we are not really ready to

answer this question yet However to awaken some doubt in the readers mind let me

note that most of the traditional arguments for constituency were based on whether

certain groups of words were treated as a unit by a transformation this b eing taken

as evidence of constituency These transformational motivations no longer exist in a

theory like LFG Other classical constituency tests give very equivocal results see

for instance the thorough discussion in Chapter of Miller References to this

undermotivation of cstructure app ear in the LFG literature Kaplan and Zaenen

a note how the to ols of LFG leave the cstructures underdetermined and

suggest motivating cstructure on the basis of proso dic structure So there is at least

some reason to b elieve that the ab ovewordlevel cstructure might b e disp ensible by

enriching some other level to cover the remaining purp oses that it serves

The heart of my prop osal is to realize that ALSs syntax should b e equated with

LFGs fstructure Why is this so Recall the treatment of the p ossessive sux of

ernini in the ALS syntax tree in c The p ossessive sux ni app ears prep osed

as an NP sister of the head noun stem This is not b ecause it has this p osition in the

surface syntax or in any plausible mo dication of it to undo A movements such as

Wh movement and extrap osition but b ecause it expresses the function Possessor

and that is the canonical p osition of p ossessors ie where lexical p ossessors go

Thus Sado cks syntax is ab out representing functional relationships and not surface

syntax and thus it is prop erly the analogue of fstructure in LFG

Let us consider one more example the treatment of a transitive verb Sado ck

forthcoming emphasizes how a transitive mo o d sux on a verb eg mmanga

Going in the other direction note the following observation of Sado ck fn If

one wanted the morphological words could b e combined into a single structure by supplementing

the rules of the morphology with the rule MS ! W Such a rule makes the claim that there is no

interesting morphological constituency ab ove the level of the word which may or may not turn out to b e true

Chapter Inuit West Greenlandic

past sub ordinate mo o d third p erson singular external category rst p erson singu

lar internal category counts as b oth a third p erson singular ergative term and a

rst p erson singular absolutive term in the clause as shown in a for the verb

takummanga when he saw me using the lexical entry for the sux from b

a S

NPerg NPabs V

IC s IC s taku

b mmanga syntax NPerg NPabs

semantics ARG ARG

morphology INFMOOD PSUB EC s IC s

The kind of representation in a has no parallel with the lexicality preserving

representations of LFGs cstructure where takummanga necessarily app ears under a

V no de but it is directly parallel to LFGs fstructure Sado cks analysis that verb

inections can b e NPs in syntax even though they are suxes in morphology

is implemented in LFG by having the suxes optionally introduce PREDs at f

structure for various grammatical functions Bresnan and Mchombo Simpson

If one ignores the unimp ortant representational dierences a is virtually

identical to what LFG would prop ose as the partial fstructure induced from the

lexical entry for takummanga roughly

pred pro

case erg

core

pers

num sg

pred seeh i

mood psub

pred pro

case abs

pivot

pers

num sg

There is a little extra information represented in the ALS syntax that do esnt app ear

in the fstructure it is pretty much the information from the LFG cstructure that

Chapter Inuit West Greenlandic

was missing in the ALS morphology namely the linear order of phrasal constituents

and p erhaps some extra levels of phrase structure embedding such as a VP no de

Note however that the linear precedence of ALSs syntax isnt always reecting

linear order as again the treatment of ni illustrates but Sado ck has prop osals for

limiting the divergences Given this need for divergences its not clear that this is

a b etter place for such information than cstructuremorphology The issue mainly

revolves around whether constraints on the word order of lexical constituents should

b e written in categorial terms as in GPSG LP rules Gazdar et al or in

functional terms or p ossibly b oth I will not try to resolve this issue here

This leaves semantics Equating ALSs semantics with LFGs astructure should

o ccasion little debate Both levels indicate the semantic arguments of a clause and

the relationships in which they stand Sado ck fn notes how his se

mantics do es the work of the principles of Functional Uniqueness Completeness and

Coherence in LFG without requiring the stipulation of external wellformedness con

ditions Arguably this desideratum is still not provided by LFGs astructure but it

is provided by the kind of semantics envisioned by Dalrymple et al At any

rate since semantics is not here my primary fo cus this matter need not b e debated

further Both ALSs semantics and LFGs astructure allow an essentially identical

statement of the argument structure constraint on p ossible binders of reexives and

innitival controllers in Inuit for example As I will show later this use of a form

of syntacticized argument structure for determining binding closely parallels work in

HPSG in which binding theory is run o the argument structure args formerly

called subcat such as Pollard and Sag and Iida et al

The conception of the lexicon is also very similar in all of ALS LFG and HPSG

a place where the morphological syntactic and semantic prop erties of phonological

forms can b e stated The remainder of the theory of ALS deals with information

ow and constraints on mismatches b etween levels This latter area has b een less

developed in LFG than other theories but there seems no formal inability to generate

the correct constraints At any rate these principles impinge little on the discussion

b elow and thus I will ignore them

Chapter Inuit West Greenlandic

The dissolution of a debate Given the ab ove understanding of how the levels

of ALS and LFG should prop erly b e equated let us reconsider the debate b etween

Grimshaw and Mester and Woo dbury and Sado ck My basic contention

is that with the p ersp ective of hindsight this debate should never have taken place

While my analysis could b e criticized as a reconstruction the basic result is that

with the passage of time advocates of a syntactic analysis of p olysynthetic forms

have nevertheless had to deal with their wordhoo d as Sado ck do es with ALS while

advocates of a lexical analysis have had to provide mechanisms whereby morpho

logically complex words are in some resp ects dierent from morphologically simple

words as in Iida et als account of Japanese causatives or the kind of account

of binding in Inuit that I consider here The issue seems to b oil down to a gener

alization of what Grimshaw and Mester fn said ab out case marking

Multiple case marking might b e attributed to syntactically derived verbs only In

a lexical theory on the other hand multiple case marking might b e a prop erty of

morphologically derived verbs only In other words the more complex structures of

derived forms can result from recursion in the morphology rather than recursion in

the syntax

Grimshaw and Mester Grimshaw and Mester consider the

passive antipassive equi and double transitive p ostbases in Labrador Inuttut

Smith suggested that the equi and double transitive p ostbases should have

a syntactic analysis so that at some level of structure such p ostbases are essentially

like an English verb like want or make which introduces a clausal or VP complement

and that there is then subsequent verb raising or clause union ie the deep struc

ture representation is roughly as suggested in c Grimshaw and Mester GM

argue that rather these forms should b e generated by lexical rules along the same

lines as the lexical analysis of the passive Bresnan b and antipassive prop osing

the rules in

For example the version of LFG Grimshaw and Mester assumed used lexical rules rather than

X morphology and had no notion of astructure

The data are identical to the West Greenlandic data given at the b eginning of this section except for one wrinkle that is mentioned in the next fo otnote

Chapter Inuit West Greenlandic

a guma rule equi sub ject p ostbases CDM

VS VgumaS

y x y

where the under brace represents binding

b kqu rule double transitive p ostbases CDM

VS VkquS O

y x y

The notation in a is supp osed to indicate that p ostbases of the guma class intro

duce a new logical argument but this argument binds the previous sub ject argument

and hence the old sub ject argument b ecomes inexpressible which is denoted by the

at the grammatical function level In b the ax introduces a new sub ject

and the old sub ject b ecomes the ob ject

GM suggest that a lexical account is preferable b ecause it most simply accounts

for the phenomena that show that these verbs are words it explains why morpholog

ically complex verbs are identical to other verbs in grammatical b ehavior including

esp ecially a general observance of the principle of functional uniqueness and it makes

correct predictions concerning rule interactions

Woo dbury and Sado ck Woo dbury and Sado ck WS react

to the ab ove account suggesting that where the two approaches make clearly dierent

predictions the facts of Eskimo languages supp ort a syntactic treatment not G Ms

lexical treatment p I will pro ceed carefully through their argumentation

The rst p oint concerns functional uniqueness GM suggested that morphologi

cally complex verbs had the same syntax phrase structure congurations agreement

and case marking patterns as other simplex verbs a p oint not automatically ex

plained by a multiclausal syntactic account WS note that this is only partially

GM only consider the case where the stem is intransitive since Smith rep orts that

double transitive suxation is restricted to intransitive stems Woo dbury and Sado ck note

that double transitive suxes can b e added to transitive verbs in all other describ ed varieties of

Eskimo as noted for West Greenlandic ab ove the construction then b ecomes more akin to the

causatives and p ermissives in other languages for which detailed lexical analyses have b een given eg Alsina and Butt

Chapter Inuit West Greenlandic

true While functional uniqueness clearly applies in this domain to allow only one ab

solutive and one ergative argument the use of multiple axation do es allow multiple

displaced terms whereas no simplex verb has these case marking patterns a

shows double antipassivization leading to two mo dalis NPs in Labrador Inuttut and

b shows two terminalis NPs resulting from suxing two double transitive p ost

bases to a transitive verb stem in Central Alaskan Yup ik

a Angutik annamik takukqujijuk siitsimik

manabs womanmod seeantipwantantipindintrsg squirrelmod

The man wants the woman to see the squirrel Labrador Inuttut

b Uum pillruanga Jimamun tangurrarnun

thiserg dopastindsgsg Jimterm b oyplterm

teguvkarniluku qalqapaka Central Alaskan Yup ik

takeletsayinfsg axeabssg

This p erson sp oke to me and said that Jim let the b oys take my axe

WS are correct that this data go es against a strict interpretation of the claim

that complex verbs should only have the same case frames as simplex verbs but

this data do es not force a syntactic treatment Rather it can b e noted that the

morphological suxing of p ostbases licenses displaced terms and recursive suxation

licenses multiple displaced terms Thus recursion in morphological structure can

replace the p ostulation of recursive syntactic structure For the core roles the claim

of functional uniqueness is here and elsewhere maintained there remains at most one

absolutive and one ergative NP and at most two crossreferencing case markers

Noun incorp oration raised by WS as a more telling argument for a syntactic

approach is actually not a problem in LFG once we accept that it is the fstructure

that captures the syntactic nature of Eskimo incorp oration that WS observe

While this is not the place for a complete analysis of noun incorp oration the outlines

of an analysis are presented b elow this analysis is not materially dierent to the

ones suggested in Bresnan and Simpson We want to explain how

For detailed discussion of noun incorp oration in Inuit see among others Sado ck

Chapter Inuit West Greenlandic

an incorp orated noun can have a mo difying adjunct in the mo dalis case a or a

p ossessive mo dier as in b

a Suulut ataatsimik ammassattorp o q

Suulutabs onemod sardineeatindsg

Suulut ate one sardine

b Tuttup neqitorpunga

reindeergen meateatindsg

I ate reindeer meat

The essence of the analysis is a cstructure rule that makes all parts of an NP

optional

NP NP N N

POSS ADJ

CASE CASE

NUM NUM

and that lexical verbs can b e formed by adding certain p ostbases to nouns socalled

noun incorp oration

A a V N

vrblz stem

OBL

mod

b V V A

Together with the lexical entries for tor and fvpgoq in and a few other

straightforward rules and lexical items this will allow a syntactic analysis of

Presuming that such an account is indeed the one we want While Sado ck

argues for the p ossibility of p ossessor stranding Bittner presents a mo del that do es not allow

p ossessor stranding emphasizing that it is very restricted as Rischel noted and suggest

ing that the o ccasional examples involve an idiom or other exceptional constructional p ossibilities

Sado ck pc questions the p ossibility of explaining the observed examples as idioms not

ing that p ossessorstranding is pro ductive The sketch analysis I present here will allow p ossessor

stranding in line with Sado cks mo del

However LFGs principle of completeness will prevent headless NPs in normal circumstances

Thus this account explains without further stipulation why a stranded p ossessor can app ear with an

incorp orated head while an NP cannot otherwise consist of simply a genitive case p ossessor NP

thanks to Jerry Sado ck for questioning me on this p oint

38 For the miniature grammar I am considering the following will suce

Chapter Inuit West Greenlandic

Greenlandic noun incorp oration at fstructure b while maintaining a lexicality

resp ecting cstructure a

a tor A PRED useeathPIVOTOBL i

vrblz mod

OBL CASE MOD

mod

b voq A PIVOT NUM SG

PIVOT PERS

PIVOT CASE ABS

a S

NP NP V

A N N V

poq A Suulut N A N

in stem vrblz stem

mik ataatsi tor ammassat

i a S ! NP V

fpivotjcorejobl g

mod

b N ! N Af

stem case

case abs

c mik Af case mod

case

Suulut N pred Suulut

stem

ataatsi N pred one

stem

ammassat N pred sardine stem

Chapter Inuit West Greenlandic

case abs

num sg

pivot

pers

pred Suulut

pred eath i

pred sardine

case mod

num sg

pers

obl

mo d

pred one

case mod

adj

num sg

pers

Just as in WSs syntactic analysis the incorp orated noun is treated as logically an

instrumental ie mo dalis CDM case ob ject at the level of fstructure

Various other of the p oints raised by GM and WS either dont distinguish

b etween a lexical and a syntactic account as WS observe or can equally b e

accounted for lexically with a more lib eral understanding of what LFG provides The

fact that reexives and innitives can lo ok to the sub ject of the base verb for a con

troller in cases of double transitive axation do es not indicate that there must b e a

phrasestructural sub ject for b oth the verb stem and the double transitive ax Given

that the logical sub ject of passives can also b e the controller of reexives and inni

tives the data suggests clearly that binding is rather sensitive to argument structure

see b elow WS ob ject to GMs account which allowed free recursive application

of lexical rules suggesting that certain p ossibilities are restricted and suggesting that

the lexical theory fails to account for such idiosyncracies in the ordering of word

formation pro cesses However it seems that while no account of such restrictions

was given by GM a lexical account should b e b etter able to explain pro cesses of

ordering and grammaticization that restrict p ossible combinations than any syntactic

account WS do not actually oer an account either Certain ordering restrictions

may well b e semantic and equally explainable under either theory Similarly the ob

servation that there are axes that can only attach to underived intransitive verbs

Chapter Inuit West Greenlandic

and not ones resulting from passivization or antipassivization is not very telling

once it is noticed that this ax always comes b efore any other p ostbases when it

is used Woo dbury and Sado ck citing Reed et al this can b e

explained by a purely morphological restriction

The only evidence that clearly seems to sit uneasily with what we are calling

the lexicalist analysis is the existence of p ostinectional complementation as in the

Central Alaskan Yup ik example in

Liissaquna taiguurtuq

Lisaabsthisoneabs comeindsgutterindsg

Lisa uttered This one is coming

This structure indeed seems to violate the desired restrictions on functional unique

ness there are two absolutive NPs and is a rare case in Eskimo where derivational

axes follow the normally wordnal inectional morphology of the verb However

as WS themselves note such p ostinectional complementation do es not result in

the same phonological pro cesses as normal suxation and it seems that an analysis

of urtuq as an enclitic is quite app ealing here Under such an analysis this sentence

would b e biclausal at cstructure and fstructure and urtuq would simply lean on

the preceding word

NP S V

encl

urtuq Liissaq NP V

una taiguq

In summary Eskimo derivational morphology clearly requires some notion of re

cursion but it seems that the implementation of recursion can b e either at the phrasal

level or at the level of morphology There is a clear sense in which inectional suxes

can act as pronouns and semantic predicates in Eskimo languages but this can b e

39 0

A demonstrative cliticizes to the preceding word in Central Alaskan Yup ik Thanks to Jerry Sado ck for explaining this to me

Chapter Inuit West Greenlandic

appropriately captured at the level of fstructure or astructure in LFG To the extent

that LFG or HPSG remains a lexical theory there do not seem any obstacles to

treating the supp osedly syntactic prop erties of certain suxes within them

Notions of lexicalism It is nevertheless the case that LFG and HPSG end up

having to use a notion of the lexicalist hypothesis that is somewhat weaker than that

implied in early work That is one cannot maintain the strongest version of the

lexicalist hypothesis given in

Syntactic rules cannot make reference to any asp ects of wordinternal structure

Anderson

b ecause morphologically complex words can have dierent sub categorization or bind

ing prop erties from simple words and that is presumably b ecause of some asp ect

of their internal structure However we can maintain a weaker version whereby all

derivational and inectional pro cesses are p erformed prior to lexical insertion and are

invisible to syntax but the syntax can see the informational structures pro duced by

the morphology using a sophisticated form of feature p ercolation An information

oriented version of the lexicalist hypothesis is then saying

The syntax can see all features of full words but do es not have access to how

these features were built up inside the morphology

It can then b e accepted that in some cases these informational structures may b e of a

type pro duced only by morphologically complex words and so there is not complete

observance of the principle of Structure Preservation that WS discuss

The ab ove denitions of lexical integrity give a kind of informationtheoretic char

acterization of lexical integrity syntactic rules are restricted in terms of how much

information ab out morphological structure that they have access to Bresnan and

Mchombo suggest that the way to approach lexical integrity is rather by not

ing that words are built out of dierent structural elements and by dierent principles

of comp osition than syntactic phrases This follows the prop osals of Simpson

that the lexical integrity hypothesis is that constituent structure pro cesses are blind

to the internal structure of lexical categories Complete words undergo lexical in sertion at the terminal no des in phrase structure trees and the categorial comp onent

Chapter Inuit West Greenlandic

cannot move or refer to individual morphemes But functional information is allowed

to p ercolate throughout the informational representations of a sentence eg the

fstructure which can give the app earance of movement

The essence of Bresnan and Mchombos approach is that the grammar of a

language does p ossess a morphology separate from its syntax with its own elements

of structure and rules of comp osition They wish to capture that structuredep endent

syntactic rules do not apply to morphological elements morphemes while functional

information can ow freely through the morphology and syntax They discuss ve

tests for their notion of lexical integrity extraction conjoinability gapping inbound

anaphoric islands and phrasal recursivity The second last inbound anaphoric is

lands seems to t less well with their denition than the others since it seems less

clearly to do with word structure although Bresnan and Mchombo attempt to inter

pret it in this way But overall the claim is that grammar includes a comp onent that

determines syntactic constituency and that its principles dier from those that apply

to morphemes in terms of their elements ordering constraints and the availability of

recursivity See also Sells for similar arguments for distinguishing words and

morphemes rather than treating b oth uniformly via functional pro jections within

Japanese and Korean

Thus in many ways the ab ove prop osed LFG and ALS analyses are much more

similar than has b een made clear in the literature But are there still remaining

clear dierences that would help us decide b etween them The data that Bresnan

and Mchombo and Sells provide serve to motivate a level of syntactic

constituency that observes lexical integrity arguing against the reservations expressed

near the b eginning of this section Any such constraints are at present imp erfectly

rendered in ALS b ecause its syntactic level is mixing lexical items and morphemes

they could p erhaps b e captured by extending the level of morphology ab ove the

word The p otential weak link of Sado cks forthcoming prop osals is that the

mixture of phrase structural and functional information in the level of syntax predicts

However Simpson do es claim that the indexing of anaphoric information is blind to the internal

structure of words b ecause of bracket erasure in the theory of Lexical PhonologyMorphology

whereas I accept that morphologically complex forms can and sometimes do pro duce dierent anaphoric binding p ossibilities from any simple verb

Chapter Inuit West Greenlandic

a certain coupling which may not always exist On the other hand to the extent that

the coupling exists Sado cks prop osal would b e strengthened This matter requires

further study but some further discussion app ears in Section

Conclusions

In this section I have shown that double transitive and equisub ject p ostbases must

b e treated as forming complex predicates I have shown that it is not necessary

to form such complex predicates in the syntax Indeed given their lexicality it

seems more desirable to form them in the morphology It has b een argued elsewhere

Rosen Alsina that complex predicate formation is an argument structure

op eration Thus we should exp ect these p ostbases to b e sensitive to relations at the

level of argument structure and this is indeed what we found So these p ostbases

give us information ab out Inuit argument structure and do not provide evidence for

surface grammatical relations as Dixon argued

The Inuit Innitive

Inuit has a mo o d morpheme llu often referred to as the innitive which is used

in certain verbal complements and more commonly in adverbial clauses either with

a meaning like while or to indicate sequential actions Complement uses are shown

b elow with a transitive complement a and an intransitive complement b

a Miiqqat Juuna ikiussallugu niriursuipput

childrenabs erg Juunaabs helpfutinfsg promiseindintrpl

The children promised to help Juuna

qitissallutik niriursuipput b Miiqqat

abs dancefutinfpl promiseindintrpl children

The children promised to dance

Note that the gapp ed NP in the complement must b e the A or the S NP This could b e

taken as an indication of syntactic accusativity but in the last chapter I showed how

But also as the gerundive BokBennema the app ositional mo o d Woo dbury a and the contemporative Fortescue Sado ck forthcoming

Chapter Inuit West Greenlandic

control is generally sensitive to the more semanticallyoriented notion of asub ject

Stating that the controllee must b e the highest asub ject of the innitival clause also

gives a p erfectly satisfactory account of the p ossibilities shown in I will prop ose

that the correct constraint on use of the innitive is as in

The innitive mo o d indicates overlapping reference b etween the highest a

sub ject of the innitival clause and the minimal acommanding asub ject of

the clause in which the innitive is embedded

If the innitival clause contains a passive then my argument structure based

account of passivization predicts control b etween the closest asub ject of the main

clause and the derived S of the innitive the highest asub ject of the innitival clause

if b eing the controllee was not restricted to the highest asub ject of the sub ordinate

clause then the oblique agent would also b e a p ossible controllee contrary to the

facts This prediction is conrmed as shown in the structure of which is

shown in

Hansi isirpuq Kaalamillu takuniqarluni

Hansinom comeinindsg Kaalatabland seepassinfsg

Hans came in and was seen by Kaalat

pred comein

h i

pivot pred Hans

pred b eseen

pivot

xadj

h i

obl pred Kaalat

grs

I state the constraint in terms of overlapping reference rather than coreference b ecause examples

from traditional narratives suggest that only a partwhole relationship b etween the two NPs is

actually required The distinction can generally b e ignored while reading this section but see a for an example This issue is further discussed in Bergsland and Sado ck forthcoming

Chapter Inuit West Greenlandic

comein

asb Hans

pass

asb

then

see

xarg asb Kaalat

arg

For the controller of the innitive there is p ositive evidence that we are dealing

with asub jects rather than grammatical relations The asub ject of an innitive

clause must always b e controlled by the immediately higher asub ject under which it

is embedded With simple predicates this amounts to coreference with the higher AS

NP but if the higher verb contains a double transitive ax an innitival mo difying

the meaning of the stem will have its sub ject coreferent to the absolutive which is

the asub ject of the stem a the structure of which is shown in b

a atisalirsurluni uqursaqquaa

getdressedinfsg putonwarmclothesaskindtrsgsg

She told him to get dressed and put on warm clothing

pred asktoputonwarmclothes

pred pro

pivot

gend masc

pred pro

core

gend fem

pred getdressed

xadj

pivot grs

Chapter Inuit West Greenlandic

ask

asb she

arg him

putonwarmclothes

asb

xarg

getdressed

while

asb

Similarly if the next higher verb is passive an innitive complement can b e coreferent

with the higher verbs logical sub ject agent In a there is coreference b etween

the lower asub ject of tie up the logical sub ject and the asub ject of prevent

pikinnaviirlugu qilirsurniqarpuq a uumasuq

kickaboutpreventinfsg tieuppassinditrsg animal

The animal was tied up by someb o dy preventing it from kicking

j i j

ab out

pred b etiedup

h i

pivot pred animal

pred preventtokickabout

h i

pivot pred pro

xadj

h i

core pred pro

grs

pass

asb animal

tieup

asb pro

arg

prevent

xarg

asb

arg pro

and

kickabout

xarg

asb as

Chapter Inuit West Greenlandic

Further notes on the innitive

The contemporative mo o d endings of innitive verbs agree only with the absolutive

argument regardless of the transitivity of the verb and unlike participles the in

nitival ending do es not vary with the transitivity of the verb Such an agreement

pattern would b e quite unusual if Inuit were merely morphologically ergative but

makes more sense given that the absolutive argument is the pivot The contempora

tive mo o d endings are shown in

West Greenlandic Contemporative

sg pl

st l lunga l luta

nd l lutit l lutik

rd l lugu l lugit

th l luni l lutik

The contrast b etween the third p erson and the anaphoric fourth p erson ending is

discussed in detail in the next subsection It often app ears as if the third and fourth

p erson endings distinguish the transitivity of the innitive verb the fourth p erson

endings app earing with intransitive verbs and the third p erson endings o ccurring

with transitive verbs b ecause anaphors are also sensitive to astructure However

b ecause binding can b e long distance unlike control of innitives there is not an

exact match see c for a fourth p erson ending on a transitive innitive

Neither absolutive nor ergative asub jects app ear in innitival complements such

as those shown in presumably b ecause these are complements any overt NP

rexpression in such a p osition would violate Principle C However with adverbial

uses of innitives although there must b e control as indicated the overt NP can

app ear in either the innitival clause or in the higher clause or it can b e prodropp ed

in b oth Thus in adverbial innitival clauses overt NPs serving as any of A S or O

are licensed Further examples are shown in

And they app ear to b e misinterpreted in this way by BokBennema

The ergative case marking on arnap in b makes it clear that it b elongs to the innitive clause and not the main clause which is intransitive

Chapter Inuit West Greenlandic

a niviarsiaq sikkirluni kiinanngua nuiratannguarpuq

girlabs blossominfsg facelittlesg app earallthesamelittlesg

The girl ie the willow herb blossoming her little face app eared at last

b arnap atisassat irrurlugit irinarsurpuq

womanerg clothesabs washinfpl singindintrsg

While the woman washed the clothes she sang

i i

Binding Phenomena

Introductory exemplication

Before showing the plausibility of an argumentstructure based account of binding in

Inuit let me briey introduce the elements that such an account should cover Inuit

has an overt anaphor noun immi as in

a uqarpuq Hansi imminut ikiurtariaqartuq

sayindintrsg Hansiabs selfterm helpmustipartsg

He said that Hansi must help himself

i j

b imminut malugilerp o q nakuanngorluni

selfterm noticeb eginindsg strongp ersonb ecomeinfsg

He b egan to notice himself himself b ecoming a strong p erson

i i i

There is also an anaphoric adverbial namminiq which acts as an emphasizer

namminiq takuaa

selfemph seeindtrsgsg

She saw it herself

i i

There are overt pronouns forms of the demonstrative system used for already

mentioned forms such as taassuminnga in However most frequently pronouns

are dropp ed but pronominals can also b e co ded by agreement suxes as discussed

b elow

Juunap Kaali taassuminnga uqaluttuuppaa

Juunaerg Kaaliabs demsgmod tellindtrsgsg

Juuna told Kaali ab out him

i j

ij k

Chapter Inuit West Greenlandic

Other NPs rexpressions will show Principle C eects although these are not further

discussed in this section

Precisely matching the anaphorpronoun division in lexical forms Inuit has in

the notional third p erson a distinction in agreement axes used for b oth verbal

agreement and the p ossessors of nouns b etween reexive and pronominal agreement

The pronominal agreement is conventionally termed third p erson agreement and it

can b e analyzed as having the same obviative prop erties as lexical pronouns The

reexive agreement is conventionally termed fourth p erson agreement and it can b e

analyzed as having the same binding prop erties as lexical reexives Both third and

fourth p erson suxes can agree with an overt lo cal NP However there is widespread

prodrop and overt NPs agreeing with fourth p erson suxes are restricted to places

where no Principle C violation would result basically adverbial clauses eg a

Possessed nouns agree with their p ossessor and the distinction b etween third

and fourth p erson endings in Inuit gives the following contrast unlike the normally

ambiguous English translation in a

a Hansip ernini asavaa

Hansierg sonsgabs loveindsgsg

Hansi loves self s son

i i

b Hansip ernera asavaa

Hansierg sonsgabs loveindsgsg

Hansi loves hisher son

This same thirdfourth p erson contrast app ears on sub ordinate verb forms indicative

and other main clause verb forms have only a third p erson

a iserami Kaali innarp o q

enterpsubsg Kaaliabs liedownindsg

When he went in Karl lay down

i i

In some varieties of Eskimo fourth p erson suxes only app ear on adverbial sub ordinate verbs

BokBennema fn but in Greenlandic they also o ccur on complement sub ordinate verbs

Chapter Inuit West Greenlandic

b iserat Kaali innarp o q

enterpsubsg Kaaliabs liedownindsg

When he went in Karl lay down

Finally Inuit has lexical pronominalreexives imminirmi and namminirmi

As suggested by Bittner these forms can b e analyzed as simultaneously

ob eying the conditions on anaphors and pronouns and hence they act as a necessarily

long distance reexive

Following Bittner I will argue in the next subsection that all of these

elements can b e accounted for under a uniform set of binding conditions Very roughly

I will argue that a reexive must b e b ound by an accessible asub ject a pronominal

must b e free of the immediately higher asub ject and the pronominal reexive must

b e b ound by a higher asub ject that isnt the nearest one Such an account will

predict the data shown in

a Kaali uqarpuq Pavia imminit

Kaaliabs sayindintrsg Paviaabs selfabl

anginirusinnaanngitsuq

bigcmpbecannegipartsg

Kaali said that Pavia couldnt b e taller than self

i j ij k

b Kaali uqarpuq Pavia taassumannga

Kaaliabs sayindintrsg Paviaabs demsgabl

anginirusinnaanngitsuq

bigcmpbecannegipartsg

Kaali said that Pavia couldnt b e taller than him

i j

ij k

And earlier by Mohanan for the similar pronominal anaphor taan in Malayalam

Pronominalreexives are innovative forms dating from the s and still not used by all sp eak

ers Bittner Bittner suggests her analysis is appropriate for all sp eakers that use this form

However there are some apparent counterexamples in the literature For example Fortescue

gives i where namminirminit is acting as a short distance reexive equivalent to imminit

i namminirminit anivuq

selfobvsgabl comeoutindintrsg

She came out of her houseplace

However it is quite p ossible that the long distance reexive usage has stabilized among the younger

sp eakers with which Bittner has worked and in general this section adopts her generalizations on the distribution of reexive forms

Chapter Inuit West Greenlandic

c Kaali uqarpuq Pavia imminirminit

Kaaliabs sayindintrsg Paviaabs selfobvabl

anginirusinnaanngitsuq

bigcmpbecannegipartsg

Kaali said that Pavia couldnt b e taller than self

i j ij k

However unlike Bittner I will suggest that these conditions op erate on argument

structure Some comparison of the similarities and dierences with Bittners own

account can b e found in Section

Asub jects and coterms

In Section I argued following Kro eger that reexive binding in Tagalog

is based on argument structure prominence In this section I will show the adequacy

of an argument structure based account of binding in Inuit Thus the fact that the

absolutive NP is the surface pivot as argued ab ove do es not of itself aect binding

p ossiblilites

Inuit is a language in which the class of p ossible binders is constrained cf b

The binder of an anaphor must b e an asub ject Thus in simple sentences a reexive

fourth p erson ending can b e b ound by an A or S asub ject but a fourth

p erson ending cannot b e b ound by an O NP even though it is the pivot

a palasip nulii tuquppaa

priesterg wifesgabs killindtrsgsg

The priest killed his wife

i i

kill pred kill

h i

asb priest

core pred priest

wife

pred wife

arg

h i

pivot

poss self

spec pred self

grs

a imminut uqarvigaanga

selfterm sp eaktoindintrsg

I sp oke to myself

Chapter Inuit West Greenlandic

sp eakto pred sp eakto

h i

asb I

pivot pred I

h i

arg self

obl pred self as

grs

Anaanami Piita nagligijana Inuktitut

mothersgerg Piitaabs lovesgsg

His mother loves Piita

i i

This gives us the rst constraint on reexives a reexive must b e b ound by an

acommanding asub ject This is in line with the universal binding theory suggested

in the previous chapter Formally the denitions I will adopt are

a An argument acommands an argument i do es not include and

every astructure that contains all instances of contains all instances of

b An argument is an asub ject if it is the least oblique argument at any level

of the argument structure

The denition in b dep ends on the denition of obliqueness that I prop osed in

Chapter terms are less oblique than obliques and within these two groupings

obliqueness is given by a thematic hierarchy so that minimally the agent or exp eri

encer b ecomes the asub ject of a simple verb The astructures in b and b

then clearly satisfy these conditions in each case the binder is an asub ject that

acommands the reexive

In these denitions an argument is any value in an argument structure and is not to b e

understo o d as in contrast to adjunct

I phrase the denition in terms of all instances of to allow for cases where app ears in

multiple places in the argument structure due to the fusion that o ccurs in passives and causatives

Alternatively it could b e stated so as to refer sp ecically to the highest o ccurrence of In this situ

ation and in general when arguments are unied it seems that only the highest instance of a group

of unied items counts as visible This same issue turns up in the functional uncertainty based theory

of long distance dep endencies of Kaplan and Zaenen b The theory given there overgenerates

in cases of functional control where verbs like seem introduce equations like "SUBJ "XCOMP

SUBJ since the functional uncertainty path can end at either the grammatical function of the overt

NP or at the second grammatical function that is controlled The account needs to b e restricted so

that only the higher p osition containing the overt NP is visible as can b e done by adding a term to the functional uncertainty equation

Chapter Inuit West Greenlandic

Anaphors in Inuit are long distance there need not b e binding in any particular

domain and so in principle a reexive element can b e b ound by any acommanding a

sub ject This is illustrated b elow for a lexical reexive a the adverbial namminiq

b and for the fourth p erson ending on an innitive c

a qutsavigiumallirpaa maanna tikillugu imminik

b egratefulforfeellikeindtrsgsg now until selfmod

paarsisimammat

lo okafterantipperfpsubs

He felt grateful that he had lo oked after him up till now

i j i

b Maalia uqarpuq Hansip namminiq

Maaliaabs sayindintrsg Hansierg selfemph

ikiurtariaqaraani

helpmustpartsgsg

Maalia said that Hansi had to help her herself

i j i i

Maalia said that Hansi himself had to help her

i j j i

c savaatimi ilisaraluni miirturvigilirmanni

sheepplsg recognizeinfsg bleatatb eginpsubplsg

misigaaq qullililirluni

feelindsg tearswellbegininfsg

When his sheep recognizing him b egan to bleat at him he felt tears

i i i i

coming to his eyes

Consider the verb ilisaraluni recognizing him in c There is necessary coref

erence b etween its asub ject the sheep and the asub ject of the next higher verb

miirturvigilirmanni b egan to bleat at Section but the fourth p erson marker

marking the O argument is coreferent not with this asub ject but the asub ject of

the next verb up qul lililirluni ie he

So far the conditions on reexives in Inuit that we have examined are in line

with what we exp ect from the theory of binding outlined in Chapter However

Inuit has an additional restriction against coreference b etween coterms two argument

are coterms if they are b oth terms within the same grstructure clause This is

not universal but some such restriction on coreference with nearby arguments is

Cf b oth the Tagalog example a and the binding of the Basque recipro cal shown in a

Chapter Inuit West Greenlandic

extremely common

This constraint is needed to explain the binding p ossibilities of the agreement

suxes on sub ordinate verbs Consider the example in

isergimmani Kaalip Hansi eqippaa

visitpsubsgsg Kaalierg Hansiabs hugindsgsg

When he visited him Karl hugged Hans

i j j ik

According to what has b een said so far this sentence should b e able to have either

the reading shown where the fourth p erson visitee is b ound by the asub ject of the

matrix sentence or the translation When he visited himself Karl hugged Hans

i i j

where the fourth p erson visitee is b ound by the lo cal asub ject the visitor yielding

a simple reexive However this latter reading is actually imp ossible even when a

more pragmatically plausible sentence is chosen This can b e captured by p ostulating

a languageparticular constraint which disallows an anaphoric element in Inuit from

b eing b ound by a coterm

Additionally it is well known that in West Greenlandic and other forms of Es

kimo the paradigm of the lexical reexive is defective The reexive lacks direct case

forms there is no absolutive form immi which would allow the sentence in

This meaning must b e expressed by intransitivizing the verb and optionally adding a

terminalis case form of the anaphor as shown in

palasip immi tuquppaa

priesterg selfabs killindtrsgsg

The priest killed himself

i i

a palasi imminut tuquppuq

priestabs selfterm killindintrsg

The priest killed himself

i i

kill pred kill

h i

asb priest

pivot pred priest

h i

arg self

obl pred self as

grs where the ergative asub ject binds the absolutive

Chapter Inuit West Greenlandic

In intransitivized forms like the S NP is an asub ject which acommands the

oblique argument but the oblique is not a coterm of the S NP so even with a general

constraint on binding coterms the S NP can bind the oblique Thus the restriction on

the binding of coterms in Inuit cannot b e directly tested with lexical reexives On

the other hand the coterm restriction is not only consistent with the distribution of

lexical anaphors but provides a functional motivation for the lexical gap as suggested

by Sado ck forthcoming Assuming that the coterm condition applies uniformly note

that a sentence such as would b e imp ossible even if there were direct case forms

of the lexical reexive Indeed direct case forms of the lexical reexive could never

b e used in main clauses They could b e used in sub ordinate clauses but here they

would almost always b e redundant since their meaning is already captured by fourth

p erson axes andor the innitive mo o d Thus assuming that the coterm condition

applies uniformly the lexical gap in the reexive paradigm is motivated if these forms

existed they would b ear almost no functional load

Is this additional constraint correctly stated in terms of a condition on coterms

Dalrymple do cuments a number of cases where anaphors must b e free within a

small domain She characterizes these constraints in LFG in terms of a Coargument

Disjointness Condition which says that the anaphor must b e free of all coarguments

of the same pred I could restate my coterm constraint in terms of the Coargument

Disjointness Condition by suggesting that al l the oblique cases of Inuit are semantic

cases that introduce their own pred but I doubt that this is the correct solution

The p ossible disjunction we are concerned with is b etween a case marking case sux

which might yield an fstructure like a and a semantic case sux which would

yield an fstructure like b

b a

h i

pred onh i

obl pred squirrel

h i

obl

obj pred squirrel

The only place where they could nonredundantly signal a binding relationship is b etween the

sub ject of a transitive innitive and a higher asub ject that is not the immediately higher asub ject Such circumstances o ccur rarely

Chapter Inuit West Greenlandic

In the cases Dalrymple do cuments there is a distinction b etween case marking

prep ositions or cases which disallow the reexive as in these cases the argument still

counts as an argument of the main predicate and semantic prep ositions or cases

which allow the reexive as the reexive is now an argument of the prep osition or

semantic case This is shown for Marathi aapan in and Norwegian seg in

b oth of which Dalrymple analyzes as reexives that ob ey the Coargument Disjointness

Condition

a Jane aaplyaasi bad badte

Jane with self mutters

Jane mutters towith herself

i i

b Janene aaplyaakartaa saad i ghet li

Janeerg for self sari b ought

Jane b ought a sari for herself

i i

a Jon fortalte Ola om seg

Jon told Ola ab out self

i j ij

b hun kastet meg fra seg

She threw me from self

She threw me away from self

i i

To extend the use of the Coargument Disjointness Condition to Inuit one would have

to say that all uses of the oblique cases were semantic However this do es not app ear

to b e the case Consider again the intransitivized reexive in a or a semantically

ditransitive verb like

a Juunap Kaali imminik uqaluttuuppaa

Juunaerg Kaaliabs selfmod tellindtrsgsg

Juuna told Kaali ab out self

i j

b Juunap Kaali taassuminnga uqaluttuuppaa

Juunaerg Kaaliabs demsgmod tellindtrsgsg

Juuna told Kaali ab out him

i j

ij k

Chapter Inuit West Greenlandic

It seems that many of the uses of oblique cases in Inuit are quite syntacticized but

nevertheless the binding conditions in Inuit are sensitive to whether we are dealing

with a term or an oblique argument so that imminik can b e b ound by Juuna in

a Thus a condition on coterms seems appropriate

An argument structure based account

I am now ready to present the complete conditions on reexive binding in Inuit

I will use the following auxiliary denitions in addition to those already introduced

a Two grammatical function and are coterms if b oth and are terms

core roles of the same predicate

b An accessible asub ject is an acommanding asub ject which is not a coterm

c Minimality is given by the ordering imp osed by acommand

The binding constraints on reexives pronominals and pronominal reexives regard

less of whether they are lexical items or agreement suxes are then

a A reexive must b e b ound by an accessible asub ject

b A pronoun must b e free of the minimal accessible asub ject and coterms

c A pronominalreexive must satisfy b oth a and b

Stating the obviation condition on pronominals in terms of the minimal accessible

asub ject explains the b ehavior of third p erson endings on sub ordinate verbs Such

endings include the closest upstairs logical sub ject among their obviation targets For

example in the dep endent past mo o d which has b oth ergative and absolutive agree

ment markers we have already seen that fourth p erson marking of the absolutive

As a technical aside note that even the use of semantic cases as is b eing suggested in the text

would not allow an appropriate command relationship to b e stated on grammatical relations rather

than astructure eg by using fcommand Dalrymple The problematic cases are oblique

asub jects as in and which would not command their anaphors See Section for

further discussion Note also that the obviation condition on pronominals is not simply a coargument

disjointness condition Indeed this obviation condition falls outside the system of p ossible obviation conditions hypothesized by Dalrymple

Chapter Inuit West Greenlandic

do es not indicate coreference with the lo cal asub ject the ergative NP pro duc

ing a simple reexive but rather must indicate coreference with a higher asub ject

as shown in a Similarly the obviative third p erson absolutive ending indicates

noncoreference with the next higher asub ject the minimal accessible asub ject and

not just with the coterm ergative NP b

a Juunap Kaali tatigimmani tuqqissimavuq

Juunaerg Kaaliabs trustpsubsgsg staycalmindintrsg

Because Juuna trusted Kaali he stayed calm

i j j

b Juunap Kaali tatigimmagu tuqqissimavuq

Juunaerg Kaaliabs trustpsubsgsg staycalmindintrsg

Because Juuna trusted Kaali he stayed calm

i j

This follows from the binding conditions stated ab ove For b the structure is as

in and in particular the obviative absolutive ending ensures that j k since

pro is the minimal accessible asub ject

pred stay calm

stay calm

pivot pro

asb pro

pred trust

trust

i h

asb Juuna

because

pivot pred Kaali

adj

i h

arg Kaali

core pred Juuna

grs

Let us turn now to the cases that show that all asub jects are p ossible binders

and not just AS NPs sub jects on a syntactically accusative account These data

come from more careful analysis due mainly to Bittner but see also Sado ck

forthcoming and Woo dbury a Recall that there are two sources of asub jects

which are not AS NPs One is derived verbs containing double transitive suxes

that is causative and similar suxes the other is passives According to the binding

theory that I have developed b oth the oblique agent and causee arguments of these

derived verbs should b e able to bind anaphors that they acommand but a passive

agent cannot bind the surface sub ject for example b ecause it do es not acommand

it The highest lo cation of the patient is not acommanded by the agent

passh lo okafterh ii

Chapter Inuit West Greenlandic

Indeed the patient acommands the agent in a passive

Chapter presented an example of a reexive that could b e b ound by either

asub ject of a higher passive verb This example is rep eated b elow

Na ja Tobiasimit uqaluttuunniqarpuq taassuma

Na jaabs Tobiasabl tellpassindintrsg demsgerg

itigartissimagaani

turndownperftpartsgsg

Na ja was told by Tobias that he had turned self down

j i k

pred b e told

h i

pivot pred Na ja

h i

obl pred Tobias

pred turndown

h i

pivot pred pro

comp

h i

core pred pro

grs

pass

asb Na ja

tell

asb Tobias

arg

xarg

turndown

asb pro

arg

arg re

The accessible asub jects for the fourth p erson ending on the sub ordinate verb are

asb and asb Note that the coterm disjointness condition prevents the anaphor

from b eing b ound by asb which would mean that he turned himself down In

k k

the present context however the imp ortant p oint is that the fourth p erson anaphor

can b e b ound by either asub ject of the higher passive verb

Another text example of a pronominalreexive b eing b ound by a passive oblique asub ject is the somewhat complicated

Chapter Inuit West Greenlandic

taamaattunik illuliortalerput

suchplmod housemakehabituallybeginindintrpl

nunaqataasunillu imminerminnit sujuariaqisutut

neighborplabland selfobvplabl b everyadvancedipartequ

isigineqalerllutik

regardpassbegininfpl

They started regularly to build such houses and they b egan to b e regarded

i i

by the neighbors as someb o dy who was much more advanced than selves

j j

The second part of this sentence the innitival clause has the astructure repre

sentation shown in

pass

asb pro

regard

asb neighbors

arg

pro

xarg

b e very advanced

arg

asb pro relmod

arg repro

Our theory predicts that imminermi the reexive pronominal must b e free of the

minimal accessible asub ject asb but that it must b e b ound by some higher a

sub ject asb or asb In the text example it is b ound by the oblique passive

agent asb

The astructure based account is also supp orted by examples including verbs with

double transitive axes The complex argument structure of anginirusinnaanngin

nirarpaa shown in a allows an oblique reexive to b e b ound by either asub ject

b whereas a reexive asso ciated with a simplex verb with the same surface

arguments can only b e b ound by the ergative b

a sayhasb couldnotb ebiggerhasb imminit ii

Chapter Inuit West Greenlandic

b Kaalip Pavia imminit

Kaalierg Paviaabs self abl

anginir u sinnaa nnginnirar p a a

big cmp becan neg say indtr sgsg

Kaali said that Pavia couldnt b e taller than self

i j ij

a tellhasb imminit i

b Juunap Kaali imminik uqaluttuuppaa

Juunaerg Kaaliabs selfmod tellindtrsgsg

Juuna told Kaali ab out self

i j

The example in shows a double transitive ax added to an intransitive verb

with the result that the surface absolutive is also an asub ject that can bind reexives

When a double transitive sux is added to an already transitive verb the lower a

sub ject the causee surfaces as a terminalis case oblique My account predicts that

this asub ject should also b e able to bind reexives that it acommands just like the

logical sub ject of passives Example conrms this prediction

Aalup Paviamut Suulut savimminik kapiqquaa

Aaluerg Paviaterm Suulutabs knifesgmod stabaskindsgsg

Aalut told Pavia to stab Suulut with his knife

i j k

ijk

The astructures for complex derivation morphology that were motivated earlier

also explain the following binding contrast

a Kaalip Juunamut irnini tatigisurivaa

Kaalierg Juunaterm sonsgsg trustthinkindtrsgsg

Kaali thinks that Juuna trusts his son

i j

b thinkhK self s son trusthJ ii

Examples of this sort are given by Fortescue and Bittner but it must b e

p ointed out that Sado ck forthcoming rep orts that his consultants failed to accept binding by the

terminalis asub ject even though his own theory predicts it as well This may just b e b ecause out of context the ergative is a much more prominent p ossible binder

Chapter Inuit West Greenlandic

a Kaalip Juuna irniminik

Kaalierg Juunaabs sonsgsgmod

tatiginnissurivaa

trustantipthinkindtrsgsg

Kaali thinks that Juuna trusts his son

i j ij

b thinkhK J antiph trusth self s soniii

In the second argument of the double transitive sux is fused with the logical

ob ject and so the lower asub ject cannot bind inside it b ecause it fails to acommand

it In antipassivization has o ccurred inside the double transitive sux This

means that the second argument of the double transitive sux fuses with the single

direct argument of the resulting intransitive verb Thus in this case the lower asub ject

do es acommand the p ossessive reexive and binding is p ossible as shown

Evidence against certain alternatives

Surface ccommand Binding in Inuit cannot b e captured in terms of a surface c

command constraint It is dicult to prove this in general since many dierent phrase

structures can b e assumed but I b elieve the examples discussed b elow are sucient

to prove that this approach is not viable If we assume that adverbial clauses are

adjoined in phrase structure then adverbial clauses are problematic

Juuna imminut saammat Elsep qiviarpaa

Juuna selfterm turnpsubsg Elseerg lo okatindtrsgsg

When Juuna turned toward her Else lo oked at him

i j

In if the adverbial clause is adjoined to the matrix IP then Elsep would

not ccommand the anaphor but can bind it This example can b e countered by

suggesting that adverbial clauses are daughters of IP or that ccommand can see

through the segments of an adjunction structure However the following sentence is more problematic

Chapter Inuit West Greenlandic

Perip Cecilie aappillassurivaa Bjarkep

Perierg Cecilieabs blushthinkindtrsgsg Bjarkeerg

iqimmani

embracepsubsgsg

Per thought that Cecilie blushed when Bjarke embraced self

i j ij

Here the main clause verb contains a double transitive ax and when as here

the adjunct semantically mo dies the embedded stem aappil las blush the reexive

fourth p erson ending can b e b ound by either the surface ergative or absolutive of the

main clause Such a sentence could b e problematic for two reasons Firstly note that

the adverbial clause has b een extrap osed after the verb and if such extrap osition is

analyzed as adjunction to a higher functional pro jection such as CP then the solutions

mentioned previously cease to work

Separately we have to determine the surface syntactic structure of sentences with

verbs with double transitive axes If we assume a biclausal structure along the lines

of Baker then this example cannot b e explained b ecause Cecilie will fail to c

command the adverbial clause If we assume a mono clausal structure then examples

like are problematic There is no account of why the absolutive ccommands

the oblique in one structure but not the other Similarly the mono clausal structure

cannot explain why asub ject obliques can bind anaphors in sub ordinate clauses while

other obliques in the same case and presumably the same p osition cannot

Thus I conclude that any attempt to do binding purely o phrase structure must

resort to certain abstractions such as the p ostulation of biclausal structures for the

level on which binding is dened and the availability of certain movement op erations

such as extrap osition b etween this level and the surface form

Linear precedence While antecedents most commonly precede lexical anaphors

this is by no means required See example And of course fourth p erson

agreement suxes regularly precede their antecedents for example in Not

only would an account based on linear precedence thus require a reordered level but

a satisfactory binding theory stated in terms of linear precedence seems unavailable

b ecause binding is restricted to asub jects whereas necessarily there would b e many

cases of other rexpressions preceding anaphors

Chapter Inuit West Greenlandic

Surface grammatical relations I have argued that not all cases of binding can

b e explained in terms of surface grammatical relations Even if we adopted a syntac

tically accusative account in which the A and S NPs were the sub ject there would

b e no explanation of how certain O NPs and certain ablative and terminalis case

obliques namely those that are asub jects on my account are also p ossible binders

In particular note that the absolutive arguments of simple ditransitive verbs and the

absolutives of causatives must b e analyzed as b earing the same surface grammati

cal relation as indicated by agreement relativizability etc and yet their binding

prop erties dier cf

Thus an account in terms of what are truly surface grammatical relations can

not b e adopted However if causatives and passives were allowed to b e functionally

complex multiclausal at the level of grammatical relations then given further rep

resentational assumptions it could b e maintained that everything which can bind

a reexive is a sub ject These assumptions include that a passive oblique agent is

at some level a sub ject rather than just b eing base generated as an oblique as in

many current versions of generative grammar Just such an assumption can b e seen

in the work of Guilfoyle et al and Bittner Within LFG an account

of morphological causatives in Japanese which p osited a multiclausal fstructure was

adopted by Ishikawa and Saiki I am not aware of any work within

LFG that has p osited a multiclausal fstructure for the basic passive construction

of a language but a biclausal representation for the Japanese adversative passive is

adopted by Ishikawa and Saiki and its general adoption is p ossible

within the underlying architecture

However a large b o dy of work in LFG Manning Alsina Butt

and other frameworks Aissen and Perlmutter Rosen Iida et al

has argued that causatives and passives are actually mono clausal at the level of

grammatical relations and that an account as envisioned in the last paragraph cannot

b e maintained Similar arguments can b e made for the mono clausality of causatives

in Inuit and they were briey given at the b eginning of Section agreement case

marking relativization and word order argue that Inuit passives and causatives are

mono clausal at the level of grammatical relations Indeed the lines of the argument

Chapter Inuit West Greenlandic

are much clearer in syntactically ergative languages where prominence at the levels of

argument structure and grammatical relations are disso ciated as we have seen Thus

I conclude that no account in terms of grammatical relations is p ossible

Against a disjunctive account I have argued for a uniform account in which

anaphors can b e b ound by all and only accessible asub jects while pronominals must

b e free of the closest accessible asub ject One could argue that the complexities

of my argument structure representation could b e avoided by using a disjunction

and saying that a reexive may b e b ound by either a logical sub ject or by the sur

face absolutive sub ject of passives But such an account would actually have ma jor

disadvantages compared with the account presented here Firstly it would fail to

explain the overwhelming tendency for all asub jects to b ehave alike in all languages

in constructions with derivational morphology cf Section

Secondly a purely disjunctive account of antecedence of reexives do es not account

for why the logical sub ject cannot bind an anaphor inside the theme of the passive

ataatani Juunamit tatiginiqarpuq

fathersgsg Juunaabl trustpassindintrsg

His father is trusted by Juuna

i i

Some ad ho c condition would b e needed to rule out this case but it follows automat

ically on my account since the agent of a passive do es not acommand the theme as

can b e seen in

passhtheme lo okafterhagent ii

Bittner provides further imp ortant evidence against a disjunctive account

by lo oking at items that are sub ject to obviation namely pronouns third p erson

endings and the pronominal reexive In a simple clause an obviative form must

b e disjoint from the lo cal asub ject which is the A or S but in a passive clause

such a form need only b e disjoint from the oblique agent asub ject since this is the

minimal acommanding asub ject and can b e coreferent with the S NP Consider the sentences in

Chapter Inuit West Greenlandic

a umiarsuarmiut akunnirit sisamat siniriirmata

sailorplabs hours four sleepperfpsubpl

itirput

wakeupindintrpl

The sailors after they had slept for four hours woke up

j jk

b umiarsuarmiut akunnirit sisamat siniriirmata Juunap

sailorplabs hours four sleepperfpsubpl Juunaerg

itirsarpai

wakeupindtrsgpl

Juuna woke up the sailors after they had slept for four hours

i j

ijk

c umiarsuarmiut akunnirit sisamat siniriirmata

sailorplabs hours four sleepperfpsubpl

itirsarniqarput

wakeuppassindintrpl

The sailors were woken up by someb o dy after they had slept for

j i ijk

four hours

For a morphologically simple verb a shows that the third p erson sux on the

dep endent verb indicates that this verbs sub ject must b e disjoint from the absolu

tive asub ject of the matrix verb as we saw b efore despite this b eing pragmatically

o dd The same disjoint from minimal accessible asub ject reading o ccurs quite nat

urally this time in b there must b e noncoreference with the ergative NP but

coreference can o ccur with the absolutive NP or any other salient individual The

astructure of b is shown in

wake up

asb Juuna

arg sailors

asb pro

after

duration four hours

The imp ortant contrast is then to compare the matrix passive c with a

In c the astructure of which is indicated in the dep endent verbs sub

ject may b e coreferent with the absolutive surface sub ject of the matrix verb since

Chapter Inuit West Greenlandic

the obviative ending only requires its referent to b e free of the immediately sup e

rior asub ject which is the suppressed agent of the passive asb at least on one

attachment p ossibility for the adjunct Coreference with the higher S or another

individual is freely p ermitted This follows if the third p erson marker indicates that

the argument is free of the minimal accessible asub ject as on the nested argument

structure account It do es not follow under a disjunctive theory of p ossible binders

pass

asb sailors

wake up

asb proarb

arg

xarg

asb pro

after

duration four hours

Indeed it is hard to come up with a statement of the facts if we assume a simple

conception of argument structure with passives simply resulting from the suppres

sion of the highest argument as in Bresnan and Zaenen The pronominal

could neither b e required to b e free of NPs that are b oth the logical sub ject and the

grammatical sub ject of the appropriate clause nor NPs that are either the logical

sub ject or the grammatical sub ject of the appropriate clause Rather the pronominal

must b e free of the logical sub ject which is in the same clause if the pronominal is an

oblique but is in the next higher clause if the pronominal is a core argument Such

a constraint is clearly less elegant and natural than the binding theory that results

under my conception of argument structure

Similar considerations arise with the pronominal reexive in sentences with verbs

with double transitive suxes Thus in the pronominal reexive imminirmi

is p ossible despite the fact that it is b ound by the A argument of the same clause

precisely b ecause the clause involves a verb with a double transitive sux and so the

pronominalreexive is not b ound by the minimal accessible asub ject which is the

absolutive NP

Peter Sells pc notes that there app ears to have b een some slippage of reference here as we

Chapter Inuit West Greenlandic

Larsip iqqarsariasia Juulup imminirminilluunniit

Larsierg intelligencesgsgabs Juuluterg selfobvableven

pitsaunirusurilirpaa

excellentbecmpbethinkinchindtrsgsg

Juulut b egan to think that Lars intelligence was even more excellent than

obvself

This follows on my account b ecause the argument structure is as in

thinkhJ bemoreexcellenthasb imminirmi ii

i i

Because of the nested argument structures imminirmi is free of the minimal acces

sible asub ject asb but b ound by the higher asub ject Juulut as required

Other alternatives This subsection has shown that certain intuitively simpler

theories of binding cannot b e maintained Of course the the account I presented is

not the only extensionally adequate way to characterize the binding p ossibilities of

Inuit Indeed it draws on the account of Bittner Ch and I b elieve the two

accounts to b e extensionally equivalent I would however suggest that the recognition

of the levels of grammatical relations and argument structure suggested here yields

a simple intuitive account whereas the complex structures of Bittner oer

no particular explanatory advantage The account is also similar to that of Sado ck

forthcoming Both of these accounts are further discussed in Section

Imp eratives

The addressee of imp eratives is the AS argument One can have Passive imp eratives

and antipassive imp eratives where the addressee is the derived S The condition on

the addressee of imp eratives is exactly the same as for the controllee of innitives it

is the highest asub ject of the clause that is construed as controlling the event

would exp ect ::: than self s intelligence However this do esnt aect the binding p ossibilities or the p oint b eing made

Chapter Inuit West Greenlandic

Expressive adjectival p ostbases

Woo dbury a describ es for Central Yup ik Eskimo the b ehavior of what

he termed expressive adjectival postbases EAPs I presume this construction also

o ccurs in other forms of Eskimo An expressive adjectival p ostbase can b e added to

a noun and it then mo dies the head

qayarrlugaq General Central Yupik GCY

kayakfunky

funky old kayak

But they can also b e added to a verb in which case they mo dify the logical sub ject

AS of transitive and intransitive verbs bc

a arnaq uterteurlurtuq GCY

womanabs returnp o orindsg

The woman p o or thing returned

b Apayam tangurraat ikayurlurai GCY

Apauaqsgerg b oyplabs helpp o orindsgpl

Apayaq p o or thing help ed the b oys

However Woo dbury notes that when such a p ostbase is added to the stem of

a transitive verb b efore further suxing it with a double transitive p ostbase the

EAP mo dies the absolutive argument and not the optional oblique terminalis case

embedded logical sub ject

Apayamun ikayurluqaasqaqa tangurraat GCY

Apayaqtermsg helpp o orpleasetellindsgpl b oyplabs

I asked Apayaq to help the b oys p o or things

I asked Apayaq p o or thing to help the b oys

This is in contrast to the b ehavior we saw ab ove for control of innitives and the an

tecedent of reexives Within the analysis developed here I would prop ose that EAP

data should b e analyzed by saying that EAPs are sensitive not only to astructure

but a combination of astructure and surface grammatical relations p erhaps due to

the pragmaticdiscourse role of EAPs An EAP attached to a verb stem will mo d

ify the highest argument at astructure which is grammatically linked as a direct argument

Chapter Inuit West Greenlandic

Summary

In this section I have shown that complex predicate forming derivational suxes

control of innitives all asp ects of binding theory and the determination of the

addressee of imp eratives is sensitive to argument structure in Inuit This is what

would b e exp ected in the framework which I laid out in Chapter Such phenomena

should not b e taken as establishing surface grammatical relations in Inuit or in other

languages

Approaches to Inuit

This section examines certain treatments of Inuit syntax from the eighties and nineties

In doing so it exemplies various other approaches to ergativity generally ones more

in the tradition of recent generative grammar than those examined in Chapter

Johnson on Central Arctic Eskimo

Johnson

Johnson argues for the morphological ergativity of Central Arctic Eskimo

suggesting that phenomena which select the absolutive participial relativization un

marked case always agrees with verb are actually markedness prop erties which

select the unmarked NP and so these prop erties shouldnt b e used to dene Sub ject

Rather the addressee p osition of the imp erative control of coreferential deletion

with sub jectequicomplementizers and binder of reexives consistently pick out

a Sub ject relation which is the ergative NP of transitive clauses and the absolutive

NP of intransitives Thus she argues that Central Arctic Eskimo has a nominative

accusative system of grammatical relations

I agree with Johnson on the basic division of prop erties but I have argued that her

unmarked NP is actually the pivot grammatical sub ject while her sub ject should

b e analyzed as the asub ject the least oblique argument at a level of argument structure

Chapter Inuit West Greenlandic

Central Arctic Eskimo

As discussed in the following section Marantz presents a reanalysis of the data

from Johnson claiming that Central Arctic Eskimo is syntactically ergative

while other varieties of Eskimo such as Greenlandic are morphologically ergative

It is thus imp ortant to nail down exactly what is meant by Central Arctic Es

kimo Johnson says that the eldwork was done in Rankin Inlet North

West Territories The ma jority of Inuit in Rankin Inlet sp eak the Aivilik dialect Do

rais and most of the rest Kivallirmiutun the Carib ou dialect studied by

Johns Dorais refers to work by Johnson on the phonology of lo calizers in the

Rankin Inlet Aivilik dialect and the examples include the phoneme s lacking from

Kivallirmiutun and so it is reasonable to conclude that Aivilik is the dialect Johnson

describ es Aivilik app ears to b e related most directly to North Ban and the

other Eastern Inuktitut dialects Dorais and so use of the term Central

Arctic Eskimo is somewhat misleading as this term is also applied to the Western

Inuktitut dialects Inuinnaqtun and Natsilik Dorais While Eastern Inuit

dialects are innovative and dier in certain resp ects from other forms of Inuit there

is little reason to accept Marantz conclusions of a fundamental typological dierence

b etween Aivilik and other forms of Inuit The ma jor syntactic phenomena of the

dialects are identical whereas the dierences Marantz account dep ends on are very

minor and rest on doubtful data

Marantz and Levin

Marantz develops the thesis that syntactic ergativity results from an alter

native assignment of thematic roles to Dstructure p ositions This thesis is further

developed by Levin who principally examines Dyirbal Warlpiri and Basque

concluding that the rst is syntactically ergative in Marantz sense while the latter

two are morphologically ergative but syntactically accusative

It would take us to o far aeld to review the entire of Marantz which

presents a rather hetero dox version of early GB so here I will consider his work only

in so far as it b ears on the nature of ergativity ie Chapter Marantzs treatment

Chapter Inuit West Greenlandic

of ergativity is in some ways quite traditional He accepts Dixons binary division of

ergativity into two sorts morphological ergativity vs syntactic ergativity although

he do es not use these names

For the variation in case marking that leads to morphological ergativity he sim

ply notes that there is a variation b etween two case marking systems which he terms

Type A and Type B With a syntactically accusative language Type A gives mor

phological accusativity and Type B gives morphological ergativity via a stipulative

rule that determines Case marking based on the transitivity of a clause For a syn

tactically ergative language Type A case marking would b e ergative case marking

while Type B case marking would give surface accusative case marking as seen in

the following chart based on Marantz Table Note that Marantz

regards absolutive as the same case as nominative and ergative as the same case as

accusative

Syntactically

NominativeAccusative Ergative

A B A B

S NOM NOM ABS NOM ABS NOM

A NOM ACC ERG ACC ERG NOM

O ACC NOM ABS NOM ABS ACC

One immediate problem is that no language has ever b een found that is syntactically

ergative but predominantly uses Type B case marking but Marantz attempts no

explanation of this This apparent problem is discussed further in Section The

bulk of Marantzs account is an analysis of syntactic ergativity

Most of the problems in Marantz arise from a problematic choice of what

the parameter is that distinguishes syntactically ergative languages from syntactically

accusative ones Marantzs theory assumes a distinction b etween internal and external

arguments Arguments internal to the VP at dstructure are assigned their thematic

roles by the verb but the external argument is assigned its thematic role by the entire

predicate the semantics of the VP Marantz sees syntactic ergativity as arising from

a choice of one of two generalizations for the argument structure of transitive verbs

as in adapted from Marantz

Chapter Inuit West Greenlandic

a Accusative languages

agent roles assigned by predicates

themepatient roles assigned by verbs

b Ergative languages

agent roles assigned by verbs

themepatient roles assigned by predicates

Since with an intransitive verb the one role is always assigned by the predicate this

parameter essentially mirrors Dixons parameter of whether a language chooses to

group AS vs O or OS vs A However whereas Dixon sees this grouping as a

surface pivot Marantz saw this grouping as o ccurring at Dstructure apparently at

o dds with b oth Dixons notion of a deep sub ject and the facts

Syntactically ergative languages do not have a dierent argument

structure

Since under Marantz hypothesis the externalinternal argument distinction is re

versed in syntactically ergative languages one would exp ect that they would show

the opp osite b ehavior on the criteria that Marantz uses to motivate sub jectob ject

asymmetries Marantz never discusses this issue but the data scarcely supp orts his

theory

His rst argument is that varying the logical ob ject of a verb can alter the

semantic role assigned to the logical sub ject X took the money vs X took a nap

while varying the sub ject can never change the semantic role assigned to the ob ject

This argument at b est seems to have a quantitative ring Varying the sub ject of a

transitive verb however simply do es not pro duce a range of predicates on ob jects

similar to the range of predicates on sub jects pro duced by varying the ob jects of

transitive verbs pp but to the extent that it can b e motivated in English I

would exp ect that the same result would o ccur in all languages including syntactically

ergative languages b ecause in general the themepatient argument has a greater role

in determining the nature of an event Tenny Dowty This is not what

Marantz theory predicts for syntactically ergative languages however But let me

turn immediately to the idiom argument which is closely related and more ob jectively testable

Chapter Inuit West Greenlandic

The second line of evidence concerns the prep onderance of ob jectidioms over

sub jectidioms in English Indeed the latter have b een claimed not to exist but there

are a few fairly solid examples such as The vultures appear to be circling FBI Di

rector Wil liam Sessions US News and World Report noted by A ManasterRamer

Nevertheless there is a very clear asymmetry here although it may p erhaps b etter

b e explained by other means At any rate this asymmetry casts doubt on rather

than conrming the prop osal in For Dyirbal Dixon pc notes that idioms

in general are absolutiveverb comp ound idioms such as the examples in while

no ergativeverb idioms have b een noted

a munumadal

arse throw

to give up some task to chuck it in

b bungubanal

knee break

to b end over fold

c mala wugal

hand give

to give a hand

d miyayyambul

smile pull

to make someone laugh

Examples like these are the opp osite of what would b e predicted by the prop osal in

Paul Kiparsky lectures has suggested that the explanation lies in terms of prominence

on the thematic hierarchy rather than in the internalexternal argument distinction Nunberg et

al argue that there are essentially semantic reasons for the observed distribution of idiom

patterns among other things the proverbial nature of idioms leads them to favor inanimates which

are less commonly agents They argue that idioms therefore provide no evidence for sub jectob ject

asymmetries at the level of syntax

Most idioms in Dyirbal seem to b e comp ounds of this sort but since almost all theories predict

that external arguments cannot form verb comp ounds the basic argument go es through unchanged

See Levin for discussion of other idioms involving inalienable p ossession of theme arguments but b eware that her examples and b oth contain copying errors

Chapter Inuit West Greenlandic

A third line of evidence not considered by Marantz is noun incorp oration It

is generally accepted that only internal arguments can incorp orate Therefore if

Central Arctic Eskimo were syntactically ergative in Marantz sense it should allow

incorp oration of agentive arguments but not themes However this is not the case

As Johnson shows Central Arctic noun incorp oration is similar to that of

other varieties of Eskimo in incorp orating internal arguments

Piita tuktusiuqpuq

Piitaabs carib oulo okforindsg

Peter is lo oking for a carib ou

Marantz syntactic evidence and a rebuttal

What evidence then did Marantz provide for the syntactic ergativity of Dyirbal and

Central Arctic Eskimo He suggested that many of the arguments standardly given

for the syntactic ergativity of Dyirbal such as pivot chaining on absolutive argu

ments are not arguments either way b ecause his theory makes no predictions ab out

these phenomena which Marantz suggested was a go o d thing since languages inde

p endently dier on what basis they use for topic chaining and he quoted contrastive

data from Yidin He do es make the following series of predictions though

Marantzs theory predicts that it is always the SUB his notation for the VP

external argument at the level of syntactic structure that is the controlled element

However he suggests that there are not control constructions in any of the go o d

candidates of syntactically ergative languages

Dative shift If the result of dative shift in a Type A Accusative language is

This claim is somewhat surprising but no arguments are presented for it Both the Eskimo

innitive and the Dyirbal purp osive construction in its jussive complement use have b een regarded

as such a construction If they were so regarded Dyirbal is consistent with Marantz analysis while

Eskimo in general would not b e I have not b een able to check the facts for Central Arctic Eskimo

in particular but would imagine that they are the same in relevant resp ects as the facts in other

dialects I argue later that Marantz is right in concluding that the Dyirbal construction

is not a control construction Section Levin argues that the double transitive

0 0

ik show a control pattern indicating that Yup ik is syntactically ergative in the p ostbases of Yup

sense of Marantz Ergativity Hyp othesis However this argument involves analyzing these suxes as

ob jectcontrol verbs which seems wrong cf fn and totally ignoring the equisub ject p ostbases which falsify her account

Chapter Inuit West Greenlandic

that there are two NPs in accusative case then the result of dative shift in a Type B

Accusative language ie a morphologically ergative language should b e that there

are two NPs in absolutive case However if the language is syntactically ergative

Marantz argues from a series of assumptions that I will not go through that the result

must b e a lexical entry of the following sort

tuni V giveagent goal log sub j transitive

That is the verb exceptionally assigns semantic roles to b oth the agent and goal

but it is passivelike in b eing log sub j and so the goal b ecomes the SUBJ by

stipulation and the theme must b e licensed by another case assigning mechanism

This approach makes dative shift in a syntactically ergative language totally unlike

dative shift in an accusative language but Marantz suggests that this theory makes

the following two predictions in interpreting these predictions it is crucial to note

that under the hypothesis that Central Arctic is syntactically ergative Marantz is

using passive to describ e what I have b een referring to as the Eskimo antipassive

First the theme should b e expressed in the same way as displaced sub jects are in

passive constructions Secondly since the dative shifted form is already log sub j

such a form should not passivize Marantz suggests b oth of these prop erties hold in

Central Arctic Eskimo the displaced theme is marked in the mo dalis which Johnson

and Marantz called the comitative

Neither of these predictions extend over to Dyirbal the displaced theme in dative

shift constructions is always in the instrumental case even in dialects and construc

tions where the displaced sub ject of a passive must app ear in the dative and such

dativeshifted forms can passivize Marantz suggests that Dyirbal lacks a real dative

shift construction and that the app earance of dative shift simply results from certain

verbs having multiple argument structures one corresp onding to the basic form and

one to the dative shifted form By p ostulating two argument structures for such

Dyirbal verbs they are not counterexamples to Marantzs theory but this result is

achieved in a rather uninteresting way

See Levin for a more thorough discussion of problems with Marantz analysis of verbs of giving in Dyirbal

Chapter Inuit West Greenlandic

On closer scrutiny however it seems unlikely that even Central Arctic supp orts

Marantz conclusions Marantz like Johnson b efore him argues that the form

in a is basic and that b is then the result of dativeshift

a angutiup titiraut nutararmut tunivaa

manerg p encilabs childterm giveindsgsg

The man gave a p encil to the child

b angutiup titirautimik nutaraq tunivaa

manerg p encilmod childabs giveindsgsg

The man gave the child a p encil

This is p ossible in Central Arctic b ecause the verbal forms are apparently the same

Note however that this do es not extend to most varieties of Inuit where b is

the basic case frame and a is clearly derived For example in West Greenlandic

the basic form is as in with the goal in the absolutive and crossreferenced on

the verb while the theme app ears in the mo dalis

Juunap miiqqat atuakkamik nassippai

Juunaerg childplabs b o okmod sendindtrsgpl

Juuna sent the children a b o ok

Most ditransitive verbs then have another Case frame marked by suxation of uti

where the abs argument ab ove b ecomes term and the mod argument b ecomes abs

Juunap atuagaq miiqqanut nassiuppaa

Juunaerg b o ok childplterm sendutindtrsgsg

Juuna sent the children a b o ok

Use of uti is not restricted to ditransitive verbs it is a common if sometimes se

mantically irregular argument adding ax Fortescue Woo dbury b

It most commonly adds a theme role although it can add a b enefactive role to certain

semantic classes of verbs

Therefore from Inuitinternal criteria it seems most likely that Central Arctic

has merely lost a certain overt morpheme but that this still is not a dativeshift con

struction but rather its reverse Marantz argument also dep ends on the assumption

that in a syntactically accusative language the second ob ject will always remain

Chapter Inuit West Greenlandic

in the normal ob ject case rather than b eing demoted into an oblique case but I also

doubt that this is true in general

Finally other parts of Marantzs theory predict an ambiguity whereby lexical

reexive forms can have either a reexivereciprocal or a passive meaning cf Spanish

Las estatuas no se besan Statues dont kiss each other or One do esnt kiss statues

However b ecause of the dierent role assignments in syntactically ergative languages

the ambiguity should rather b e that a sentence meaning He washes himself should

also mean He washes something unsp ecied This prediction seems to b e conrmed

with the Dyirbal reexive riy

Central Arctic Eskimo like all other varieties of Eskimo do esnt have a sp ecial

lexical reexive form Rather the intransitivized form of the verb is used with a reex

ive word app earing in the mo dalis case which Marantz refers to as the comitative

Nevertheless Marantz suggests that these intransitivized forms also have a passive

meaning and provide remarkable supp ort for his analysis The central contrast is

that as well as a reexive meaning there is supp osed to b e the following contrast in

the second meaning of intransitivized forms b etween Central Arctic and Greenlandic

Eskimo reecting the fact that only the former is syntactically ergative

a angut takuvuq Central Arctic

manabs seeindsg

The man sees something

b Tigianaq takuvuq Greenlandic

foxabs seeindsg

The fox was seen

However the claims for Greenlandic are disputed by Johns Firstly as

Marantz himself notes the contrast in at b est holds for a subset of verbs in

the languages Many other transitive verbs either have no intransitivized form or

Note however that the nonreexive use of this morpheme what Dixon terms the false reexive

diers from constructions like the Spanish one cited in that an instrumental case NP can still b e

used to mark the actor as also in the antipassive what Marantz called the passive In the Spanish

imp ersonal se construction the agent cannot b e expressed it must simply b e one So it is not

clear that the constructions are comparable

As opp osed to the terminalis which is more usual in Greenlandic cf

Chapter Inuit West Greenlandic

the intransitivized form is always a stativeinchoative preserving the theme or an

unsp eciedob jectdeletionlike form preserving the agent in b oth of these languages

just as in English I broke the door The door broke vs I fought the Romans I fought

An example of each type in Qairnirmiut is shown b elow see Woo dbury Johns

for further discussion of verb classes in Eskimo languages

a Jaaniup niqi tamuajaa

Jaanierg meatabs chewpartsgsg

Jaani chews the meat

b Jaani tamuajuq

Jaaniabs chewipartsg

Jaani chews

a arnaup kaapi kuvijaa

womanerg coeeabs spillpartsgsg

The woman spilt the coee

b kaapi kuvijuq

coeeabs spillipartsg

The coee spilt

And so Marantz claims that this parametric dierence b etween Green

landic and Central Arctic o ccurs only in a remaining residual class of verbs The

fact that Inuit intransitives are not sp ecically reexive forms is already problematic

for Marantz predictions as noted by Levin who p oints out that the in

transitive conjugation is found with verbs of all semantic classes and that it cannot

b e asso ciated with the feature T that Marantz would asso ciate with a purely

reexive form Moreover Marantz only ever names one member of the putative

residual class taku see as in However it seems that this supp osed contrast

do es not actually exist Johns suggests that her Greenlandic consultant

translated takuvoq as He sees something in line with Central Arctic and not as

He was seen Example b is originally from Swadesh Johns cites a

The feature T indicates following Burzio that a verb do es not assign a thematic role to its sub ject

Chapter Inuit West Greenlandic

communication from Woo dbury suggesting that Swadesh manufactured this exam

ple Indeed the example do es not seem to b e present in Kleinschmidt the

source for Swadeshs sketch grammar Further Marantz gave as his example of

how in Greenlandic reexives b ehave much as in the Spanish with an ambiguously

passive meaning

Piniarto q to qupp o q

hunterabs killindsg

The hunter was killed or

The hunter killed himself

However BokBennema cites Johns and her own consultant as sug

gesting that this sentence can only have a reexive reading and not the rst middle

reading given ab ove In summary a putative change in the intransitivization class

of one or two Inuit verbs seems scarcely sucient evidence on which to build a theory

of language

Johns lists some other more general criteria for rejecting Marantz analy

sis It would imply that there are cognate morphemes in Greenlandic and Central

Arctic which are passive in one and antipassive in the other and vice versa Such a

fundamental reanalysis is p erhaps not imp ossible but it seems unlikely the dialects

are partially mutually intelligible Further Johns notes that a theory like Marantz

is incompatible with any version of a Semantic Bo otstrapping hypothesis Grimshaw

Pinker that dep ends on there b eing a basic mapping b etween semantic

arguments and their Dstructure p ositions or in more syntactic terms the theory is

a fundamental violation of Bakers Uniformity of Theta Assignment Hyp oth

esis Note that my own Inverse Grammatical Relations analysis do es not violate

these hypotheses although it do es then allow inverse mappings b etween argument

structure and grammatical relations which a child would have to learn

63 That is the verb should b e placed in the class of purely transitive verbs

Chapter Inuit West Greenlandic

Three GB analyses

This section lo oks at three recent GB analyses of Inuit and ergativity more generally

I assume a certain familiarity with the versions of GB that were in vogue when I was

writing this See Nap oli for background

BokBennema

BokBennema rejects the hypothesis of two unrelated types of ergativity morpholog

ical and syntactic and suggests following Payne that ergativity results from

the inability of a verb to assign structural Case to its direct ob ject an analysis which

would allow parallels to b e drawn with other parts of the grammar where problems

arise with Case such as causatives Thus she b egins a mo dern trend of syntacticiz

ing what was previously thought of as morphological ergativity The result in Inuit

is three p ossible strategies i expressing the agent as an oblique ii expressing the

theme as an oblique and iii exceptional case marking

BokBennema argues that the passive and antipassive constructions of Inuit result

from taking options i and ii ab ove while option iii yields the canonical transitive

construction She suggests that absolutive case is like nominative as it is unmarked

and always present mo dulo prodrop Thus agreement with an absolutive argument

is like sub ject agreement and absolutive should b e assigned by I This suggests that

the direct ob ject NP scrambles out of the VP to a p osition within the domain of

In where Case is assigned BokBennema suggests that assigning nominative to the

direct ob ject is p ossible b ecause another element in I which is also present in the

homophonous p ossessive agreement markers assigns Case to the ergative sub ject

She suggests that Ergative case on intransitive sub jects could b e ruled out either by

making nominative case assignment compulsory or by making the agreement feature

incompatible with intransitive verbs

BokBennema has a third argument for aligning nominative and absolutive p

suggesting that in a number of morphologically ergative languages

nonnite clauses have sp ecial restrictions they always have to b e intransitive in

Mayan languages and the complements of ob jectcontrol verbs a subset of the

Chapter Inuit West Greenlandic

double transitive suxes of Section have to b e intransitive in Inuit She cites

data such as the following from Central Arctic Eskimo

a Arnaup nutaraq pro titirauti nanirquvaa

i i

womanerg childabs p encilabs ndtellindsgsg

The woman tells the child to nd a p encil

b Arnaup nutaraq pro titirautimik nanisirquvaa

i i

womanerg childabs p encilmod ndantiptellindsgsg

The woman tells the child to nd a p encil

The argument here seems to b e based on exploring the theory internal consequences

of a preconceived notion of what the structure of these sentences is On the assump

tion that the double transitive suxes pro ject indep endently as verbs BokBennema

exp ects sentences where a stem has b een suxed by a double transitive sux to b e

biclausal and then tries to make theoretical capital out of the fact that a is

bad whereas it should b e p ossible on a biclausal ob jectcontrol analysis But as

we have already discussed and as BokBennema also notes it is not that double

transitive p ostbases cannot b e suxed to transitive stems it is simply that the em

b edded causee asub ject must then app ear as a terminalis oblique as was shown in

ab Another example from Central Arctic is shown b elow

angutiup Piitamut qimmiq tauturquvaa

manerg Peterterm dogabs lo okatwantindsgsg

AThe man wantstells Peter to lo ok at the dog

There is nothing in this pattern of data that is unique to ergative languages

Exactly the same observation could b e made ab out the morphological or otherwise

mono clausal causatives of many other languages For example in Turkish the correct

form is a with an oblique dative causee not b styled on a ab ove

BokBennema prop oses dividing each of the two classes of axes we discussed in Section

into two the double transitive suxes into ob ject control verbs and causativeECM verbs and the

equisub ject suxes into raising and equi verbs But this analysis seems to b e supp orted mainly by

considering translation equivalents in Europ ean languages rather than by lo oking at Inuitinternal

criteria as she acknowledges on p For the former pair of classes Bittner fn

observes that I am not aware of any evidence in WG Inuit to justify this dierence in the analysis and suggests that a treatment of some of the verbs in this class as ob ject control verbs is problematic

Chapter Inuit West Greenlandic

a Disci mektubu m udure imzalatt

dentistnom letteracc directordat signcauspast

The dentist made the director sign the letter

b Disci m uduru mektubu imzalatt

dentistnom directoracc letteracc signcauspast

The dentist made the director sign the letter

Indeed Comrie notes that this is the most common pattern for causatives

crosslinguistically Use of an oblique agent might suggest that the stem has un

dergone intransitivization but this issue mirrors similar continuing discussion in

accusative languages such as whether the French faire par construction involves

passivization Kayne For these and other reasons I nd BokBennemas

discussion of observing lexical integrity versus employing indep endent syntactic pro

jection pp very unsatisfactory The facts presented seem little related to

the identication of absolutive case or to the analysis of ergativity in general

BokBennema suggests the following structures for simple intransitive and transi

tive clauses at Sstructure the NP of an intransitive clause can have the Dstructure

p osition of either of the NPs in the transitive clause dep ending on whether the verb

is unergative or unaccusative

BokBennema attempted to distinguish a class of ob ject control verbs from causativeECM

verbs but this seems unsustainable as noted in the previous fo otnote BokBennema could have

made some capital from the fact noted in fo otnote that in Labrador Inuttut double transitive

suxes attach only to transitive stems although she app ears not to have noticed this and illustrates

the ab ove argument with Central Arctic and Greenlandic forms However it is also common

crosslinguistically for the formation of morphological causatives to b e restricted to intransitive stems

Note that the Inuit terminalis is semantically closest to a dative case indeed some authors

call it that rather than a semantically instrumentalablative case of the sort that is common for

passive agents

67 b is the structure of

Chapter Inuit West Greenlandic

a IP

NP I

VP I Kuunuk

Kuunukabs

V puq

indsg

aallar

leave

b IP

I NPerg

Jaakup

Jaaku

I NPabs

arnaq

franskiq

OP I

French

paa

woman

indsgsg

VP O

sar

freq

NP VP

NP V

angirlaat

bringhome

The parallelism b etween p ossessive and actor agreement is captured by the as

sumption that Inuit NPs also have an inectional head Her equivalent of DP is called NIP

Chapter Inuit West Greenlandic

NIP

NPposs NI

NP NI

genitive

BokBennema suggests that I assigns b oth ergative and absolutive Case The ergative

argument moves to Sp ec IP it is now in the same A p osition as intransitive sub jects

which gives a natural explanation of the generally nominativeaccusative anaphoric

binding facts but not the ability of logical sub jects of passives to bind reexives that

motivated the astructure account of binding discussed in Section It is also in

the same Sp ecHead relationship as is used for ergativegenitive assignment in NIPs

and so if we assume that mood within I also licenses genitive case the case marking

of the ergative argument is explained

Where then do es the ob ject go Assuming that ob jects are generated adjacent

to the verb the canonical word order

Adv erg abs obl Adv Verb

suggests that the absolutive has scrambled left away from its Dstructure p osition

next to the verb

Further BokBennema rep orts Bittners observation that the direct ob ject NP

in examples like cannot take narrow scop e dierent women each time with

resp ect to the frequentative op erator

Jaakup arnaq franskiq angirlaattarpaa

Jaakuerg womanabs Frenchabs bringhomefreqindsgsg

Jaaku often brings home a French woman

BokBennema uses this fact to argue that the ob ject must b e outside of the scop e

of these op erators at Sstructure Since some of these op erators app ear in I as well

as O on her analysis this indicates that the absolutive argument must scramble to

adjoin to I an unusual landing sp ot but BokBennema claims to have evidence for

68 This is true even of less pragmatically loaded examples as discussed in Section

Chapter Inuit West Greenlandic

adjunction to intermediate pro jections in Inuit noun phrases as well This movement

is necessary b ecause the absolutive argument needs Case It cannot get Case from

the verb by hypothesis but it can b e assigned nominative by In in a pro cess of

AdjunctHead agreement which BokBennema suggests might also b e needed to

assign absolutive in predicate nominal constructions

Quite apart from the generally unapp ealing disjunctive assignment of nominative

Case and adjunction to I this account seems to make selective use of Bittners

scop e observations While BokBennema correctly rep orts Bittners observations for

the scop e of transitive absolutives she do es not consider Bittners rep ort that the

ergative argument can take wide or narrow scop e with resp ect to sentential op erators

If we follow standard practice and have an op eration of Quantier Raising but not

Quantier Lowering this would suggest that the Ergative argument would have to

app ear lower at Sstructure Thus if the argument from scop e ab ove is to b e taken

seriously these structures need mo dication

The Mo dalis is Accusative Case hypothesis In the nal chapter of her b o ok

BokBennema presents a dierent hypothesis that Mo dalis is really an accusative case

marker assigned by the verb While the optionality of mo dalis arguments and the

usual presence of si or another antipassive derivational morpheme initially makes

the standard antipassive analysis lo ok attractive BokBennema suggests it has three

aws

It do esnt explain why antipassives are only found in ergative languages

Restrictive theories of valencychanging rules that only aect the highest or

external argument would not allow an antipassive rule that demotes ob jects

Assuming the biclausal analysis for intransitive verbs to which double transitive

suxes have b een attached that we discussed ab ove antipassivization of these

forms do es not work as her antipassive analysis would predict

She nds the claims of Postal for restricted constructionsp ecic antipassive rules in various accusative languages unconvincing

Chapter Inuit West Greenlandic

She suggests that analysis of antipassivization as incorp oration of an abstract noun

by Baker solves some of these problems but do esnt explain why antipassives

only o ccur in ergative languages

Her prop osal is then that all uses of the mo dalis see the list in are in fact

cases of an accusative case assigned by the verb In other words the mo dalis is

analyzed as a direct case She suggests this clearly explains why antipassives only

o ccur in ergative languages this is normal nominativeaccusative case assignment

The problem with double transitive suxes is solved since ECM verbs regularly

assign accusative to sub jects She thus suggests that the socalled antipassive suxes

are really a form of raising verb that can license accusative case

I nd this analysis generally unconvincing Saying that mo dalis is accusative

would greatly complicate the statement of the binding theory esp ecially in dialects

like Central Arctic where the mo dalis is regularly used to form the intransitivized

reexives of transitive verbs

angut ingminik kapivuq

manabs selfmod stabindsg

The man stabb ed himself

If the mo dalis is a direct case this binding pattern should b e imp ossible according

to the binding theory developed earlier or the similar theories of Bittner and

Sado ck forthcoming

The claim of the functional uniqueness of direct casesfunctions would also cease

to hold Recall from Section that a clause can have multiple mo dalis NPs due to

multiple suxation of antipassive suxes example a The statement of basic

word order would b e complicated now there would b e a core argument the mo dalis

mixed in among the oblique NPs as would b e verb agreement antipassive forms take

intransitive agreement suggesting that they have only one direct argument but now

they would actually have two direct arguments Finally it is unclear why passive

could not apply to an antipassive form promoting the mo dalis NP

Chapter Inuit West Greenlandic

Johns

Johns analysis of Inuktitut in particular the Qairnirmiut Carib ou dialect of Baker

Lake to the west of Hudson Bay makes the argument that the app earance of erga

tivity is just an epiphenomenon resulting from the interaction b etween language

particular lexical features and universal principles p She essentially imp orts

the nominalist tradition Section into a GB framework carefully building

up an analysis of sentences in the participial mo o d which I summarize b elow The

analysis dep ends on three assumptions i the inability of the verb to pro ject a VP in

Inuktitut ii that the morpheme jaq is consistently a passive participle morpheme

with nominal category in all its o ccurrences and iii the set of functional categories

p ostulated for Inuktitut

The Passive nominal A passive participle morpheme jaqtaqgaq may

attach to the stem of transitive verbs yielding a nominal whose referential index is

the internal argument

kapijaq mikijuq

stabpasspartabs smallintrpartsg

The stabb ed one is small

Johns assumes that a role inherited by derivational morphology here the agent

need b e realized i there is a licensing mechanism acknowledging that this is a

violation of the Criterion

The copula attaches to nominals creating a verb to which can b e attached in

transitive agreement and this general pro cess in this particular case pro duces the

Inuktitut passive

nanuq angutimit kapijau juq

p olarb earabs manabl stabpasspartcopintrpartsg

The b ear was stabb ed by the man

lit The p olar b ear is the by the man stabb ed one

Here the copula or some higher no de can optionally license expression of the agent

argument as an adjunct

Chapter Inuit West Greenlandic

The p ossessive phrase Possessive phrases are analyzed as a functional pro jection

Agr essentially equivalent to D This pro jection can o ccur ab ove any NP includ

ing the passive nominal yielding either a regular p ossessive phrase or the transitive

relative construction which translates as the one that agent Ved as appropriate

a angutiup qimmia

manrel dogsg

the mans dog

b angutiup kapijaa

manrel stabpasspartsg

the one that the man stabb ed

lit the mans stabb ed one

a AgrP

Agr NP

angutiup

manrel

NP Agr

N a qimmiq

dog sg

b AgrP

NP Agr

angutiup

manrel

NP Agr

N a kapijaq

stabpasspart sg

This analysis is app ealing b ecause the same case Relative marks p ossessors and

agents of active sentences in Inuit and the agent crossreferencing on the verb is

identical to p ossessive crossreferencing on the verb The sp ecier gets a p ossessive

role in a and the agent role in b Note that it cannot receive the theme

Chapter Inuit West Greenlandic

role in b as this is already assigned as the passive nominals referential index a

theoretical observation supp orted by the data

The transitive clause The participial mo o d transitive clause results from em

b edding the transitive relative construction in a further Agr functional pro jection

AgrP IP

Agr NP

abs

nanuq

AgrP Agr

N V

p olarb earabs

Agr NP

angutiup sg

manrel

NP Agr

a N

kapijaq

stabpasspart

Its not clear what forces the second head movement presumably morphological

constraints Note also that only deverbal forms can app ear under Agr and not

other NPsAgrP s Intransitive verbs are embedded only b eneath Agr This ex

N V

plains why patient and intransitive verb agreement are similar We can simply say

that absolutive case is assigned to the sp ecier of Agr Johns suggests that the fact

that absolutive agreement app ears outside ergative agreement in nonzero forms is

also explained The suggestion is that the literal translation of is roughly The

p olar b ear is the mans stabb ed one although there is no overt copula morpheme

here as there is with the passive

Its around this p oint that things get messy To account for the default order of

NPs in Inuit sentences ERG ABS see Section Johns suggests that there is

movement of the ergative NP which is actually required by what has b een said so far

citing Baker it is assumed that traces of X do not assign Case so following

70 But recall the caveats presented in Section

Chapter Inuit West Greenlandic

the head movement of Agr to Agr the ergative NP must move so that it can still b e

N V

Case marked Johns suggests that the ergative NP moves to adjoin to AgrP she

assumes that an agreement element can govern through another agreement element to

which it is morphologically adjoined though this seems novel I see several problems

here rstly Baker presents empirical supp ort for this prop osal in a frame

work that did not use the many functional pro jections that result from the SplitIn

hypothesis of Pollock and later work This constraint cannot b e maintained

in a more recent analysis such as Johns with multiple functional categories that

verbs and other predicates move up through These items cannot lose their Case

assigning ability every time they move Secondly its not clear that the adjoined

Agr should b e able to assign Case as suggested and if the Agr complex is assigning

N N

two cases it is not clear what determines which one is assigned where Thirdly it is

quite unclear whether Johns analysis correctly captures even the basic nominative

accusative binding facts of Inuit let alone the more complicated data observed in

Section Johns argues that the ergative NP can bind inside the absolutive NP

b ecause it moves to adjoin to a p osition ab ove the absolutive NP at sstructure But

given her structures it is quite unclear how Sp ec AgrP of an intransitive verb

and an NP adjoined to a transitive AgrP form a natural class for binding theory or

why adjunction to a higher p osition aects binding anyway Moreover the sugges

tion of A movement of the ergative argument should p erhaps create weak crossover

violations which it do esnt this issue is discussed in Section b elow

The discussion so far has dealt solely with the participial mo o d This mo o d is very

common in main clauses in the form of Inuit that Johns studied but less common or

The motivation for this claim the need for p ossessor ascension in cases of incorp oration in

Mohawk stands in contrast to examples of allowed p ossessor stranding in Eskimo cited by Sado ck

although Bittner has questioned their pro ductivity

For example if French nite verbs move to T at Sstructure as in Pollock then under

the assumption b eing considered in the text they would not b e able to assign Case to their ob jects

Either some form of extended transparency must b e assumed in the spirit of Grimshaws

extended pro jections or alternatively the prop osals of Chomsky imp ose a much weaker

requirement merely that at some p oint in the derivation the verb must pass into or through AgrO

and that the ob ject NP must pass into or through Sp ec AgrO

And similar problems would arise for identifying the controllee of innitives which is also basi cally nominativeaccusative Section

Chapter Inuit West Greenlandic

not used in main clauses elsewhere At any rate there is an alternative the indicative

mo o d with transitive endings vaqpaq

angutiup nanuq kapivaa

manrel p olarb ear stabindtrsgsg

The man stabb ed the b ear

The transitive indicative has none of the nominal prop erties of the passive participial

that we have gone over it cannot o ccur in passive sentences passive nominals nor

the transitive relative construction Nevertheless Johns suggests that sentences with

this form are exactly the same but that the transitive indicative has the sp ecial

prop erty of requiring to attach to Agr she suggests it has the feature Declarative

which has to app ear in Agr This thesis app ears to have a certain diachronic truth

Section but b oth for the indicative and verbal uses of the participial mo o d I

nd it has little app eal as a synchronic analysis There is really no reason to think

that the ergative NP is functioning as a noun p ossessor in verbal clauses indeed b oth

the preferred and other p ossible word orders indicate otherwise

An unaccusativeunergative split Johns do es not explore all the consequences

of her prop osals for the basically GB framework in which she writes For example

she seems committed to a lexical treatment of the passive via the passive nominal

morpheme rather than a syntactic treatment where the underlying logical ob ject

moves to the surface sub ject p osition Indeed she do es not seem to have available

to her any notion of movement from a Dstructure VP internal p osition to a surface

sub ject p osition and so she could not represent a split b etween unaccusative and

unergative intransitive verbs in this way either

Johns in fact argues that there is no evidence for a split b etween

unergative and unaccusative intransitives in Inuit as do es BokBennema

It is p erhaps true that a binary unaccusativeunergative distinction is not very well

founded see Zaenen but I think that Inuit do es p ossess evidence of the sort

that has b een used to motivate unergative and unaccusative classes of intransitives

It has already b een mentioned in the discussion of Marantz around example

Chapter Inuit West Greenlandic

that of verb stems that can b e used transitively and intransitively some pre

serve the agent in the intransitive use and others the theme In intransitive uses of

the latter verbs it would seem app ealing in a GB framework to say that the theme

of verbs that preserve the theme is generated VP internally at Dstructure This

could b e supp orted from evidence from noun incorp oration The general pattern in

languages is that only themelike internal arguments tend to incorp orate Mithun

summarizes some of her results as follows If a language incorp orates Ns

of only one semantic Case they will b e patients of transitive Vswhether the lan

guage is basically of the ergative accusative or agentpatient type If a language

incorp orates only two types of arguments they will b e patients of transitive and in

transitive Vs Thus the p oint of current interest is that while Sado ck sp eaks

only of ob ject incorp oration Bittner and Bittner and Hale forthcoming

b show that incorp oration of unaccusative sub jects is in fact p ossible

a ipittunik saviiruppuq

sharpplmod knifeb egoneindintrsg

There are no more sharp knives

b pualasunik puisinippuq

fatplmod sealarriveindintrpl

There arrived fat seals

This suggests that these axal verbs are unaccusative and in general axal verbs

can b e made into standalone verbs by suxing them to the dummy stem pi

However Johns do es not nd this convincing preferring to work in terms of asp ectual

classes

And on the basis of this Baker seeks a Case parameter that would explain this

and the intransitive agreement on verbs with incorp orated arguments in Greenlandic unlike some

other languages

A third p otential argument for unaccusativity is the alleged app earance of antipassives of un

accusative intransitive verbs Bittner contains one example of such a construction which is

rep eated in BokBennema and De Ho op However Bittner and Hale forth

coming b rep ort that the initial example was only obtained from one consultant and follow up

work with six consultants on this and other sentences with antipassives of intransitives suggests that such sentences are all ungrammatical

Chapter Inuit West Greenlandic

What is the structure of relative clauses Johns assumes that

the structure of Inuit relative clauses is actually two noun phrases in app osition

Johns presents no particular argument for this structure apart from the fact that it

works within the context of her analysis but I supp ose one could argue as follows If

we think of the relative clauses ab ove as a head noun followed by a participial phrase

then note that the participial phrase follows the head noun just like a noun which

attributively mo dies another noun

a arnaq kalaaliq

womanabs Greenlanderabs

thea Greenlandic woman

b kalaaliq arnaq

Greenlanderabs womanabs

thea female Greenlander

Secondly note that participial relatives are nominalizations This is further discussed

in Section but note here that they take nominal case endings agreeing in case

and number with the pivot

miiqqanut Juunap paarisainut

childplterm Juunaerg lo okafterreltrsgplterm

for the children that Juuna is lo oking after

Indeed participial relatives can app ear as indep endent noun phrases meaning the

one who was Ved Thus assimilating them to the class of other app ositional nomi

nals is app ealing So Johns would suggest that the structure of

qilalugaq piniartut malirsugaat

whaleabs huntergenpl chasereltrplsg

the whale that the hunters are chasing

is more accurately represented by the calque the whale the hunters chased one

recall that ergative case is the same as genitive case

77 This also the p osition of Fortescue

Chapter Inuit West Greenlandic

However Bittner argues that a head noun and a following relative clause

are actually a single clausal nominalization a nominalized internallyheaded relative

clause rather than there b eing an app ositional structure of a head noun followed by

a nominalized participial relative She suggests the phrase structure shown in

D DP

NP qilalugaq D

N D N

reltr plsg

t VP

VP DP

erg

piniartut

t V

Bittner provides one fairly comp elling argument for this clausal structure namely

that the ergativegenitive NP can bind a reexive on what Johns prop osed as the

head noun

nirisani Maakkap saniminut ilisimasaa

fo o dsgsg Maakkaerg sidesgterm putreltrsgsg

qimmirsuup aallaruppaa

dogbigerg runawaywithindtrsgsg

A big dog ran away with her fo o d which Maakka had put by her side

i i i

This would b e totally unexp ected if we were simply dealing with two nominals in

app osition and so I conclude that an analysis via a clausal nominalization is correct

Some further discussion on the treatment of clausal nominalizations within the type

of framework that I am assuming can b e found in Chapter

Bittner is interpreting the case of piniartut here as functioning within a clause ergative not as mo difying a noun genitive

Chapter Inuit West Greenlandic

Extension to other languages Towards the end of her pap er Johns tries to

extend her mo del to Dyirbal referring to certain suggestions of Hale This

section is quite misguided Hales pap er itself is problematic see Dixon

Dixon and the references cited in the latter but Johns use of it

is worse She suggests that the l of Lconjugation Dyirbal verbs is a passive par

ticiple morpheme This is what Hale suggests as the diachronic origin of the

Lconjugation in Australian languages but Hale argues that this is not

the correct synchronic analysis of languages including Dyirbal He argues that the

l has b een reanalyzed as a conjugation class marker and gives as evidence that

the conjugations have b ecome partially inconsistent with strict sub categorization in

Dyirbal ab out of transitive stems are in the Lconjugation the rest in the Y

conjugation and ab out of intransitive verbs are in the Lconjugation Dixon

It is clear that an analysis of l as a synchronic passive participle mor

pheme cannot b e maintained for Dyirbal for this and other reasons such as the

existence of the Dyirbal antipassive

Bittner

Bittner is a work of such scop e and depth that I cannot p ossibly summarize

it adequately in a few pages but I will try and outline some of the ma jor lines of the

analysis and their assumptions Bittner also provides a Casetheoretic explanation of

morphological ergativity in West Greenlandic Inuit Again the verb cannot assign

accusative Case and so the ob ject must raise to a higher pro jection to receive Case

But the details of her framework are rather dierent Bittner attempts to motivate

an abstract level of syntax for which she maintains the name sstructure suggesting

that this level determines structural Case assignment agreement syntactic binding

relations and the minimum scop es of quantiers it is the default LF This level is not

required to have any particular relationship to the surface form and indeed Bittner

makes lib eral use of PF movement op erations for what she regards as syntactically

unimp ortant movements various pro cesses of adjunction and extrap osition

Arguing that absolutive agreement is outside ergative agreement when they are

not fused and that absolutive agreement is fused with switch reference morphology

Chapter Inuit West Greenlandic

which is usually in Comp Finer Bittner suggests that absolutive agreement

is in C while ergative agreement is in I She assumes a version of the VPinternal

sub ject hypothesis where at Dstructure sub jects are the distinguished adjunct of a

lexical head a p osition I will denote as Adj VP and suggests that the absolutive

argument of transitive verbs is generated in Sp ec VP and that the instrumental

theme argument of a ditransitive verb is generated as the complement of V She

assumes that only the complement sp ecier and distinguished adjunct of lexical heads

are A p ositions so Sp ec IP for instance is an A p osition

The basic structures for transitive and intransitive sentences that Bittner assumes

are shown in Bittner has the argument of all intransitive verbs originat

ing in Adj VP but Bittner email Dec suggests that unaccusative sub jects

could b e base generated in Sp ec VP and then b e forced to move through the dis

tinguished adjunct of VP p osition by invoking the Extended Pro jection Principle

Chomsky

a CP

C IP

q I DP

I VP Juuna

Juunaabs

vu DP VP

i i

indtr

suli

work

Juuna works

Note that neither of the premises here is unproblematic The order of agreement axes varies as

discussed in Section Both ergative and absolutive agreement suxes on sub ordinate verb forms

distinguish b etween proximative and obviative forms so it is not clear why absolutive agreement

in particular should b e in Comp and moreover these suxes app ear to act more like unbounded reexives than a switch reference system Section

Chapter Inuit West Greenlandic

b CP

C IP

i I DP

I VP miiqqat

childpl

VP KP pa

i i

indtrsg

DP V Juunap

Juunaerg

V KP

a junaarnirmik uqaluttuup

accidentmod tellab out

Juuna told the children ab out the accident

As further evidence for placing ergative agreement in I she notes that with inni

tive verbs in llu there is only agreement with absolutive arguments

a Miiqqat Juuna ikiussallugu niriursuipput

childrenabs erg Juunaabs helpfutinfsg promiseindintrpl

The children promised to help Juuna

b Miiqqat qitissallutik niriursuipput

children abs dancefutinfpl promiseindintrpl

The children promised to dance

Bittner suggests that this agreement pattern supp orts her theory of agreement where

by absolutive agreement is controlled by Comp and ergative agreement by In one

suggests that just as in English a nonnite In do es not agree But this equation

as well is not unproblematic since note that in a there is an overt absolutive

NP app earing in Sp ec IP and the standard prediction of GB control theory is that

an overt NP cannot b e licensed in Sp ec IP in innitive clauses

As shown ab ove the morphologically unmarked absolutive NPs are analyzed as

Kless DPs but other nominals are analyzed as KPs The K trivially licenses the

Kase was introduced into GB as a feature by Fukui and prop osed as a functional

Chapter Inuit West Greenlandic

DP complement but the K of structural cases is deemed to b e empty so that it in

turn has to b e antecedentgoverned by a head in a Caseassigning conguration so

as to satisfy the ECP Bittner stresses how her theory of Case diers from b oth

the categoryoriented theory of early GB and the p ositionoriented theory of more

recent work For Bittner Case assignment only o ccurs in a more elab orate structural

conguration informally a head can only assign Case to an NP that it governs if

it can also see another NP which can serve as a case comp etitor for the rst NP

This means that heads can gain and lose their Case assigning p otential dep ending on

what NPs are generated in or move through their pro jections The prop osal seems

to oer very nice results for Inuit but lo oks increasingly forced when extended to

other language types it requires null determiners adjoined to V heads to provide a

Case comp etitor in accusative languages and so on Note also that Case is assigned

under government and not by the now usual mechanism of Sp ecHead agreement

Bittner do es not try to derive the unmarked surface word order at Sstructure

She notes

Out of context the preferred order of arguments in clauses is ergnom

oblV This app ears to b e derived by fronting the ergative sub ject over

the nominative ob ject at PF without consequences for any of the syntactic

or semantic phenomena discussed in this study

Given the scop e facts that she rep orts Section one cannot get b oth the word

order erg abs and scop e absolutives have wide scop e right at Sstructure One

needs to p ostulate discrepancies b etween Sstructure and either PF or LF and Bittner

chooses the former of course she also needs and uses Quantier Raising to account

for scop e ambiguities but nevertheless this makes the Sstructure the default LF

However the movements en route to PF which Bittner prop oses in various places

app ear to weaken the empirical base of the theory and Bittner never suggests what

interface conditions of the phonetic comp onent require such movements

category ab ove DP similar to CP ab ove IP by Lamontagne and Travis

Emptiness is here an abstract notion since on Bittners analysis all Ks have surface phonological

content if their complement is lexical and dont if it is a pro There is thus no correlation b etween these overt phonological dierences and the notion of emptiness for purp oses of the ECP

Chapter Inuit West Greenlandic

For binding purp oses Bittner suggests that arguments of a verb reconstruct down

to their highest Ap ositions inside the VP since A movement do es not count for

binding and she analyzes Sp ec IP as an A p osition Binding is limited in Inuit

to NPs that are distinguished adjuncts of a lexical pro jection The coverage of the

binding theory I b elieve to b e identical to the one that I oered earlier essentially

b ecause the notion of the VP pro jection in recent GB work mirrors the notion of

argument structure In Bittners work passives and double transitive suxes are an

alyzed as higher verb pro jections and thus b oth the embedded and higher asub jects

of these constructions on my analysis are NPs in Adj XP of one of two lexical

pro jections on her analysis However Bittner cannot state binding theory in terms

of ccommand b etween the anaphor and its antecedent since in many cases such as

anaphors and agreement markers in adverbial clauses the reconstructed sub ject will

b e to o low in the tree to ccommand or even mcommand the anaphor To solve

this problem Bittner prop oses to use accessible sub ject paths which is essentially a

system of coindexation that heads up from a sub ject to the head of the extended pro

jection in which it app ears The head of a sub ject path is then required to ccommand

the antecedent I see this as an essentially technical solution without any particular

explanatory merit Basically the sstructure is not of itself providing the dominance

relations that her account requires Control of the controlled sub ject of innitives is

also stipulated to b e by a sub ject via the Extended Control Condition A controlled

element is b ound by an accessible sub ject path in its binding domain

The result is two notions of Sub ject The items in Sp ec IP which receive

82 0

This is consistent with the original denition of A and A p ositions given the VPinternal

sub ject hypothesis Sp ec IP never receives a role and so should b e an A p osition but most other

recent work wishes to maintain Sp ec IP as functionally an Ap osition in languages like English

and so many p eople now work in terms of Lrelated p ositions which includes Sp ec IP A p otential

0 0

aw with regarding Sp ec IP as an A p osition is that movement to A p ositions is predicted within

GB to cause weak crossover violations This issue is discussed further in Section

Bittners explanation of why the ergative cannot bind the lo cal absolutive in sub ordinate clauses

compare the discussion around is quite technical As mentioned the absolutive agreement

is analyzed as b eing in C A stipulation in the denition of a path ccommanding a no de p

states that a path can only ccommand a no de if the head of is not an extended pro jection

of the no de Thus although the sub ject path heads up to the top of the extended pro jection CP it

can never actually bind something in C As far as I have b een able to tell this stipulation has no other use

Chapter Inuit West Greenlandic

absolutive Case provide one natural class The other notion of sub ject is things that

are the distinguished adjunct of a lexical head These two notions of sub ject are

essentially the same two notions that I try to motivate in the present work and indeed

my analysis is generally compatible with and in places inspired by the analyses of

Bittner The question that remains in my mind however is whether the theoretical

backdrop of Bittners analyses are helping or hindering her prop osals For example

as discussed immediately ab ove constructing the binding theory in terms of sub ject

paths so that there is something that ccommands anaphors seems essentially a

technical feat rather than the framework in any way illuminating what is going on

Parallels

All of these three analyses have at least a few commonalities Each places the ab

solutive in a high p osition within the pro jection of In In Bittner and Johns

analyses the absolutive NP is in Sp ec IP or its equivalent AgrP on Johns anal

ysis In contrast BokBennema adjoins the absolutive NP to I Although it is not

readily apparent I b elieve all three analyses present basically the same theory of

what ergativity is namely that the verb cannot assign accusative case in ergative

languages BokBennema says this explicitly while in Johns analysis this follows

from the claim that verbs cannot pro ject a VP in Inuit and so two arguments can

only b e expressed by using passive participle forms Bittners analysis is essentially

the same as BokBennemas except that she reverses the markedness wrongly I b e

lieve given that accusative languages are much commoner than ergative ones On

her analysis no verb is inherently able to assign Case and verbs only gain the ability

to assign case if a case comp etitor a D is adjoined to them This adjunction o ccurs

in accusative languages but not in ergative languages like Inuit The accounts dier

in their treatment of binding BokBennema and Johns attempt to capture binding

This is similar to the prop osal of Guilfoyle et al

This markedness reversal leads to another o dd markedness reversal in the analysis of passives

in accusative languages Bittner and Hale forthcoming b The addition of a passive morpheme has

to b e analyzed in their framework as the removal of the adjoined D that normally licenses verbs to assign accusative case in accusative languages

Chapter Inuit West Greenlandic

asymmetries in terms of congurational relationships in the higher functional pro jec

tion part of the tree Bittner do es binding mainly in terms of relationships within

lexical mainly VP pro jections I thus see her account as quite similar to b oth the

account developed here and the one in Sado ck forthcoming

Bobaljik and his precursors

The ma jority of analyses of ergativity equate Absolutive with Nominative and then

sometimes Accusative with Ergative with the result that intransitive sentences have

the same analysis in b oth ergative and accusative languages while transitive sentences

are analyzed dierently in the two language types Bobaljik describ es him

self as reviving the idea of Levin and Massam that we should rather equate

Nominative and Ergative and also Accusative and Absolutive This has the result

that transitive sentences are analyzed the same in ergative and accusative languages

whereas intransitive verbs are analyzed dierently Essentially ergativity then means

that all intransitive verbs have ob ject agreement with their actant Bobaljiks pro

p osal thus reduces to the suggestion that intransitive sub jects in ergative languages

are like ob jects that is the AbsolutiveSasOb ject analysis that was discussed in

Chapter

Levin and Massam

Levin and Massam consider the treatment of morphologically ergative languages as

suming Marantzs division into morphologically and syntactically ergative languages

Their theory of Case is that nomerg case is always assigned by I and accabs case

is always assigned by V The app earance of morphological ergativity dep ends on

a parameter of whether I or V must always b e a Caseassigner The sub ject is al

ways in Sp ec IP but V can assign Case at a distance in morphologically ergative

languages

The Niuean data Levin and Massam discuss are apparently problematic for Bobaljik as

he has the intransitive sub ject remain in ob ject p osition whereas Levin and Massam dep end on all sub jects b eing in the sub ject p osition but I have not further investigated the situation in Niuean

Chapter Inuit West Greenlandic

They consider Niuean Austronesian Tongic subgroup sentences with sentential

complements and observe that in some the sub ject takes ergative case and in most

absolutive without apparent dierences in word order They explain how this con

icts with previous theories of Casemarking of Sbar arguing that Sbar must b e

able to b e Casemarked but that some Sbars have inherent Case They then exam

ine how this conicts with Burzios generalization if construed to hold of sentential

complements arguing that T A do es not always hold but is derivable from Case

Theory in certain contexts while A T is conrmed and can b e derived from their

Case Theory They suggest that the Casemarked Sbars cannot b e NPs b ecause

then they would b e extraction islands and they are not However this argument

is not very strong since many languages eg Quechua Manning Chiche wa

Bresnan a exhibit similar nominalized sentential complements whose exter

nal distribution is that of NPs This phenomenon remains a general problem for

syntactic theory

Bobaljik

Bobaljik adopts the kind of articulated IP tree shown in and suggests that

while in transitive sentences A always moves to Agr and O always moves to Agr for

intransitive sentences in an accusative language S moves to Agr while in an ergative

language S moves to Agr This movement is forced by which Agr is active

Where T assignment of a role to the sub ject and A Case assignment to the ob ject

This notion of active must b e distinct from the notion of strong vs weak features that app ears in other Minimalist work

Chapter Inuit West Greenlandic

Agr P

Sp ec Agr

Agr TP

Sp ec T

T Agr P

Sp ec Agr

Agr VP

NP V

V NP AS

These dierent movements yield the dierent Casemarking patterns since nomina

tive ergative Case is checked by Agr while accusative absolutive Case is

checked by Agr He acknowledges that nominative and absolutive resemble each

other as b eing the Case that is always at least abstractly realized and as b eing

generally the morphologically unmarked case but this is seen just as a reection of

which Agr is active the always active Agr would tend to b e less marked p erhaps

by Economy

Bobaljik never makes clear whether he sees his claims as applying to what have

b een termed morphologically ergative languages or syntactically ergative languages

or b oth Of the languages discussed Basque Niuean Abkhaz Inuit Nisgha the

ma jority have standardly b een categorized as morphologically ergative Bobaljik

makes two arguments for his analysis

But we saw in Chapter that there are reasons to b elieve that b oth Inuit and Nisgha should

actually b e classed as syntactically ergative at least according to the Inverse Grammatical Relations hypothesis that I have advanced

Chapter Inuit West Greenlandic

Argument One The main argument for this prop osed structure comes from bind

ing theory Bobaljik suggests that in ergative languages just as in accusative lan

guages the A argument binds the O argument in reexives and recipro cals For

example in Basque

mutilek elkar ikusi dute

b oyserg each othernom see auxsgAplE

The b oys saw each other

Bobaljik following Anderson takes this as evidence that the ergative NP

asymmetrically ccommands the absolutive NP at the level at which the Binding

Theory applies Inuit lacks basic transitive reexives it intransitivizes the verb but

the same p oint can b e illustrated with the reexive form of the p ossessive marker

compare a and

Further Bobaljik cites Chomsky in suggesting that Binding Conditions

as conditions on interpretation should apply at LF the interpretive interface and

hence suggests that it is not the case that Binding can b e satised at an earlier level of

derivation with the ccommand relationships then undone by subsequent movement

This command relationship must exist at LF and the default assumption is that the

ergative also ccommands the absolutive at Sstructure sp ellout and so absolutive

cannot p ossibly b e assimilated to nominative Note that this use of LF directly

contradicts Bittners analysis where the absolutive must b e higher than the ergative

at least in the default LF to explain the wide scop e reading of absolutives While

not directly committed to a level exactly like either sstructure or LF in general I

side with Bittner in b elieving that binding is b est dened on an intuitively deep er

level which I term argument structure

Finally Bobaljik argues against the alternative account which equates nominative

and absolutive case as follows If the absolutive argument raises to Sp ec IP to

get Case and then the ergative argument raises by an A movement to some higher

p osition either to capture the ErgAbsV word order facts of Inuit or in order to also

explain the binding facts mentioned ab ove as in Johns then one would exp ect

Weak Crossover eects that do not o ccur This prediction and the question of weak crossover in general will b e further discussed later in this section

Chapter Inuit West Greenlandic

Argument Two Bobaljik notes that in accusative languages NPs in Sp ec Agr

cannot b e marked nominative or trigger agreement in tenseless contexts innitives

and gerunds and so must usually b e realized as PRO Given Bobaljiks ergativity

hypothesis the obvious prediction is that in tenseless contexts in ergative languages

the ergative argument should b e unable to app ear or trigger verb agreement while

the absolutive argument should b e able to app ear and trigger agreement in b oth

transitive and intransitive sentences

While noting that many ergative languages lack suitable innitive forms Bobaljik

claims that this prediction is supp orted by the Inuit llu construction He suggests

that this construction is like an innitive has agreement only with the absolutive

argument and allows expression of an absolutive but not an ergative argument within

the same clause citing examples like

a Miiqqat ikiussallugit niriursuivutit

childrenabs helpfutinfplabs promiseindsg

You promised to help the children

b aggissallutit niriursuivutit

comefutinfsg promiseindsg

You promised to come

c niviarsiaq sikkirluni kiinanngua nuiratannguarpuq

girlabs blossominfsg facelittlesg app earallthesamelittlesg

The girl ie the willow herb blossoming her little face app eared at last

Bobaljik takes this construction as evidence for his analysis saying that in

innitivals just as in English Agr is defective in a nontense environment and hence

it can neither license Case nor agreement In other words he claims that in a llu

innitive there will b e neither agreement with A nor will A b e licensed to app ear

overtly with ergative Case The agreement facts are correct but the claims ab out

Though this is not the case in Portuguese of course Rap oso Portuguese innitival

clauses allow overt nominative sub jects and the verb agrees with them This is generally analyzed

as resulting from a case assigned to the CP p ercolating to the sub ject or a Case originating in C

b eing assigned to the sub ject

Example b corrects a typo in Bobaljiks a The translation of c has also b een altered in accord with my understanding of Bergsland

Chapter Inuit West Greenlandic

overt NPs are just wrong as discussed in Section neither an S nor an A can

app ear in complement uses of the innitive while in adverbial uses like c overt

NPs functioning as any of A S or O may app ear Hence this argument collapses

Bobaljik

The rst half of Bobaljik reworks Bobaljik bringing it into accord with

the available linguistic data while the second half deals with two topics how to

recapture the notion of Sub ject within this framework and how to deal with active

stative languages in which some intransitive verbs take an ergativelymarked actant

So in the rst section Bobaljik deletes discussion of Niuean as an ergative language

with an accusative binding theory suggesting that the reexive there is used more

as an emphatic marker He cites discussion by Keenan of Samoan where the

reexive lava is principally an emphatic marker but can b e used with either the

ergative or absolutive of a transitive to indicate coreference based more on precedence

than dominance See also Mosel who argues for this element in Samoan not

b eing an anaphor at all

Regarding the Inuit innitive without overtly correcting his earlier claims Bobal

jik now suggests that we should fo cus primarily on the agreement morphology as

suming that the relations expressed by this morphology are the essential relations

of the clause whereas consideration of the distribution of lexical NPs is p otentially

misleading p The theoretical basis for this is suggesting that Inuit falls in with

the class of languages whose lexical NPs have b een analysed as adjuncts while the

agreement morphemes are actually the arguments of the verb Jelinek Baker

This prop osal do dges the counterevidence from the distribution of overt NPs

in innitive clauses but it makes a number of further predictions which Bobaljik do es

not go on to discuss I will take up this agenda b elow

The next section deals with how to recapture the notion of deep sub ject grouping

A and O under Bobaljiks analysis recall that under his analysis the A argument

moves to Sp ecAgr while the S moves to Sp ecAgr Bobaljik suggests this class

is needed to determine the antecedent of reexives and the controller and controllee

Chapter Inuit West Greenlandic

in controlled construction like the Inuit llu innitive The essence of Bobaljiks

prop osal is that we make checking the features of T the equivalent of the Extended

part of the Extended Pro jection Principle that every clause must have a sub ject

If this feature checking is required then an A argument and an S argument in an

Accusative language will move rst to Sp ec TP and then to Sp ec Agr while an

S argument in an ergative language will move to Sp ec Agr and then to Sp ec TP

Being in Sp ec TP conveys sub ject prop erties There are a number of theory internal

issues here many Minimalist analyses regard the sub ject as moving straight to

Sp ec AgrS bypassing TP and if Ts Nfeatures are strong as Bobaljik assumes

then an S and an O will actually b e in dierent p ositions at Sp ellOut This might

b e detected by word order facts for instance by using adverbs that adjoin to AgrP

but the available evidence actually suggests that all absolutive NPs are in the same

p osition at Sstructure Section However given the current state of ux of

the Minimalist Framework I will not pause to investigate these issues in detail The

crucial p oint is that this analysis gives Bobaljik a congurational way of grouping A

and S as well as the congurational way of grouping S and O that was previously

prop osed I also want to have two notions of sub jectho o d but I b elieve it is necessary

to dene asub ject over argument structure as I have done and an analysis that simply

groups surface A and O will b e insucient

The nal section addresses the fact that a number of Ergative languages have

the p erhaps unexp ected prop erty that some intransitive verbs mark their single ac

tant Ergative rather than Absolutive Basque Georgian etc Bobaljiks solution

is to adopt Hale and Keysers suggestion that unergatives are underlyingly

transitive In cases where there is Ergative marking on the S argument there is

analyzed as b eing an unincorp orated ob ject either abstract or arguably overt in

certain Basque forms whereas in other cases like Inuit where all intransitives clearly

Though again recall the evidence from Section that these actually app ear to b e prop erties

of asub jects not just surface As and Os I have not yet seen a minimalist account of passive or

ECM constructions but I b elieve Bobaljiks prop osals would not extend correctly to these cases

The section do es not address the question of why and whether these languages should b e analyzed

as Ergative rather than Active but under Bobaljiks analysis arguably this is just a terminological question

Chapter Inuit West Greenlandic

act as intransitives the analysis is that unergatives are underlyingly still transitive

but that their ob ject incorp orates into the verbal head as Hale and Keyser suggest

Note that Bobaljiks prop osal amounts to a very strong semantic claim ab out lan

guages such as Basque namely that all and only those intransitive verbs that take

an Ergative argument are unergative verbs which underlyingly have the syntactico

semantic structure suggested by Hale and Keyser But I b elieve a closer examination

of almost any such split intransitive language will b elie this claim Most commonly

the class of single actant verbs taking the ErgativeActive marking is either much

smaller than the semantic class of unergatives or conversely much larger and the

Stative class is fairly small and often closed Merlan For the case of Basque in

particular it has b een argued that there is not a p erfect match b etween semantically

unaccusativeunergative verbs and the choice of ergativeabsolutive case marking

Ortiz de Urbina Clearly meaning and case marking are correlated but

only imp erfectly

Why Absolutive should b e group ed with Nominative

Morphologically

Bobaljiks analysis diers from most others in identifying nominative case with erga

tive case rather than employing the usual identication of nominative case with ab

solutive case However I b elieve that at a morphologicalmorphosyntactic level

the right choice is to identify Absolutive and Nominative for at least the reasons laid

out here

Markedness It is a basic observation that nominative or absolutive is the case that

is always registered on some term of a verb mo dulo prodrop and quirky case mark

ing in Accusative and Ergative languages resp ectively and almost exceptionlessly

they are the morphologically unmarked cases there are a few exceptional languages

with marked nominatives but the absolutive seems to b e exceptionlessly unmarked

The idea of nominative and absolutive b eing the same unmarked category app ears in Trubetzkoy

and was probably just assumed by various earlier authors as the term absolutive was something of a late invention

Chapter Inuit West Greenlandic

These observations can b e extended nominative and absolutive are the cases nearly

always used for citation and for the topic of equational or copular sentences Dixon

Ordering of verbal agreement In languages with dual agreement with b oth

axes on the same side of the verb the general pattern is that ob ject markers

are closer to the verb than sub ject markers in Accusative languages for example

in Bantu languages Chiche wa ChiMwini Haya Muskogean languages Chick

asaw and Cho ctaw Aztec Michoacan Sierra Kunparlang NonPamaNyungan

Australia In contrast in languages with ErgativeAbsolutive verbal cross refer

encing it is the ergative ax that app ears closest to the verb while the absolutive ax

is outside it for example in Inuit Abkhaz Mayan Jacaltec Quicheand Tzutujil

The value of this evidence dep ends on ones theory To someone adopting the as

sumptions of Baker and the SplitIn Hyp othesis ordering of axes gives very

imp ortant information ab out not only morphology but the syntactic structure the

ordering of functional pro jections Under the new morphologychecking framework

of Chomsky the imp ortance of this evidence is marginalized but nevertheless

I will take it as indicating something imp ortant at a morphological level

Plausibility of Historical Change Languages are not statically ergative or ac

cusative There are fairly well examined cases of languages changing their character

from ergative to accusative and vice versa There is also a fairly intuitive account

of how this can happ en dating to at least Kurylowicz Kurylowiczs basic

observation is that if an accusative language has a passive that allows expression of

an oblique agent and for whatever reasons prefers passive verb forms to the p oint

that it stops using transitive verb forms as seems to b e happ ening in Maori then

the language has almost b ecome ergative This transition is complete when the pre

vious marker of an oblique agent is reinterpreted as the marker of a core grammatical

ChiMwini Haya Chickasaw and Cho ctaw are from Murasugi Chiche wa and

Kunparlang from Bresnan and Mchombo Aztec from Merrield et al

96 Murasugi

Chapter Inuit West Greenlandic

relation and the old passive marker is reanalyzed p erhaps as a transitive marker

The same account can b e told in reverse for an ergative language that prefers the

antipassive b ecoming accusative The imp ortant thing to note is that in this account

the old nominative b ecomes the absolutive in the new system and vice versa There

is no similarly app ealing story in which the old licenser of nominative case b ecomes

the new licenser of ergative case

While it is certainly not the case that all ergative systems arise from passives cf

Garrett Section in general ergatives and accusatives arise from marked

cases while absolutives arise from the unmarked case nominative

Semantic Scop eSp ecicity As Bittner discusses if any argument in a

sentence has a general presupp osition of sp ecicity or is almost always interpreted

as a wide scop e argument then it is the nominative in Accusative languages and

the absolutive in Ergative languages This again suggests that it is nominative and

absolutive that form a natural class

Weak crossover and the Pronominal Argument Hyp othesis

The ow of ideas in this subsection is rather complex In part this is b ecause I am

discussing two incompatible prop osals of Bobaljiks Bobaljik argues that cer

tain analyses of Inuit should b e ruled out b ecause they would predict weak crossover

violations that do not o ccur an argument rep eated in Bobaljik while a later

section of Bobaljik suggests that Inuit is a pronominal argument language in

the sense of Jelinek which would imply that Inuit should exhibit no weak

crossover eects at least with NP arguments The Pronominal Argument Hyp othe

sis in turn can b e argued for and against from a number of p oints of view I argue

following Baker and against Bobaljik that the existence of pronominal

arguments cannot b e simply correlated with overt morphology As well as predicting

the absence of weak crossover eects on NPs adopting the pronominal argument hy

p othesis has b een used to explain various things such as the app earance of free word

order and free pro drop and rich case and agreement systems Jelinek Sp eas

But in general these phenomena seem p o orly correlated In this discussion I

Chapter Inuit West Greenlandic

can but touch the surface of these issues See Bresnan b and Ortiz de Urbina

for two recent contrasting treatments of weak crossover the latter deal

ing with Basque another ergative language and Austin and Bresnan for a

detailed defence of the view that the phenomena that have b een handled under the

pronominal argument hypothesis are not in fact phenomena that covary

Weak Crossover Bobaljik suggests that a technical problem with certain

previous analyses of ergativity is that the derivations they prop ose should lead to weak

crossover violations that do not in fact app ear to o ccur For example recall that on

Johns analysis the ergative argument moved by A movement to adjoin to a

p osition ab ove the absolutive argument But if this were the case Bobaljik argues

then a sentence like a should b e ruled out as a weak crossover violation for the

same reason that b is ruled out in English

anaanani t nagligijana a Piitaup

i i

Piitaerg motherposssgreflabs lovesgEsgA

Piita loves his mother

i i

b Who do es his mother love t

i i i

While the analysis of Bittner avoids this problem by suggesting that the

ergative NP is fronted at PF note that under Bittners analysis Sp ec IP is an

0 0

A p osition so the p otential problem for Bittners analysis is that A movement of

the absolutive NP to Sp ec IP should lead to weak crossover violations in certain

circumstances But in fact it is acceptable for the p ossessor of an ergative to b e

coindexed with the absolutive NP that moves over it

Ilaani niviarsiaraq arnaata t ap eraa

Once girlabs mothersgsgerg asksgsg

Once her mother asked the girl

i i

The details of why it is ruled out dep end on the account of weak crossover used but standardly

a weak crossover violation results when a coindexed term app ears b etween an op erator the result

0 0

of A movement and its trace Also b ecause the ergative NP is b eing p ostulated to move by A

movement to its default p osition note that the result is that all cases of binding the p ossessor of

an absolutive by an ergative should b e ruled out and not only cases when the ergative NP is a

Whword or a quantier

Chapter Inuit West Greenlandic

To b olster his case Bobaljik notes that clear weak crossover eects have

b een noted in certain ergative languages Basque and Nisgha He do es not examine

Inuit but nevertheless to ok the ab ove to b e an argument in favor of his own anal

ysis in which there was no A movement However for separate reasons to do with

sustaining his argument from innitive clauses Bobaljik suggests that Inuit

is a pronominal argument language which would predict that there would never b e

any weak crossover eects in Inuit sentences with NP arguments for quite separate

reasons Thus I will return to the discussion of weak crossover after a discussion of

the pronominal argument hypothesis

The Pronominal Argument Hyp othesis There has b een a long tradition in

Native American linguistics dating at least to Boas which suggests that in certain

languages the agreement axes of verbs are actually pronominal arguments of the

verb stem while optional lexical NPs are merely mo difying adjuncts Jelinek

rst brought this argument into the generative fold dubbing it the Pronominal Ar

gument Hyp othesis Under such an account it is the pronominal crossreferencing on

the verb that exhibits congurational structure while lexical NPs are just adjuncts

which can app ear freely p ositioned The most interesting recent discussions of this

prop osal are to b e found in the work of Baker which I summarize b elow

Bobaljik claims that it is signicant that the Inuit innitive agrees with

S and O but not A b ecause under the assumption that a Jelinekstyle analysis

is appropriate for Inuit this shows us that an S or O argument is p ermitted with

the innitive as witnessed by the overt agreement while an A argument is not

allowed it must b e PRO which is silent b oth lexically and as a piece of agreement

morphology Now this argument only go es through on the strong and p erhaps

naive assumption that there is an absolute correlation b etween a morpheme b eing

overt and its serving as a pronominal argument syntactically and semantically There

are various arguments against this for example you encounter problems when you

try to extend a Jelinekstyle analysis to other Australian languages such as Jiwarli

which totally lack overt verbal crossreferencing Austin and Bresnan but the

nicest argument against this p osition that I know is from Baker and I will

Chapter Inuit West Greenlandic

summarize it briey here

Bobaljiks claims are vitiated as so on as you accept that some cases of pronominal

arguments must b e zero morphemes as is already accepted by Jelinek for the

analysis of the Warlpiri third p erson arguments and in Inuit to o we could simply

suggest that all innitival mo o d ergative markers happ en to b e zero morphemes But

Baker makes a much more sophisticated argument than that from Mohawk He

considers the forms in and asks whether the ob ject markers in the neuter and

femininezoic genders bc can b e analyzed as zero morphemes given that the

agreement ax in such cases is the same as for an intransitive sub jecta

a teknunyakhwa I dance intransitive

b knuhwes I like it neuter

c knuhwes I like her femininezoic

d rinuhwes I like him masculine

His answer is that while the femininezoic ob ject marker seems to b e a zero mor

pheme all the available evidence indicates that the neuter is not a zero morpheme

but rather neuter is denoted by the absence of a morpheme When a verb is placed

in the stative asp ect intransitives and verbs with neuter objects undergo an inversion

pro cess where the sub ject agreement morpheme is replaced by an ob ject agreement

morpheme whereas forms with an ob ject agreement marker remain unchanged

a tewaknunyahkwv I have danced intransitive

b waknuhweu I have liked it neuter

c knuhweu I have liked her femininezoic

d rinuhweu I have liked him masculine

Generally Mohawk verbs mark sub ject number but this number marking is neutral

ized if there is an ob ject Crucially femininezoic ob jects cause neutralization of

number marking just like masculine ob jects but neuter ob jects do not neutralize

98 See also Nordlinger for a similar argument from Wambaya

Chapter Inuit West Greenlandic

sub ject number marking Thirdly the agreement morphology can only mark two

actants and this results in the curious restriction that for ditransitive verbs the di

rect ob ject must b e neuter This can b e easily explained if we assume that Mohawk

verbal morphology can express no more than two arguments but that the neuter is

not represented in the morphology whereas the femininezoic is expressed as a zero

morpheme See Baker for data justifying these last two arguments

Baker then shows that syntactically a third p erson neuter actant b ehaves in every

way identically to other pronominal agreement actants Binding theory free word or

der and comitative mo dier p ossibilities are exactly the same with third p erson neuter

ob jects as with any other ob jects The only conclusion is that a third p erson neuter

pronominal is somehow asso ciated with the appropriate forms in Mohawk in the syn

tax even though morphologically there is nothing there not even a zero morpheme

Thus evidence of overt or nonovert morphology cannot b e taken as denitive evi

dence of whether there is a pronominal Hence facts ab out overt morphology can at

most b e suggestive they cannot b e taken as decisive evidence as Bobaljik

would wish

Baker adopts Jelineks pronominal argument approach for the analysis of

Mohawk but go es b eyond the standard diagnostics that Jelinek used to motivate her

approach to noncongurationality in Warlpiri namely free word order discontinuous

constituents and free null anaphora to which can b e added lack of movement trans

formations lack of pleonastic NPs and use of rich Case and verbal agreement systems

Sp eas Baker examines and nds supp ort for Jelineks analysis by lo oking

at disjoint reference conditions a is go o d despite other evidence for Binding

Condition C op erating in Mohawk data from extraction from lexical NPs as op

p osed to sentential complements and the absence of weak crossover sub jectob ject

asymmetries b is go o d unlike the translation

a Wathayake Sak raoashare

factdupsSbreakpunc Sak MsPknife

He broke Sak s knife

i i

Chapter Inuit West Greenlandic

b Royatakehnhas Sak raoashare

MsOhelphab Sak MsPknife

Sak s knife helps him

i i

a Uhka wateshakonorukwanyu raoskare

who factdupMsSFsOkisspunc MsPfriend

Who kissed his girlfriend

i i

b Uhka wateshakonorukwanyu akoskare

who factdupMsSFsOkisspunc FsPfriend

Who did her b oyfriend kiss her

i i

Weak Crossover revisited Reconsidering in this light Bobaljiks analysis Bobal

jik seems forced to say that at least Basque and Nisgha for which he do cuments weak

crossover eects are not pronominal argument languages On the other hand while

his ob jections to certain other analyses of Inuit as b eing wrong b ecause they would

predict that sentences in the normal word order should b e weak crossover violations

can technically b e maintained Bobaljik now seems committed to the p osition that

Inuit should not display Weak Crossover phenomena Moreover he is also predicting

similar disjoint reference and extraction nonasymmetries of the sort Baker describ es

A preliminary assessment suggests that the data actually split which would sup

p ort Austin and Bresnan Bittner rep orts normal Principle C

b ehavior in Inuit so Inuit do es not have this prop erty of noncongurationality My

own discussions with one I nupiaqsp eaker suggested that standard weak crossover ex

amples were acceptable to her For example contrasting with the English data in

Even though Basque at least demonstrates many of the other prop erties that are classically

asso ciated with the noncongurationality that the pronominal argument hypothesis allows free

word order and a rich case and agreement system Ortiz de Urbina and elsewhere

100

Note also that the use of discontinuous constituents is extremely limited in Inuit and the order

of agreement axes discussed in Section varying but normally with the absolutive marker

outside the ergative would app ear to preclude these axes b eing regarded as the congurational

part of the language

101

However I should immediately qualify this datum by p ointing out that Maria Bittner email

Jun rep orts getting inconsistent results when she tested weak crossover examples So more thorough examination of this issue is clearly necessary

Chapter Inuit West Greenlandic

the I nupiaq data is in

a Who help ed his mother

i i

b Who did his mother help t

i i

a Kia aakani ikayuqpaung

Whoerg mothersgabs helpinterrtrsgsg

Who help ed her mother

i i

b aakangan ki na ikayuqpaung

mothersgerg whoabs helpinterrtrsgsg

Who was help ed by hisher mother

i ij

In a fourth p erson morphology Section means that the sentence can only

have a coreferent meaning as shown In b the p ossessor can b e another individ

ual but imp ortantly it can also b e Who There is no weak crossover violation

Sado ck forthcoming

Sado ck forthcoming outlines an autolexical syntax treatment of ALS which I have

already done to some extent in Section and then pro ceeds to provide a treat

ment of binding theory The account he provides is very similar to and in some ways

inspires the account presented here Binders of reexives must satisfy a syntactic

condition they must asymmetrically ccommand the reexive and a semantic con

dition they must b e an asub ject The account diers in choosing the syntax rather

than the argument structure as the level at which command is dened In conjunction

with prop osing the phrase structure shown in this gives Sado ck a neat account

of what I captured through the coterm condition ergatives cannot bind absolutives

102

Inuit lacks overt Whmovement but weak crossover eects should still b e seen b ecause the Wh

word moves at LF Compare for instance the similar weak crossover data for English quantiers in

i and see Bobaljik

i a Everyone help ed his mother

i i

b His mother help ed everyone

i i

103 Note that b is active unlike its gloss

Chapter Inuit West Greenlandic

b ecause they do not asymmetrically ccommand them while b oth ergative and ab

solutive asub jects asymmetrically ccommand oblique NPs recall that for Sado ck

pronominal agreement app ears as an NP at the level of syntax

NP NP VP

erg abs

NP V

obl

The indep endent evidence for this phrase structure is not strong although Sado ck

tries to make the b est case for it and some additional representational assumptions

seem to b e required such as that adverbial clauses are generated as a daughter of S

rather than b eing adjoined Nevertheless the structure seems to account neatly for

the Inuit data However I b elieve that this approach cannot b e argued to have general

crosslinguistic application and further that there is a p ossible problem lurking in the

Inuit data

I showed ab ove Section that in Toba Batak an antecedent need not com

mand an anaphor at the level of syntax but must command the anaphor at the level of

argument structure More simply note that general adoption of a condition of asym

metric ccommand would prevent the p ostulation of even a partially at structure of

the sort Sado ck suggests in all those languages such as Basque or Tagalog where core

arguments can bind other core arguments that are more oblique at the level of argu

ment structure For example this theory of binding would not b e able to license the

reexive in given the phrase structure suggested by Kro eger

which is shown in

Sinaktan ngbabae angkaniyang sarili

dvhurt genwoman nomher self

Athe woman hurt herself

104

See Kro eger for a defense of this somewhat unusual phrase structure The present

argument dep ends only on the claim that there is a at structure for the NPs following a verb

Chapter Inuit West Greenlandic

Sp ec I

I S

V NP NP

Sinaktan ngbabae angkaniyang sarili

Thus the incorp oration of syntactic structure command relationships seems unsatis

factory within a crosslinguistic theory of binding

Finally there is one dierence b etween the predictions of my account and Sado cks

My account following Bittner predicts that a passive agent should b e able

to bind another oblique argument of the verb since it is an acommanding asub ject

However this should b e imp ossible on Sado cks account since the oblique agent

of passives do es not asymmetrically ccommand other obliques since they are b oth

daughters of VP The data on this p oint deserves further investigation but Bittner

gives an example that supp orts this binding p ossibility

a Jaaku Piitamit imminik uqaluttualiamik nassinniqarpuq

Jaakuabs Piitaabl selfmod storymod sendpassindintrsg

Jaaku was sent a story ab out self by Piita

i ij j

b passhJ sendhP selfstoryii

105

She notes however that some sp eakers fail to allow binding by the ablative NP but argues that this is due to the greater salience of the absolutive NP

CHAPTER

Lezgian

If the Philippine Eskimo and Mayan language families have b een the cause for much

linguistic debate by providing apparently conicting notions of sub jectho o d other

languages have proved problematic by providing little or no evidence for the surface

grammatical relation sub ject In this chapter I consider one such language Lezgian

and how it could b e treated within the framework that I have b een developing see

Van Valin for a similar consideration of the related language Archi Certain

NPs the ergative and absolutive NPs of transitive and intransitive verbs resp ectively

and dative exp eriencers clearly have a certain prominence in Lezgian but this is

mainly with resp ect to phenomena that I am regarding as sensitive to argument

structure There seems to b e very little that cannot b e captured in terms of a

structure prominence and use of the termnonterm distinction One could argue for

a very weak SO pivot purely on the basis of case marking or a very weak SA pivot

on the basis of certain control facts Hasp elmath notes although maybe they can b e

otherwise explained This unfortunate indeterminacy over whether an argument has

prominence solely b ecause it is an asub ject or whether it should also b e treated

as a grammatical sub ject has b een noted for other languages that lack a rich set

of valence changing op erations See for instance Inman for discussion of

Collo quial Sinhala Ultimately we would hop e to b e able to settle these issues but

in the meantime the existence of such languages provides a strong reason for carefully

distinguishing sub jectho o d tests that are tests of asub jectho o d from ones that are

Chapter Lezgian

tests of grammatical sub jectho o d

Basic Background on Lezgian

The earliest written Lezgian literature dates from the second half of the nineteenth

century and the Russian general Uslar also provided a go o d grammatical description

of the language during this p erio d From the s on there is work in Russian on

Lezgian most notably work by Gadziev in the forties and fties Unfortunately

this work is inaccessible to me and this chapter is based entirely on work on Lez

gian written in English principally Mel cuk and the excellent grammar by

Hasp elmath

Genetic Aliation

Lezgian is a member of the Lezgic branch of the NakhoDaghestanian family its clos

est relatives including Archi and Tabasaran sp oken in southern Daghestan inside

Russia and northern Azerbaijan

Case Marking and Case Markedness

All nouns pronouns and demonstratives have absolutiveergative case marking The

absolutive app ears to b e a bare stem eg sew b earabs The ergative and all

the other cases are then formed from an oblique stem eg sewre b earerg

sewrez b eardat The ergative also has a zero ending but is based on the

oblique stem Alternatively one could say that the ergative is simply zero and

that the absolutive is subtractive The ergative is used only to mark the agent of

transitive verbs it apparently sometimes marked instrumentals in older forms of the

language Hasp elmath The other cases are a dative and a genitive and

originally lo cative cases arranged in a slightly defective three by ve paradigm Many

of the other cases have sub categorized uses dative exp eriencer sub jects p ostelative

stimuli p ostessive patients adelative sub jects of ze can etc

In these and other examples I follow Hasp elmath in using a p erio d to separate the bare stem from the oblique stem ending and a hyphen to mark other case endings

Chapter Lezgian

Agreement

There is no agreement The only p ossible qualication to this statement is that nouns

mark number and so nominal predicates have the same number as their sub jects in

copular clauses

Suwan ceher gzaf muqajat ha jwanar ja

mountaingen goatplabs much careful animalpl cop

Mountain goats are very careful animals

However I would suggest that this is not a grammaticized agreement system but

rather that hajwanar animals is plural simply b ecause we are talking ab out more

than one goat that is b ecause of a fact ab out the world There is some further

discussion of this phenomenon in Section

Among the Daghestanian languages Lezgian is exceptional in lacking agreement

and so it is p erhaps worth briey considering the comparative evidence Proto

Daghestanian apparently had a noun class system and the verb agreed with the

OS in class that is there was an ergative agreement pattern This pattern of class

marking and agreement is preserved in all of the mo dern languages except certain

members of the Lezgic subfamily namely Lezgian Agul and the southern dialect

of Tabasaran Kibrik Kibrik though it seems to also b e decaying in

Avar it is present only on vowelinitial verb stems Kibrik Archi Lezgic

subfamily is typical it has a system of four noun classes and the verb agrees with

the OS in class and number The situation in Tabasaran also Lezgic subfamily is

complex Kibrik There are three sets of p erson and number agreement mark

ers for verbs which I will term actor undergo er and dative Intransitive verbs can

take the actor or undergo er agreement axes in what app ears to b e a uid splitS

system using the terminology of Dixon Two actant verbs agree with an

actor using the actor suxes or an exp eriencer using the dative suxes and in certain

cases there is then an optional second agreement marker either undergo er or dative

agreeing with an undergo er or recipient argument Additionally the verb marks a col

lapsed version of the protoDaghestanian class system distinguishing only two classes

and again agreeing with the OS This class agreement system is totally lost in the

Chapter Lezgian

southern dialect Thus the old ergative agreement system is b eing overlaid by a uid

activestative system The b eginnings of a p ersonnumber agreement system can also

b e seen in Akhvakh Kibrik

Word Order

Sentences are normally head ie verb nal although esp ecially in sp eech one or

more constituents of the clause may app ear after the verb The most common word

order seems to b e SVAOV but Hasp elmath suggests that this is just

a consequence of a preference for putting given information b efore new information

and the fact that sub jects are generally given information Thus word order should

p erhaps b e given the sort of pragmatic treatment argued for in Austin forthcoming

Heavy constituents can b e moved to either end of a sentence The immediately pre

verbal p osition seems to have something of a sp ecial status questioned constituents

usually but not necessarily app ear here But note that this is a clauseb ound re

ordering although the questioning of elements in sub ordinate clauses is allowed in

Lezgian the questioned constituent remains within its own clause

Termhoo d

Because of the lack of valency alternations and agreement it is very dicult to assess

termho o d in Lezgian Circumstantial evidence can b e had from the distribution

of Lezgian valence patterns For what Hasp elmath calls the standard

valence patterns it is noteworthy that they divide into two groups intransitive and

transitive valence patterns All the intransitive patterns have an absolutive argument

and zero or one other oblique case argument and the transitive patterns have b oth

an ergative and an absolutive argument and then similarly zero or one oblique case

arguments This might suggest that the absolutive and ergative are terms The

remaining nonstandard valence patterns are genitive arguments of certain verbal

idioms dative exp eriencer sub ject verbs and noun awunxun dob e comp ounds

Dative exp eriencer sub jects can b e distinguished from other oblique case arguments

2 I use here Dixons notation for word order since it is not clear what the pivot is in Lezgian

Chapter Lezgian

that could b e regarded as asub jects and so I will suggest that they should also b e

regarded as terms

Valence changing morphology

Valence changing morphology is very limited in Lezgian There is a causative mor

pheme ar Hasp elmath but it is less than fully pro ductive For verbs with

which it do es o ccur with one or two exceptions it adds an ergative argument ie

moving the verb from an intransitive to a transitive valence pattern The ax ar

also app ears with a limited number of transitive verbs but here it causes no dierence

to the verbs meaning or valence pattern There is also a p eriphrastic anticausative

made from the stem of weak transitive verbs and the auxiliary xun b ecome b e It

is used to denote pro cesses strictly without an external agent as opp osed to merely

backgrounding the agent This limited range of voice changing op erations app ear to

provide very little evidence ab out the structure of Lezgian

Phenomena exp ected to b e sensitive to argument

structure

In this section I will provide evidence from idioms imp eratives binding and control

in complement and adverbial clauses to show that in Lezgian to o these phenomena

seem to b e sensitive to argument structure

Binding Theory

In the vast ma jority of cases the antecedent of a reexive is an asub ject That is the

antecedent can b e an ergative A NP a an absolutive S NP b or a dative

exp eriencer sub ject c or it can b e in various other roles that can still plausibly

b e analyzed as asub jects such as the adelative argument of xun b e able d

Lezgian verbs are divided into strong verbs stressed on a thematic vowel and weak verbs

stressed on the stem and lacking a thematic vowel This construction is only available for weak verbs

Chapter Lezgian

a Kamala wicin bubadin ruza qacuna

Kamalerg selfgen fathergen rieabs takeaor

Kamal to ok his fathers rie

i i

b Rus g uzguda wiciz kiligna

girlabs mirroriness selfdat lo okaor

The girl lo oked at herself in the mirror

i i

c Adaz wicin roman patal k une gaji xtin delilar

hedat selfgen novel for youallerg giveaop like detailplabs

kimizmaj

lackingimpfcontpast

For his novel he still needed details like the ones you gave

i i

d Q uzu qullugcidiwa j anzax gila wicin kir exirdal qwan

old employeeadel only now selfgen thought endsress until

luhuz xana

sayinf canaor

Only now could the old employee express his thoughts completely lit

i i

up to the end

Lezgian has a number of interesting arguably idiomatic constructions where one

of the apparent actants is a genitive case mo dier of a noun Several such expressions

are formed with the noun rik heart An example is shown in

Seliman rikela j p enzer aqaliz alatna

Selimgen heartsrel erg window closeinf falloaor

Selim forgot to close the window

lit For him to close the window fell o from Selims heart

On the assumption that such constructions have a normal argument structure along

the lines of

forgethforgetter forgotten i

but idiomatic syntactic expression then this construction can b e used to distinguish

things sensitive to asub jectho o d from things that are sensitive to a surface grammat

ical relation It is thus imp ortant to note that the genitive NP in this construction can control reexives

Chapter Lezgian

Adan rikel wicin stxajar xkwezwa

shegen heartsress selfgen brotherpl returnimpf

She remembers her brothers

Antecedence is not strictly limited to sub jects Sometimes a recipient can bind a

theme or a theme can bind a lo cative

Axundova Basirazni Ab duselimaz cpin qullugar

Axundoverg Basirdatand Ab duselimdat selvesgen jobpl

mubarakna

congratulateaor

Axundov congratulated Basirand Ab duselim on their new jobs

i i

Hasp elmath suggests that the absolutive NP of a transitive sentence can

only b e the antecedent for a reexive when the antecedent is animate and no other

p ossible antecedent exists in the clause But this constraint seems doubtful since

later p Hasp elmath shows a recipro cal whose antecedent is the absolutive of a

transitive sentence and inanimate

Ada kusar sadsadaw agudna

sheerg piecesplabs oneabsoneadess approachaor

She put the pieces together lit close to each other

i i

At any rate an absolutive noun cannot bind into the ergative or dative exp eri

encer argument in its own clause

a Wici Ali gatazwa

selferg Aliabs b eatimpf

Himself is b eating Ali

i i

The recipro cal is expressed by a reduplicated form of ceb themselves or here sad one In

this reduplicated form the rst comp onent app ears in the case of the antecedent and the second

in the case required by the syntactic p osition of o ccurrence of the recipro cal This is of interest

mainly b ecause it is a counterexample to the restrictive theory of information sharing presented in

Pollard and Sag if their binding theory is assumed to have universal applicability where

only information of type index p erson number and gender is shared b etween antecedents and

anaphors In Lezgian case must b e transmitted as well but in the current architecture of HPSG

there is no small bundle of information that includes information on p erson number gender and case

the minimal containing sort is actually local See Hasp elmath for further examples

of this construction See Kathol for an HPSG agreement theory that do es group CASE

with other agreement features although it app ears not to make any predictions ab out agreement p ossibilities in cases of anaphoric binding

Chapter Lezgian

b Wiciz Alidin rusar akuna

selfdat Aligen daughtersplabs seeaor

Himself saw Ali s daughters

i i

The examples I have presented so far are consistent with an astructure based ac

count of binding However Hasp elmath argues that binding in Lezgian is

not actually syntactically determined but dep ends rather on pragmatics He writes

In contrast to a number of particularly well studied languages esp ecially English

where coreferential omission and reexivization have b een shown to b e sub ject to

welldened grammatical constraints Lezgian do es not show such clear grammatical

conditioning of rules of coreference The controller and the target of reexivization

and coreferential omission seem to b e pragmatically rather than syntactically deter

mined He is concerned with such exceptional examples as the following in the rst

of which the absolutive NP binds inside the ergative NP and in the second of which

the apparent antecedent is the p ossessor of an oblique NP

a Tagixalu wicin kirri akaznawaj

Tagiuncle selfgen thoughtplerg tormentprfpast

Uncle Tagi was tormented by his thoughts

lit Self s thoughts tormented Uncle Tagi

i i

b I x urerin q uzuburun gafara j ceb inriq

here these villageplgen oldsbstplgen wordplinel selves

Dagustandaj atajbur ja

Daghestaninel comeaopsbstpl cop

According to the old p eople of these villages they came here from

i i

Daghestan

The data are thus overall quite reminiscent of data on zibun binding in Japanese

the antecedent of zibun can usually but not always b e describ ed in terms of sub ject

ho o d what I am viewing as astructure prominence Gunji Katada In

any such situation there are two choices either one can adopt a disjunctive theory of

antecedence where antecedence normally ob eys syntactic constraints but certain ex

amples are pragmatically licensed by an alternative mechanism or one can maintain

a single set of binding conditions which necessitates relying on as yet not very devel

op ed theories of pragmatics The ma jority of the literature on binding in Japanese

Chapter Lezgian

has adopted the rst course maintaining that antecedence of zibun is a test for sub

jectho o d while acknowledging that there are certain exceptional p erhaps logophoric

cases However Iida argues strongly for Japanese that a unied account

should b e maintained She argues that use of the reexive is principally determined

by the speakers perspective the vantage p oint from which the sp eaker decides to

view the event which is often but not always that of the sub ject although she sup

plements this with a simple syntactic condition Her condition in the formalism of

HPSG see Pollard and Sag is that zibun cannot ocommand its antecedent

This constraint is basically equivalent to saying that zibun cannot acommand its

antecedent in my framework This is a much weaker constraint than saying that the

antecedent must acommand the anaphor but if applied to Lezgian it would never

theless rule out examples like while p ermitting examples like Further

Iidas notion of sp eakers p ersp ective quite clearly motivates the use of the reexive

in examples like

I am not really in a p osition to decide b etween these two alternatives But on

either theory whatever syntactic constraints that there are on binding are stated at

the level of argument structure The theories dier mainly in whether most examples

are treated purely in terms of astructure prominence with a pragmatic theory for

the remaining cases or whether all examples are treated by the conjunction of a

weak condition in terms of astructure prominence and a strong theory of pragmatic

prominence

Kibrik presents some reexization facts from various other Daghestanian

languages All of these languages clearly deserve further study but for Tabasaran

he writes note that actant reexivization is thus controlled by the highestranked

semantic role in the hierarchy given ab ove and the same seems to b e true of Akhvakh

Such observations supp ort the astructure based account of binding which I have b een

prop osing However if examples like those in should also turn up in other

NakhoDaghestan languages on closer examination then deciding b etween the two

alternatives discussed ab ove could turn out to b e a recurring problem

Chapter Lezgian

Imp eratives

The addressee of an imp erative or a negative imp erative which uses a dierent

prohibitive verb form is the asub ject most commonly the ergative NP of a transitive

verb a the absolutive NP of an intransitive verb b or a dative exp eriencer

asub ject a

Sa zinsinin clenar zugura wa abur a

dat one kindgen constituentpl ndimpv and theyabs

galkurzawaj so juzrin kanikaj car cugu

connectimpfpart conjunctionplgen b elow lineabs drawimpv erg

Find the co ordinate constituents and draw a line b elow the conjunctions

that connect them

sa b Ja Farid

abs comeimpv pt Farid

Farid come

The addressee is commonly deleted as in the ab ove examples but it can app ear

apparently without any sp ecial emphasis

Wuna bagislamisa buba

youerg forgiveimpv father

Forgive me father

In contrast the absolutive argument of a transitive verb cannot b e the addressee

of the imp erative even when this would not b e completely implausible semantically

Hasp elmath

Ja Allah quli p olicijadi jaq

pt Allah quli p oliceerg catchimpv abs

Allahquli let the p olice catch you

Finally the genitive idiom construction again suggests that the constraint is prop erly

stated in terms of argument structure rather than some surface syntactic p osition

b ecause an imp erative can b e formed with the genitive p ossessor as the addressee

Chapter Lezgian

Cun ci qaqan qawal hik ksudajta

gen weabs wegen high ro ofsress how sleepfutpastcnd

rikel xkwas

heartsress returnimpv

Remember how we used to sleep on our high ro of

Controlled adverbial clauses

Lezgian has a series of converbs that can b e used to form adverbial clauses and for a

limited degree of clause chaining Here I will fo cus on the aorist converb Hasp elmath

states that it is used to express chains of actions carried out by the same

sub ject This is true as a rst order generalization a few exceptions are discussed

b elow In particular when the converb is used to express sequential action the a

sub ject of the adverbial clause is gapp ed regardless of whether it is the ergative NP of

a transitive a the absolutive NP of an intransitive b or a dative exp eriencer

a Nacalnikdi sehnediz e qecna caz wiridaz cuxsagul

directorerg absstagedat gooutaoc wedat allsbstdat thanksabs

lahana

sayaor

The director came onto the stage and thanked all of us

ruza qacuna taraz aqaxiz baslamisna b Kamala

ergrieabs takeaoc treedat climbinf b eginaor Kamalerg

Kamal to ok a rie and started climbing up a tree

c Wun am takuna xkwemir ha

youabs datheabs negseeaoc returnprohib pt

Dont come back without having seen him

This is true of the overwhelming number of examples although there are the o dd

exceptions Hasp elmath notes but suggests that it is only p ossible

when the ergative argument is inanimate and therefore not a typical sub ject

D uja qilajqiliz jalawri quna kuzwa

worldabs endinelenddat ameplerg holdaoc burnimpf The world completely engulfed in ames is burning

Chapter Lezgian

lit ames having caught it from end to end

Also in the following sp ecial temp oral construction there is no control Hasp elmath

Am fena wad warz ja

sheabs goaoc ve month cop

She went ve months ago

lit She having gone it has b een ve months

Control of innitival complement clauses

Lezgian has a variety of syntactic forms for complement clauses see Hasp elmath

Ch The three main strategies are innitive masdar and substantivized

participle complements The latter two types of complements are clausal nominaliza

tions Here I will just examine innitival complements

Controllee

Most verbs that select an innitival clause complement control the asub ject of that

complement Again this asub ject can b e the ergative of a transitive verb a

the absolutive of an intransitive verb b or a dative exp eriencer c

a Zawaj a b ejaburciwal exiz zezwacir

Iadel erg that shameabs b earinf canimpfnegpast

I could not b ear that shame

b Didediz seherdiz z kanzawa

motherdat abs towndat goinf wantimpf

Mother wants to go to town

c Nabisataz wicin gul akwaz kanzawa

Nabisatdat dat selfgen husbandabs seeinf wantimpf

Nabisat wants to see her husband

The masdar is the traditional term in Caucasian studies for the verb form used in event nominalizations

Chapter Lezgian

This is consistent with my general theory that the controllee should b e an asub ject

There is further discussion of innitival complements b elow in Section

According to the predictions of GB the lack of tense should prevent an innitive

from assigning nominative case and thus licensing one of its arguments But in

Lezgian an innitive can app ear with all of its sub categorized arguments It is rather

a prop erty of the upstairs verb whether there is control While the verbs ab ove control

the asub ject of the innitive the verb kice is afraid tothat also takes an innitive

complement but only optionally controls the asub ject of the innitive a shows

control bc show full clausal complements

a Selimaz galaz raxaz kicedaj wicin bubadiq

Selimdat fatherpoess with talkinf afraidfutpast abs selfgen

Selim was afraid to talk to his father

i i

b Rusaz gadadiz wic akwaz kicezwa

girldat b oydat selfabs seeinf fearimpf

The girl is afraid that the b oy will see her

c Adaz i zeherlu qiler bagda awaj

hedat these p oisonous headplabs gardeniness b einpart

qusari nez kice xana

birdplerg eatinf afraid b eaor

He was afraid that the birds in the garden might eat these p oisoned heads

Bobaljik fn suggests that the cases of overt ergatives in innitive

complement clauses in Lezgian presented by Murasugi are not a problem

for his theory b ecause the constructions in question are nominalizations not true

innitive clauses While this is p erhaps true of the examples Murasugi presented

the same cannot b e said for the form ending in iz which is called the innitive here

following Hasp elmath These innitive clauses do not have the external syntax of

NPs and app ear only i as complements to certain verbs ii in purp ose clauses and

iii in adverbial clauses similar to the ones in which the Inuit innitive app ears

This distribution is quite similar to that of other verb forms recognized as innitives

Actually Murasugis examples a and c are aorist converbs while b is a sub

stantivized participle none are masdar forms as Bobaljik asserts Of these only b is truly a nominalization but none are what I would want to term an innitive form

Chapter Lezgian

Controller

Most verbs that take innitive complements can b e thought of as sub ject control

verbs since the asub ject of the higher clause is identied with the asub ject of the

complement but note that the controlling asub ject can b e in an oblique case a

dep ending on the sub categorization frame of the higher verb There are also ob ject

control verbs where the controller of the complement is an oblique in the main clause

a C ukwera gadadiw zenginiz e qeciz tazwacir

C ukwererg b oyadel abs battledat gooutinf causeimpfnegpast

C ukwer did not allow the b oy to go to the battle

waz b Qabustana cungur jagiz cirnaj

youdat erg cungur hitinf teachaorpast

Qabustanerg

Qabustan taught you to play the cungur a string instrument

In general the controller needs to b e determined semantically as argued by Sag

and Pollard and Pollard and Sag Consider in particular the Lezgian

examples shown in which again involve idioms with rik heart Here the con

troller of the innitives sub ject is a genitive p ossessor of an oblique sup erelative NP

Cul qacuz ci rikela j alatna a

erg b eltabs takeinf wegen heartsrel falloaor

We forgot to take a b elt along

lit For us to take a b elt fell o from on our heart

p enzer aqaliz alatna b Seliman rikela j

erg window closeinf falloaor Selimgen heartsrel

Selim forgot to close the window

Such a construction would b e most problematic for any syntactic theory of controller

determination However it seems rather less unusual in the context of the semantic

theory of control prop osed by Sag and Pollard and Pollard and Sag

Under such an analysis in this idiomatic construction the genitive mo dier of rik

heart would b e an experiencer just like any other experiencer and given that

the main clause introduces an orientation type control relation then the experi

encer can indeed must b e coindexed with the gapp ed sub ject of the innitive The

Chapter Lezgian

Lezgian construction ab ove would then b e quite parallel to genitive controllers which

Sag and Pollard discuss which only o ccur in NPs in English

a Sandys promise to Tracy to leave the party early caused quite an uproar

b The vow of the sp eakers to ght for ab ortion rights was met with applause

Indeed further supp ort for an essentially semantic theory of controller selection come

from the observation that an English version of the Lezgian idiom

It slipp ed out of Susans mind to ring her p o diatrist

must mean that Susan is supp osed to ring her p o diatrist and cannot mean that she

intended to get her secretary to ring the p o diatrist

Idioms

Idioms in Lezgian favor internal patientive arguments as in other languages Some

examples from Hasp elmath app ear b elow

a caradan sejiniz gil jargi awurla xalqdin

aliensbstsggen thingdat hand long doaoptemp p eoplegen

wilik b ejabur zeda

b efore disgraceful b ecomefut

When one steals someone elses things one disgraces oneself b efore the

p eople

b Za mustulux gajidaz genze gunug xiwe

Ierg go o dnews bringaopsbstsgdat kerchief givemsd neckiness

qunwajdi tir

holdprfpartsbst coppast

I had promised to give a kerchief to the one who brings the go o d news

c A qwaz ja juldasar zaq jab akala

stopimpv pt comradepl Ipoess earabs attachimpv

Stop comrades listen to me

Another common pattern for idioms is where the free slot is the p ossessor of an

absolutive or oblique argument of the verb These idiom patterns seem in accord

with the notion of argument structure b eing prop osed here They show that Lezgian

do es not t with the notion of syntactic ergativity employed in Marantz

Chapter Lezgian

Is there a surface pivot

This section will discuss whether there is a surface pivot in Lezgian and if so what it

is Mel cuk argues that absolutive SO NPs in Lezgian are pivots indeed the

only terms In the rst subsection I present his arguments I am not aware of any

other arguments for the SO NPs having pivot status in Lezgian Mel cuksprop osal

is argued against by Job and more carefully by Hasp elmath and

the next subsection contains corrections and counterarguments to Mel cuks analy

sis based mainly on arguments from Hasp elmath I conclude that the case for an

SO pivot in Lezgian is extremely weak The third subsection shows that several

phenomena that have b een used in other languages for determining pivothoo d are

syntactically neutral in Lezgian that is they do not distinguish any subset of S A

and O The fourth subsection examines what evidence there is for an SA pivot in

Lezgian

The case for an SO pivot Mel cuk

Mel cuk argues that the correct analysis of Lezgian is that all sentences are

intransitive genuine transitive verbs do not exist in Lezgian but rather all verbs are

stative with an optional agentive complement Thus he suggests that the correct

calque for is roughly along the lines of Caused by Ali the dog died

7 0 0

Mel cuk is a revised version of Mel cuk based exclusively on data from Gadziev

8 0

Mel cukis vague on precisely what analysis is to b e given to the ergative NP suggesting that

it is an agent complement or maybe even a circumstantial where by circumstantial Mel cuk

means roughly what I mean by an oblique However it would seem that an analysis as an oblique

could scarcely b e maintained since as Mel cukobserves there is a clear class of pure intransitive

verbs which do not allow an ergative argument even when adding such a role could not b e viewed

as semantically anomalous under Mel cuksanalysis

i a ziv c razva

snowabs meltpres

The snow is melting

b raqini ziv c razva

sunerg snowabs meltpres The sun is melting the snow

Chapter Lezgian

Alidi k ic qena

Alierg dogabs

killaor

Ali killed the dog

For Mel cukthis analysis is an argument that Lezgian has no ergative construction

as he denes the term However here we can concentrate on his claim that the

absolutive NP is always the grammatical sub ject or pivot In this subsection I

will review the arguments Mel cuk makes for this analysis and the problems and

counterarguments that he attempts to nullify Ultimately though I b elieve that his

analysis is unconvincing but I will defer criticism of Mel cuks analysis to the next

subsection

In the commonest sort of one actant sentence the single argument is in the ab

solutive and this argument must b e the grammatical sub ject according to Mel cuks

inductive denition of grammatical sub ject Mel cuk then adduces three arguments

for why the absolutive NP in what are traditionally viewed as transitive sentences

should also b e regarded as the grammatical sub ject

In all sentences in Lezgian an ergative NP can b e omitted without the sen

tence b eing incomplete even in the absence of context whereas deleting an

absolutive NP is p ossible only when the sentence is embedded within a suitable

context

Mel cuk rep orts Gadzievs claim that coreferential omission in co ordination is

only p ossible when clauses of the same transitivity are conjoined Mel cuk

suggested that this in no way supp orts grouping the A and S together in Lezgian

This is true but as Mel cuk observes it also in no way supp orts grouping

together O and S

With predicative complement uses of substantivized participles there is oblig

atory or optional agreement with the absolutive NP regardless of whether it

is the S or the O

Thus Mel cukconcluded that the absolutive NP should b e regarded as the grammat ical sub ject in all sentences

Chapter Lezgian

Mel cuk noted that there are no voice alternations in Lezgian which is at least

consistent with his p osition that all sentences in Lezgian are intransitive He then

pro ceeded to deal with ve ob jections to his analysis to which we might add a sixth

new data from Kibrik that is discussed in a fo otnote Mel cuk

Mel cuksanalysis implies that the ergative case has a heavy semantic load due

to the actions of X despite b eing a direct case but Mel cukb elieves this not to

b e a problem since it seems typical of Lezgian for cases to carry considerable

semantic cargo es suggesting that for example the adelative what Mel cuk

called the ablative can alternate with the ergative to express involuntary or

indirect causation

Lezgian has a number of verbs that only take an ergative case actant If the

ergative in such sentences were regarded as the grammatical sub ject this would

weaken the claim that the ergative in other sentences is not the grammatical

sub ject but Mel cuksuggests that he is happy to regard such sentences as having

no grammatical sub ject claiming that there are other sub jectless sentences in

Lezgian for weather predicates and with certain arbitrary sub jects

Work by Talibov had suggested that Lezgian imp erative forms agree in p erson

with their understo o d or overt sub ject but Mel cuk convincingly argues that

Talibov was actually collapsing three dierent verbal mo o ds which Mel cuk

also Hasp elmath refers to as the imp erative the optative and the hor

tative and then claiming that the resulting amalgam represented an agreement

system While the imp erative is semantically restricted to second p erson sub

jects b oth the optative and the hortative can o ccur with a sub ject of any

p erson although the hortative is somewhat o dd with a second p erson sub ject

The optative and hortative have distinct meanings of desiring something for

someb o dy and of hoping that the addressee agrees that something should take

place

The situation in Lezgian seems to closely mirror that found in Basque see for instance Laka

Many but not all such verbs derive from comp ounds of a noun awun do

Chapter Lezgian

Lezgian has a causative sux and one might b e tempted to argue that such

causative verbs must b e transitive But Mel cuk argues that while axing a

causative to what have traditionally b een viewed as intransitive verbs pro duces

an apparently transitive verb it is noteworthy that adding the sux to

what have traditionally b een viewed as transitive verbs pro duces little or no

dierence in meaning

a ac uqun

to sit down

b ac uqarun

to seat someb o dy

a aldatmisun

to cheat dup e

b aldatmisarun

to cheat dup e

Because of the b ehavior of the causative ax on socalled transitive verbs

Mel cuksuggests that the meaning of the ax is not a causative as in Turkish

say where it pro duces a three actant verb when added to a transitive verb

but rather it means b ecause of some external eorts and that this meaning

comp onent serves to license an ergative NP with verbs which are otherwise

purely intransitive ones where an ergative NP is normally disallowed

0 h

Mel cuknotes data from Kibrik that with the control verbs kan want and k c e

b e afraid of it is the absolutive of an intransitive the ergative of a transitive

and the dative exp eriencer of an exp eriencer sub ject verb that is controlled and

gapp ed in the lower clause Mel cuksuggests that these data do not undermine

his analysis and that he would treat them semantically the gapp ed PRO

Kibriks data is from the Khlut dialect but the same facts hold in other dialects Kan and

k ce corresp ond to kan and kice in standard Lezgian as transliterated by Hasp elmath into roman

letters There is further data relating to the adverbial clause use of the innitive called by Mel cuk

the circumstantial gerund where Kibrik apparently suggests that the main clause actant is always

gapp ed in favor of a coreferential NP o ccurring in the adverbial clause But the opp osite pattern is also clearly p ossible cf Hasp elmath

Chapter Lezgian

controller is the Agent role where he intends Agent as a hyperrole covering

Actors Causers Exp eriencers and Perceivers

Lastly Mel cukdiscusses the p ossible ob jection that he is imp osing an ergatives

are passives analysis on the language but dismisses this as a misunderstanding

His analysis acknowledges sentences like as the basic communicatively

neutral sentences of Lezgian merely suggesting that they resemble standard

passives in transitivity and use of an agent complement

Thus Mel cuk concludes that there is no ergative construction in Lezgian All

sentences are intransitive and the absolutive NP is always the grammatical sub ject

or pivot in pure intransitives in sentences with ergative agents or adelative ab

lative indirect agents and also in sentences with what others might view as dative

exp eriencer sub jects

The case against an SO pivot

Labile verbs and NPomissibility

Hasp elmath argues against Mel cuks claim that the ergative NP is always

omissible suggesting that in general Lezgian has only limited contextuallylicensed

omission of NPs of all cases He rather supp orts the older claim of Gadziev that

there is just a small class of labile verbs that have two valence patterns one with and

one without an ergative argument such as regun which can b e erg grind abs or

abs is ground Hasp elmath suggests three tests that show that certain verbs are

labile in this manner

The negative of labile verbs is ambiguous b etween an agentive sense and a

nonagentive sense

Indijada kaler reqizwac

Indiainess cowplabs killdieimpfneg

In India cows dont die intransitive frame of labile verb

In India cows are not killed free omission of the sub ject

Chapter Lezgian

This ambiguity which o ccurs only with labile verbs shows that the caused by

agent meaning is contained in one of the frames of the verb and not simply

provided by the ergative case sux as Mel cukhad suggested

As discussed ab ove the imp erative addressee is the absolutive argument of

intransitive verbs and the ergative argument of transitive verbs The absolutive

argument of transitive verbs cannot b e the imp erative addressee even when

the ergative NP is omitted However with labile verbs either the ergative NP

a or the absolutive NP in the verbs intransitive use b can b e the

imp erative addressee even when this is not very plausible on semantic grounds

a Ja G ulmehamed gete xux

pt G ulmehamed p ot breakimpv

G ulmehamedbreak the p ot

b Ja gete xux

pt p ot breakimpv

Pot break

Mel cuknoted the involuntary agent construction where an involuntary agent

can b e marked in the adelative a and argued that an advantage of his

analysis is that if the regular construction in b is analyzed as an intransitive

verb with an agentive complement then these two structures are quite parallel

whereas on the standard analysis the agent in one is an oblique but it is

the sub ject in the other On Mel cuks analysis the only dierence is the case

marking of the NP with the ergative case meaning caused by and the adelative

meaning indirectlyaccidentally caused by

a Zamiradiwa j gete xana

Zamiraadel p otabs breakaor

Zamira broke the p ot accidentallyinvoluntarily

b Zamiradi gete xana

Zamiraerg p otabs breakaor Zamira broke the p ot

Chapter Lezgian

However Hasp elmath suggests that the right generalization is the one presented

by Gadzievthat an adelative involuntary agent is simply an additional oblique

added to a clause The parallel ab ove is mistaken and can only b e pro duced with

the true labile verbs An adelative NP cannot normally replace the ergative NP

of transitive verbs a but rather an adelative NP can b e added to clauses

with intransitive verbs b and even clauses with transitive verbs if the

ergative NP can b e interpreted nonagentively c Thus the adelative NP

is not parallel to the ergative NP

a Taibatawa j rak aq a jna

Taibatadel do orabs op enaor

Taibat accidentally op ened the do or

b Maisadiwa j c ukwer qurana

Maisaadel owerplabs wiltaor

Maisa involuntarily allowed the owers to wilt

c Didediwa j p erdedi ca j quna

motheradel curtainerg reabs catchaor

Mother accidentally caused the curtain to catch re

For other verbs Mel cuk noted cases where the ergative NP was omitted and

given a generic human interpretation one and tried to combine these cases with

labile verb cases to suggest that the ergative NP was always omissible But as well

as reestablishing the class of labile verbs Hasp elmath notes p that it is not

only ergative NPs that can b e omitted with a resulting generic human interpretation

but all asub jects That this is equally p ossible for absolutives as well as ergatives is

clearly seen in an example like where in turn an ergative and an absolutive are

admitted and b oth receive a generic interpretation

caradan sejiniz gil jargi awurla

erg aliensbstsggen thingdat hand long doaoptemp abs

xalqdin wilik b ejabur zeda

p eoplegen b efore disgraceful b ecomefut

When one steals someone elses things one disgraces oneself b efore the p eople

Chapter Lezgian

This discussion undermines Mel cuks rst argument and also his rst counter

argument listed ab ove Mel cuks second argument regarding co ordination is really

arguing from absence of evidence However there has apparently b een diachronic

change in co ordination p ossibilities or at any rate the mo dern Lezgian usage that

Hasp elmath observes diers from what Mel cukrep orts from Gadziev Hasp elmaths

data if anything supp ort an SA pivot as I will discuss later I now turn to Mel cuks

nal argument for an SO pivot

Agreement of substantivized participles

Mel cuk presents a paradigm of data designed to show that a predi

cate substantivized participle may agree with an absolutive case NP and must if the

absolutive NP is animate whereas agreement with an ergative NP is imp ossible Part

of this paradigm is shown in

a Zun qsan kela jdi ja

Iabs well learnpartnomlzsgabs b epres

I am a welleducated p erson

b C un qsan kela jburdi ja

weabs well learnpartnomlzplabs b epres

We are welleducated p eople

c C na sa ktab kela jdibur

weerg one b o okabs readpartnomlzsgabsnomlzplabs

b epres

We have read a b o ok

d C na gzaf ktabar kela jdibur ja

weerg one b o okabs readpartnomlzsgabsnomlzplabs b epres

We have read many b o oks

However this data is at o dds with what is rep orted by Hasp elmath

He suggests that there are two p ossibilities In one b oth the sub ject and the

substantivized participle complement are in the nominative and the sub ject and the

Chapter Lezgian

substantivized participle complement agree in number as in other nominal copula

constructions recall

Cun gzaf ktabar kela jbur ja

weabs many b o okplabs readaopsbstpl cop

We are ones who read many b o oks

Here the structure is presumably a copular construction b etween two NPs as in

Each of the two NPs joined by the copula is in the absolutive and additionally ktabar

is absolutive as this is the exp ected case of the theme argument of kelun read

gzaf ktabar kela jbur ja Cun

many b o okplabs readaopsbstpl cop weabs

np np

We are ones who read many b o oks

The use of the copula with substantivized participial clauses is indicative of the

fact that these clauses have the external distribution of NPs in Lezgian Among other

tests this is conrmed by their app earance as the complement of prep ositions and

in the p ossessor slot re p ostp ositions Like several other p ostp ositions in Lezgian

quluq b ehind after takes a genitive argument for spatial uses and a sup erela

tive argument for temp oral uses As Hasp elmath notes if substantivized

participial clauses are NPs then there is no dierence syntactically or semantically

b etween cases of a p ostp osition with a lexical noun complement ab and cases

of a p ostp osition with a substantivized clausal complement c whereas otherwise

a disjunctive account would b e required

a dawedila j quluq

after warsrel

after the war

am galatna b Sa timil kwalaxdila j quluq

after sheabs tireaor one little worksrel

After a little work she b ecame tired

c Fu t urdalaj quluq daxdi Nadiraw daftar

fo o d eataopsbstsrel after daderg Nadiradess noteb o ok

giz tuna

bringinf causeaor

After they had eaten dad made Nadir bring a noteb o ok

Chapter Lezgian

Example shows a substantivized participle app earing as a genitive case p osses

sors of another noun itself a lexicalized masdar form

ktabdin qismet kelzawajdan qatunila j

b o okgen destinyabs readimpfptpsbstsggen considermsdsrel

aslu ja

dep ending cop

A b o oks fate dep ends on the readers considerations

These tests seem sucient to conclude that substantivized clauses have the external

b ehavior of NPs

The other initially o dd structure is shown in

Cna gzaf ktabar kela jdi ja

weerg many b o okplabs readaopsbstsg cop

We have read many b o oks

Here the surface arguments are those of the base verb kelun takes a frame of erg

abs in alaqun b e able to takes a frame of srel abs and the substantivized

participle never agrees b eing in the default singular form

Macala j wuc xajitani alaqzawajdi ja

Macsrel whatabs indef b eableimpfpartsbst cop

Mac can do anything

The details of this second construction which Hasp elmath terms the verbal predicate

substantivized participle are slightly obscure it often but not always seems to b e

used for contrast or fo cus I would like to suggest that there is nothing verbal here

but rather the structure of this construction is

Cna gzaf ktabar kela jdi ja

weerg many b o okplabs readaopsbstsg cop

We have read many b o oks

Everything but the copula is a substantivized clause which the copula predicates of

a null expletive The form of the substantivized participle is always singular as the

null expletive is singular as are the English expletives it and there There seems to

b e indep endent evidence for such copula constructions with null expletive sub jects

compare the examples in Hasp elmath

Chapter Lezgian

a Tama serin tir

forestiness co ol coppast

It was co ol in the forest

b Daglara isatdani hak ja

mountainpliness noweven thus cop

In the mountains it is like that even to day

At any rate the examples in ab can b e analyzed as the rst type of predica

tive substantivized participle and the forms in cd with a singular form of the

participle can b e analyzed as the second type of predicate substantivized participle

The crucial case for Mel cukturns out to b e the alleged p ossibility of plural agreement

on d However Hasp elmath email Apr rep orts that his consul

tants nd this sentence ungrammatical Thus regardless of the structure prop osed

for these constructions there is not a prop erty of agreement here that groups S and O

against A

Other concerns

Several of Mel cuks other arguments are also less than fully convincing The lim

ited pro ductivity of the causative morpheme esp ecially its restricted app earance on

certain transitive verbs limits the strength of that argument Mel cuks

suggestion that sub jectless sentences denoting environmental physiological and psy

chological states are quite common in Lezgian stands in contrast to Hasp elmaths

observation that there are no verbs that take no arguments weather verbs

b eing expressed as marf qwazwa rain falls etc Most of the examples Mel cukgives

are of adjectives in copular constructions Copular constructions can have only a null

expletive sub ject as discussed ab ove but Mel cuks argument dep ends cru

cially on saying that this is also the case for verbs which seems to b e incorrect On the

other hand in this dissertation I essentially accept his argument that evidence from

Actually a could ambiguously also b e of the second type

Further Hasp elmath notes that it is unclear where such examples come from since they do not

app ear in Gadzievsworks supp osedly the only source Mel cukused p

Chapter Lezgian

controllee selection is not indicative of a surface pivot but rather that it indicates

prominence at argument structure

Sometimes pivotsensitive phenomena that are neutral in Lezgian

This section examines relative clauses scop esp ecicity and quantier oating In

some languages these phenomena seem to b e sensitive to a surface pivot but none of

them seem to distinguish a pivot relation in Lezgian

Relative clauses

Lezgian has b oth participial relative clauses and a correlative construction but I

will only examine the former here From a morphological p oint of view Lezgian

participial relatives resemble those of Eskimo but the similarities do not run very

deep as Lezgian allows any role to b e gapp ed in the relative and allows longdistance

relativization For example a and b show relativization of the ergative and

the absolutive resp ectively from an aorist participle Datives oblique arguments and

lo cative and temp oral obliques can also b e relativized on

a Q fej jac zanawurri reqe kukwarna

goawayaop bullabs wolfplerg wayiness tearaor

The bull which had gone away was killed by wolves on the way

b Pacahdin xazina cunuxaj ugrijar cun ja

kinggen treasuryabs stealaop thiefplabs weabs cop

We are the thieves who stole the kings treasury

Relativization may b e long distance and the relativized element can b e in either a

nite or nonnite sub ordinate clause In this case only the top level verb app ears in

the participle form and the embedded verbs are in their normal form for example

the nite aorist in

Musadi atana lahaj muhman ci xalu tir

Musaerg comeaor sayaop guestabs wegen uncle coppast

The guest that Musa said had arrived was our uncle

Chapter Lezgian

Where disambiguation is necessary the reexive may b e used as a resumptive

pronoun

cun wicikaj raxazwaj kas

weabs selfsbel talkimpfpart man

the man were talking ab out

It might b e thought that the dierence b etween Eskimo and Lezgian participial

relatives results from a category dierence Eskimo participles are nouns whereas

Lezgian ones are not The suggestion would b e that the pro cess of nominalization

xes a certain role of the verb to b e the referential index of the noun However this

suggestion cannot b e maintained

Just like adjectives Lezgian participles can b e turned into nouns by substantiviza

tion

a qacu green

b qacudi green one

a kelaj aorist participle of read

b kelajdi who ever read the one who read

An adjective participle or a genitive NP must b e substantivized in this way to serve

as the head of a noun phrase However even this pro cess of substantivization do es

not select a certain argument role of the predicate as the referential index of the noun

Thus a future participle is the actor in a and the undergo er in b

a Gzaf qin qadajda gzaf tabni ijida

much oathabs takefutpartsbstsgerg much lieabsalso dofut

He who swears a lot also lies a lot

b Nedajdini alukdajdi b es

eatfutpartsbstsgand putonfutpartsbstsg enough

zezwacir

b eimpfnegpast

There was not enough to eat and to put on

Chapter Lezgian

Thus a limitation on which roles can b e relativized on as in Inuit cannot b e at

tributed to the relativization strategy b eing to use participial nominalizations since

Lezgian participial nominalizations show no constraints on which role is relativized

on Lezgian is thus quite problematic for an account such as Johns

Semantic scop e and DenitenessSp ecicity

A noun phrase in either the ergative or absolutive and I presume other cases can

b e either sp ecic or nonsp ecic This is illustrated for A S and O in

galaz a Ci x ure sa itimdi dawedila j g uguniz wiciq

selfpoess with wegen villageiness one manerg warsrel after

nemserin disehli xkanaj

Germanplgen womanabs bringbackaorpast

A man in our village had brought a German wife with him after the

war

b Jargi Alidi adawa j pul qacuna

tal l Alierg headel moneyabs takeaor

The tal l Ali to ok the money from him

a Sa x uraj masa x uruz swas tuxun patal ilcijar

one village other villagedat brideabs bringmsd for matchmakerpl

qweda

comefut

Matchmakers come to bring a bride from one village to another village

b Bazanaxar ulkwe turfandikaj x uz frontdiz

brotherinlawpl countryabs stormsbel preserveinf frontdat

fena

goaor

The brothersinlaw went to the front to protect the country against the

storm

a Inal aburun wilik cetin carcar a qatnawa

here theygen infront dicult waterfal labs app earprf

Here in front of them a dicult waterfal l has app eared

Chapter Lezgian

b Peq er wacun qerexdiw acuqna

crowplabs rivergen bankadess sitaor

The crows alighted on the bank of the river

Given the lack of passiveantipassive type op erations in Lezgian this is hardly surpris

ing There is not an Englishlike system of articles but indeniteness can b e explicitly

marked by the addition of sa one however this is not required cf a a

Hasp elmath This use of sa as an indenite marker is quite grammati

cized sa is also used as an indenite pronoun as part of a general demonstrative

interrogativeindenite pronoun system including such forms as sana somewhere

and sanra in some places The presence of sa can b e treated as a diagnostic for an

indenite reading

It is denitely not the case that the absolutive must take scop e over VPlevel

op erators as Bittner rep orts for Greenlandic Examples of the sort shown

in with absolutives inside the scop e of negation are very common

a Za ci muqwaraburukaj sadazni sa xabarni

Ierg wegen relativesbstplsbel onedateven one newseven

ganacir

giveaornegpast

I didnt give any information to any of my relatives

x y informationx relativey giveI x y

b Zuwa lahana kanda cara awac

selferg sayaoc mustfut choice b eneg

I have to say it myself there is no choice

x choicex

Not unexp ectedly other NPs can also take scop e over an absolutive NP In a

an ergative has scop e over an absolutive in the embedded clause b shows a

This is not a problem for Bittner as she allows the option of transparent ergative languages

where the absolutive do es not raise at Sstructure indeed she suggests that they are more common

than raising ergative languages See for instance the discussion of Warlpiri in Bittner and Hale

forthcoming a

The scopal judgements here are based on my interpretation of the translations of sentences in

Hasp elmath Hasp elmath do es not address questions of scop e himself Nevertheless since

these examples and most of the examples in this section are from texts which allows contextualized translation I consider them quite reliable evidence

Chapter Lezgian

sub elative taking scop e over an absolutive

a Za harda qwe tar akurun teklifzawa

Ierg everysbstsgerg two treeabs plantmsd prop oseimpf

I prop ose that everyone plant two trees

b Hardakaj kwalinjiqan ijesi xanwa

everysbstsgsbel housegendaygen ownerabs b ecomeprf

Each one has b ecome the owner of a house and family

Thus in Lezgian one cannot supp ort a high p osition for the absolutive at Sstructure

on the basis of Sstructure b eing the default LF as Bittner suggests for Inuit

Indeed if anything the examples in Hasp elmath suggests that the absolutive

prefers narrow scop e

Quantier oat

There are two forms of quantier oat one in existential sentences and another

restricted to universal quantiers but they are not restricted to certain grammatical

relations in a way that would indicate a pivot

The evidence for a weak SA pivot

Co ordination

There are two forms of clausal co ordinate conjunction in Lezgian the native particle

and the b orrowed ni which suxes to the rst word of the last conjunct clause

particle wa Joining clauses and other constituents with the conjunction wa and is

I have replaced the apparent typo ttar with tar in example a Note also that numerals

co o ccur with singular nouns in Lezgian

Hasp elmath has the rst constituent of the last conjunct clause I am unsure if

or how he intended constituent to dier in interpretation from word but it is certainly not the

case that ni app ears after the end of the rst top level constituent for example it app ears after

the rst word of a relative clause in i

i Zi pab azarlu ja a jalrizni kiligdaj kas awac

Igen wife sick cop childpldatand lo okfutpart p erson b eneg My wife is sick and there is no one to lo ok after the children

Chapter Lezgian

an innovation in Lezgian wa was b orrowed apparently in this century from Turkic

and still rarely o ccurs in the sp oken language where sequential event chaining is

regularly achieved by using aorist converbs Hasp elmath

Hasp elmath also Mel cuk cites Gadzievs rep orts from the s that

omission of a shared actor in co ordination is only p ossible if the clauses are all

transitive or all intransitive However Gadzievnoted that this rule was not always

followed in the s and Hasp elmath rep orts that the rule certainly no longer ap

plies Today omission under co ordination is p ossible with sub jects The shared NP

stands in the case required by the rst conjunct a shows co ordination of an abs

S NP and an erg A NP b shows the reverse Kiparsky lecture notes

suggests that the fact that this sentence is unambiguous shows that there cannot

b e coreferential omission of an absolutive nonasub ject c shows coreferential

deletion b etween a dative exp eriencer and an ergative argument its unclear whether

it should b e termed an A or an S minetun ask b eg takes the frame erg dat

Regarding c Hasp elmath maintains that the main verb of the second conjunc

tion takes no sub ject but only a clausal argument The verb baslamisun b egin takes complements

where the complement sub ject is overtly expressed Thus a literal translation of i would b e It

b egan that the children were playing That this is the correct syntactic structure can b e seen from

the fact that the casemarking of the NP that expresses the b eginner varies according to the valence

pattern of the complement verb cf the Ergative b eginner in ii However the b ehavior of this

verb under co ordination as shown in the text and also Hasp elmath a makes it

seem unlikely that Hasp elmaths analysis is correct Rather it seems more likely that baslamisun

must b e analyzed as a raising verb or some form of complex predicate forming verb that preserves

the case of its complements sub ject This b ehavior is known from phenomena like quirky case in

Icelandic Andrews

i Ajalar qugwaz baslamisna

childplabs playinf b eginaor

The children b egan to play

ii Nabisata wicin ktab keliz baslamisna

Nabisaterg selfgen b o ok readinf startaor

Nabisat started to read her b o ok

If Hasp elmaths analysis were correct the account of co ordination b eing discussed in the text would

b e greatly complicated Further evidence for a raising analysis comes from examples bc in

Hasp elmath where the meaning suggests that the b eginning event only scop es over the innitive verb and not the aorist converb clauses as Hasp elmaths bracketing implies

Chapter Lezgian

a Gada xtana wa ktab qacuna

b oyabs returnaor and b o okabs takeaor

The b oy returned and to ok the b o ok

b Bubadi Ahmed gatana wa q fena

fathererg Ahmedabs b eataor and goawayaor

Father b eat Ahmed and father went away

c Culaw n ukrez kice xana wa minetiz baslamisna

black birddat afraid b ecomeaor and b eginf b eginaor

The black bird b ecame afraid and b egan to b eg

While the older pattern allowed conjunction only in cases of case matching the

mo dern pattern app ears to supp ort the recognition of a sub ject relation Given

that coreferential omission in co ordination has generally b een held to work o surface

pivots Dixon Kro eger rather than o semantic classes this suggests that

ergative A NPs absolutive S NPs and dative exp eriencers act as the grammatical

sub ject of sentences ie that the language is syntactically accusative However

there are some reasons to b e cautious Hasp elmath cautions that b ecause

the co ordination construction is a nonnative device used mainly in translations

and b o okish Lezgian it is clear that syntactic arguments based on it cannot b e

very strong In particular it needs to b e established that this is a syntactic clause

reduction rule and not just contextually licensed prodrop compare the distinction

motivated by Kro eger and my own discussions of Tagalog and Inuit

co ordination If all examples like b are unambiguous this suggests a syntactic

constraint but on the other hand Hasp elmath email Apr emphasizes that

coreferential deletion in Lezgian shows little syntacticization and seem to b e largely

pragmatically controlled A little evidence for this p osition comes from where

b oth the A and O arguments of the second clause are gapp ed

K une k u rik alaj igitdin sikil

youallerg youallgen heartabs b eonpart herogen pictureabs

cugu wa ci zurnaldiz ra qura

drawimpv and wegen journaldat sendimpv

Draw a picture of your favorite hero and send it to our journal

Chapter Lezgian

In summary co ordination is suggestive of an SA pivot in Lezgian but further re

search is required

Evidence from complementation for an SA pivot

Hasp elmath suggests one other sub jectho o d test which picks out the class

of ergative A NPs absolutive S NPs and dative exp eriencers For this subsection let

me call those NPs sub jects assuming that the language is accusative This test is

switchreference with kan want complements Earlier I showed that kan can take

a controlled innitival complement with control of the asub ject bc Another

example is in

Nabisataz ktab keliz kanzawa

Nabisatdat erg b o okabs readinf wantimpf

Nabisat wants to read a b o ok

Hasp elmath suggests that the controllee in this innitive construction must b e a

sub ject Not only can there not b e control of the absolutive NP of a transitive or

dative exp eriencer verb ab but there cannot even b e control of other asub jects

c Rather in such cases and in cases when there is no coreference whatso ever

the complement must b e expressed using an aorist converb form de

a Musadiz didedi sehierdiz ra quriz kanzawa

Musadat mothererg towndat sendinf wantimpf

Musa wants to b e sent to town by his mother

b Nabisataz wic guluz akwaz kanzawa

Nabisatdat selfabs husbanddat seeinf wantimpf

Nabisat wants to b e seen by her husband

c Nabisataz wilikan gul wicin rikela j alatiz

Nabisatdat former husbandabs selfgen heartsrel falloinf

kanzawa

wantimpf

Nabisat wants to forget her former husband

Incidentally a similar switchreference pattern in complementation also o ccurs in Basque with

the roughly corresp onding verb nahi desire Ortiz de Urbina

Chapter Lezgian

d Nabisataz wilikan gul wicin rikela j alatun

Nabisatdat former husbandabs selfgen heartsrel falloaoc

kanzawa

wantimpf

Nabisat wants to forget her former husband

lit Nabisat wants her former husband to disapp ear from her heart

e Nabisataz rusa ktab kelna kanzawa

Nabisatdat girlerg b o okabs readaoc wantimpf

Nabisat wants her daughter to read a b o ok

However there are two p ossible ob jections to this as a grammatical sub jectho o d test

Firstly c is the only example that Hasp elmath gives to argue that there cannot

b e control of other asub jects in innitival complements This example seems to me

somewhat unconvincing since rikelaj alatun means something like to accidentally

forget and is thus uncontrolled and nonvolitional It may just b e that an innitival

complement of kan needs to b e an event over which the wanter has some control

Secondly even if the constraint as stated is correct these facts could b e captured

by simply saying that the controllee must b e an asub ject and a term This is more

consistent with my general theoretical p osition that the controllee is principally de

termined by asub jectho o d The result would b e motivation for the termnonterm

distinction in Lezgian but no motivation for a grammatical relation of sub ject

Conclusion

The choices are to say that Lezgian has no pivot as Foley and Van Valin

argue for certain languages to say that there is an absolutive pivot evidenced solely

by case marking or to say that the language has an accusative pivot based on the

arguments presented ab ove even though they are more suggestive than conclusive

At the moment I think it is most reasonable to suggest that Lezgian traditionally had

no pivot just a termnonterm distinction but for sp eakers who use coreferential

omission in co ordination as outlined ab ove a weak SA pivot has developed through

contact with Russian and other accusative languages

Chapter Lezgian

Regardless of the correct analysis of Lezgian this latter situation where a lan

guage has morphological ergativity but an accusative SA pivot is one that must b e

countenanced in any crosslinguistic study While the Inverse Grammatical Relations

analysis that I have illustrated for Inuit seems appropriate for certain language fami

lies there are other languages with ergative morphology for which there is no reason

to establish an SO pivot cf Ortiz de Urbinas arguments for the accusativity

of Basque even though some of them are based on language features that I would

deem sensitive to argument structure or Walmatjari which was briey mentioned in

Chapter

While such a state of aairs is problematic for certain frameworks there is

no diculty in simultaneously providing for ergative case marking and accusative

grammatical relations in a framework such as LFG or HPSG It can simply b e

sp ecied that the sort of transitiveverb takes an erg abs case marking pattern while

the sort of intransitiveverb takes an abs pattern and the sort of experiencerverb

takes a dat abs pattern In most cases as p erhaps in Lezgian such phenomena arise

for essentially diachronic reasons and we should not exp ect a synchronic structural

account of such case marking patterns This issue is discussed further in Chapter

For instance the minimalist framework of Chomsky app ears to p ostulate a very close

connection b etween syntactic p osition and case assignment although see Bobaljik for one

attempt at how to capture ergative case marking at the same time as an accusative notion of sub ject

CHAPTER

Concluding discussion

This chapter briey considers the treatment of Dyirbal within the framework I have

b een developing and then ends with some general concluding remarks

Dyirbal

So far I have said rather little ab out Dyirbal which has b een esp ecially in the genera

tive literature the paradigm case of a syntactically ergative language How well do es

it and related Queensland languages t in with the class of socalled syntactically

ergative languages that I have b een exploring

Evidence compatible with an SO pivot

The most commonly cited evidence for the syntactic ergativity of Dyirbal is co ordi

nation clause chaining A series of clauses can b e co ordinated only if each clause

shares the SO NP The coreferent NP in the second and subsequent clauses can

but need not b e gapp ed A couple of examples are given in exhaustive exem

plication app ears in Dixon This data is what I would exp ect if the

I will conne my remarks to Dyirbal but sometimes more limited syntactically ergative fea

tures of a kind compatible with my prop osal are also found in a number of other languages of the

Queensland area including Yidin Kalkatungu Warrgamay Bandjalang north New South Wales

and p erhaps Warungu See Dixon

Like many languages Dyirbal has no overt word for and but co ordination is shown by proso dy

as well as through the freedom of gapping normally required absolutive arguments from clauses

Chapter Concluding discussion

SO NPs are the surface pivot

a bayi yara yanu bangun yibingu buran

iabsth manabs gonfut i iergth womanerg seenfut

The man went and the woman saw him

b bayi yara bangul gubingu mundan bayi yara

iabsth manabs iergth gubierg bringnfut iabsth manabs

bangun jugumbiru balgan

i iergth womanerg hitnfut

The man was brought here by the gubi shaman and he was hit by the

woman

Similarly an SO pivot is used for relativization The role of the head noun in

the relative clause must b e S or O

a bayi yara miyandanu yanu

iabsth manabs laughrel gonfut

The man who was laughing went

b balan yibi bangul yarangu miyandanuru buran

i iabsth womanabs iergth manerg laughrelerg seenfut

The man who was laughing saw the woman

bangu yugungu gunbanuru bangul yarangu nayguna

ivergth treeerg cutrelerg iergth manerg Iacc

birriju balgan

almostemph hitnfut

The tree which the man had cut nearly fell on me

One cannot directly relativize on an NP that is not the pivot of the relative clause

In order for a relative clause to b e formed relativizing on the NP in A function

antipassivization must b e used so that it b ecomes a surface pivot as in

bayi yara jilwalnanu bagun gudagu yanu

iabsth manabs kickantiprel i idatth dogdat gonfut

The man who kicked the dog went

3 Although a passive translation is used b oth clauses are in the unmarked active voice

Chapter Concluding discussion

There are some other more minor phenomena in Dyirbal that are sensitive to the

pivot For example the particle warra indicates that an event concerned the wrong

p erson or thing and Dixon states that it always qualies a pivot argument

a bala yugu warra nudin

ivabsth tree wrong cutnfut

They cut down the wrong tree

b bayi warra miyandanyu

iabsth wrong laughnfut

Hes the wrong p erson to laugh

Such phenomena are not predicted from anything I have said but within my analysis

seem quite consistent with an SO pivot

An apparent problem

However an apparent problem for my theory is the existence of sentences such as

those in

a bayi yara walngarranyu nabaygu

iabsth manabs wantnfut bathepurp

The man wanted to bathe

b bayi yara walngarranyu buralnaygu bagun

iabsth manabs wantnfut seeantippurp i idatth

yibigu

womandat

The man wanted to see the woman

c bayi yara walngarranyu bangun yibingu burali

iabsth manabs wantnfut i iergth womanerg seepurp

The man wanted the woman to see him

a yabu numangu gigan banagaygu

motherabs fathererg tellnfut returnpurp

Father told mother to return

Chapter Concluding discussion

b nayguna bangul gubingu gigan bagul wanalgu

Iacc iergth gubierg tellnfut idatth b o omerangdat

banuljingu yaranunyjingu wugalnaygu

igenthogendat mangenogendat giveantippurp

The gubi told me to give the mans b o omerangs to him

c na ja bayi yara gigan gubingu mawali

Inom iabsth manabs tellnfut gubierg examinepurp

I told the man to b e examined by the gubi do ctor

Dyirbal constructions such as these have b een analyzed in the generative literature

as cases of complementation with equiNP deletion or a controlled PRO Ander

son Levin BokBennema Under such an analysis

walngarranyu want in is similar to logical sub ject control verbs from Eu

rop ean languages while gigan tell in is similar to logical ob ject control

verbs Note now the b ehavior of the supp osed complements they display an appar

ently ergative pattern in a and a the pivot S is controlled in b and

b antipassivization has o ccurred so that the logical sub ject has b ecome an S and

again the pivot is controlled Examples c and c show that if the purp osive

verb form is not antipassivized then it is the pivot O that is controlled not the A

Dixon It is imp ossible for this construction to o ccur with control of the

A p osition

This is an apparent exception to my general theory where control should always

work o argument structure which results universally in an accusative pattern of

control regardless of a languages choice of pivot It app ears that Dyirbal deviates

by b eing to o ergative in that even control works on an ergative basis

Now it is p ossible that I may have to accept that the controllee is sometimes the

pivot As noted in Chapter there are some minor patterns of control in Tagalog

where this seems to b e the case and I left unresolved similar facts in Toba Batak

However I b elieve that I can maintain the strongest version of my theory in the

light of these Dyirbal facts b ecause here we are actually not dealing with controlled complements at all

Chapter Concluding discussion

This dierent b ehavior of what are semantically complements at least in an En

glish sp eakers mind is probably a consequence of some more general features of

Australian languages In many Australian languages clause embedding is avoided

and instead sub ordinate clauses are adjoined Hale Dixon notes for

example that there are no clausal nominalizations in Dyirbal He also states the view

implicit in Dixon that Dyirbal has no complement clause constructions

The grammar of Dixon do es not distinguish the ab ove complement clause

use Rather it provides a unied treatment of the ending liygu the choice of

form dep ending on the conjugation class of the verb This ending is referred to by

Dixon and glossed by me as the purp osive inection The ab ove use of it app ears

not particularly common It is used much more frequently to explain the goals of a

previous action for example in

anyja bayi yanu yurigu barrgangu jurrganaygu

ptcl iabsth gonfut kangaroodat wallabydat sp earantippurp

Then he went out to sp ear kangaroos and wallabies

In all uses of this construction there must b e coreference b etween the SO of the

main clause and the SO of the liygu clause

But I think it would also b e completely mistaken to regard all liygu clauses as

equivalent to the innitives of purp ose of various Europ ean languages reinterpreting

the previous examples as Father sp oke to mother in order that she would return

etc A clear demonstration that something dierent is going on app ears in the

following text example In the text the rst man had hidden sharp pieces of quartz

where two women used to sit and then this sentence follows

banum balagarra nurbanyu nyinayarranyu gunbali bangu

thabl twopeopleabs returnnfut sitstartnfut cutpurp ivergth

The two women returned from there and started to sit down only for them

the sharp quartz pieces to cut the women

Clearly in this example the li purp form is in no sense serving as head of a purp ose

clause Rather I think that this form should b e analyzed more as a clause chaining

4 I intend adjoined here in a nontechnical sense

Chapter Concluding discussion

construction Dixon refers to this ending as o ccurring on implicated verb

complexes and this seems much nearer the mark The socalled purp osive ending

can b e employed on any clause providing it refers back to a previous event that set

the stage for it Dixon

The action referred to by an implicated VC is only p ossible by virtue of

an event referred to by a previous sentence of the discourse having taken

place either the event has b een p erformed as a necessary preliminary

to the intended implicated action or the implicated action is a natural

but p erhaps unplanned consequence of the event

It is in fact p ossible for the first sentence in a topic chain to have an

implicated VC although this happ ens relatively infrequently In such a

case the necessary implicating earlier event has taken place but has not

b een referred to in the discourse

Among other evidence the sheer textual frequency of this verb form suggests that it

is not appropriate to compare it with purp ose clauses from Europ ean languages For

example in Text XV Dixon sub ordinate clauses with the purp osive

inection o ccur in of the pivot chains lo osely sentences sometimes

more than once In another of the sentences the main verb is in the purp osive

form

Secondly it is not the case that the SO NP of the purp ose clause is necessarily

gapp ed although it usually is Dixon All that is required is coreference

a is a dramatic example of this from a creation myth where the b oil is the child

b is a plainer example where a noun marker is rep eated in the purp osive clause

Pace Levin who describ es them as structures of obligatory control

Dyirbal noun markers indicate noun class and regularly accompany nouns as in many of the

examples ab ove they can also indicate demonstrativelike contrasts but the there series which

app ears in all the examples in this section can b e used without any demonstrative sense They can also b e used by themselves functioning like a pronoun as here

Chapter Concluding discussion

a anyja bangul burrubay julman bayi nyalnga

ptcl iergth b oilabs squeezenfut iabsth childabs

mayiyarraygu

comeoutb eginpurp

He squeezed the b oil with the result that a male child came out

b bayi yara walmanyu bayi bagun jugumbilgu

iabsth manabs getupnfut iabsth i idatth womandat

balgalnaygu

hitantippurp

The man got up to hit the woman

Thirdly if we take a verb of the promise class of Sag and Pollard then on

a semantically based theory of controller selection the controller or determinant of

the complement event should b e the committor which is the asub ject of the main

clause for example in English where it is Sandy who will leave

Sandy promised Tracy to leave the party early

However if a purp osive clause is attached to a clause with such a verb in Dyirbal the

necessary coreference relationship is still b etween the SO NPs of b oth clauses and

not with the logical sub ject committor A sentence like a where the agent

is not expressed in the complement clause could b e used to convey The woman

threatened to sp ear the man where the A of the main sentence is interpreted as

coreferent with the A of the purp ose clause as well as the O NPs of the two clauses

b eing coreferent but this is not a control relationship another actor can b e explicitly

mentioned as in b

a bayi yara bangun yibingu ya jijarran bagali

iabsth manabs i iergth womanerg threatennfut sp earpurp

The woman threatened the man that he would b e sp eared

b bayi yara bangun yibingu ya jijarran bangul

iabsth manabs i iergth womanerg threatennfut iergth

gubingu bagali

gubierg sp earpurp

The woman threatened the man that the gubi would sp ear him

Chapter Concluding discussion

Thus I would like to conclude that we are dealing with a form of clause serialization

construction and that is the reason why it is sensitive to the pivot as in co ordination

rather than sensitive to astructure relations

Dyirbal do es have some other phenomena that reveal the nature of its argument

structure and suggest that it uses the same notions of astructure prominence as all

other languages The addressee of an imp erative is the maximal asub ject including

in cases of antipassivization

a ninda bayi yara balga

younom iabsth manabs hitimpv

Hit the man

b ninda bagul yaragu balgalna

younom idatth mandat hitantipimpv

Hit the man

Incorp orationidioms work as in other languages as was discussed in Section

Unfortunately there is little or no evidence from binding since Dyirbal uses a de

transitivizing reexive verb form

Hence I would like to conclude that Dyirbal also ts in with the class of syntac

tically ergative languages that I have examined so the name of the class was well

chosen Dyirbal has b oth a clear SO pivot at the level of surface relations and

evidence for an argument structure that follows universal lines

This has b een but an outline of a p ossible treatment of Dyirbal clause structure A complete

treatment do es have to recognize that there is a tighter link b etween a purp ose clause and its

host than b etween two co ordinated pivot chained clauses Also like most subsequent authors I

have not examined the p ossibility of recursive what Dixon called iterative purp ose clauses Dixon

Indeed it is not clear that these have ever b een given a satisfactory treatment

in any generative framework the analysis of Dixon was not kosher even by the standards of transformational grammar as it crucially relied on a countercyclical deletion op eration p

Chapter Concluding discussion

The Oblique Analysis of Dyirbal

As an alternative to my Inverse Grammatical Relations analysis BokBennema

and Kiparsky prop ose that the explanation for the unusual prop erties of Dyir

bal is that the SO absolutive argument is the only term while A ergative ar

guments are actually obliques This analysis the Oblique Analysis I referred to in

Chapter is adopted by Kro eger in an attempt to explain the purp orted

control facts of Dyirbal that I have just discussed Kro egers theory was that a con

trollee must b e an Actor and a term and the unusual prop erties of Dyirbal followed

from the fact that Actors are not terms in that language In this section I briey

consider but reject this alternative analysis

Arguments for and against

The fact that the ergative is homophonous with the instrumental in Dyirbal has b een

taken as supp ort for the Oblique Analysis But since the ergative is also homophonous

with the instrumental or another oblique case in many other languages this argument

cannot b e very strong Moreover Dixon suggested several grounds for why ergative

and instrumental must b e distinguished despite their homophony At argument

structure A NPs and instrumentals have dierent constructional p ossibilities Dixon

While b oth may b e promoted so that they b ecome the pivot it is via

dierent morphemes an A NP may b ecome pivot through antipassivization nay

nay while an instrumental NP b ecomes the pivot through use of an applicative

morpheme mbal Also Dixon notes a constraint where a sentence can

have at most one instrumental case NP but sentences may freely contain an ergative

NP and an instrumental NP

A second argument for the Oblique Analysis is that the ergative NP can b e freely

omitted in Dyirbal sentences like other oblique case NPs whereas a pivot NP can

8 0

Kro eger also attributes this analysis to Mel cuk however I b elieve this to b e

inaccurate I would classify Mel cuks analysis of Dyirbal as an instance of the Inverse Analysis

While Mel cuk balks at calling the ergative a grammatical ob ject calling it instead an agentive

complement he very clearly states pp that he do es regard regular two actant verbs in Dyirbal as transitive in contrast with his analysis of Lezgian

Chapter Concluding discussion

only b e omitted within a context a sentence ie pivot chain without an expressed

pivot is viewed in isolation as incomplete On the Oblique Analysis this could b e

taken as showing that ergative NPs function like other obliques On the Inverse

Grammatical Relations analysis this b ecomes a sp ecial prop erty of pivot NPs related

to sp ecicity

There are no other particularly convincing arguments for the p osition The fact

that with ditransitive verbs the third argument is always an oblique is consistent

with the Oblique Analysis but many languages lack double ob ject constructions

including all ergative languages of which I am aware I think that much of the app eal

of this analysis has b een a certain theoretical convenience one do es not need to

recognize inverse mappings which are problematic for many theories Things that

would normally happ en to sub jects in other languages but dont happ en to the A

argument in Dyirbal cease to b e problems b ecause now it is the O that is the sub ject

of an intransitive sentence But using this as a criterion b egs the question

Despite the absence of verb agreement and a causative form of verbs there are

several pieces of evidence that show quite clearly that Dyirbal has transitive and

intransitive verbs and other data that are suggestive of this distinction

Dyirbal has one more clause chaining construction the nurra construction

This is similar to a switch reference marker it shows that the pivot of the

nurra clause is necessarily coreferential with the A NP of the main clause it

also has a semantic comp onent and can only b e used when two events follow

one up on the other

bala yugu bangul yarangu madan waynyjinurra

ivabsth stickabs igenth manerg thrownfut gouphillimmed

The man threw the stick and then immediately went uphill

If the ergative NP is a term then this construction is consonant with switch

reference systems in other languages On the other hand this construction

would seem most unusual if the ergative NP were an oblique

9 A number of these arguments are prop osed indep endently in Johnson

Chapter Concluding discussion

Verbs in Dyirbal are strictly transitive or intransitive that is there are not am

bitransitive or labile verbs as in Lezgian Dixon Under the oblique

analysis this would have to b e describ ed by saying that some verbs disallow

an agentive oblique presumably for semantic reasons But this seems unlikely

since Dixon notes that there are pairs of verbs which seem se

mantically identical except that one is transitive and the other intransitive such

as walmanyu to wake up intr walmbin to wake up tr and jananyu to

stand up intr and jaran to stand up tr

Connected with the previous p oint is the existence of two conjugation classes

one of which has mainly transitive verbs and the other of which has mainly

intransitive verbs On the oblique analysis these classes would have to b e

analyzed as b eing purely diachronic detritus since all verbs are regarded as

intransitive

The treatment of case marking on pronouns would b e completely anomalous

on the oblique analysis Dyirbal has nominativeaccusative marking on pro

nouns see b elow But on the oblique analysis we would have to say

that sub ject pronouns app ear in one form with verbs which license an agentive

oblique and in another form with verbs that do not license an agentive oblique

Further this second form the nominative under other analyses is also used

when pronouns app ear as agentive obliques The nominative case is clearly

the unmarked case in the pronoun system despite certain irregularities in the

singular forms and Kiparsky predicts that the unmarked nominative

or absolutive case should not b e able to o ccur marking obliques and so the

pattern of pronominal case marking cannot b e explained

Under the Inverse Analysis the Dyirbal antipassive is an antipassive Under

the Oblique Analysis it app ears to b e an otherwise unattested advancement

op eration and it is not clear what distinguishes it from the applicative so that

the antipassive can only b e used to advance certain instrumental case NPs

while the applicative can only b e used to advance the rest

Chapter Concluding discussion

When the Dyirbal applicative morpheme is attached to intransitive verbs it

often has a comitative meaning

a bayi yara nyinanyu

iabsth manabs sitnfut

The man is sitting down

b balan jugumbil bangul yarangu nyinayman

i iabsth womanabs iergth manerg sitapplicnfut

The man is sitting down with the woman

On the assumption that Dyirbal has transitive verbs this is an instance of

the typologically common comitative construction On the oblique analysis

axation of mbal applic would b e a mysterious pro cess that introduced

a new actant the comitator as the sub ject while making the previous sub ject

an agentive oblique Further the calque for on the oblique analysis seems

rather unlikely the woman was satdownwith by the man

Dixon observes that adverbs are also transitive or intransitive and

must agree in transitivity with verbs On the oblique analysis this would again

b e mysterious we would have to either nd a purely semantic account or

suggest that adverbs are agreeing with the p ossibility of the verb taking an

agentive oblique

Conclusions

I prop ose that the null hypothesis should b e the following Dixon All lan

guages distinguish b etween clauses that involve a verb and one core noun phrase in

transitive clauses and those that involve a verb and two or more core NPs transitive

clauses including ditransitive as a subtype I do not nd the arguments presented

ab ove sucient to rule out the null hypothesis in fact I think they strongly supp ort

it Thus I prop ose to accept an analysis of Dyirbal the Inverse Grammatical Re

lations analysis that maintains that Dyirbal has transitive sentences and to reject the Oblique Analysis that denies this

Chapter Concluding discussion

Such a conclusion is also supp orted by application of Occams razor I have shown

in Chapters and that there are a number of languages which are syntactically erga

tive and in which evidence from agreement causatives and binding shows clearly that

there are transitive verbs Such languages cannot b e treated by the Oblique Analysis

in fact Kiparsky gives as a piece of supp ort for an Oblique Analysis of Dyir

bal the prediction that syntactic ergativity can only o ccur in languages without verb

agreement The Inverse Grammatical Relations analysis is needed indep endently for

these languages It is thus clearly more economical to conclude that Dyirbal should

also b e analyzed via the Inverse Grammatical Relations analysis than to once again

invent a new class of which it is the only member recall Section

Observations p erhaps not requiring structural explana

tions

There are several things that some accounts of ergativity have attempted to explain

ab out which I have said very little Let me briey address several of them here

Ergative case marking in syntactically accusative languages

The app earance of ergative morphological markings in a syntactically accusative lan

guage might initially seem unmotivated and strange I would like to suggest that

such markings are p ossible for two reasons one diachronic and one semantic

Morphological ergativity emerges easily diachronically If an ergative case marking

situation arises from a passive made obligatory as has b een suggested for the p erfect

asp ect in the Indic branch of IndoEurop ean Anderson but the A and S roles

reassert themselves as the grammatical sub ject or pivot as well as the asub ject then

morphological ergativity results and ergative case marking will remain for essentially

diachronic reasons This case marking must then simply b e stipulated in the lexical

forms of verb classes just as exp eriencer sub jects often take dative case in various

languages

There is an imp ortant general p oint lurking here Many current generative anal

yses attempt to explain all case marking facts by synchronic syntactic distinctions

Chapter Concluding discussion

However as has b een observed by Givon Cole et al and Estival and

Myhill it is frequently the case that synchronic morphology reects the syntax

of earlier forms of the language b ecause in general syntactic change precedes mor

phological change Cole et al show how dative exp eriencers can move from

b eing obliques to having sub ject prop erties while retaining their oblique case mor

phology and the same seems to happ en with ergative NPs that arise from obliques

It is not insightful but simply wrong to try to include the historical syntax of a

language within its synchronic description Thus a synchronic grammar should not

seek to explain all case marking facts since some of them will reect the history of

the language and b e arbitrary from a synchronic p ersp ective

However sometimes it seems that such a diachronic residue can b e reinterpreted

and the choice of case can have a coherent semantic basis This is suggested for Hindi

by T Mohanan and for Urdu IndoEurop ean Pakistan by Butt and King

They argue that in these languages the ergative app ears on grammatical

sub jects in the p erfect which exhibit conscious choice giving contrastive pairs such

as the following from Urdu

a anjum royii

Anjumnom cryperf

Anjum cried

b anjumne royaa

Anjumerg cryperf

Anjum cried on purp ose

For instance I think it is incorrect for the theories of Bittner and Bittner and Hale forth

coming b to try to predict the case marking on displaced terms in Inuit based on phrase structural

congurations of which category Casebinds the NP or which category the NP is a complement of

Such a theory cannot explain the regional historical and semantic variation in Case marking of

displaced terms in Inuit briey discussed in For instance Johnson suggests that the

ablative is usually used for passive agents in Central Arctic Eskimo but that the terminalis is used

with a semantically determined class of verbs There is no reason to suggest that the phrase struc

ture of passive sentences varies dep ending on the case marking of the passive agent but suggesting

dierent phrase structures would b e the only way to t this data into the account of Bittner and

Hale

The dierent verb forms result from the fact that verbs only agree with nominative NPs in HindiUrdu

Chapter Concluding discussion

The situation thus b ecomes very similar to that found in the subtype of activestative

languages where case marking on S NPs is determined semantically on a caseby

case basis such as Acehnese Eastern Pomo and TsovaTush NakhoDaghestanian

Georgia Holisky languages which Dixon carefully distinguishes

as Fluid S languages In such cases it is imp ortant to recognize that case marking

is not an arbitrary system for distinguishing terms Cases have a meaning a p oint

most strongly argued by Wierzbicka Where cases do not reect surface

grammatical roles they reect some other system of distinctions generally with a

semantic basis and such cases of morphological ergativity need not b e viewed as ar

bitrary or unexp ected any more than the existence of dative case marked exp eriencer

sub jects

The absence of syntactically ergative morphologically accusative

languages

Several p eople have faulted Marantz analysis see ab ove for predicting four

types of case marking in particular the fourth case where a language is syntacti

cally ergative but has Type B case marking app ears not to exist Levin Bok

Bennema Kiparsky Of course a claim of the nonexistence of Type B

case marking in syntactically ergative languages should not b e made to o strongly

For example languages like Dyirbal that are syntactically ergative but have NPsplit

ergativity do have Type B case marking on pronouns

a nana banaganyu

weplnom returnnfut

We returned

b nyurra nanana buran

youplnom weplacc seenfut

You all saw us

c nana nyurrana buran

weplnom youplacc seenfut

We saw you all

Chapter Concluding discussion

Note that despite the dierence in case morphology the language maintains an SO

pivot as can b e seen in the following example of co ordination see Dixon

for further examples

nyurra nanana buran banaganyu

youplnom weplacc seenfut returnnfut

You all saw us and we returned

Nevertheless Type B case marking in syntactically ergative languages is not very

widespread whereas Type B or morphologically ergative case marking is quite com

mon in syntactically accusative languages I susp ect that the reason for this has

mainly to do with the diachronic paths by which languages shift b etween ergativity

and accusativity As Dixon Ch argues the paths b etween ergativity and

accusativity and back are not symmetrical One of the consequences of this seems

to b e that morphological ergativity is quite widespread while Type B case marking

in syntactically ergative languages is quite limited

The absence of antipassive in syntactically accusative languages

Another fact that some analyses for example BokBennema Ch attempt

to capture is that antipassive is apparently restricted to ergative languages p erhaps

it is even the case that antipassive is restricted to syntactically ergative languages

This is disputed by Postal who endeavors to nd instances of antipassive in

morphologically and syntactically accusative languages but it seems that at most

these languages have limited or constructionsp ecic analogues of antipassive such

as Unsp ecied Ob ject Deletion John ate the burgerJohn ate or the conative alter

nation Mary kicked the ballMary kicked at the ball and not the general antipassive

op erations of some ergative languages

However I would like to suggest that p erhaps no structural explanation of this

observation is needed since a functional explanation app ears to suce Dixon has

12 y

In contrast coreferential omission in co ordination in Yidin dep ends solely on morphological case

Dixon In a number of other languages such as Hindi Mohanan and Marathi Joshi

coreferential omission in co ordination is p ossible only if the NPs agree in b oth grammatical

relation and case So a universal theory must allow either or b oth case and grammatical relation to determine p ossibilities for coreferential omission in co ordination

Chapter Concluding discussion

observed that an antipassive is necessary in syntactically ergative languages b ecause

of the prop ensity of humans to wish to talk for a while ab out the actions of a certain

Actor Conversely there is extremely little need for an antipassive in a syntactically

accusative language it is not needed to feed the pivot relation in such cases and

antipassive in general has no useful consequences for argument structure sensitive

op erations such as binding theory and derivational morphology b ecause the asub ject

remains the asub ject Its only role would b e to background the patient but in

general more limited mechanisms such as Unsp ecied Ob ject Deletion seem to suce

for this purp ose

With these p oints out of the way let me turn to what I have done

Conclusions

This dissertation has tried to motivate and substantiate a new typology of ergative

languages The typology maintains two classes namely syntactically and morpholog

ically ergative languages but greatly expands the class of languages that are classied

as syntactically ergative to include Philippine and other mixed pivot languages

This new classication results from a prop osal ab out levels of linguistic representa

tion I have argued that syntactic theory must recognize two levels a level of surface

grammatical relations and a level of syntactic argument structure While others have

argued for a level of argument structure separate from b oth the surface syntax and

a notion of thematic roles or thetastructure syntactically ergative languages show

most clearly that these two levels b oth exist and have their own prominence relation

ships

I have argued that universally there are features of language that are sensitive to

each of these levels My conception of syntactic ergativity is more limited and hence

the class of languages it covers is broader b ecause I only exp ect certain pro cesses

the ones sensitive to surface grammatical relations to b ehave dierently in a

syntactically ergative language Other pro cesses like binding and control pro cesses

A language with thoroughgoing ergative syntax must have an antipassive derivation to feed its SO pivot otherwise it could not op erate Dixon

Chapter Concluding discussion

of construal to use Kiparskys term are universally sensitive to argument structure

which has a roughly accusative nature This captures the generalization attributed

to Ken Hale in Miller that in a transitive construction with an agent and a

theme in a language with nominal reexives it is always the theme that is enco ded

as a reexive pronoun coindexed with the other agent

This claim diers markedly from previous work which has regarded a canonical

syntactically ergative language as one in which everything works on an ergative

absolutive basis eg Dowty leaving mixed pivot languages as something of

an anomaly Other work has emphasized the prevalence of mixed pivot languages

eg Van Valin but has provided no constraints on what patterns of mixed

pivot b ehavior are p ossible Kazenin suggests an implicational hierarchy al

lowing the cut o for features that are syntactically ergative to b e placed anywhere

along that hierarchy The present theory predicts many fewer p ossibilities for mixed

pivot languages If a language is syntactically ergative all phenomena sensitive to

grammatical relations should b e ergative or neutral If a language is syntactically

accusative all phenomena sensitive to grammatical relations will b e accusative or

neutral Phenomena sensitive to argument structure will always have an accusative

like nature The predictions app ear to b e largely conrmed although there remain

one or two problems that require further study eg control in Toba Batak

Johnson argues that all languages are syntactically accusative so as to

maintain a universal assignment of initial grammatical relations in RG and b e

cause pro cesses such as causatives inchoatives and applicatives app ear to refer to a

syntactically accusative level of organization But b oth these desiderata can b e satis

ed by adopting a level of argument structure as I have outlined reinforcing the need

for a level of argument structure that has b een motivated on indep endent grounds

by much recent work Alsina Grimshaw Rosen Adoption of a level

of argument structure allows us to recognize that ergative languages can b e b etter

explained by allowing an Inverse Grammatical Relations analysis Such an analysis

can explain Johnsons own observation p that pro cesses of relativization

topicalization coreferential deletion and question formation regularly fail to apply

to ergative NPs whereas Johnsons prop osal for dealing with these data was

Chapter Concluding discussion

selfconfessedly ad ho c

The p ossible alternatives seem less desirable Recognizing no extra level as in the

work of Johnson or Marantz has b een shown to b e quite problematic

Trying to work in terms of thematic roles rather than recognizing argument structure

app ears to b e untenable This was explicitly illustrated for the case of Toba Batak

and similar considerations would also make this approach unworkable in Inuit The

main remaining alternative is to keep two levels but to dene them dierently This

is how I interpret Bittners approach Congurational relationships within the

VP are one level of prominence while nal p ositions at sstructure give the other

level of prominence But this ends up predicting that what I have called grammatical

sub ject prop erties are actually prop erties of an A p osition This causes theoretical

problems even for Inuit since Bittner must p ostulate that NPs reconstruct down from

this p osition for purp oses of binding but not for purp oses of semantic interpretation

However the larger typological problems come from consideration of other languages

such as Tagalog where Kro eger has argued convincingly that the ang marked

NP has the prop erties of a grammatical sub ject and not the prop erties of an NP

in an A p osition a p osition normally reserved for NPs with a particular discourse

function

Adopting a level of argument structure at which constraints such as binding the

ory are dened ends up lending supp ort to a lexicalist theory of syntax A clear

distinction b etween pivots and asub jects as can b e motivated most clearly in syn

tactically ergative languages argues against a syntactic approach which tries to lump

these together as prop erties of NPs that are sub jects at some level In syntactically

ergative languages it can b e seen clearly that accounts of binding based on either

surface phrase structure congurations or surface grammatical relations are unsatis

factory I have shown that an argumentstructurebased treatment of binding and

derivational morphology b ecomes a necessary and satisfactory solution However

once argument structure is recognized as a separate level there is then nothing to b e

gained by p ostulating op erations of verb movement to generate complex derivational

forms and so the simpler lexicalist theory is to b e preferred for reasons of economy

Sources of Examples

c Bittner Morin and Tiou

a Johnson a Janussen p

b Johnson b Janussen p

a Bittner a constructed

b Sado ck forthcoming a Ortiz de Urbina i

c Bobaljik b b Ortiz de Urbina

a Bittner ii

b Bittner c Ortiz de Urbina

a Johns a ii

a De Wolf Anderson

a Andersen a Dixon Anderson

Anderson

b Andersen a Dixon Anderson

Hudson Dixon

a England a b

b England a b Ortiz de Urbina

c England a c Dixon

d England a b Kro eger ac

a England b adapted from Foley and

Campana a Van Valin abd

b England b Schachter

Campana a Schachter

c Campana c a Schachter a

d England b b Schachter b

Campana c Schachter b

England a a Kro eger c

a Comrie b Kro eger a

b Comrie a Fortescue p

Comrie BokBennema

a Comrie b Bittner a part

Sources of Examples

Postal ab etc b Comrie

Jackendo Schachter

myself a Emmorey

Bittner a b Emmorey

BokBennema a c Emmorey

adapted from Fortescue a Emmorey

p b Emmorey

a BokBennema b Sugamoto

b BokBennema b Sugamoto

a Bittner a Schachter

b Bittner a Tarpent

c Bittner b a Schachter

a Bittner a part b Schachter

b BokBennema p c Schachter a

from Fortescue a Schachter

p b Kro eger b

c Bittner a Fortescue p

Kro eger b Mey

Kro eger from a BokBennema b

text recorded by Blo omeld adapted from Fortescue

Kro eger a p

Fortescue p Bittner b

Edna Paneatak MacLean a Bittner

consultation b Bittner

Bittner b a Iida et al

Bittner c b Shibatani b

Bittner a Iida et al

Bittner Iida et al

a proverb a Perlmutter

b Ivan Sag pc b Perlmutter

a Carden b a a Kiparsky from

b myself Panchatantra

a Bittner b Kiparsky from

b Bittner Manu

myself a Marantz

myself b Mohanan b

Kleinschmidt p myself

De Ho op Dalrymple cf

Bittner Postal

Edna Paneatak MacLean Jackendo cf

Sources of Examples

a Woo dbury and Sado ck consultation

a Bittner

b Woo dbury and Sado ck b Bittner

b from Sado ck Bittner

p Bittner

Woo dbury and Sado ck Bittner

b Bittner

a Bittner Bittner

b Bittner a Fortescue p

Fortescue p b Bittner a

a Fortescue p Sado ck forthcoming

a Bittner from a Payne Ho op er Bay

b o ok Yup ik

a Bergsland p Fortescue p

Retranscrib ed by me a Enel p

b Bittner b b Enel p

a Fortescue p a Bittner

b Thalbitzer p b BokBennema p

Fortescue p a Fortescue p

Bittner a Fortescue p

a Sado ck forthcoming b Bittner

b Sado ck forthcoming c Bittner a

a Sado ck forthcoming a Bittner

b constructed b Bittner

Bittner c Bittner b

fn i Fortescue p d Bittner c

a Mey a Bittner b

a Fortescue p b Bittner a

Bobaljik b c Bittner c

a Fortescue p d Bittner b

b Fortescue p Smith p

c Fortescue p a Sado ck forthcoming

Bobaljik b a adapted from a novel by

Sado ck forthcoming Hans Anthon Lynge

constructed a Smith p cited by

a constructed Grimshaw and Mester

a Dalrymple one Woo dbury and Sado ck

option

b Dalrymple one b Woo dbury and Sado ck

option

Sources of Examples

Woo dbury a a Dalrymple

Marantz b b Hellan a

a Johns c a Bittner

b Johns c b Bittner

a Johns a a Bittner a

b Johns a b Bittner c

Marantz Bittner b

a BokBennema a Bittner

b Johnson b Bittner

BokBennema b b Bittner

Johnson Fortescue p

a Comrie a Bittner a

b checked with G uven a Bittner

G uzeldereconsultation Bittner

BokBennema cf Bittner

Bittner b Bittner b

Marantz b from Bittner ac

Marion Johnson pc Bittner

Johns Woo dbury a

Johns a Woo dbury a

Johns b Woo dbury a

a Bittner a a Dixon p

b Bittner and Hale forthcoming b Joan Bresnan lectures

b a from Dixon pc

Fortescue p c Joan Bresnan lectures

Bittner p fn from Dixon pc

Bittner b part d Joan Bresnan lectures

Bittner from Dixon pc

a Bittner BokBennema

b Bittner Johnson p

Bobaljik a Marantz from

a Bobaljik a Johnson The

b Bobaljik a free translation was altered

c Bobaljik b from by Marantz it should b e the

Bergsland p p encil

Dierently transcrib ed by Bittner a

me Bittner b

a Bobaljik a a Marantz a

b Bobaljik b b Swadesh p

Sources of Examples

b Hasp elmath Sado ck forthcoming xx

a Hasp elmath a drawn from Mller ed

b Hasp elmath

a Sag and Pollard b Baker

myself Baker

a Hasp elmath a myself

b Hasp elmath c Edna Paneatak MacLean

c Hasp elmath consultation

fn ia Mel cuk a fn imyself

fn ib Mel cuk b Schachter

Mel cuk a Bittner

Hasp elmath Hasp elmath b

Hasp elmath a Hasp elmath b

a Hasp elmath b Hasp elmath

b Hasp elmath c Hasp elmath b

a Hasp elmath b d Hasp elmath

b Hasp elmath c Hasp elmath

c Hasp elmath Hasp elmath

Hasp elmath a Hasp elmath b

a Mel cuk b Hasp elmath a

b Mel cuk c Hasp elmath

c Mel cuk g a Hasp elmath a

d Mel cuk g b Hasp elmath b

Hasp elmath a Hasp elmath b

Mel cuk i b Hasp elmath a

Hasp elmath Hasp elmath c

Mel cuk i Hasp elmath b

a Hasp elmath a Hasp elmath b

b Hasp elmath b a Hasp elmath a

c Hasp elmath b Hasp elmath b

Hasp elmath c c Hasp elmath

Hasp elmath one Hasp elmath

option of Mel cuk h Hasp elmath

Hasp elmath a a Hasp elmath

Hasp elmath one b Hasp elmath a

option of Mel cuk h c Hasp elmath c

a Hasp elmath a a Hasp elmath

b Hasp elmath b Hasp elmath c

a Hasp elmath a c Hasp elmath

b Hasp elmath a Hasp elmath

Sources of Examples

b Dixon Hasp elmath

Dixon Mamu Hasp elmath

dialect a Hasp elmath c

Dixon b Hasp elmath

a Dixon a Hasp elmath a

b Dixon p ortion

a Dixon b Hasp elmath a

b Dixon a Hasp elmath a

c Dixon b Hasp elmath b

a Dixon a Hasp elmath a

b Dixon b Hasp elmath a

c Comrie a Hasp elmath b

Dixon p part of b Hasp elmath b

Text XV a Hasp elmath b

Dixon p Text b Hasp elmath a

XV fn i Hasp elmath a

a Dixon p part of a Hasp elmath a

Text XV fn i Hasp elmath

b Co oreman fn ii Hasp elmath

Sag and Pollard a b Paul Kiparsky lecture notes

a Dixon from Martin

b Dixon Hasp elmath pc

a Dixon c Hasp elmath c

b Dixon Hasp elmath

Dixon Hasp elmath a

a Dixon a Hasp elmath

b Dixon b Hasp elmath b

Butt and King c Hasp elmath

a Dixon d Hasp elmath

b Dixon e Hasp elmath a

c Dixon a Dixon

Dixon b Dixon

a Dixon

Bibliography

Abusch D The scop e of indenites Natural Language Semantics

Aissen J L and D M Perlmutter Clause reduction in Spanish In D M

Perlmutter Ed Studies in Relational Grammar Chicago University

of Chicago Press Earlier version published in

Allen W S Transitivity and p ossession Language

Alsina A Predicate Composition A Theory of Syntactic Function Alternations

PhD thesis Stanford

Alsina A and S Joshi Parameters in causative constructions In L M Dobrin

L Nichols and R M Ro driguez Eds Papers from the th Regional Meeting

of the Chicago Linguistics Society Vol Chicago Linguistics So ciety

Andersen T Ergativity in Paria Nilotic OVS language Lingua

Anderson S R On the notion of sub ject in ergative languages In C N Li

Ed Subject and Topic New York Academic Press

Anderson S R On mechanisms by which languages b ecome ergative In C N

Li Ed Mechanisms of Syntactic Change Austin University of Texas

Press

Anderson S R Wheres morphology Linguistic Inquiry

Andrews A D The representation of case in Mo dern Icelandic In J Bresnan

Ed The Mental Representation of Grammatical Relations Cambridge

MA MIT Press

Bibliography

Andrews A D The ma jor functions of the noun phrase In T Shop en Ed

Language typology and syntactic description Vol Clause structure

Cambridge Cambridge University Press

Andrews A D and C D Manning Information spreading and levels of

representation in LFG Technical Rep ort CSLI CSLI Stanford CA

Austin P Transitivity alternations in Australian Ab original languages To

app ear in P Austin and B Blake Eds Theorectical linguistics and Australian

Aboriginal languages

Austin P Causatives and applicatives in Australian Ab original languages ms

La Trobe University and Stanford ersity

Austin P forthcoming Word order in a free word order language The case of Jiwarli

ms La Trobe University To app ear in Language

Austin P and J Bresnan Noncongurationality in Australian Ab original

languages ms La Trobe University and Stanford University

Baker M C Incorporation A Theory of Grammatical Function Changing

Chicago IL University of Chicago Press

Baker M C On some sub jectob ject nonasymmetries in Mohawk Natural

Language and Linguistic Theory

Baker M C Pronominal inection and the morphologysyntax interface In

M Ziolkowski M Noske and K Deaton Eds Papers from the th Regional

Meeting of the Chicago Linguistics Society Vol Chicago Linguistics

So ciety

Beghelli F D BenShalom and A Szab olcsi When do sub jects and ob jects

exhibit a branching reading In The Proceedings of the Twelfth West Coast

Bibliography

Conference on Formal Linguistics Stanford CA Stanford Linguistics Asso ciation

Beghelli F T Stowell and A Szab olcsi Predicate and distribute Paper

presented at the Third CSLI Workshop on Logic Language and Computation

Bell S J Cebuano Subjects in Two Frameworks Blo omington IN Indiana

University Linguistics Club

Bergsland K A grammatical outline of the Eskimo language of West Green

land mimeo Oslo Available from Interdocumentation Co Ag Zug Switzer

land

Bittner M Agreement and anaphora in West Greenlandic Eskimo ms London

Bittner M On the semantics of the Greenlandic antipassive and related

constructions International Journal of American Linguistics

Bittner M Canonical and Noncanonical Argument Expressions PhD thesis

University of Texas at Austin

Bittner M Ergativity binding and scop e ms Rutgers University

Bittner M Case Scope and Binding Dordrecht Kluwer Academic

Bittner M and K L Hale forthcoming a Ergativity Towards a theory of a

heterogeneous class ms Rutgers University and MIT To app ear in Linguistic

Inquiry

Bibliography

Bittner M and K L Hale forthcoming b The structural determination of case and

agreement ms Rutgers University and MIT To app ear in Linguistic Inquiry

Blake B J Tagalog and the ManilaMt Isa axis La Trobe Working Papers in

Linguistics

Bobaljik J D Nominally absolutive is not absolutely nominative In J Mead

Ed The Proceedings of the Eleventh West Coast Conference on Formal Lin

guistics Stanford CA Stanford Linguistics Asso ciation

Bobaljik J D On ergativity and ergative unergatives In C Phillips Ed

Papers on Case and Agreement II MIT Working Papers in Linguistics Vol

BokBennema R Case and Agreement in Inuit Berlin Foris

Bonet i Alsina M E Morphology after Syntax Pronominal Clitics in Romance

PhD thesis MIT

Borsley R Phrasestructure grammar and the Barriers conception of clause

structure Linguistics

Bresnan J Ed a The Mental Representation of Grammatical Relations Cam

bridge MA MIT Press

Bresnan J b The passive in lexical theory In J Bresnan Ed The Mental

Representation of Grammatical Relations Cambridge MA MIT Press

Bresnan J Exercises and problem sets Lexical functional grammar ms

Stanford University

Bibliography

Bresnan J a Category mismatches Paper presented to the plenary session of

the Annual Conference on African Linguistics

Bresnan J b Linear order vs syntactic rank Evidence from weak crossover

To app ear in the Pro ceedings of CLS

Bresnan J and J Kanerva Lo cative inversion in Chiche wa A case study of

factorization in grammar Linguistic Inquiry

Bresnan J and S A Mchombo Topic pronoun and agreement in Chiche wa

Language

Bresnan J and S A Mchombo The lexical integrity principle Evidence

from Bantu ms Stanford University and University of California at Berkeley To

app ear in Natural Language and Linguistic Theory

Bresnan J and A Zaenen Deep unaccusativity in LFG In Proceedings

of the Fifth Biennial Conference on Grammatical Relations University of

California San Diego

Burzio L Italian Syntax a GovernmentBinding Approach Dordrecht D Rei

del

Butt M and T H King Semantic case in Urdu In L M Dobrin L Nichols

and R M Ro driguez Eds Papers from the th Regional Meeting of the Chicago

Linguistics Society Vol Chicago Linguistics So ciety

Butt M J The Structure of Complex Predicates in Urdu PhD thesis Stanford

Campana M A Movement Theory of Ergativity PhD thesis McGill

Carden G a Logical predicates and idiolect variation in English Masters

thesis Harvard

Bibliography

Carden G b A note on conicting idiolects Linguistic Inquiry

Cena R M Patient primacy in Tagalog Paper presented at the LSA Annual

Meeting Chicago IL

Cho Y Y and P Sells A lexical account of inectional suxes in Korean

ms Stanford University To app ear in Journal of East Asian Linguistics

Chomsky N Syntactic Structures The Hague Mouton

Chomsky N Aspects of the Theory of Syntax Cambridge MA MIT Press

Chomsky N Lectures on Government and Binding Dordrecht Foris

Chomsky N Some Concepts and Consequences of the Theory of Government

and Binding Cambridge MA MIT Press

Chomsky N Barriers Cambridge MA MIT Press

Chomsky N Some notes on economy of derivation and representation In

R Freidin Ed Principles and Parameters in Comparative Grammar

Cambridge MA MIT Press

Chomsky N A minimalist program for linguistic theory MIT Occasional

Papers in Linguistics Number

Chung S Case Marking and Grammatical Relations in Polynesian Austin

TX University of Texas Press

Cole P W Herb ert G Hermon and S N Sridhar The acquisition of

sub jectho o d Language

Comrie B Ergativity In W P Lehman Ed Syntactic Typology

Austin TX University of Texas Press

Bibliography

Comrie B Degrees of ergativity Some Chukchee evidence In F Plank

Ed Ergativity Towards a theory of grammatical relations London

Academic Press

Comrie B Language Universals and Linguistic Typology Oxford Basil Black

well

Comrie B Causative verb formation and other verbderiving morphology In

T Shop en Ed Language typology and syntactic description Vol Grammati

cal categories and the lexicon Cambridge Cambridge University Press

Co ok K W The search for sub ject in Samoan In R Blust Ed Currents

in Pacic Linguistics papers on Austronesian languages and ethnolinguistics in

honour of George W Grace Pacic Linguistics C Canberra Reseach

School of Pacic Studies ANU

Co oreman A Ergativity in Dyirbal discourse Linguistics

Craig C G The Structure of Jacaltec Austin TX University of Texas Press

Dalrymple M The Syntax of Anaphoric Binding Stanford CA CSLI

Dalrymple M J Lamping and V Saraswat LFG semantics via constraints

In Proceedings of the Sixth Conference of the European Chapter of the Association

for Computational Linguistics University of Utrecht

De Guzman V P Ergative analysis for Philippine languages An analysis

In R McGinn Ed Studies in Austronesian Linguistics Athens OH

Ohio University Center for Southeast Asia Studies

De Ho op H Case Conguration and Noun Phrase Interpretation PhD thesis

Groningen

Bibliography

De Wolf C M Voice in Austronesian lanugages of Philippine type In M Shi

batani Ed Passive and Voice Amsterdam John Benjamins

Dixon R M W The Dyirbal language of North Queensland Cambridge

Cambridge University Press

Dixon R M W A Grammar of Yidiny Cambridge Cambridge University

Press

Dixon R M W Ergativity Language

Dixon R M W Complement clauses and complementation strategies ms

Australian National University

Dixon R M W Ergativity Cambridge Cambridge University Press

Dorais LJ The Canadian Inuit and their language In D R F Collis Ed

Arctic Languages An Awakening Paris Unesco

Dowty D R Grammatical relations and Montague Grammar In P Jacobson

and G Pullum Eds The nature of syntactic representation Dordrecht

Reidel

Dowty D R Thematic protoroles and argument selection Language

Durie M A Grammar of Acehnese on the basis of a dialect of North Aceh

Dordrecht Foris

Durie M The grammaticization of number as a verbal category In V Niki

foridou and M V Clay Eds Proceedings of the Berkeley Linguistics Society

Vol

Bibliography

Emmorey K The intonation system of Toba Batak In P Schachter Ed

Studies in the Structure of Toba Batak UCLA Occasional Papers in Lin

guistics Number

EncM The semantics of sp ecicity Linguistic Inquiry

Enel C Elements de Grammaire et de Vocabulaire de la Langue Ouest

Groenlandaise Elements of the grammar and vocabulary of West Greenlandic

Paris Museede lHomme

England N C a Ergativity in Mamean Mayan languages International

Journal of American Linguistics

England N C b A Grammar of Mam a Mayan Language Austin TX

University of Texas Press

Estival D and J Myhill Formal and functional asp ects of the development

from passive to ergative systems In M Shibatani Ed Passive and Voice

Amsterdam John Benjamins

Farkas D F Quantier scop e and syntactic islands In R A Hendrick C S

Masek and M F Miller Eds Papers from the Seventeenth Regional Meeting of

the Chicago Linguistics Society Chicago Linguistics So ciety

Finer D The syntax of switchreference Linguistic Inquiry

Fodor J D and I A Sag Referential and quanticational indenites Lin

guistics and Philosophy

Foley W A and R D Van Valin Jr Functional Syntax and Universal Gram

mar Cambridge Cambridge University Press

Forsberg V A p edagogical grammar of Tb oli Studies in Philippine Linguistics

Fortescue M West Greenlandic London Cro om Helm

Bibliography

Fortescue M The historical source and typological p osition of ergativity in

Eskimo languages ms University of Cop enhagen

Fukui N A Theory of Category Projection and Its Applications PhD thesis

MIT

Garrett A The origin of NP split ergativity Language

Gazdar G E Klein G K Pullum and I A Sag Generalized phrase structure

grammar a theoretical synopsis Oxford Basil Blackwell

George L M Ergativity and Relational Grammar In E Kaisse and J Han

kamer Eds Papers from the Fifth Annual Meeting North Eastern Linguistic

Society

Gerdts D B Antipassives and causatives in Ilokano Evidence for an erga

tive analysis In R McGinn Ed Studies in Austronesian Linguistics

Athens OH Ohio University Center for Southeast Asia Studies

Gibson J D and S Starosta Ergativity east and west In P Baldi Ed

Linguistic Change and Reconstruction Methodology Berlin Mouton de

Gruyter

Gil D Quantier scop e linguistic variation and natural language semantics

Linguistics and Philosophy

GivonT Historical syntax and synchronic morphology An archaeologists eld

trip In Papers from the Seventh Regional Meeting Chicago Linguistics Society

Givon T Syntax A FunctionalTypological Introduction Vol I Amsterdam

John Benjamins

Grimshaw J Form function and the language acquisition device In C L

Baker and J L McCarthy Eds The Logical Problem of Language Acquisition

Cambridge MA MIT Press

Bibliography

Grimshaw J Argument Structure Cambridge MA MIT Press

Grimshaw J Extended pro jection ms Brandeis University

Grimshaw J and RA Mester Complex verb formation in Eskimo Natural

Language and Linguistic Theory

Guilfoyle E H Hung and L Travis Sp ec of IP and Sp ec of VP Two sub jects

in Austronesian languages Natural Language and Linguistic Theory

Gunji T Japanese Phrase Structure Grammar A UnicationBased Approach

Dordrecht Reidel

Hale K L The passive and ergative in language change The Australian case

In S A Wurm and D C Laycock Eds Pacic Linguistic Studies in Honour of

Arthur Capell Canberra Linguistics Circle of Canberra

Hale K L The adjoined relative clause in Australia In R M W Dixon Ed

Grammatical Categories in Australian Languages Canberra Australian

Institute of Ab original Studies

Hale K L and S J Keyser Some transitivity alternations in English Tech

nical Rep ort MIT Center for Cognitive Science Lexicon Pro ject Working

Papers Boston MA

Harris A C Georgian Syntax A Study in Relational Grammar Cambridge

Cambridge University Press

Hasegawa K The passive construction in English Language

Hasp elmath M On the question of deep ergativity the evidence from Lezgian

Papiere zur Linguistik

Hasp elmath M A Grammar of Lezgian Berlin Mouton de Gruyter

Bibliography

Hellan L Anaphora in Norwegian and the Theory of Grammar Dordrecht

Foris

Hewitt B G Georgian Ergative or active Lingua

Hindley J R and J P Seldin Introduction to combinators and calculus

Cambridge Cambridge University Press

HirschbuhlerP VP deletion and acrosstheb oard quantier scop e In J Puste

jovsky and P Sells Eds Proceedings of the Twelfth Annual Meeting of the North

Eastern Linguistic Society Graduate Linguistics Students Asso ciation

University of Massachusetts Amherst

Holisky D A The case of the intransitive sub ject in TsovaTush Batsbi

Lingua

Hudson J Walmatjari Nominativeergative or nominativeaccusative In Pa

pers in Australian Linguistics No Canberra Department of Linguistics

Research School of Pacic Studies

Iida M Context and Binding in Japanese PhD thesis Stanford University

Iida M A unied theory of zibunbinding ms CSLI Stanford

Iida M C D Manning P ONeill and I A Sag The lexical integrity of

Japanese causatives ms Stanford University Paper presented at the th meeting

of the Linguistic So ciety of America

Inman M V Semantics and Pragmatics of Col loquial Sinhala Involitive Verbs

PhD thesis Stanford

Ioup G Some universals for quantier scop e In J P Kimball Ed Syntax

and Semantics Vol New York Academic Press

Ishikawa A Complex Predicates and Lexical Operations in Japanese PhD

thesis Stanford

Bibliography

Jackendo R Semantic Interpretation in Generative Grammar Cambridge

MA MIT Press

Jackendo R Semantic Structures Cambridge MA MIT Press

Jake J Why Dyirbal isnt ergative at all Studies in the Linguistic Sciences

Janussen E Handbog i grnlandsk grammatik Handbook of Greenlandic gram

mar Nuuk Pilersuik

Jelinek E Empty categories case and congurationality Natural Language

and Linguistic Theory

Jensen J T and A Johns The morphosyntax of Eskimo causatives In D B

Gerdts and K Michelson Eds Theoretical Perspectives on Native American

Languages Albany State University of New York Press

Jesp ersen O Philosophy of Grammar London George Allen and Unwin

Job M Ergativity in Lezgian In F Plank Ed Relational Typology

Berlin Mouton

Johns A Grammatical relations and case assignment in Eskimo Cahiers

linguistiques dOttawa

Johns A Transitivity and Grammatical Relations in Inuktitut PhD thesis

University of Ottawa

Johns A Deriving ergativity Linguistic Inquiry

Johnson D E Ergativity in universal grammar ms IBM T J Watson

Research Center

Johnson M R Ergativity in Inuktitut Eskimo in Montague Grammar and

in Relational Grammar Blo omington IN Indiana University Linguistics Club

Bibliography

Joshi S Selection of Grammatical and Logical Functions in Marathi PhD

thesis Stanford

KalmarI The antipassive and grammatical relations in Eskimo In F Plank

Ed Ergativity Towards a theory of grammatical relations London

Academic Press

Kaplan R M and A Zaenen a Functional precedence and constituent struc

ture In C Huang and K Chen Eds Proceedings of ROCLING II Taipei

Republic of China

Kaplan R M and A Zaenen b Longdistance dep endencies constituent struc

ture and functional uncertainty In M Baltin and A Kro ch Eds Alternative

Conceptions of Phrase Structure Chicago University Press

Katada F The LF representation of anaphors Linguistic Inquiry

Kathol A On agreement in HPSG ms Ohio State University

Kayne R S French syntax The transformational cycle Cambridge MA MIT

Press

Kazenin K I Split syntactic ergativity toward an implicational hierarchy

Sprachtypologie und Universalienforschung

Keenan E L a Remarkable sub jects in Malagasy In C N Li Ed Subject

and Topic New York Academic Press

Keenan E L b Towards a universal denition of sub ject In C N Li Ed

Subject and Topic New York Academic Press

Keenan E L Anaphora invariants and language universals In D Bates Ed

The Proceedings of the Tenth West Coast Conference on Formal Linguistics

Stanford CA Stanford Linguistics Asso ciation

Bibliography

Keenan E L and B Comrie Noun phrase accessibility and universal gram

mar Linguistic Inquiry

Kibrik A E Canonical ergativity and Daghestan languages In F Plank Ed

Ergativity Towards a theory of grammatical relations London Academic

Press

Kibrik A E Towards a typology of ergativity In J Nichols and A C

Woo dbury Eds Grammar inside and outside the clause Some approaches to

theory from the eld Cambridge Cambridge University Press

Kibrik A E Constructions with clause actants in Daghestanian languages

Lingua

Kiparsky P Morphology and grammatical relations ms Stanford University

Kitagawa Y Subjects in Japanese and English PhD thesis University of

Massachusetts Amherst

Kleinschmidt S P Grammatik der gronlandischen sprache mit theilweisem

einschluss des Labradordialects Berlin G Reimer

Ko opman H and D Sp ortiche The p osition of sub jects Lingua

Krieger HU and J Nerb onne Featurebased inheritance networks for com

putational lexicons In E J Brisco e A Cop estake and V D Paiva Eds

Inheritance Defaults and the Lexicon Cambridge Cambridge University

Press

Kro eger P Phrase Structure and Grammatical Relations in Tagalog Stanford

CA CSLI Publications

Bibliography

0 0 0

Kurylowicz J Ergativnost i stadial nost v jazyke Ergativity and the stadial

theory of language Bul letin of the Academy of Sciences of the USSR

Also published as La construction ergative et le developpement stadial du lan

gage Annali del la Scuola Normale Superiore di Pisa reprinted in

Kurylowicz J Esquisses Linguistiques I DeuxiemeeditionMuenchen W Fink

Lab ov W Sociolinguistic Patterns Philadelphia PA University of Pennsyl

vania Press

Laka I Unergatives that assign ergative unaccusatives that assign accusative

In J D Bobaljik and C Phillips Eds Papers on Case and Agreement I MIT

Working Papers in Linguistics Vol

Lamontagne G and L Travis The syntax of adjacency In M Crowhurst

Ed Proceedings of the West Coast Conference on Formal Linguistics Vol

Stanford CA Stanford Linguistics Asso ciation

Larsen T W Functional correlates of ergativity in Aguacatec In Proceedings

of the Seventh Annual Meeting of the Berkeley Linguistics Society

Larsen T W The syntactic status of ergativity in Quiche Lingua

Larsen T W and W M Norman Correlates of ergativity in Mayan grammar

In F Plank Ed Ergativity Towards a theory of grammatical relations

London Academic Press

Legendre G W Raymond and P Smolensky An optimalitytheoretic ty

p ology of case and grammatical voice systems In Proceedings of the Berkeley

Linguistics Society Vol

Levin B C On the Nature of Ergativity PhD thesis Department of Electrical

Engineering and Computer Science MIT

Bibliography

Levin J and D Massam Surface ergativity Casetheta relations reexamined

In S Berman Ed Proceedings of North Eastern Linguistics Society XV

Lewis D Scorekeeping in a language game Journal of Philosophical Language

Ludlow P and S Neale Indenite descriptions In defense of Russell Lin

guistics and Philosophy

Manning C D Imbabura Quechua relative clauses and other nominalizations

ms Stanford University

Manning C D Romance is so complex Technical Rep ort CSLI Stan

ford University Stanford CA

Manning C D LFG HPSG and CG Acronyms and logics ms Stanford

University Paper presented at the Second CSLI Workshop on Logic Language

and Computation

Marantz A P On the Nature of Grammatical Relations Cambridge MA

MIT Press

McClendon S Ergativity case and transitivity in Eastern Pomo International

Journal of American Linguistics

McCloskey J Clause structure ellipsis and prop er government in Irish Lingua

Mel cukI A Grammatical sub ject and the problem of the ergative construction

in Lezgian In C Bernard Ed Studies in the Languages of the USSR

Carb ondale IL Linguistic Research

Mel cukI A Dependency Syntax theory and practice Albany State Univer

sity of New York

Bibliography

Merlan F Split intransitivity functional opp ositions in intransitive inection

In A Woo dbury and J Nichols Eds Grammar Inside and Outside the Clause

New York Cambridge University Press

Merrield W R C M Naish C R Rensch and G Story Laboratory Man

ual for Morphology and Syntax Huntington Beach CA Summer Institute of

Linguistics Fifth edition

Mey J The cyclic character of Eskimo reexivization Acta Linguistica

Hafniensia

Miller B W Noncongurationality in Tagalog PhD thesis University of

Michigan

Miller P H Clitics and Constituents in Phrase Structure Grammar PhD

thesis Utrecht To app ear with Garland New York

Mithun M The evolution of noun incorp oration Language

Mithun M Implications of ergativity for a Philippine voice system In B Fox

and P J Hopp er Eds Voice Form and Function Amsterdam John

Benjamins

Mohanan K P Grammatical relations and anaphora in Malayalam Masters

thesis MIT

Mohanan T Causativization in Malayalam ms Stanford University

Mohanan T Arguments in Hindi PhD thesis Stanford University

Morin YC and E Tiou Passive in Burushaski In M Shibatani Ed

Passive and Voice Amsterdam John Benjamins

Mosel U Transitivity and reexivity in Samoan Australian Journal of Lin

guistics

Bibliography

Mosel U and E Hovdhaugen Samoan Reference Grammar Oslo Scandina

vian University Press

Mulder J a Syntactic ergativity in Coast Tsimshian Smalgyax Studies in

Language

Mulder J b The viability of the notion of sub ject in Coast Tsimshian Canadian

Journal of Linguistics

Murasugi K G Crossing and Nested Paths NP Movement in Accusative and

Ergative Languages PhD thesis Department of Linguistics and Philosophy MIT

Nap oli D J Syntax Theory and Problems New York Oxford University

Press

Nordlinger R The case of the rd p erson ob ject in the Wambaya auxiliary

University of Melbourne Working Papers in Linguistics

Nunberg G I A Sag and T Wasow Idioms Language

Ortiz de Urbina J Parameters in the Grammar of Basque Dordrecht Foris

Payne J The decay of ergativity in Pamir languages Lingua

Payne T E Role and reference related sub ject prop erties and ergativity in

Yup ik Eskimo and Tagalog Studies in Language

Perlmutter D Imp ersonal passives and the unaccusative hypothesis In Pro

ceedings of the Berkeley Linguistics Society Vol

Perlmutter D M Deep and Surface Structure Constraints in Syntax New

York Holt Rinehart and Winston

Bibliography

Perlmutter D M Ed Studies in Relational Grammar Vol Chicago IL

University of Chicago Press

Perlmutter D M The inadequacy of some monostratal theories of passive In

D M Perlmutter and C G Rosen Eds Studies in Relational Grammar Vol

Chicago IL University of Chicago Press

Pinker S Language Learnability and Language Development Cambridge MA

Harvard University Press

Pinker S Learnability and Cognition The Acquisition of Argument Structure

Cambridge MA MIT Press

Plank F Ergativity syntactic typology and universal grammar Some past and

present viewp oints In F Plank Ed Ergativity Towards a theory of grammatical

relations London Academic Press

Pollard C and I A Sag Anaphors in English and the scop e of binding theory

Linguistic Inquiry

Pollard C and I A Sag HeadDriven Phrase Structure Grammar Chicago

IL University of Chicago Press

Pollock JY Verb movement universal grammar and the structure of IP

Linguistic Inquiry

Postal P Crossover Phenomena New York Holt Rinehart and Winston

Postal P M Antipassive in French Lingvistic Investigationes

Pullum G K Word order universals and grammatical relations In P Cole

and J M Sado ck Eds Syntax and Semantics Volume Grammatical Relations

New York Academic Press

Bibliography

Rap oso E Case theory and IntoComp The inected innitive in Europ ean

Portuguese Linguistic Inquiry

Reed I O Miyaoka S A Jacobson P Afcan and M Krauss Yup ik Eskimo

Grammar Fairbanks AK Alaska Native Language Center University of Alaska

Riehemann S Word formation in lexical type hierarchies A case study of

baradjectives in German Masters thesis T ubingen

Rigsby B NassGitksan An analytic ergative syntax International Journal

of American Linguistics

Rischel J Some characteristics of noun phrases in West Greenlandic Acta

Linguistica Hafniensia

Rischel J Derivation as a syntactic pro cess in Greenlandic In F Kiefer

Ed Derivational processes Sto ckholm Research Group for Quantitative

Linguistics

Rosen C and K Wali Twin passives inversion and multistratalism in

Marathi Natural Language and Linguistic Theory

Rosen S T Argument Structure and Complex Predicates PhD thesis Brandeis

University

Sado ck J M Noun incorp oration in Greenlandic A case of syntactic word

formation Language

Sado ck J M Autolexical syntax A prop osal for the treatment of noun

incorp oration and similar phenomena Natural Language and Linguistic Theory

Sado ck J M Some notes on noun incorp oration Language

Sado ck J M Autolexical Syntax A Theory of Parallel Grammatical Repre

sentations Chicago University of Chicago Press

Bibliography

Sado ck J M forthcoming Reexive reference in West Greenlandic To app ear in

Contemporary Linguistics University of Chicago Dept of Linguistics

Sag I A and D Go dard Extraction of de phrases from the French NP In

M GonzalezEd Proceedings of the North Eastern Linguistics Society Vol

Sag I A and C Pollard An integrated theory of complement control

Language

Saiki M On the co ordination of gapp ed constituents in Japanese In Pa

pers from the st Regional Meeting of the Chicago Linguistics Society

Chicago Linguistics So ciety

Schachter P The sub ject in Philippine languages Topic actor actortopic

or none of the ab ove In C N Li Ed Subject and Topic New York

Academic Press

Schachter P Referencerelated and rolerelated prop erties of sub jects In

P Cole and J M Sado ck Eds Syntax and Semantics Volume Grammatical

Relations New York Academic Press

Schachter P Semanticrolebased syntax in Toba Batak In P Schachter

Ed Studies in the Structure of Toba Batak UCLA Occasional Papers

in Linguistics Number

Schmerling S F A categorial analysis of Dyirbal ergativity Texas Linguistics

Forum

Sells P Thematic and grammatical hierarchies Albanian reexivization In

H Borer Ed Proceedings of the Seventh West Coast Conference on Formal

Linguistics Stanford CA Stanford Linguistics Asso ciation

Sells P Korean and Japanese morphology from a lexical p ersp ective ms

Stanford University To app ear in Linguistic Inquiry

Bibliography

Shibatani M Voice in Philippine languages In M Shibatani Ed Passive

and Voice Amsterdam John Benjamins

Silverstein M Hierarchy of features and ergativity In R M W Dixon Ed

Grammatical Categories in Australian Languages Canberra Australian

Institute of Ab original Studies

Simpson J H Aspects of Warlpiri morphology and syntax PhD thesis MIT

Simpson J H Warlpiri MorphoSyntax Dordrecht Kluwer

Smith L R An analysis of axal verbal derivation and complementation in

Labrador Inuttut Linguistic Analysis

Sp eas M J Phrase Structure in Natural Language Dordrecht Kluwer Aca

demic

Starosta S A Pawley and L Reid The evolution of fo cus in Austronesian

In A Halim L Carrington and S A Wurm Eds Papers from the Third Inter

national Conference on Austronesian Linguistics Pacic Linguistics Series C

No Canberra Department of Linguistics Research School of Pacic

Studies

Sugamoto N Reexives in Toba Batak In P Schachter Ed Studies in

the Structure of Toba Batak UCLA Occasional Papers in Linguistics

Number

Swadesh M South Greenlandic Eskimo In C Osgo o d Ed Linguistic

Structures of Native America Vol New York Viking Fund Publications

in Anthropology

Tarpent ML Ergative and accusative A single representation of gramatical

relations with evidence from Nisgha Working Papers of the Linguistics Circle of

Victoria

Bibliography

Tenny C L Grammaticalizing aspect and aectedness PhD thesis MIT

Thalbitzer W Eskimo In F Boas Ed Handbook of American Indian Lan

guages Vol Smithsonian Institute Bureau of American Ethnology

Trager G L An outline of Taos grammar In C Osgo o d Ed Linguistic

Structures of Native America Vol of Viking Fund Publications in Anthropology

New York Viking Fund

Trask R L On the origins of ergativity In F Plank Ed Ergativity Towards

a theory of grammatical relations London Academic Press

Trechsel F R A Categorial Fragment of Quiche PhD thesis University of

Texas at Austin

Trubetzkoy N S Le rapp ort entre le determine le determinant et le deni

In Melanges de linguistique oerts a Charles Bal ly Geneva Georg et Cie

Van Valin Jr R D Grammatical relations in ergative languages Studies in

Language

VelaquezCastilloM The semantics of Guarani agreement markers In L A

Sutton and C Johnson Eds Proceedings of the Seventeenth Annual Meeting of

the Berkeley Linguistics Society

Weir E M H Incorp oration in Nadeb In D L Payne Ed Amazonian

Linguistics Studies in Lowland South American Languages Austin

TX University of Texas Press

Wierzbicka A The case for surface case Ann Arb or MI Karoma

Wierzbicka A Case marking and human nature Australian Journal of Lin

guistics

Bibliography

Wilkins W Thematic structure and reexivization In W Wilkins Ed Syn

tax and Semantics Vol Thematic Relations San Diego Academic

Press

Williams E Argument structure and morphology The Linguistic Review

Williams E Thematic Structure in Syntax Cambridge MA MIT Press

Woo dbury A C a Greenlandic Eskimo ergativity and Relational Grammar In

P Cole and J M Sado ck Eds Syntax and Semantics Volume Grammatical

Relations New York Academic Press

Woo dbury A C b The Greenlandic verbal sux ut Interactions of linguistic

form and grammatical function In Proceedings of the Third Annual Meeting of

the Berkeley Linguistics Society

Woo dbury A C Study of the Chevak Dialect of Central Yup ik Eskimo PhD

thesis University of California Berkeley

Woo dbury A C a Marginal agent clauses in Central Alaskan Yupik Eskimo

Internal and external syntax In W H Eilfort P D Kro eb er and K L Peterson

Eds Papers from the Parasession on Causatives and Agentivity at the st

Regional Meeting of the Chicago Linguistics Society Vol Chicago

Linguistics So ciety

Woo dbury A C b Noun phrase nominal sentence and clause in Central Yupik

Eskimo In A Woo dbury and J Nichols Eds Grammar Inside and Outside the

Clause New York Cambridge University Press

Woo dbury A C and J M Sado ck Axal verbs in syntax A reply to

Grimshaw and Mester Natural Language and Linguistic Theory

Bibliography

Zaenen A Unaccusativity in Dutch Integrating syntax and lexical semantics

In J Pustejovsky Ed Semantics and the lexicon Dordrecht Kluwer

Zubizarreta M L The relation b etween morphophonology and morphosyntax

The case of Romance causatives Linguistic Inquiry

Index

antipassive A see primitives Dixons

and equisub ject suxes A A see primitives Kibriks

i t

Dyirbal acommand

Inuit astructure see argument structure

Nisgha asub ject

applicatives accessible

Dyirbal AA p ositions

app ositional Inuit

Inuit see innitive Inuit Abkhazian

Archi accusative

argument pro jection subsystem

argument structure Acehnese

Autolexical Syntax activestative languages

Basque

actor

binding theory see reexives

Philippinist

Burushaski

adverbs

Dyirbal

case

agreement

direct

Archi

Inuit

Burushaski

Lezgian

Inuit

Marantz theory

lack of in Lezgian

oblique

Tabasaran

Catalan

Warlpiri

causatives

agreement order

Inuit see double transitive suxes

Inuit

Lezgian

Aguacatec

Turkish

Aivilik see Central Arctic Eskimo

Cebuano

Akhvakh

Central Alaskan Yup ik

Albanian

ambitransitive verbs see labile verbs

Central Arctic Eskimo

anaphors see reexives

Index

Yidin Central Sib erian Yupik

core role see terms Chamorro

coreferential omission in co ordination ChiMwini

see co ordination Chiche wa

coterms Chukchi

cyclic transformations Coargument Disjointness Condition

combinator

dative shift

complex predicates

alleged in Central Arctic Eskimo

Basque

Inuit

Dyirbal

conditional

dative sub jects

Inuit

deniteness see sp ecicity

conjugation classes

derived S see primitives Dixons

Dyirbal

ditransitives see also dative shift

conjunction see co ordination

Inuit

contemporative

double transitive suxes

Inuit see innitive Inuit

Inuit

control

Dyirbal

alleged in Basque

alleged in Dyirbal

Chukchi

Eastern Pomo

HPSG

ECM verbs see double transitive suf

Inuit

xes

Lezgian

ECP

Mam

equiNP deletion see also control

Tagalog

equisub ject suxes

Toba Batak

and antipassive

converbs

and passive

Lezgian

ergative

co ordination

case marking

Basque

subsystem

Central Sib erian Yupik

ergativity

Chukchi

historical origin of

Dyirbal

morphological

Hindi

split

Inuit

syntactic

Lezgian

Trasks typology

Marathi

Eskimo see Inuit Yup ik

Nadeb

exp eriencer sub jects see dative sub jects

Tagalog

expressive adjectival p ostbases

Index

innitive Extended Pro jection Principle

Inuit

uid S languages

Lezgian

Four Relations Analysis

Portuguese

fourth p erson

intransitive analysis see Oblique Anal

Inuit

ysis

functional uncertainty

Inuit see also Central Arctic Eskimo

Inuktitut I nupiaqLabrador In

generics see sp ecicity

uttut Qairnirmiut

Georgian

topic see topic Inuit

gerundive

Inuktitut

Inuit see innitive Inuit

Inverse Analysis

Gitksan

Inverse Grammatical Relations analy

grstructure

sis

grammatical relations

involuntary agent construction