Radical decomposition and argument structure

(1)

Patrycja Jab lo´ nska

A Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Department of Linguistics

University of Tromsø

(2)

(3)

Acknowledgements

Thisdissertationwouldneitherhavebeenbegunnornishedwithoutanum-

ber of people who contributed in various ways. First and foremost, I am

deeply indebted to my supervisor, Tarald Taraldsen, whose way of think-

ingabout linguistics isreected on every page ofthis thesis, withtheusual

disclaimers. In developing the ideas included in this work I have beneted

greatlyfromMichalStarke'sinspiringnanosyntaxseminar. Iwouldalsolike

tothankElisabetEngdahl,Bo»enaRozwadowskaandToreNessetforagree-

ingtobethemembersofmydoctoralcommittee. Lastbutnotleast,Ithank

SvetoslavMarinovwithloveforalltheL A

T

E

X-inghelpandhisunfailingbelief

inme.

(4)

(5)

List of Abbreviations vi

List of Tables viii

1 Introduction 1

1.1 Event decomposition . . . 1

1.2 The lexicon . . . 20

1.3 The

Θ

^-Criterion^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ³⁶

1.4 Deconstructing splitintransitivity . . . 39

1.5 Pylkkänen's staticfusion asaparameter . . . 47

1.6 The statusof Passiveina Voicesystem . . . 50

1.7 Peeling. . . 54

1.8 Overviewof thethesis . . . 60

I Themes 63 2 Event Decomposition and conjugation classes 65 2.1 Inter-language variation . . . 65

2.1.1 Dening the problem . . . 65

2.1.2 Autonomous root hypothesis (Déchaine, 2003). . . 74

2.1.2.1 LillooetSalish(St'át'imcets) . . . 74

2.1.2.2 English . . . 76

2.1.2.3 ModernHebrew . . . 77

2.1.2.4 PlainsCree . . . 78

2.1.3 Problems withDéchaine(2003) . . . 80

2.2 Intra-language variation. Themes . . . 83

2.2.1 Dening conjugation classes inPolish. . . 83

2.2.2 -i/y- stems . . . 88

2.2.3 Inchoative -n- stems . . . 105

2.2.4 -ej- stems . . . 109

2.2.5 More ondegree achievements . . . 113

2.2.6 Semelfactives . . . 121

2.2.7 Summary: dierent typesofThemes . . . 126

(6)

2.2.8 SecondaryImperfective . . . 127

3 Split Intransitivity 145 3.1 Thereexive markerand conjugation classes. . . 145

3.1.1 Preliminaries . . . 145

3.1.2 TheinteractionbetweenThemesandthereexivemarker160 3.1.3 Ananalysis . . . 168

3.1.4 Cross-linguisticevidencefor multiple Subjectpositions 173 3.1.5 Other reexive-marked derivations . . . 182

3.1.6 Conclusion . . . 194

3.2 ImpersonalPassive . . . 196

3.2.1 Problemswith -NO/TO.Overviewof literature . . . . 196

3.2.2 Lavine on -NO/TO . . . 203

3.2.3 -NO/TOand conjugation classes . . . 215

3.2.4 Conclusion . . . 228

3.2.5 Excursuson Turkish . . . 232

II Event Separators 237 4 Constructions involvingparticipial morphology 239 4.1 `Polyfunctionality' ofparticipial morphology . . . 239

4.1.1 Stative vseventive passive . . . 239

4.1.2 Resultative . . . 241

4.1.3 Eventive passive vsPerfect Tense . . . 244

4.1.4 Emonds(2000) . . . 251

4.2 -NO/TOrevisited. . . 258

4.2.1 Category statusvsexternal distribution . . . 263

4.3 Event Separator -n/t-ina lavishinsertion system . . . 269

4.3.1 Interaction withconjugation classes . . . 269

4.3.2 Levelof attachment . . . 278

4.3.3 ACC Case assignment . . . 278

4.3.4 Dierent places inf

seq

^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ²⁸¹

4.3.5 `DoublePassives' revisited . . . 284

4.3.6 Availability and morphological shape of the external argument . . . 288

4.3.7 Compatibility withthereexive clitic. . . 293

4.3.8 Semantic restrictionson theexternalargument . . . . 297

4.3.9 Degreesof idiompreservation . . . 303

(7)

5 Out-of-controlas init-less Causative 315

5.1 Preliminaries . . . 315

5.2 Cross-linguistic evidencefor OOC . . . 316

5.3 Dative Reexive Construction . . . 325

5.4 Analysis . . . 328

5.4.1 Init-less Causative . . . 329

5.4.2 The reexive marker . . . 336

5.4.3 The typology ofreexive-marked constructions . . . . 345

5.4.4 Comparison withparticipial constructions . . . 354

5.4.5 Looseends. Accidental vsabilityreadings . . . 356

6 Conclusion 361 A Conjugation classes and their paradigms 371 A.1 I THEME

high

^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ³⁷¹

A.1.1 -i- stems . . . 371

A.1.2 -y- stems . . . 371

A.1.3 -aj- stems . . . 372

A.1.4 semelfactive-n- stems . . . 373

A.2 IITHEME

low

^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ^. ³⁷³

A.2.1 inchoative -n-stems . . . 373

A.2.2 deadjectival-ej- stems . . . 374

A.3 IIIOTHERTHEMES . . . 374

A.3.1 -owa-stems . . . 374

A.3.2 -a- stems . . . 375

A.3.3 -e- stems. . . 375

Bibliography 377

(8)

List of Abbreviations

ABS Absolutive f Feminine

AC Anti-causative FI Faire Innitive

ACC Accusative fut Future

ACT Active Intransitivizer

(Salish)

FOC.A/S FocusSubject

ADESS Adessive FP Functional Projection

ADHORT Adhortative GEN Genitive

adj Adjectivizer G&M Guasti and Moro

AGR Agreement IMM Immediate (Salish)

AOR Aorist IMPF Imperfective Auxiliary

(Salish)

ART Article INCH Inchoative

AS Argument Structure inf Innitivalsux

AUT Autonomous Intransi-

tivizer(Salish)

INSTR Instrumental

B&K Butt andKing IRC Impersonal Reexive

Construction

CAC The Co-Ax Con-

straint

LCS Lexical Conceptual

Structure

CAU Cause Transitivizer

(Salish)

LSR LexicalSemanticRepre-

sentation

CAUS Causative LOC Locative

CL: Classier (Tariana) L&R Levin and Rappaport

COLL.DET Collective Determiner MDP Minimal Distance Prin-

ciple

CON Compound Connector m Masculine

cop copula NACT Non-Active

C&S Cardinalettiand Starke NOM Nominative

C&Sh Cardinaletti and Shlon-

sky

n nominalizer

DA Degree Achievement (NON)VIS (Non-)visual(Tariana)

DEF Denite NUM.CL: Numeral Classier

deg Degree neg Negation

DET Determiner neut Neuter

dim Diminutive nf Non-feminine (Tariana)

distr Distributive (non)vir (Non)virile

DM DistributedMorphology nom nominalization

DRC Dative Reexive Con-

struction

OBJ Objective Case

EC ElsewhereCondition OBL Oblique (Salish)

ERG Ergative OC Obligatory Control

(9)

EXIS Existential OOC Out-of-Control

emph Emphatic PART Partitive

PASS Passive perf Perfectivizing prex

POSS Possessive p.prt Past Participle

pf Perfective pst Past

pl Plural prt Participle

pr.prt Present Participle PROG Progressive

pref Prex pres Present

REC.P Recent Past REM.P Remote Past

re Reexive re.poss ReexivePossesive

REP Reported(Evidential in

Tariana)

RT Reexiva Tantum

sem Semelfactive Theme sg Singular

SI SecondaryImperfective CliticSI Secondary Imperfective

sux(Polish)

STA Stative SUB Subordinating

Th Theme TOP.ADV Topic Advancing Voice

TOP.NON.A/S Topical Non-Subject TRANS Transitive

VERT Vertical = boundary(Salish)

(10)

(11)

2.1 Person/Numbersuxes . . . 85

2.2 Present and Past paradigmfor pisa¢ . . . 85

2.3 Motion VerbsI . . . 102

2.4 Semelfactives . . . 122

2.5 Typology ofThemes . . . 126

2.6 Motion VerbsII . . . 138

4.1 Swedishpassive andperfect participles . . . 246

4.2 Present Tense conjugation zosta¢ . . . 262

4.3 PastTenseconjugation zosta¢ . . . 262

4.4 Lack of-n/t- prtwith`unaccusatives'. . . 270

4.5 -n/t- prtwithSI of`unaccusatives' . . . 270

4.6 -n/t- prtwithhigh Themestems . . . 271

4.7 Subjecthood tests. . . 293

4.8 Semanticproperties of externalarguments . . . 299

5.1 Malagasy prexes . . . 318

5.2 inherent OOC verbs . . . 322

5.3 DRC- summaryof properties . . . 329

5.4 Semanticrestrictions onargument inreexive constructions . 338 5.5 Syncretism within themasculine declension . . . 339

5.6 Spell out ofthereexive clitic invarious constructions . . . . 345

6.1 Comparison of ESand thereexiveclitic . . . 367

A.1 non-Past Tense . . . 371

A.2 PastTense. . . 371

A.3 non-past Tense . . . 372

(12)

A.10Past Tense. . . 374

A.11non-PastTense . . . 374

(13)

Introduction

1.1 Event decomposition

One of the most prominent debates in the generative literature, starting

mostprobablywithChomsky (1970)concerns thedivisionoflaborbetween

processesandrulesapplyingintheLexiconandprocessesrelevanttoSyntax.

Theproblemisespeciallyacuteinthedomain ofargumentstructure. Thus,

researchersworkingwithinthelexicalisttradition,andendorsingtheLexical

Integrity Hypothesis (cf. di Sciullio and Williams (1987)), still debate how

much information should be associated with the lexical item. The kind

of information that is being considered includes the notion of a thematic

grid, the number and type (i.e. internal vs external) of arguments, the

thematic labels of participants, etc.. As research has progressed, lexical

entries have grown more and more specic. At the same time it has been

noticed that the `fat' lexical entries run into obvious problems related to a

considerableexibilityofthe argument structurecongurationsthatagiven

verb(read: root)canoccurin. (1)isfromBorer (2005). (2)isanexampleof

a systematic argument frame change that came to be known as`causative-

inchoative alternation' andwill bea major topic ofthis dissertation.

(1) a. The factoryhornssirenedthroughout theraid.

b. Thefactory hornssirenedmiddayand everyone broke for lunch.

c. The policecarsirenedthe Porsche toa stop.

d. The policecarsirenedup to theaccident site.

e. The policecarsirenedthe daylight out of me.

(Borer, 2005:(7))

(2) a. Pat broke thewindow.

b. The window broke.

This fact lead some researchers to postulate an elaborate module of rules

taking place in the lexicon, which relate to the syntactic component via

(14)

`linkingrules' (cf. inter alia Levin and Rappaport(1995), Reinhart (1996),

Reinhart (2002)). To take an example, in Levin and Rappaport's (1995)

systemtheverbbreak hasa complexbieventive Lexical Semantic Represen-

tation (LSR).Theauthors taketheintransitiveform ofthis verbtoariseas

a result of binding the external cause at the level of mapping from LSR to

Argument Structure. Thisis schematicallyrepresented in(3):

(3) Intransitive break

LSR [[x do-something] cause [ y become BROKEN]]

↓

Lexical binding

∅

Linkingrules

↓

Argument structure

< y >

(4) Transitive break

LSR [[x do-something] cause [ y become BROKEN]]

Linkingrules

↓ ↓

Argument structure x

< y >

However, as argued in Borer (2005), the theory in which lexicon is a sep-

arate module with its own block of rules, (which, incidentally, have to be

extrinsically ordered), and which has to communicate with syntax via an-

other block of `linking rules' essentially amounts to duplicating the same

type ofinformation inthe lexicon,aswell asina syntactictree. Moreover,

thelexicon conceived of in thisway operates withexactlythe same kind of

vocabulary that thesyntax does(i.e. reference to external and internal ar-

guments, passivization, reexivization,etc.). From thisperspective itseems

that assuming only one module is the null hypothesis (cf. also Williams

(toappear)). Finally,if itturns out thatthe syntactic principlesoperating

outsidethedomain ofaword (whateverthedenitionofthelatter) arealso

relevant for the word-internal `domain',there seemsto beenough reasonto

take syntaxto be responsible for handling whathas traditionallybeen seen

aslexicalrules.

Idiosyncrasy of the lexicon

Itneedsto beemphasized, however,thatbeingderivedinsyntax isnotmu-

tually exclusive with being listed in the lexicon. Thus, the usual lexicalist

objection to deriving (some of) argument structure alternations in syntax

is high degree of semantic idiosyncrasy and distinct phonological processes

seen e.g. inso-called lexical causativesand passives, asopposed to syntac-

tic causatives and passives (cf. Wasow (1977), Shibatani (1976), Hale and

Keyser (2002), Travis (2000b) inter alia). This objection might in fact be

orthogonaltothenumberofmodulesassumed. Thepropertyofidiosyncratic

(15)

thereofforeventivepassivesisillustratedbelowforChichewa,whichusestwo

dierent morphemes to express twodierent passives: -idw for theeventive

passive and -ik for thestativepassive.

(5) a. Chimanga

corn

chi-ku-gul-idwa

AGR-PROG-buy-PASS

ku-msika.

at-market

`Cornis being bought at themarket.' (no idiomaticreading)

b. Chimanga

corn

chi-ku-gul-ika

AGR-PROG-buy-STAT

ku-msika.

at-market

`Cornis cheap at themarket.' (idiomatic reading)

c. Chaka

year

chatha

last

chimanga

corn

chi-na-lim-idwa.

AGR-PROG-cultivate-PASS

`Last yearcorn wascultivated.'

d. Chaka

year

chatha

last

chimanga

corn

chi-na-lim-ika.

AGR-PROG-cultivate-STAT

`Last yearcorn wasbountiful.'

DubinskyandSimango (1996:756)

As observed, however, in Marantz (1997), the very fact that the eventive

passive cannotreceive anyidiomaticreadingsseemsto argueagainstequat-

ingthe buildingblocksofsyntaxwith`phonologicalwords',since one would

expect that every `phonological word' has a potential to develop idiomatic

readings,asit necessarilybelongs to the realmof lexicon.

Although in the present work adjectival or stative passives will only be of

secondary interest, it is still hoped that this dissertation will adduce some

evidence for the syntactic nature of the argument structure manipulations.

To wit, I will take a dierent stand with respect to the contrast between

(5a) and (5b). I will argue, contrary to Marantz and all other accounts

postulating a xed domain to which idiomatic readingsare restricted, that

the`specialmeanings' are infactpossible,although much morerarely,with

eventive passives. A good illustration is an idiomatic example: The die is

cast. Being `frozen' inthepassive testies to theidiomatic character of the

phrase. One might argue that this is in fact a stative passive. Yet, a sim-

ilar idiom is available in Polish, but the use of auxiliary zosta¢ (`become')

indicates theeventive nature ofthe passive:

(6) Ko±ci

dice

zostaªy

become

pst.3pl.non−vir

rzuco-n-e.

throw-PASS-pl.nonvir

`The dieiscast.'

One more piece of reasoning that is missing is motivating the connection

between eventivity and the type of auxiliary. I think this can be done by

indicatingthedierenceintimeadverbialinterpretation: ifthereisanevent

present in (6), the adverbial will pick up this event and specify it's time

(16)

the adverbial will provide a time frame at which the state held. In the

last case, if the predicate happens to be an individual-level predicate, the

adverbialwill coerceit intoa stage-levelpredicate, asin(7):

(7) YesterdayJohn wasintelligent.

Totheextentthatcoercionispossible,theimplicationin(7)isthatatsome

timeeitherprecedingorfollowing yesterday John displaysa lesserdegree of

intelligence.

Consider now (8):

(8) Wczoraj

yesterday ko±ci

bones

zostaªy

become

pst.3pl.nonvir

rzuco-n-e.

throw-PASS-pl.nonvir

`The diewascast yesterday.'

Theinterpretation of (8)is clearly eventive, i.e. the state of `beingthrown'

holds ofthe dieforeverafterthe event tookplace (cf. Parsons' 1990 Resul-

tant State).

Onemore pieceof evidencefor theeventivityinducing propertyof zosta¢ is

negation. Assuming,asisratherstandardinPolish,thataprexalnegation

on the participle induces an adjectival/stative interpretation, we see that

whenever the negated participle occurs in the context of zosta¢ (`become'),

themeaning ofthelattershiftsto`remain'. Inotherwords,adjectivalinter-

pretationof the participleis incompatiblewiththeAuxiliary useof zosta¢:

(9) Prawa

rights

zostaªy

become

pst.3pl.non − vir

nie-ograniczo-n-e.

neg-restrict-PASS-pl.

`The rights remained/*became unrestricted.'

Sincethe interpretationof the `Auxiliary' doesnot haveto shift in(8)from

`become' to `remain', I conclude that (8) is an eventive passive. Conse-

quently,theidiomaticreadings, ashard asthey might beto nd,are infact

availableforeventivepassivesaswell. Iftheidiomaticityof(6)isnodierent

fromthe idiomaticityofkick the bucket, andthe burdenofproofison those

that would like to argue otherwise, then it becomes harder to delimit the

domain of `special meanings' (cf. also section 4.3.9 for more discussion on

idioms).

Similarly, the claim in McGinnis (2000) to the eect that idioms are not

available with category-external causatives does not seem to be borne out.

Although McGinnis' claimconcerns so-called `outer' causative in Japanese,

itseemsthatItalian causativein(10)isof thesamestatus, asindicatedin-

teralia by its predictable periphrasticnature, almostalways compositional

andpredictable meaning,andthevaguenatureofcausation. Yet,idiomsare

alsoavailablewiththistypeofcausative,asshownin(10)(TaraldTaraldsen,

(17)

(10) far

make ridere

laugh i

the polli

chickens

Italian

`todo something ridiculous'

Therefore,Iwillassumeinthepresentworkthatthedegreeoffrequencywith

which certainitemsorphrasesaresubjectto semantic driftisdependenton

the portion of the syntactic tree that they lexicalize: the lower the portion

of the tree, the more often an itemwill be listed, and hencemore prone to

undergo semantic shifts. The demonstration of this particular point, how-

ever,is beyondthescopeof thepresentwork.

Why not Aktionsart?

Following the neo-constructionist view of argument structure alternations

(e.g. Baker (1988), Ritter and Rosen (1998), Marantz (1997), van Hout

(1996), Borer (1998), Ramchand (1997), Ramchand (2003)), I will decom-

pose the information traditionally taken to reside in a lexical item into a

hierarchy of syntactic projections. I will also adopt a post-Davidsonian ap-

proach to semantics, where the term isdened inRamchand (2003)asthe

theoretical position whereby verbs do not have events in their `argument

structure',but where eventsand subeventscorrespondingto theinterpreta-

tionofverbal headsareutilised intheexpression ofcompositional semantic

meaning. (Ramchand, 2003:fn 10). The decomposition of themacro-event

into subevents will not, however, be based on the aspectual properties of

the structure (contra e.g. van Hout (1996), Borer (2005)). Moreexplicitly,

contrary to much of recent aspectual literature (cf. e.g. Verkuyl (1972),

Dowty(1979),Tenny (1987), Krifka (1992)),Iwill not assumetheso-called

compositionalapproachtoaspect,wherethepresenceofaspeciedquantity

ofan argument Xresults ina particular (i.e. telic)aspectualinterpretation

of the verb. In other words, the answer to the question posed by Borer

(2005): whetherthe syntacticallyrelevant argumentalrolesarereducible to

aspectualroleswillbenegative. Firstly,thepriviledgedroleofthequantized

internal argument inSlavic hasbeen shown inSchoorlemmer (1997) not to

holdfor at least one wayof theimplication : thepresence of thequantized

DPdoesnotinduceatelicor perfectiveaspectualinterpretationoftheverb,

i.e. (11),with aquantized internal argument, isatelicand imperfective 1

:

(11) Janek

Janek jadª

eat

pst.3sg.m

t¦

this

ACC

kanapk¦

sandwich

ACC

(przez

(for póª

half

godziny).

hour)

`Janekatethis sandwich (forhalfan hour).'

1

Forthepurposesofthisdissertation,itisnotcrucialtodistinguishbetween(a)telicity

onthe onehand,and(im)perfectivityonthe other,as inthecases consideredhere,the

predicates pattern similarly with respect to all tests, independentlywhether these are

perfectivity of telicity tests. Therefore, I do not wish to engage in a debate whether

perfectivity canbereducedto telicity ofnot. Thereaderisreferred to avastliterature

(18)

Moreover, as argued inRamchand (2003), even inEnglish there is a group

of verbs like push which get interpreted atelically in spite of the presence

of a quantized internal argument, as e.g. in John pushed the cart for half

an hour. In other words, I subscribe to Ramchand's position where she

takes `the telicity eects in the class of creation/consumption verbs with

quantized objectsto besemanticentailmentsandnot encoded inthelexical

determination of the verb or its syntactic reexes' (Ramchand (2003:13)).

These semantic entailments, I assume, are a general scenario in Polish in

cases where the verb is perfective. Since the present work is not about

aspect however, the reader is referred to Ramchand (2003) for elaborated

arguments why the mapping from objects to events is only relevant for a

subsetofso-called`directarguments'. Toanumberofherarguments,letme

justadd oneconsideration froma Polish perspective.

Let us take the Borerian system(Borer, 2005) asan example of aktionsart

approach to thematic roles 2

. Borer (pp. 19-20, ibid.) discusses and rejects

Reinhart's argument against aspectual analysis relating to the existence of

unaccusative atelic predicates. Borer uses Degree Achievements like yellow

andreddentoshowthattheseareinfactvariablebehaviorverbs: theycanbe

interpretedasan activity,inwhich casethereading of(12)issimultaneous,

or theycan beinterpreted astelic, andthe reading is sequential:

(12) Theapple yellowed and reddened.

Thistypeofexibilityisobviouslypredictedbytheradicallyconstructionist

system of Borer. Yet, it comes with an additional prediction that thesole

argument of (12) on the activity reading will display behaviors associated

with an external argument. This is because in Borers system, when Asp

Q

isnot projected, the sole argument can only moveto theSpecier ofEvent

Phrase,andbeassignedaninterpretationoforiginator withallthesyntactic

consequencesof thisunergative structure.

Now, inthisdissertation Ishowthatpreciselythegroupofverbsofthetype

of (12) (referred to as Degree Achievements) in Polish are both atelic and

imperfective, while at the same time they display clearly unaccusative be-

havior(i.e. intherelevant sense ofa given levelof`unaccusativity'- cf. the

discussioninsection1.4). Theiratelicityismanifested incompatibilitywith

a for X time adverbialin(13). Insection 2.2.8 Iargue thatthey behave as

unaccusativepredicates inthe sensethattheyresistSecondaryImperfective

formation (cf. (14)). Further substantiationof their unaccusative nature is

provided in chapter 3, where incompatibility with the reexive marker, as

well as ungrammaticality in Impersonal Passive (cf. (15) and (16) respec-

2

Thechoiceofthisparticular systemiscrucialsinceBorer (2005),as opposedtoe.g.

Verkuyl(1972), doesderive thepresenceof onlyone-way implicationinSlavic: the DP

inSpec,Asp

Q

^P^is^alwaysinterpretedasquantizedbyvirtueofSpec-headagreementwith theprexlocated intheheadofAsp

Q

^P;^yet,^the^presence^of^the^DP^is^not^necessary^for

(19)

tively) isinvoked.

(13) Jabªko

apple

N OM.neut

czerwieniaªo

redden

pst.3sg.neut

przez

for dwa

two

tygodnie/*w

weeks/*in

dwa

two

tygodnie.

weeks

`The applereddened for twoweeks.'

(14) *przy-±lepn-ywa-¢,

pref-blind-SI-inf,

*wy-pi¦kn-ywa-¢

pref-beaut-SI-inf

intended: `tobegetting blind,to begetting beautiful'

(15) *czerwienie¢

redden

si¦,

re,

*±lepn¡¢

get.blind si¦,

re,

*pi¦knie¢

get.beautiful si¦

re

(16) *Czerwienia-no/

redden-NO/

*lepni¦-to.

*get.blind-NO

intended: `Therewasreddening/ gettingblind.'

The conclusion from the preceding discussion seems to be that the aspec-

tual properties of the predicate need to be established independently from

the eventstructure where particular argumentsreceive semantic interpreta-

tion. Inthepresentwork Iwillrefer tothelatterregionof theclauseasthe

`thematic domain'. That is not to be taken as implying that no Aspectual

headscanappearwithinthisdomain(cf. alsothediscussioninTravis(2005)

onthedistinctionbetween lexical, functional,and event-related categories).

Theterm `thematicdomain'referstoalowregionoftheclause(roughlythe

domainof A-movement).

Dynamic boundary of l-syntax/ rst phase syntax

Certain fundamental assumptions relating to the status of this thematic

domain also require elucidation. A variety of terms has been employed in

the literatureto refer to this domain,which traditionallyhasbeen taken to

belongtothelexicon,twomostknownofthosetermsbeingHaleandKeyser's

(2002)l(exical)-syntax andRamchand'srst phasesyntax. Theleading idea

behind postulating this separate sub-module of syntax is the observation

thattheinformationassociatedwithparticularlexicalitemsissubjecttothe

sameprinciplesthatgovernthesyntacticcomponent,whileatthesametime

displayingcertaindierencesalludedtoabove(e.g. idiosyncraticcharacter).

Thus, Hale and Keyser (ibid.) argue that denominal verb as e.g. bag or

(20)

(17) bag theapples

V

bag

P

DP

theapples P

P N

Theassumptionin(17)isanecessarytightrelationbetween(18a)and(18b) .

(18) a. putthe apples inthebag

b. bagtheapples

Conation is argued to be subject to the Head Movement Constraint (cf.

Travis(1984)), asconrmed byimpossibilityto `skip'Pon thewayto V:

(19) *bag theapples in

Extending Hale and Keyser's line of argumentation, Travis (2000b) and

Travis(2005)providesfurtherevidence forthe syntacticlookingconstraints

on the structure of the lexical items. Her argument is based on the for-

mationof lexicalcausatives being subject to the DoublyFilled Voice Filter

(originally from Sportiche (1998:273)) in Tagalog. The main claim is that

thelexical as well as syntactic causative morpheme pag- is only allowed to

surface overtly in case the argument has moved out of the Specier of the

projection where pag- is located. Otherwise the morpheme has to delete.

Ashasbeen mentioned before, it isthe null hypothesis thatifthe l-syntax,

as well as s-syntax can be described in similar terms, they have to be so

described.

Yet, ifthere weren't any dierencesbetween l-syntax and s-syntax,thedis-

cussion would never arise in the rst place. And as I mentioned before,

the dierences most often discussed in this context include semantic and

phonological idiosyncrasy, as well as lack of productivity of l-syntax. Se-

mantic idiosyncrasywas illustrated in(5)for the`lexical' passive. The lack

ofproductivity isexemplied in(20), where arrive cannotundergo `lexical'

causativization inspite of its beingunaccusative (cf. its ability to occurin

apresentationalthere sentence in(21), aswellasasaprenominal participle

in(22) 3

.

(20) *Markarrived somepeople.

3

Bothofthesetestsmightinfactbechallengedasunaccusativitydiagnostics. Iemploy

`arrive'asit seemstobethe mostcertaincandidate foranunaccusativeverb inEnglish

inviewof the problemswithEnglishunaccusativity diagnostics(cf. specicallysection

(21)

(21) There arrived some people.

(22) recently arrivedguests

In a radical system which does not assume a boundary between l-syntax

and s-syntax (20) is predicted to be possible, contrary to facts. Assuming

then that there exists a boundary between the somewhat `unruly' syntax

and the syntax proper, a question arises with respect to the locus of this

boundary. Thisisthepointonwhichdierenttheoriesdiverge. ForHaleand

Keyser (2002) the external argument can only be introduced inthe syntax

proper (i.e. s-syntax) bya functional projection. Functionalprojectionsby

hypothesis do not occur within l-syntax. In Travis (2000b) and (2005) the

boundaryis instantiated byEvent Phrase(EP)locatedon topof thehigher

VPshell, asin(23).

(23) Travis' l-syntax domain

EP

E VP

DP V'

V AspP

AspP VP

DP V'

V PP

Inthissensel-syntaxincludesanexternalargumentintroducedbythehigher

verbal head (or by Asp in certain circumstances - cf. chapter 5). Travis'

system shares this assumption with Ramchandian tripartite structure for

the `rstphasesyntax'representedbelow.

(22)

(24) Ramchandian tripartitelowerdomain

ν

^P

INITIATOR

ν

^'

ν

^VP

UNDERGOER V'

V RP

RESULTEE R'

R XP

All of the aforementioned theories represent a `static' approach to the l-

syntax-s-syntax boundary. The main proposal in the present dissertation

willrelateto a `dynamic'wayof conceivingof theboundary inquestion. In

otherwords,Iwill adoptthepartoftheproposalinTravis'systemthatpos-

tulates the existence of certain morphological devices delimiting one event

domain. ThemorphologicalexponentsinquestionwillbelabeledEventSep-

arators (ES) and will be instantiated by participial morphology in Polish.

However, due to a dynamic nature of ES merger, in dierent constructions

dierent congurations might arise, yielding variousdegrees ofthepresence

ofan external argument. In thissense, itseemsto me, theconict between

Haleand Keyser'son theone hand,and Travis'and Ramchand's systemon

theotherhand seemsto beresolved.

Inspite ofthe dierencesexistingbetween Travis' andRamchand's systems

(e.g. the Aspectual projection intervening between thetwo verbal shells in

theformervsthelackofsuchaprojection inthelattersystem),manyofthe

aspectsofbothcouldbecompared. Outofnecessitytokeeptheterminology

consistent, but also due to the fact that Ramchandian systemseems to be

better equipped to handlelexical prexation inSlavic (dueto theexistence

of a special projection - Result Phrase (RP)) - I will adopt the tripartite

divisionin(24)withcertain crucialmodications.

Basic assumptions concerning decomposition

According to Ramchand (2003),

ν

^P întroduces ^the ^causation êvent ând ^li-

censesdierenttypesofexternalarguments. VP,ontheotherhand,species

(23)

or process. The latteris a denitional propertyof every dynamicverb. Fi-

nally,RPgivesthe`telos'or`resultstate'oftheeventandlicensestheentity

that comes to hold the result state. Both Causation event, as well as Re-

sultevent are taken to bestates, whereas theobligatoryevent isa Process.

This is meant to lead to a simpler ontology. The particular subevents are

`glued' together by means of a `cause/lead to' relation (following Hale and

Keyser (2002)) under therubric of Principle of Event Composition. Ramc-

hand's denitions oftwo derived predicates overeventsbased on this event

composition ruleareillustrated in(25):

(25) a. [

∃

^e

1

^,^e

2

^[State(e

1

⁾^&^Process(e

2

⁾^&^e

1 →

^e

2

^]]

→ def

^Causing(e

1

⁾

b. [

∃

^e

1

^,^e

2

^[State(e

1

⁾^&^Process(e

2

⁾^&^e

2 →

^e

1

^]]

→ def

^Result(e

1

⁾

Iwill modify that structure slightly. Firstly, Iwill assume thattheProcess

(ifany)alwaysresidesin

ν

^P.^The^reasons^for^that^particular^assumption ^are

several. Firstly, it is not clear to me in what sense a State preceding the

Transition/Process is necessarily tied to the presence of external initiator.

For instance, from the point of view of Déscles and Guentchéva (1995) the

precedingstateisacharacteristic propertyofeverydynamiceventuality(i.e.

both unergative run,aswell asunaccusative arrive).

Secondly,ifcausingsubevent isalwaysaState, thenthequestionariseshow

to distinguish between stative causative psych verbs (26a) and remaining

causatives (26b) :

(26) a. His behaviorsurprised Mary.

b. He broke thestick.

Ramchand considers psych-verbs of the `fear' type asdierent from regular

causatives in thattheir objectis Rhematic (i.e. further species the event,

but is not an `inner Subject'). Those, however, are crucially dierent from

the verb in (26a) , which passivizes and the interpretation of the object is

clearly as `undergoing' some change. One might distinguish between (26a)

and (26b)bypostulating that theProcess denoting head is present only in

(26b) ,butthisisnotanoptioninasystemwheretheProcess-denotinghead

is the obligatory nucleus of every verb. Therefore, I will assume (following

Marantz (2003)) thatthesubeventsout ofwhich amacro-event isbuilt can

beofvarioustypes. Specically,thecausing subeventcanbebothaProcess

(asinthe majorityof causative verbs) and a State(as instative causatives

in (26a) , as well as so-called `anticausatives', cf. chapter 3 for further dis-

cussion).

Although the aforementioned issues might still be considered peripheral,

there is one important consequence related to shifting the Process event

upwards (i.e. fromV into

ν

^). ^The ^side-eectôf ^this ^move îs^that ^process îs

(24)

inverbswhichlack

ν

^Paugmentation,theProcesspartisonly exceptionally available dueto theoperationof S-summing

4

. That means thattheseman-

tics of VP is Transition exclusively. Thiswill have crucial consequencesfor

the restrictions on the formation of so-called Secondary Imperfective (cf.

section2.2.8). Furthermore, I willalso delimit a group of verbs where even

the light verblayer hastransitional (and not stative or processual) seman-

tics,i.e. transitivedirectedmotionverbs. Moregenerallyspeakingthen,the

claim is that

ν

^denotes ^a ^Process ^(usually) ^or ^State ^(e.g. ^withpsych-verbs or anticausatives).

One important property distinguishes Ramchandian system from the radi-

callyconstructionist viewexpressed e.g. inBorer (2005). The latterclaims,

building on Marantz (1997) that the lexicon, or rather encyclopaedia, is a

repository ofunstructured `stu'whose relationto syntaxis completely ar-

bitrary. InRamchand's words, inthe absenceof anysyntacticinformation

atall,thelexiconreducestoanomenclaturewhosecooptionbythesyntactic

combinatoricsystemisreducedtoanaccident,oratbestamystery. (Ram-

chand, 2003:3). Since I share with Ramchand the concern to adequately

capture the restrictions on argument structure exibility, e.g. thefact that

arrive inEnglishcan neverbe usedtransitively,I willnot embrace therad-

ical Borerian view either. The way Ramchand restricts the exibility of

thesystem is by `tagging' a lexical itemwith categorial features which are

syntactically relevant. Thus, e.g. her lexicalentry for defuse is in(27):

(27) defuse: [

ν

^,^V

_i

^,^R

_i

^]

That means thatthe lexicalitem inquestion will associate its lexicalency-

clopaedic content with all the three heads: Cause (or Initiation in a more

updated terminology), Process and Result. Coindexing is taken to indicate

thatone argument checks/identies two subevents. Thisisdue to the EPP

requirement on eachof the subevents.

Thelexicalentry for ProcessIntransitive dance isillustrated in(28).

(28) dance: [

ν _i

^, ^V

_i

^]

Identifying the sole argument with the Subject of Process is forced by the

assumption that VP isan obligatory partof everydynamic verb. Ifthat is

so, however, than the sole argument of dance should show at leastsome of

thepropertiesofobjects,sinceSpec,VPisaprototypicalobjectposition. As

far asIcan see, however, noevidence to thateect isavailable, e.g. ne/en-

cliticization out of unergative verbs in Romance is impossible. It might be

contended, however, thatthene/en-cliticization iscontingent on thefateof

4

What Icall `process' here correspondsto whatis referred to as [+stages] inRoth-

stein (2004) and implies that the event is not an atomicentity, but has some internal

(25)

theargumentafterithasbeenmergedinitsinitialposition 5

. Inotherwords,

the movement from Spec,VP to Spec,

ν

^P ⁱⁿ^theRamchandiansystemmight be argued to destroy the conguration relevant for subextraction. There is

anotherargument,however,againstthepresenceofVP,althoughitconcerns

adierent setof data. Consider (29), where thepresenceof a lexicalprex

in(29b)and (29c)inducesa change inselectionalpropertieswithrespectto

the ACC object:

(29) a. Janek

Janek

N OM

biª

beat

pst.3sg.m

koleg¦/*z¦by.

colleague

ACC

^/ ^*teeth

ACC

`Janek beata colleague.'

b. Janek

Janek

N OM

wy -biª

pref-beat

pst.3sg.m

z¦by/

teeth

ACC

^/

*koleg¦.

*coleague

ACC

`Janek losthis teeth.'

c. Dyrygent

conductor

N OM

±piewaª

sing

pst.3sg.m

piosenk¦/

song

ACC

^/

*dzieciaki.

*kids

ACC

`The conductor sanga song/*thekids.'

d. Dyrygent

conductor

N OM

roz-±piewaª

pref-sing

pst.3sg.m

dzieciaki/

kids

ACC

^/

*piosenk¦.

*song

ACC

`The conductor got thekidsinto singing.'

Althoughunprexed transitive bi¢(`beat')allowsonly animateobjects,and

unergative or cognate object ±piewa¢ (`sing')allows a very restrictedset of

inanimate objects, once they are prexed, the selectional properties switch

around: wy-bi¢ can only take a very restricted set of inanimate objects,

androz-±piewa¢ animateones. IntheRamchandiansystem,the`unselected'

objectwouldbeintroduced inthe Specier ofRP,andduetotheobligatory

presenceofVPand the EPPrequirement on it,itwouldhave to remergein

Spec,VP.Ifthatisso,however,itisnotclearhowtheselectionalrequirement

of VP might be satised. One might propose alternatively that it is the

externalargumentthatistheUNDERGOER(i.e. itismergedinSpec,VP).

Yet, the truthconditions of the prexed sentences in(29)belie that sortof

explanation: Janekdidnot havetodrink asmuchasadrop ofwineinorder

toget Mariadrunk, asshownin(30) 6

:

5

Althoughembeddedinthepresenttheoryofanticausatives(cf. chapter3),thislineof

argumentationwouldnotworkeithersincesubextractionoutofanticausativesisclearly

possibleinbothItalianandFrench.Ifanticausativesareidenticaltounergativesinterms

ofthenallandingsiteoftheargumentinSpec,initP,thesubextractiontestispredicted

tocomeoutidentically,contrarytofacts.

6

Thechoiceofverbsin(29) is meanttoexcludetheanalysis whereby theunprexed

versionsdonotinvolveaVP,butinsteadare Rhematicobjects inRamchandiansystem.

Inthat case the argument would not hold, as the selectional restrictions inunprexed

(29bc)wouldonlyholdoftheRhematiccomplements. Theonlytypeofverbsthatdoes

not,as far as I cansee, change selectional properties ofobjects when prexed,is verbs

participatingincausative/inchoativealternation(cf. section 2.2.2for discussion). These

(26)

(30) Janek

Janek

N OM

u-piª

pref-drink

pst.3sg.m

Mari¦,

Maria

ACC

za

for

ka»dym

each

razem

time

wylewaj¡c

pour.out

pr.prt

zawarto±¢

content

swojego

his

kieliszka

wine.glass do

into zlewu.

sink

`Janek got Maria drunk, each time pouring out the content of his

wine glassinto thesink.'

Considering the above, if optionality of VP has to be assumed at least for

some cases,and if thereis no evidence for thepresenceof VP withunerga-

tives, it seems to be the null hypothesis that VP should also be absent in

unergatives of the dance type. Therefore, I will simply assume that VP is

missingwith unergatives of that kind. The sole argument is interpreted as

aSubject ofProcess inSpec,

ν

^P.

Finally, Ramchand's lexicalentry for a verbwhich hasa very impoverished

specication,i.e. widen (intransitive) isin(31).

(31) widen: V

Now, if transitivization amounts to causativization, then under Ramchan-

dian assumptions, it follows that only verbs which lack

ν

^-shell ^will ^be ^able

to be transitivized,asshownin(32):

(32) a. John widenedthegap.

b. * John dancedMary 7

.

This type of reasoning, however, hasto be taken with caution. Thus, e.g.

based on theungrammaticality of (33), Ramchand proposes that all of the

relevant verbs arespecied inthelexiconasin(34):

(33) *John arrived/fell/disappearedMary.

(34) arrive: [

ν _i

^,^V

_i

^,^R

_i

^]

Sheclaims that the soleDP is theinstigational forcebehindits own transi-

tion to `arrival'. That wouldimply that there arecertainsemantic nuances

thatgetlost whentranslated tootherlanguageswithclearerunaccusativity

diagnostics. Yet, from the Polish perspective, it seems especially unmoti-

vated to propose (34), since as I will try to show in this dissertation (cf.

especially in section 2.2), there is a substantial number of stems which do

not have causative equivalents, inspite of not having

ν

^-shell. ^Moreover, ^if

arrive was non-unaccusative in the sense of possessing

ν

^layer ^(cf. ^section

in order to refute the argument in the main text, one would have to assume that all

verbs otherthan the ones participating in the causative/inchoative alternation involve

Rhematiccomplements. Tothe extent that I understandRamchand's system, this was

notanintention.

7

ThecasesofJohndancedMaryoutoftheroomwouldprobablybeanalysedas`Mary'

(27)

1.4),itisnot clearwhyitdoespassall theotherof theremaining fewunac-

cusativitydiagnostics,namelythepossibilitytooccurinpresentationalthere

constructions, andattributive participle:

(35) There arrived some people.

(36) recently arrivedguests

Ofcourse,itispossiblethatpresentational`there'testisnot reallysensitive

tothepresenceof

ν

^-layer,^e.g. ⁱⁿ^Norwegian^acorrespondingconstructional- lowsunergativessuchasdanse (`dance'),sove (`sleep'), etc. (Tarald Tarald-

sen,p.c.). Furthermore, theprenominal participle testin (36)also presents

certain problems: as observed in (Pesetsky, 1995:23), many of seemingly

unaccusative verbs fail it. In this sense, however, arrive seems to be as

`unaccusative' as it could, as not only does it pass the presentational con-

structiondiagnostic, but alsothevery restrictiveprenominal participle one.

Thus, when faced with verbs such as arrive one has at least two strategies

atone'sdisposal: eithertoacceptthatthere canbe someaccidentalgapsin

causativization,which iswhattheRamchandiansysteminanycasedoesby

meansofthefeaturalspecicationon verbs,or elseto proposethattherele-

vantverbsalreadydisplayacausativetripartitestructureandhencedisallow

further(morphological)causativization. Evenifoneiswillingtoassumethat

Englisharriveisinfactasin(34),thelattertypeofstrategyfordealingwith

gapsincausativizationwouldnothelpinPolish,sincethereexistsagroupof

verbsinPolishwhichareunaccusativebyallsortsofdiagnostics(cf. chapter

2 and 3) and still do not possess causative equivalents. Some examples of

thiskind arein(37):

(37) przyby¢ (`arrive'), umrze¢ (`die'), wi¦dn¡¢ (`wilt'), sczezn¡¢ (`van-

ish'), kamienie¢ (`get stone-like'),marnie¢ (`get miserable')

Inprinciplethereisnothingwrongwithidentifyingallthreeprojectionswith

one and the same argument. Polish, in fact, is transparent inthat respect

since it has a minimal pair: przyby¢ (`arrive') and uda¢ si¦ (`go, depart'),

where the argument of the latter hasclearly more `instigational' properties

than the argument of the former. This semantic dierence correlates with

unaccusativitydiagnostics,i.e. onlythelatterisacceptableintheImpersonal

Passive (cf. section3.2and 3.2.3).

(38) a. * Przyby-t-ona przyj¦cie.

arrive-PASS-oat party

b. Uda-n-o si¦ naprzyj¦cie.

go-PASS-oreat party

(28)

Animportant detailof(38b)isthattheverbisnecessarilymarkedwiththe

reexivemarker. Iwillargueinsection3.1thatthereexive markersignals

identication of two

Θ

^-roles ⁱⁿ ^one participant. If Polish requires identication ofthe external and internal

Θ

^-role^to ^be ^licensed morphologically by meansofthe reexiveclitic,then itisonly theverbin(38b)(butnot(38a) )

thatrepresents a lexical feature specication proposedfor arrive byRamc-

hand.

8

Atthispointitisalsonecessarytotakeastandwithrespecttothenatureof

ν

^/v,^which^has^beenextensivelyusedintheliteraturestartingfromChomsky (1995b),andhasgrowntobethelocusofamultiplicityofdierentfunctions.

Someof the functionsareenumeratedbelow.

1.

ν

^/v^is^acategory-dening head,whichselects foraparticular root (cf.

Marantz(1997) and subsequent work). Thus, Marantz (1997) postu-

lates that v is only one of the three functional heads that determine

the syntactic category of a root: v's, n's, a's. I will not adopt that

assumption for two reasons. Firstly, Marantz claims that little vcon-

structs verbal meanings like `causative', `stative', etc. That seems to

suggest that themorphology displayed in thecausative variant of al-

ternating causative-inchoative verbs should be associated with little

v. Onthe other hand,healso takes Semitic templaticmorphology - a

moreplausiblecandidateforacategory-deninghead-tobeassociated

with`the little vsystem'(cf. Marantz(2001)). Clearly, thereare lan-

guageswhichpossessboth: causative augmentsderivingthetransitive

variants of alternating verbs, aswell as verbal templatic morphology.

Consider e.g. Amharic inthisrespect:

(39) Amharic

√ dkm

^- ^däkkämä ^(`be ^tired')^- ^a-däkkämä ^(`make ^tired')

Thetransitivevariantoftherelevantpredicatehastobeprexedwith

the causative morpheme a-, while the verbal templatic morphology

stays. Clearly, both the prex and the template cannot occupy one

head,unlesssomeheadsadjoinedtolittle varepostulated. Therefore,

inthepresentwork,Iwillanalysecausativemorphologicalaugmentsas

occupying

ν

^and^standing ^for ^a Process/State subevent withbleached encyclopaedic content. On the other hand, Polish, like Semitic, dis-

plays morphological verbalizers manifested in the shape of thematic

vowels. I will take these vowels to be only indirectly determining the

category of the root by spelling out the verbal functional sequence

8

Theverbin(38b)isalsolexicallyprexed,whichmakesitslexicalentryanalogousto

(29)

(henceforth, f

seq

⁾ ^(cf. ^section ^1.2 ^for discussion).

9

In this sense they

might (asinthe caseof`high'thematic vowels),but donothaveto be

associated withtheProcess/State subevent thatIuse

ν

^to ^denote.

Apartfrom this empirical consideration, there is a more fundamental

issue related to a general theory of lexical categories: V, N, A. Al-

thoughitisbeyondthescopeofthisworktoprovetheparticularpoint

(butsee chapter 4 for arguments based on Polish participial passives

andnominalizations),IbelieveMichalStarke'srecentidea(nanosyntax

seminar)tobeontherighttrack: thedistinctionsbetweenwhatwecall

`adjectives',`nouns'or`verbs'arecontingentontheregionoftheclause

beingspelledoutbyaparticular lexicalitem. Thelowestregionofthe

clausereceivestheinection perceivedas`adjectival'. Ifthederivation

continues into higher regions, `nominal' morphology occurs. Finally,

an `outgrown' noun becomes a verb. Hopefully, the investigations in

chapter 4will provide some substantiationof thisclaim.

2.

ν

îs ^the ^locus ôf ^manner ^(cf. ^Hale ând ^Keyser ^(2002)). Î ^will âdopt

that assumption with certain modications with respect to Hale and

Keyser's assumptions. Firstly,manner features (licensing manner ad-

verbials) will depend on the nature of

ν

^(i.e. ^Process ^vs ^State). ^Sec-

ondly, Hale and Keyser (ibid.) distinguish between patient-manner

andagent-manner verbs (e.g. splash vs smear respectively). Since in

the latter verbs the manner component relates to the external argu-

ment, they are ungrammatical in the inchoative variant, which lacks

anagent (cf. (40) vs(41)):

(40) a. The carssplashed mudon thewall.

b. Mud splashedonthe wall.

(41) a. Theysmeared mudon thewall.

9

Theindirectrelationshipbetweenthematicvowelsandalexicalcategoryismanifested

e.g. bythepresenceofthematicvowelsinpassiveparticiplesdisplayingessentiallyadjec-

tivalmorphology. Tothe extentthat participlesemployedineventivepassivesmightbe

arguedto be adjectivesderived fromverbs, the latter strategyseems impossible inany

neo-constructionist frameworkfor stativeadjectivalparticiples, which donotentailany

event taking place. Yet, both involve thematicvowels. Iillustratewith aminimal pair

(cf. theabovediscussionconcerningzosta¢ (`become/remain')):

(i) a. Ksi¡»ki

books nie

neg zostaªy

become

pst.3pl.non − vir

prze-czyt-a-n-e.

pref-read-TH-PASS-pl.non-vir

`Thebookshavenotbeenread.'

b. Ksi¡»ki

books

zostaªy

remain

nie-prze-czyt-a-n-e.

neg-pref-read-TH-PASS-pl.non-vir

`Thebooksremainedunread.'

Thethemevowelisglossed`TH'.Thispointsuggeststhatthepresenceofathematicvowel

cannotindicatethatagivenitemisor hasbeena`verb'atany pointinitsderivational

(30)

b. *Mud smearedon thewall.

Haleand Keyser (2002:35)

I will however assume that the manner component on a

Θ

^-assigning

head can only relate to the argument that it assigns a

Θ

^-role ^to ^(i.e.

only an external argument in the case of

ν

^). ^In ^this ^sense, ^however,

theverbsin(41)areindistinguishable. Theungrammaticalityof(41b)

is thus taken to be due to the same reason that yields (42) ungram-

matical,withinanimate mud asan external argument:

(42) *Mudsmeared butteron thecarpet.

The reader is referred to chapter 3 for a detailed discussion of the

anticausative structureinvolved in(40b) .

3.

ν

îs ^the ^locusôf âgentivity^, î.e. ^features ^relevant ^to ^the ^licensing ând

interpretation ofexternal arguments. That will be adopted only with

the reservation made above (i.e.

ν

^has ^to ^be ^a ^Process ^to ^involve

agentivity). Furthermore, Irestrictthe licensing of externalargument

to the `implied external argument', i.e. only licensing of adjunct by-

phrasesinaregularcase (see below).

4.

ν

checks/values ACC Case on the object. I will not adopt this assumption. In general, I will not adopt the Probe-Goal approach to

Case checking (cf. e.g. Chomsky (1999)). Although

ν

^seems ^to ^be

instrumentalinACCCase assignment,thisisonlydueto thefactthat

acertainlevelinf

seq

^needs^to^be^reachedⁱⁿ^order^for ^the^ACC^Case^to

emerge. This assumednotion of Case is based on the idea by Michal

Starke,whereastructuralCasehastoemergefromobliqueCaseinthe

courseofthederivation(cf. especiallysection1.7onmoreon`peeling',

aswell aschapters 4 and 5for therelevantdiscussion).

5.

ν

întroduces ân êxternal ârgument. ^That âssumption ^requires ^some

discussion. Although I arguethatthepresenceof

ν

^licenses^a ^Subject

ofProcess/State,thisisnotyettheexternalargumentthatsurfacesin

avanilla-avor active sentence. Instead theSubjectofProcess isonly

animpliedadjunct,and ifitoccurs,itsmorphological shapeisoblique

(i.e. a by-phrase). That means that I need to reject Ramchandian

EPPfeature on

ν

^.

6.

ν

ôf â^dierentâvor ^(i.e.

ν Become/Result

^,^cf. Âlexiadou ândÂnagnos-

topoulou (2003)) is present with unaccusatives. I will not adopt that

assumption. Thus, what I will sometimes refer to as `unaccusatives'

(although see the discussion in section 1.4) are structures where the

verbal stem spells out bare VP (i.e. no

ν

^system ^is ^involved). ^The

(31)

issue relating to the presence of

ν

^with unaccusatives seems to be purelyterminologicalinthefollowingsense. Inordertodistinguishbe-

tween `unaccusative' and`non-unaccusative' structuresone ofthetwo

strategies seems to be available: (i) postulating multiplicity of dier-

ent`avors'ofagivenfunctionalhead;(ii)pursuingaveryne-grained

decompositionanalysisoftherelevantstructures. Sinceinthepresent

workIadoptthelatterstrategy,thedierencebetween `unaccusative'

and`non-unaccusative'verbsinthepresentsystemwillboil downtoa

dierentnumberoffunctionalheadsinvolved. HadIchosentoassume

ν

ⁱⁿ ^the ^structure ôf închoative ùnaccusatives', one additional head

ν _n

^would^have^to ^be ^involvedⁱⁿunergativesand transitives. Sincethe crucial border that seems to emerge out of empirical investiations in

thepresent work is between inchoative `unaccusatives' (i.e. Theme

low

stems) and other verbs (i.e. Theme

high

^stems), ^I ^decide ^to ^label ^the

transitionheadinvolvedintheformerasVP

Become

ⁱⁿ^order^to^separate

itfromthelight verbsysteminvolvingmultiplicity of

ν

^'s.

7.

ν

^denes^the^rst^phase. ^This^assumption ^isrepresentativeofastrand of research building on Chomsky's (1999) proposal to the eect that

syntacticderivationsundergosemanticandphonologicalinterpretation

inincrementalchunksorphases. ApartfromC

0

andD

0

,

ν ⁰

^assigning^a

Θ

^-roleîs^claimed ^to^head â^`strong^phase'. ^The ^gistôf ^theîdea îs^that

oncea phasal head is complete, movement and agreement operations

can only target the head of the phaseand its edge (i.e. theSpecier

andadjuncts),butitscomplementis`frozen'orimpenetrable. Thishy-

pothesis,combinedwiththeempiricalworkonl-syntaxor`thedomain

of specialmeanings' instigated many accounts arguing for thephasal

statusof category-deningheads:

ν

^/v,^a,ⁿ^(cf. ^e.g. ^McGinnis ^(2000),

Arad(2003)amongmanyothers). Thus,themajordistinctionbetween

root-externalandcategory-external material inasense translatesinto

the above discussion between l-syntax and s-syntax boundary. I will

not adoptthe phasal status of

ν

êither. ^The ^rst, êmpirical^reason îs

thatcertainunpredictable morphophonologicalprocessesarestillhap-

pening above

ν

^. Ône êxample îs ^the^suppletive ^choice ôf â ^Secondary

Imperfective(SI)morphemeinPolish. TheusualSecondaryImperfec-

tive suxesare: -aj-forthe -i/y-conjugation classverbs,and -i/ywa-

forallthe remainingclasses. However,withsomemovement verbsthe

SI morphology is completely unpredictable, i.e. either suppletive or

onlyvaguelyrelatedto thenon-SI form:

(43) a. wy-j±¢ (`go out')vs wy-chodzi¢ (`go out

imp

^')

b. wy-jecha¢ (`depart') vswy-je»d»a¢ (`depart

imp

^')

(32)

Iargueinsection2.2.8 thatthe SImorphologyhosting headislocated

on top of

ν

^. ^That ^seems ^to ^indicate ^that

ν

^cannot ^be ^the ^boundary

delimiting thedomain of specialphonology,meaning, etc.

Moregenerallyspeaking,Iwill arguethatthestemisstretchable, and

the boundaryof the event (usuallybrought up inconnectionwith so-

called`rstphase')isdependent onthelevelwhere theso-calledEvent

Separator merges.

Summarizing,theonlyfunctionthat

ν

^is^fullling^under^currentassumptions isintroducingoneof thesubeventsinapossiblycomplexmacro-event. This

isdue to radicaldecomposition, where other featurestraditionally taken to

be carried by

ν

âre ^taken ôver ^by ôther ^functional projections on top of

ν

^. ^Thus, ^all ^the assumptions enumerated above concerned the projection

that I will label

ν _{N EU T}

^(or, ^more ^generally^,

ν ₁

^). ^I ^will, ^however, ^propose

thatthereis infactahierarchyoflight verbslocated ontop of

ν _{N EU T}

^that

license subsequent levels of`externality' of an argument. In thatsense, the

distinctionbetween `internal' and `external' argument will not be conceived

of as a dichotomy, but rather as a gradient property correlating with the

position ofthe DP ina functionalsequence 10

.

Italsoemerges fromtheabovediscussionthatIwillhaveto makeextensive

use of a movement from one

Θ

^-position ^to ^another. ^This ^is ^rstly ^due ^to

having three (inner) Subject positions: INITIATOR, UNDERGOER and

RESULTEE (where a Subject should really be taken as Subject of some

predication relation), and only two structural Cases: NOM and ACC. The

secondreasoninvolvesaparticularanalysisofthereexive marker,whereby

one DP needs to check two

Θ

^-features. ^This ^is^tantamount ^to ^rejecting ^the

Θ

^-Criterion^-â^topic^to^whichÎ^turnⁱⁿ^section^1.3. Âs^for^whichparticipants can get identied with which other participants (Ramchandian coindexing

convention), Itake itthatthe underlying principle mustbe someversion of

RelativizedMinimality(Rizzi(1990) - cf. section1.3, aswell as3.1).

1.2 The lexicon

Thequestion of `unpronounced'lexicalitems, or zero morphology isrelated

to the general view of the lexicon. As already mentioned, in Ramchand's

system syntactic information is represented as `tags' on lexical items. The

lexicalencyclopaediccontent isassociated withparticular functionalprojec-

tions through the tags. In cases of predictable alternations (e.g. causative-

inchoative alternation) the tagging on particular lexical items would miss

an important generalization, and therefore the existence of zero morphol-

ogy(for

ν

⁾^needs ^to ^be ^assumed. ^The^question ^is ^not ^restricted^to ^so-called

10

This,obviously,isnottantamounttogivingupthedichotomybetweenexternaland

internalargument intheSubjectvsObjectsense. InthepresentworkIamonlypreoc-

(33)

substantive(open-class)items,butconcernsalsothenatureofso-calledport-

manteaumorphemes.

OnepossibleapproachtotheseisalongthelinesofGiorgiandPianesi(1997),

where a fusion of particular syntactic projections is taken as a lexical pa-

rameter(cf. alsoBobaljik(1995)and Pylkkänen(2002)). The`staticfusion'

approachà laGiorgi andPianesi, however,facesa challengewheneverthere

occurs some othermaterial, whichsplits thetwo purportedlyfusedheads.

A dierent execution is presented in Borer (2005). Functional heads are

taken to be open values to which rangeneedsto beassigned bya varietyof

means. Oneofthesemeansinvolvesrangeassignmentbyaheadfeature(e.g.

English<pst> Tense feature), which requiressuccessive head movement of

thelexicalhead tosupportit. Borertakesthishead featureto beinasense

a-morphous(thuscomplyingwiththeWord andParadigm viewofmorphol-

ogy). Consequently, no zero-morphemes are assumed. Instead The Great

PhonologicalDispenserneedstoreturnanoutputfortherelevant structure.

Inthat kind of systemportmanteau morphemes are conceived of as double

range assigners. Consider e.g. English indenite article a. Borer assumes

the existence of innate universal functional hierarchy - an assumption that

Iwill share. Her nominal structure universally contains a Classier Phrase

(CLP) dominatedbytheQuantity Phrase(#P).

(44)

[

^#P ^<e>

_d

^[CLP ^<e>

DIV

^]]

Open value <e>

DIV

^can ^be ^assigned ^range ^by^the ^head ^feature <Plural>.

Open value <e>

d

^is ^assigned ^range ^by ^cardinals. ^Since singularity is not overtly marked on listemes in English, yet they are not interpreted as de-

faultmass, somethingmustassignrangeto <e>

DIV

^. ^Borer ^takes^it^to^be^a.

Therefore,English a is taken to be a double range assigner: a divider,and

acounter -an example ofopen valuefusion.

Although theresults thatI aimto achieve inthe present work,could prob-

ably also be restated in Borerian system, due to a controversial status of

head movement in linguistic theory (cf. e.g. Matushansky (to appear)), I

will adopt a dierent view of the lexicon. Instead, my way of conceiving

ofthe lexicon isclosest to the one presented inWilliams (2003)and Starke

(nanosyntax seminar), i.e. I assume that lexical items realize/ spell out

subsequences of the universal functional hierarchy. This is the idea that I

will often refer to under the rubric of lavish insertion. Lexical items are

hierarchically structuredbundles offeatures. How bigachunkof functional

hierarchyislexicalizedbyaparticularitemisessentiallyalexicalaccidentof

aparticularlanguage. Furthermore,buildingontheideas byMichalStarke,

Iwillalsoassumethatincertaincircumstancesalexicalitemx canspellout

a subset of the features that it lexicalizes. The relevant circumstances are

(34)

to x. Suppose the relevant lexicalitems arespecied inthelexiconas:

(45) a. x: [F

6

^... ^F

4

^]

b. y: [F

4

^... ^F

2

^]

In principle, there are two situations possible. I will refer to them as `up-

squeezing'(cf. (46))and `down-squeezing' (cf. (47)) respectively:

(46) up-squeezing

F

6

F

5 x

z

| } {

F

4

F

3

F

2 y

z

| } {

(47) down-squeezing

F

6

F

5

F

4 x

z

| } {

F

3

F

2 y

z

| } {

Inprinciple,bothofthescenariosarepossible. Theparticular executionwill

have to depend on the way lexicalization algorithm works. E.g.,ifinsertion

is early, in a bottom-up derivation, only one head will be accessible, and

thereforethe most specic lexical items should be selected. Iflexical inser-

tionislate,however, all thesyntacticheadsintherelevant subsequence are

already accessible, and hence there is a choice with respect to how specic

items arechosen. Assuming that insertion algorithm istop-down, however,

would result in the prediction that only down-squeezing is allowed. Since

thecases discussed inthis dissertation will infact involve both scenarios in

(46),aswellas(47),Iwillassumethatlateinsertionhappensnoearlierthan

when the highest projection in the thematic domain is merged (i.e. initP)

andtheinsertionalgorithmprocessesthewholetreesimultaneously. Inother

words, incases where two items inthe Numerationhave overlapping lexical

specications,asx andy above,allthe possible insertion variantsaretried,

Radical decomposition and argument structure

Patrycja Jab lo´ nska

A Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Department of Linguistics

University of Tromsø

Θ

seq

high

low

↓

∅

↓

< y >

↓ ↓

< y >

pst.3pl.non−vir

pst.3pl.nonvir

pst.3pl.non − vir

pst.3sg.m

ACC

ACC

Q

Q

Q

N OM.neut

pst.3sg.neut

ν

ν

ν

ν

∃

1

2

1

2

1 →

2

→ def

1

∃

1

2

1

2

2 →

1

→ def

1

ν

ν

ν

ν

ν

i

i

ν i

i

ν

N OM

pst.3sg.m

ACC

ACC

N OM

pst.3sg.m

ACC

ACC

N OM

pst.3sg.m

ACC

ACC

N OM

pst.3sg.m

ACC

ACC

N OM

pst.3sg.m

ACC

pr.prt

ν

ν

_i

_i

ν _i

_i

ν _i

_i

_i

ν _n

ν ⁰

ν _{N EU T}

ν ₁

ν _{N EU T}

_d