National Lrbraryof Canada

Bibliothèque nationaledu Canada

Acquisitions and Direçr,on des acquisitions etBibliographie services Branch des services bibliographiques

395 WellmQlon Streel 395. ru~ WclhnglonOttawa. Onlano Ottawa (Onlano)K1AQN-1 K1AON-1 ' ..... ',,',' ~ ..~". ,,",.. ,~, "

(\,, r•.',' ,\.,~ •• , .,..,.".... ','

NOTICE

The quality of this microform isheavily dependent upon thequality of the original thesissubmitted for microfilming.Every effort has been made toensure the highest quality ofreproduction possible.

If pages are missing, contact theuniversity which granted thedegree.

Sorne pages may have indistinctprint especially :f the originalpages were typed with a poortypewriter ribbon or if theuniversity sent us an inferiorphotocopy.

Reproduction in full or in part ofthis microform is governed bythe Canadian Copyright Act,R.S.C. 1970, c. C-30, andsubsequent amendments.

Canada

AVIS

La qualité de cette microformedépend grandement de la qualitéde la thèse soumise aumicrofilmage. Nous avons toutfait pour assurer une qualitésupérieure de reproduction.

S'il manque des pages, veuillezcommuniquer avec l'universitéqui a conféré le grade.

La qualité d'impression decertaines pages peut laisser àdésirer, surtout si les pagesoriginales ont étédactylographiées à l'aide d'unruban usé ou si l'université nousa fait parvenir une photocopie dequalité inférieure.

La reproduction, même partielle,de cette microforme est soumiseà la Loi canadienne sur le droitd'auteur, SRC 1970, c. C-30, etses amendements subséquents.

•

•

•

THE SYNTAX OF SENTENTIAL NEGATION:

Interactions with Case, Agreement, and(In)definiteness

by

Leslie J. de Freitas

A Thesis submitted to the Faculty of Graduate Studies and Researchin partial fulfillment cf the requirements of the degree of

Doctor of Philosophy

Department of LinguisticsMcGiII UniversityMontréal, Québec

JULY1993

© Leslie J. de Freitas 1993

."". National Libraryof Canada

Bibliothèque nationaledu Canada

Acquisitions and Direction des acqUisitions etBibliographie Services Branch des services bibliographiques

395 WcllJnglon Slreel 395. rue WellinQtonQnawa.Onlano Onaw;) (Onl~nè)K1AON4 K1AON4

The author has granted anirrevocable non-exclusive licenceallowing the National Library ofCanada to reproduce, loan,distribute. or sell copies ofhisjher thesis by any means andin any form or format, makingthis thesis available to interestedpersons.

The author retains ownership ofthe copyright in hisjher thesis.Neither the thesis nor substantialextracts from it may be printed orotherwise reproduced withouthisjher permission.

L'auteur a accordé une licenceirrévocable et non exclusivepermettant à la Bibliothèquenationale du Canada dereproduire, prêter, distribuer ouvendre des copies de sa thèsede quelque manière et sousquelque forme que ce soit pourmettre des exemplaires de cettethèse à la disposition despersonnes intéressées.

L'auteur conserve la propriété dudroit d'auteur qui protège sathèse. Ni la thèse ni des extraitssubstantiels de celle-ci nedoivent être imprimés ouautrement reproduits sans sonautorisation.

ISBN 0-315-94606-7

Canada

(short title)

The Syntax of Sentential Negation

•ABsTRACT

11ùs thesis undertakes to refine our understanding cf the syntactic properties ofsentential negation. The proposed analyses operate at the juncture of recentinnovations to Case, Agreement, and X-bar theories, within a Govemment andBinding framework. Case is checked in a Specifier/Head configurationwhenever possible, and agreement is analyzed as the reflex of a Case-checlàngoperation at S-structure. The proposaI that the inventory of functional categoriesavailable in Universal Grammar includes a Negation Phrase (NegP) is adoptedas a point of departure.

In the context of this investigation, certain syntactic properties are attributed tothe head and specifier of NegP. It is proposed that the specifier of NegPprovides an A-position in which NPs may be Case-checked. S-structureCase-checking is reflected in agreement marking on the negative head. Evidencefor LF Case-checking in this position is derived from the Case-licensing ofdirect objects in negated clauses in Colloquial Welsh and Russian. Definitenesseffects are analyzed as due to constraints on an additional Case-licensing optionrequired if negation blocks Case assignment under govemment. Variations inagreement patterns in affirmative and negative relative clauses in Literary andColloquial Welsh are attributed to the barrier status of the head of NegP.

•

•

•

RÉSUMÉ

Cette thèse entreprend de raffiner notre compréhension des propriétéssyntaxiques de la négation phrastique. Les analyses qui y sont proposées sesituent à la jonction d'innovations récentes dans les théories du cas, de l'accord,et de X-barre, dans le cadre de la théorie du Gouvernement et du Liage. Le casest apparié à une configuration de spécifieur-tête chaque fois que cetappariement est possible, et l'accord est analysé comme l'effet d'une opérationde vérification casuelle en structure-S. Cette thèse adopte comme point de départl'idée que l'inventaire des catégories fonctionnelles disponible dans laGrammaire Universelle inclut un Syntagme de Négation (NegP).

Au cours de cette étude, certaines propriétés syntaxiques seront attribuées à latête et au spécifieur de NegP. TI sera avancé que le spécifieur de NegP contientune position-A dans laquelle le cas des NP peut ètre vérifié. La vérificationcasuelle en structure-S se manifeste par une marque d'accord sur la tête négatif.Des arguments en faveur de la vérification casuelle en LF (pour forme logique)seront basés sur la légitimation casuelle des objets directs des propositions niéesen gallois familier et en russe. Les effets propres aux noms définis serontattribués à des contraintes sur une option supplémentaire de légitimation casuellemise en oeuvre lorsque la négation bloque l'assignation du cas dans une relationde gouvernement. Les différences de marques d'accord que l'on trouve dans lespropositions subordonnées affirmatives et négatives en gallois littéraire etfamilier seront attribués au statut de barrière de la tête de NegP.

•

•

•

ACKNOWLEDGEMENTSJ would first Iike to thank my thesis supervisor. Mark Baker. tirst ti.)r his

important contributions to this work in the form of advice and criticism. and equallyfor pmviding momentum when things bogged down. and for occasionally remindingme what my thesis was abo:lt when J wandered too far down other paths. Mark sets astandard in his own work which J could not hope to achieve. but this thesis improvedgreatly in the anempt. J have also benefited a great deal from conversations with peoplewho have suggested revisions to the analysis. helped me with the data. or just listenedpatiently to my ramblings. J would Eke to take this opportunity to thank SoniaAlberton. John Eobaljik. Maria Babyonysllev. Orly Cohen. Nigel Duftield. BrendanGillon, Steve Harlow, Paul Law. Alan Libert, Carol Neidle. Elizabeth Pearce. LjiljanaProgovaç, Alain Rouveret. Benjamin Shaer. Ur Shlonsky, Knut Tarald Taraldsen. LisaTravis, Hubert Trockenbrod. Daniel Valois. and Raffaella Zananini. Special thanks inthis regard go to M:l:ire Noonan. with whom 1 developed the basis for the analysis ofrelativization in Literary Welsh given in Chapter 2. Working with M:l:ire was bathenlightening and fun, and J look forward to future col13borations! J follow the McGilItradition of being endlessly grateful to Zofia Laubitz. for editing beyond the cali ofduty. Thanks also to Jean-François Prunet for helping with the French abstracto 1 amalso very grateful to my Welsh informants, John Williams and Ewen Edwards. and toTatyana Petrova and Evgyeni Tomov, for their help with the Russian data.

The people J want to thank here do not fall neatly into categories; many havebeen important to me not just as colleagues but as friends. For always being willing todiscuss anything under the sun, even occasionally linguistics, J would like to thankMengistu Amberber, Cathy Burns, Ted Caldwell, Mark Campana, Alan Juffs. ZOiiaLaubitz, Alan Libert, John Mathews, Alison Mitchell, and Milire Noonan. Last but fart'rom least, l'd like to thank Anna Maclachlan and Benjamin Shaer, for being wonderfulfriends, occasional sparring partners, and always supportive. Profs Nicole Domingueand G1yne Piggon also deserve special thanks for their kindness and encouragementover the years.

FinaIly, 1 would like to thank my wonderful sisters, Jennifer and Elizabeth, andmy 'extended' family; Andrew, Rodolfo, Mariela, and Thomas and Hilda G., for theirlove and encouragement.

Thomas, What cao 1 say? Words are inadequate, but perhaps appropriate in thecontext of a linguistics dissertation. Thank you for being strong when 1 wasn't, forheing encouraging when 1 felt overwhelmed, and for laughing at me, or with me, asneeded, and, in short, for everything. 1couldn't have done it without you.

The research presented here was carried out with the financiaI assistance of aSocial Sciences and Humanities Research Council of Canada Doctoral Fellowship, forwhich 1 am very gratefuI. 1 would aIso like to express my gratitude to McGilI GraduateFaculty and the McGilI Post-Graduate Students' Society for travel grants whichpermined me to present sOrne of these ideas at the 27th Regional Meeting of theChicago Linguistics Society in 1991 and at the Colloque Internationale de la Négationat l'Université de Paris X (Nanterre) in 1992.

This thesis is dedicated, with love, to my parents.

•

•

•

Inspector Tiger: This house is surrounded. l'm afraid 1 must not askanyone to leave the room. No, 1 must ask nobody...no, 1 must askeverybody to .. .1 must not ask anyone to leave the room. No one mustbe asked by me to leave the room. No, no one must ask the room toleave. 1...I...ask the room shaH by someone be left. Not. Ask nobodythe room somebody leave shaH 1. ShaH 1 leave the room? Everyone .must leave the room...as it is...with them in it. Phew. Understand?

Agatha Christie Sketch. The Complete Monty Python's Flying Circus:AIl the Words Vo/wne 1.

• TABLE OF CONTENTS

1. CHAP1"ER 1: INTRODUCTION ••••••••••••••••••••••••••••••••••••••••••••••••••••.••••••••••••••••••••• 11.1. Introductioo 11.2. A Note on the Fnunework and an Introductioo te the D.ta ......................•......... 31.3. The Syntaetic Treatment of Sentential Negation: Recent Proposais ...•......•............ 7

1.3.1. Pollock (1989): Verb-Raising and Agreement •....•...............•................. 91.3.2. Chomsky (1989): Neg.tion and Syntaetic Affix.tion ...••......•................... 151.3.3. Ri.zzi (1990): Negation as an A·-Spccifier.....••...........•........•................ 171.3.4. 0uha11. (1990): The Neg Par.uneter.....••............................••.............. 221.3.5. 1le11etti (1990): Negation as a Clitic in ltalian .......••..••......•........••...••.... 231.3.6. Lalca (1990): Neg as the Head of Sigma·Phrase ••.........••......•................• 261.3.7. Zanuttini (1991): Negation and Tense •••••••••••••••••••••••••••••••••••••••••••••••• 27

1.4. Summary 301.5. Endnotes te Cbapter 1 32

2. CHAP1"ER 2: NEGATION AND RELATIVlZATION lN LITERARY WELSH••••••••••••••• 342.1. Introduction •••••..••••••••....•••••....•••.....••......••••...........•..••......•.......•.....•... 342.2. Initial Paraliigm •......•..••••••..•.••••.....••....•••••.....••.••......••.....•....••••.....••.... 34

2.2.1. Two Relativization Strategies.•.••..•.•••.....••••••••..•....•...••......•....••.....•.• 342.2.2. Negated Relative Clauses .••••••••.••••...••••.•.......•.•.•...••.....••..........••.... 37

2.3. Some Initial Facts about Welsh Syntax •.••...•••••.........••.......•....••....••••..•••...... 382.3.1. Word Order and Constituent Structure in Welsh 382.3.2. Agreement and Null Arguments in WeIsh 412.3.3. Wh-<iuestions and Cleft Constructions 452.3.4. Picd-Piping and Complementizer Selection 482.3.5. "True" Island Constraints 492.3.6. Summary••••••••••••••••••••••••••••••••••••••••••••••••••••••.•••••••••••••••••••••••••••• 50

2.4. Previous Analyses ofRelativization in Welsh: The Emply CategoryPrinciple ••••.•••••••••••••••••••••••••••••••••••••.••••.••••••••.....•••.•.......••.....••......••.....••... 54

2.4.1. Introduction •.••••••••..•••••..••••••.••••....••••••••••..••..•••..••••...••••••..••••..... 542.4.2. Harlow (1981): The Emply Category Principle 552.4.3. Sadler (1988): A Coojoincd ECP ••••.••..••••••••••.••••.••...••••...•••••.•.•••...•.. 632.4.4. Rouveret (1990): A Barriers Approacb 672.4.5. Summary•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.• 74

2.5. A New Proposai ••••••••••••••.••••...•••••..•••...•••••••.....••••••••.••.....•.•.....•••..•.•.••. 762.5.1. Introduction ••••••••.•••••••••••..•.•••.•••••••.•.•..••••••....•••...••••.••••••..•••••.... 762.5.2. Tbeoretical FllIIIIOWOrk••••••••••••••••••••.••••••••••.•••••••.••••••.••••....•••..•••••• TI

2.6. Agreement in Relative Clauses•••••.•••••••••••...•••••••••.•.••••••.••••••••••.•••••...•••••... 882.6.1. Background•••••••••••••••••••••••••.••...••••.•••...•.•••..••.•••..••••.•••••••••••....••• 882.6.2. Deriving the Distinct COmplementizers••••••••••••••...••••.•••••..•••••..••..••••••• 912.6.3. Deriving the Indirect Pattem 932.6.4. Deriving the Direct Pattem 972.6.5. Wh-Questions and C1efls: The "Indirect" Complemeotizer withthe "Direct" Agn,ement Pattem 1002.6.6. Long-Distance Relativization lOI

2.7. Case-Licensing in Spec/NegP in Matrix and Relative Clauses 1052.7.1. Negatcd SubjectRelatives lOS2.7.2. Negatcd Object Relatives: No Agreement and a ResumptiveProooun StIategy 109

i

..

•

•

2.7.3. Negated Object Relatives: Agreement on Negation and aResumptive Pronoun Stralegy III2.7.4. On the Autonomy of Neg as a Case-Li=sing Head 1152.7.5. Other Pre-Sentential Markers in Welsh 1232.7.6. Problems with Long A-cllains 1292.7.7. Conclusion 130

2.8. EndnoleS to Chapter 2 131

3. CHAPTER 3: SUBJECT-aBJECT ASYMMETRIES AND DEFINITENESSEFFECTS IN THE GENITIVE OF NEGATION IN RUSSIAN•••••••••••••••••••••••••••••••••••••••••• 138

3.1. Introduction 1383.2. Initial Paradigm 139

3.2.1. Intransitive Sentences: D-Structure Grammatical Relations andthe Geniti"e of Negation 1403.2.2. Optionality and (In)definiteness Effecls 1423.2.3. Summary 143

3.3. A Previous Analysis: Pesetsky (1982) 1443.3.1. C-seIection ofNPs vs. QPs 1443.3.2. Case-Assi:;nment ••••••••....••••...•.••..••••.•••..••.....•••.....••••.•••••.••...••••.•• 1463.3.3. Optiona\ity and Interpretation of Genitive NPs 1483.3.4. Problems with the Treatment 149

3.4. Part 1: Structural Restrictions on the Genitive of Nel,'lltion 1513.4.1. Introduction 1513.4.2. Russian Phnlse SlnL:ture 1533.4.3. Position of NegP 1533.4.4. Case-Assignment under Go\=ent in Nep!ed Sentences 1583.4.5. Structural Constnùnls on Genitive of Negation 1663.4.6. Supporting Evidence 1763.4.7. Conclusion to Part 1 180

3.5. Part 2: Definiteness Effecls Associated with the Genitive of Negation 1823.5.1. Introduction 1823.5.2. Dealing with the Concept of (In)definiteness 1823.5.3. Definiteness and the Accusative-Genilive Alternalion 1863.5.4. The Properties of FP 1883.5.5. A Problem: Obligatory Genitive ofNegalion 1913.5.6. Indefinite NPs as Vsriables. Negation as an ()perator 193

3.6. Summary 1983.7. Endnoles to Chapter 3 198

4. CHAPI'ER 4: CASE-LlCENSING, 'EMBRACING' NEGATION, ANDDEFINITENESS IN COLLOQUIAL WELSH••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 202

4.1. Introduction 2024.1.1. Word Order in Colloquial Welsh 203

4.2. Sentential Negation in Colloquial Welsh 2064.2.1. Two Negative madcers••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 2064.2.2. Sentential Negation vs. constituent Negation 2074.2.3. Preposition Insertion in Negated Sentences 2084.2.4. The Status of the Pte- and Post-Verllal Negative Marlcers 210

4.3. Sentential Negation and Tense in Colloquial Welsh ••••••••••••••••••••••••••••••••••••••••• 2114.3.1. Infinitivals ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 2124.3.2. Past Participles••.•••••••••••••••.••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 2134.3.3. Absolute Constructions•••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 215

il

,•

•

•

4.3.4. Imperatives ........•....................................................................... 2164.3.5. COnclusioQ 218

4.4. Sentential NegatioQ and Case in CW 2194.4.1. Negate<! Relative Clauses in ew 2194.4.2. Subjoct-object Asymmetries and Case in Negated Clauses in ew 2244.4.3. Medial Negation as a Banier te Ca....-Assignmeat 2274.4.4. Case-Licensing of VSO Direct Objects 2304.4.5. (ln)definiteness Revisite<!: Case-licensing in SpecINegP '" 2314.4.6. Variable Word-order in Pembrokeshire Welsh 235

4.5. Summary...•.••....••.•......•.....•.............•.•..•............................................. 2374.6. Endnotes te Cbapter 4 ......••....•...••...............•.......................................... 238

5. CONCLUSION ••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.••••• 243

APPENDIX: NEGATION AS HEAD AND SPECIFIER OF NEGP 247

REFERENCES..........••••••••.••••..••••••.•••••••••••••.•••••.••...••••...••...••.•••....•.•.•••...••...•.•..••• 252

ili

•

•

•

1. CHAPTER 1: INTRODUCTION

1.1. INTRODUCfION

In this thesis, 1 address the problem of defining the syntactic nature of

sentential negation. Two of the major disputes concerning this issue are its

categorial status in X-bar theory and where it is generated relative to other

components of a sentence. In an attempt to elucidate these and other questions, 1

examine syntactic processes which appear to be affected by sentential negation

in Standard Literary Welsh, Colloquial Welsh, and Russian.

Languages appear to use one of three distinct devices for marlàng

sentential negation: negative verbs, negative particles, and negative derivational

morphemes. Sometimes the use of one of these devices is accompanied by a

secondary modifiQltion in the sentence being negated. Payne (1985, section 2.4)

mentions (i) changes in word order, (ü) changes in tone, (ili) neutralizations of

tense distinctions, (iv) supporting verbs, and (v) changes in noun Case, all

trlggered by negation. This thesis deals in large part with a secondary

modification not mentioned by Payne; changes in agreement patterns.

In this study, 1 focus on changes in Case, agreement, and word order

trlggered by sentential negation in Welsh and Russian. Under the theoretical

assumption that word order and agreement changes can be related to Case, 1

propose an analysis of sentential negation which accounts for bath ways in

which it can interfere with and ways in which it can facilitate case-licensing. 1

adopt the proposal that the inventory of functional projections made available

cross-linguistically includes a Negation Phrase (NegP). Facilitation of case­

licensing is related to the additional specifier position generated. Interference is

1

• related to the barrier that negation may create for verb-raising or for Case

assignment under government. However, where negation is affixal. its blocking

effect can be neutralized by syntactic incorporation derived by head-to-head

movement.

In the first chapter, a brief introduction to sorne recent proposais

regarding the syntactic nature of sentential negation is presented. This discussion

is intended te give the reader an idea of both the kinds of proposais that have

been made and the kind of argumentation used.

In the second chapter, 1 introduce the problem of negation triggering

changes in agreement and relative clause formation in Welsh, and review

previous analyses of these facts. In the context of an analysis of the role played

by sentential negation, revisions are proposed to the treatment of wh-movement

and agreement. It is argued that it is not necessary to maintain the traditional

distinction between direct and indirect strategies of relativization in that

language. Rather, the distinct complementizers and agreement patterns exhibito:d

with relativization out of different positions within the sentence reflect the S­

structure Case-licensing configurations; configurations which are altered by the

presence of sentential negation. The phenomenon of agreement surfacing on the

negative marker is analyzed as reflecting a Case-checking procedure in a

Specifier-head configuration established at S-structure, made possible by the

[+Case] specification of the head of NegP.

In the third chapter, the phenomenon of Genitive Case triggered by

sentential negation in Russian is addressed. 1 argue that this can be accounted

for under the assumption that the specifier of sentential negation provides an

extra position for LF Case-checking. Evidence is presented te support the claim

2

• that the head of NegP can Casc-license an NP independcntly of the [Case]

specification of the verb. This analysis allows us to derive the lack of verbal

agreement with NPs in the Genitive of Negation, the unmarked word order of

the Genitive-marked NPs in these constructions, and the correlations between

Case-marking and (in)definiteness in negated clauses.

In the fourth chapter, 1 extend the c!..i"lS concerning the interaction of

negation and Case by considering the syntax of sentential negation in Colloquial

Welsh (CW). The first part of this chapter addresses the relationship between

sentential negation and Tense; 1 present arguments to support the c1aim made by

Zanuttini (1991) that NegP cannot be generated in a clause that does not include

a Tense Phrase (TP). 1 then argue that the interference of negation with Case­

licensing of the direct object follows from the presence of the clause-medial

negative marker, which blocks Case-assignment under government te an NP

lower in the structure. Based on data from Pembrokeshire Welsh, 1 will then

argue that, as in Russian, the specifier of NegP provides an additional Case­

licensing position for NPs.

1.2. A NOTE ON'IBE F'RAMEwORK AND AN INTRODuCTION 1'0 THE DATA

In the framework which we will, for lack of a better terrn, refer te as

Chomskian Syntax, it is not uncommon te come across an introductery

statement in a paper to the eftect that "sorne standard version of

GovernmentlBinding theory will he assumed". As frustrating as it may be for

readers, there are certain understandab1e motivations for such a statement.

Generative linguistics is still very much in evo1ution. Many aspects of the

theory, bath peripheral and fundamental, are subject te serious inquiry and

reanalysis. And, commonly, tinkering may he carried out in one area that need

3

• not have any repercussions on other aspects of the gr4Jllmar. It is possible to

concentrate on a small area in order to determine whether it may eventually shed

light on other questions. Except where more recent innovations are adopted (i.e.

regarding the theory of Case), what follows will assume a version of syntacùc

theory as described in Chomsky (1986a).

The first paradigm to be discussed in this thesis is taken from Literary

Welsh (LW). LW is traditionally described as having two distinct relativizaùon

strategies called the direct and indirect strategies, each of which is associated

with relativizing out of certain positions within the sentence. In the direct

strategy, the particle a precedes the relaùve clause. As shown in (1)-(2), an

verb does not exluoit agreement with the relativized NP.

instance of subj~ ~ativization, there is a gap in the relativization site, and the

• (1) y dynioni a ddarllenodd [eli y llyfrthe men Fr read-3sg the bookthe men tJuu read the book

[Harlow 1981:237]

(2) *y dynioni a ddarllenasant [eli y llyfrthe men Fr read-3pl the book

In the indirect strategy there is still a gap in the relativization site which,

as in the direct strategy, cannot be filled by a pronominal, but we find

agreement appearing on the head goveming the relativization site. In this case,

the particle that precedes the relative clause is y(r).\ This is illustrated in (3)-(4),

where a periphrastic direct object is relativized.

4

•

•

•

(3) Yllongi y gwnaeth Sion ei werthu [e]ithe boat PT did John 3sgm-sellthe boat thaI John soM

[Harlow 1981:236]

(4) .y llongi y gwnaeth Sion gwerthu [e]ithe boat PT did John sen

An interesting effect is induced by negating a relative clause. In cases

which nonnally require the direct strategy, e.g. subject re1ativization, negating

the relative clause appears to force the indirect pattern; agreement with the

relativized NP subject is obligatory. This is shown in (5)-(6).

(5) Ydynioni na ddarllenasant [e]i y llyfrthe men Negread-3pl-past the bookthe men thal didn 'r reat! the book

[Harlow 1981 :237]

(6) *y dynioni na ddarllenodd [e]i y llyfrthe men Neg read-past the book

The second paradigm to be discussed cornes from Russian, where

negating a clause appears to alter the Case-assigning properties of the sentences.

If we take as an example a regular transitive verb, the direct object nonnaily

surfaces with Accusative Case. When the sentence is negative, the direct object

may surface either with Accusative or Genitive Case. This is illustrated in

(7)_(9).2

(7) ja vizu kniguI-NOM see book-ACC1see the book

s

•

•

(8) ja ne vizu knigiI-NOM Neg see book-GEN

(9) ja ne vizu kniguI-NOM Neg see book-ACC

[Neidle 1988:34]

An interesting wrinkle involves the interpretation of thesc sentences; if

an NP that could surface with Genitive Case appears instead with Accusative

Case, that NP tends to be interpreted as definite. In this way, an ambiguity with

respect 10 definiteness that is present in affmnative sentences in Russian can be

removed in a negated sentence, as shown in (10)-(11).

(10) ja ne vizu knigiI-NOM Neg see book-GEN1don 'c see aiche book.

(11) ja ne vizu kniguI-NOM Neg see book-ACC1 don'c see the book.

[Neidle 1988:34]

The final set of data that will be analyzed cornes from negation in

Colloquial Welsh, which differs in certain interesting ways from Literary

Welsh. First, it requires an additional negative marker, similar 10 French pas in

distribution and placement, and second, a 1exkal item appears preceding the

direct object in negated sentences. This item is not found in the corresponding

LW sentences, as shown in (13).

6

• (12) Phalodd John ddim *(0) 'r ardd.Neg-dug-3sg John Neg PT the gardenJohn didn 't dig the gartien.

[Jones & Thomas 1977:323]

(13) Ni phalodd John yr ardd.Neg dug-3sg John the gardenJohn didn't dig the gartien.

This set of faets interaets with the Russian data in that the preposition

does not surface in CW when the direct object is indefinite.

(14) Rosim i ddim lliw ynddo ariodNeg-put-1sg 1Neg color in-3sgm ever1didn '( put coloring in it ever.

[Awbery 1990:6)

These faets show sentential negation interaeting with Case and agreement

and suggest that this interaction is sensitive te the notion of (in)definiteness.

This work is offered as a contribution both te work on the syntactie

nature of sentential negation and to the analyses of the phenomena affected by

negation in the languages under consideration. In the context of a proposaI that

sentential negation can project its own phrase, positive revisions te the analyses

of the parachgms under discussion in WeIsh and Russian are facilitated.

1.3. THE SYNTAC11C TREATMENT OF SENTENTIAL NEGAll0N: RECENTPROPOSAIS

The syntactic status of sentential negation is currently the focus of much

attention. Until very recently, X-bar theory was questioned primarily along

certain well-defined lines; specifically, there was a debate as te whether X-bar

7

• should or ~hould not be restricted to binary branching, whether all languages

instantiate the same projections, and whether all languages reflect the same

hie..-archical ordering of projections. However, the inventory of functional heads

which could project phrases was generally held to he constant: Infl(ection), and

Comp(lementizer). When Abney (1987) proposed a new functional category,

D(eterminer) Phrase, this initiated the most recent wave of proposed revisions to

X-Bar Theory: debates as to the number and kind of functional categories made

available by Universal Grammar (UG).

In keeping with Baker's influential proposai that word-formation can

take place in the syntax (syntactic iru:orporation, Baker (1985), (1988», new

functional projections headed by affixal morphemes have been proposed to

account for morpheme ordering. Since affixation, under this treatment, can he

derived via head-to-head movement, these morphemes took on head status. The

order of morphemes within a given lexical item provided one motivation for

ge.'1erating the new phrases in a particular relation to one other.3 In other cases,

new functional projections have been proposed in order to provide a new

landing site for movement Qohnson (1990), Mahajan (1990».

Although one might expect widespread repercussions in other areas due

to the introduction of these new phrases, this is not necessarily the outcome.

While the introduction of a new head can potentially introduce an additional

barrier for XP-movement, the specifier position projected by the head can

provide a compensating escape hateh. Furthermore, if a lower head incorporates

into the head of one of these new phrases and carries it along as it is raised

further, the phrase will not constitute a barrier for antecedent govemment by

s

•

•

•

that head. This follows from the definition of minimality in the ECP, combined

with the fact that the potential intervening heads are traces of the mised head.

One of the most influential recent proposais regarding functional

categories is due to Pollock (1989): that IP, which previously illCluded both

tense and agreement, should be separated into two separate functional

categories, Tense Phrase (TP) and Agreement Phrase (AgrP). In the sanIe

article, he maintained, following Kitagawa (1986), that X-bar should include an

additional functional projection, Negation Phrase (NegP).

Of primary concem to this thesis is this proposed NegP. Below, 1 outline

proposais regarding this projection made by Pollock (1989), Chomsky (1989),

Belletti (1990), Ouhalla (1990) and Zanuttini (1991). 1 will also outline Laka's

(1990) proposai that Neg is one potential head of the functional projection

Sigma-Phrase, and Rizzi's (1990) analysis of sentential negation, in which he

proposes an analysis of Ross's Inner Island Constraints (Ross 1984) induced by

negation without involàng a NegP. In more recent work, however, he has

adopted a NegP; in fact, it plays an important role in his analysis of inversion

triggered by affective variables in English (Rizzi 1991).

The papers discussed below address more than just the role of sentential

negation. However, 1 will limit the discussion here to a few core facts that will

serve to illustrate the main claims made by these researchers with respect 10 the

syntactic status of sentential negation.

1.3.1. PoLWCK (1989): VERB-RAlSING AND AGREEMENT

Pollock proposes a more highly articulated structure for IP than had

previously been assumed, in which the formerly dual-headed phrase is separated

into a Tense Phrase (TP) and Agreement Phrase (AgrP). He also proposes that

9

• negation heads its own projection, Negation Phrase (NegP). These projections

are generated in the relative order shown in (15).

(15)

Pollock's article addresses certain surface differences between French

and English, including the placement of adverbs with respect to verbs and the

required insertion of do in English negative and interrogative sentences, as

illustrated by the sentences below.4

(16) Jean n'aime pas Marie.*John likes not Mary.John does not like Mary.

(17) Aime-t-il Marie'?*Likes he Mary'?

(18) *Jean souvent embrasse Marie.John often kisses Mary.

(19) Jean embrasse souvent Marie.*John kisses often Mary.

Pollock claims that the differences illustrated above relate in large part te

verb-movement in these languages, which is sensitive te the transparency or

10

• opaciry of Agreement (Agr). A raised verb can only assign its theta role from

inside a transparent functional category.S It is then proposed that Agr in French

is transparent, and therefore alIows a theta-assigning verb to raise to it. In

English, however, Agr is opaque, 50 V-raising is limited to non-theta-assigning

verbs: have, and be.

The claim that only non-theta-assigning verbs can raise to Agr in English

accounts for the relative order of verbs and adverbs. On the assumption that

these adverbs are generated adjoined to VP, theta-assigning verbs such as làss

should occur to the right of the adverb, as in (18), above. Non-theta-assigning

verbs raise to Agr, and appear to the left of adverbs.

1.3.1.1. NEGAnON AND DO-SUPPORT IN ENGUSB

An additional difference between French and English addressed by

Pollock is that the phenomenon of do-suPPOTt (the occurrence of a dummy verb

carrying tense features) is found only in the latter.

(20) Jean n'a pas chanté.John did not sing.-John not sang.

Pollock's analysis treats V-to-T raising as obligatory. He derives this by

the requirement that [+tense], identified as an operator, must bind a variable,

which is provided by the trace of V-movement. According to Pollock, English

sentences which do not include an auxiIiary have or he are confronted with the

problem of providing such a variable. He proposes that this requirement can be

met in affirmative sentences by a phonologically null verb do, which raises to T.

11

•

•

•

In negative sentences, a potential barrier. NegP. intervenes between T

and Agr, which induces an Empty Category Principle (ECP) violation for

raising at Logical Form (LF) after syntactic affix-Iowering. Only a lexical verb

can L-mark NegP, neutra1izing its barrierhood. sc in a negative sentence with a

null do the trace of V would not satisfy the ECP. Instead, negation forces do to

have lexical content.

In essence, then, sentential negation projects a functional phrase NegP,

the head of which constitutes a potential barrier for antecedent govemment of a

trace by a raised head. This barrier can be neutra1ized by L-marking, but L­

marking can only be effected through a lexical, phonologically 'rea!' verb, not

by the phonologically null counterpart of do which is found in ordinary

affirmative sentences.

Pollock crucially distinguishes the negative marker not from negative

adverbs such as rarely or never. He defends the special status attributed to not in

English as the head of a maximal projection, NegP, by noting that other

(negative) adverbs do not interfere with LF V-raising. Thus, while sentence (21)

below is ungrammatical without do, the sentences in (22)-(24) with a negative

adverb do not require it.

(21) *John not left the children a1one.

(22) John did not leave the children a1one.

(23) John never left the children a1one.

12

•

•

(24) John rarely left the children alone.

Thus, the principle evidence for an analysis of a negative elemen~ as ti:e

head of NegP is its interaction with head-movement.

1.3.1.2. 'EMBRAONG' NEGATION AND FIN1TE VS. NON FIN1TEVERBS

Sentential negation in French requires two markers, ne and pas.6 This

has been referred to as embracing negation, because the negative markers appear

on either side of the tensed verb. Pollock analyzes these as the head and

specifier of NegP, respectively. Since pas is in the specifier of NegP (therefore

to the left of ne, on the assumption that specifiers are ta the left in French

phrase structure), Pollock must aceount for the fuct that this order of negative

markers is never instantiated in the data. He accounts for this by attributing

clitic status to the head ne; as a clitic, it is forced ta raise to T, a position ta the

left of the specifier of NegP. Thus, the second essential claim regarding

sentential negation in French is that ne in French is a clitic.

The marker ne precedes bath finite and non-finite verbs in French.

However, pas follows the finite verb and precedes the non-finite verb.

(25) Jean D'aime pas Marie.Jean Neg-love-3sg Neg MarieJean does nollove Marie.

(26) *Jean ne pas aime Marie.

13

• (27) Ne pas sembler heureux est une condition pour écrire desromans.Neg Neg seem-[-finite] happy is a prerequisite write-[-finite]novelsNot to seem happy is a prerequisite for writing novels.

(28) *Ne sembler pas heureux [...]

He accounts for the different word order found in finite and non-finite

negative clauses by having the verb raise over Neg to T only in the former case;

in non-finite clauses, the verb remains in Agr ("short movement"). Ne, due to

its clitic nature, raises to T independently of verb-raising. These two possible

landing sites for verb-movement, made available by the proposai that IP is

separated into TP and AgrP, are crucial factors in his account. The alternative

derivations are represented in (29).

(29)

ne

t

Agr1

V -finite

There is an additional variation in French infinitival constructions. Thus,

(30a) be10w is a possible version of (30b), but (31a) is not an acceptable variant

of (31b).

14

• (30) a.b.

(31) a.b.

n'etre pas heureuxne pas etre heureuxnot 10 he happy.

*ne sembler pas heureuxne pas sembler heureuxnot 10 seem happy.

Interestingly, the distinction found between theta- and non-theta­

assigning verbs in English is found here in French infinitival constructions.

Only auxiliary verbs allow the order [ne-V-pas] in infinitivals. Pollock handIes

this distinction by analyzing [+finite] tense as strong and [-finite] tense as weak,

these being analogs for T of the notions transparency and opaqueness for Agr.

Only non-theta-assigning verbs raise to [-finite] tense. This accounts for the

difference in acceptability between the sentences above.

1.3.2. CHOMSKY (1989): NEGAnON AND SYNTAcnC AFFIXAnON

Chomsky's paper is primarily conœrned with investigating the tapic of

XO-movement as it relates ta word formation. He takes the position that

inflectional morphology derives from V-raising ta 1 and/or I-lowering ta V

(affix hopping).7 He adopts Pollock's articulated phrase structure, with a TP

dominating a NegP dominating an Agreement Phrase, but introduces a second

agreement phrase, above TP. He claims that the higher AgrP is associated with

subject agreement (AgrSP) and the lower AgrP with object agreement (AgrOP),

for example, the object agreement on past participles triggered by clitic climbing

in French (Kayne 1991). Like Pollock, he analyzes not as the head of NegP.

The relative order of projections is illustrated in (32).

15

• (32)

v

His paper does not introduce any new treatments for negation per se, but

I will briefly describe his explanation for the required do-support in negative

sentences in English, below.

• 1.3.2.1. Do-SUPPORT IN ENGUSB

Recall that Pollock maintains that V-raising is instantiated in ail English

sentences, but that this raising is invisible in the case of simple non-emphatic

affirmative sentences, which generate an empty do under Agt that then roses to

T in order ta provide a variable for the [tense] operator ta bind. In a negative

sentence, the NegP aets as a barrier for government of the trace in Agr unless it

is L-marked by a lexical (non-empty) do. Chomsky's analysis of non-emphatic

affirmative sentences does not invoke a phonologically empty do. The question

for Chomsky is why, in negative sentences, do-insertion is required, instead of

affix-Iowering followed by LF raising, eliminating the improper chain.

Chomsky claims that negation, in combination with affix-Iowering,

creates an Empty eategory Principle violation al LF. In an affirmative sentence,

the iIl-formed chair. created by affix lowering can be repaired by raising the

16

•

•

•

verb with its affIxes back ta T al LF. This LF raising is blocked by the presence

of negation. At LF, the [verb+affIxes] cannot raise from the lower Agr over

Neg ta T without leaving behind a verb trace that is not properly govemed.

Chomsky must then explain why verb-raising to Agr and then to I,

crossing Neg, is possible when the verb is an auxiliary, as in the sentence John

Ms not written books. If Neg blocks LF raising ta repair an ECP violation

created by affIx-Iowering, why does it not block overt raising ovèr Neg? He

daims that this difference results from a distinction between which traces are

deletable al LF. An Agr trace is assUIr.ed to be deletable, but a V-trace is not.

The trace dominated by Agr left by raising an auxiliary verb over Neg to I is an

Agr-trace, deletable at LF, but the trace in the same position left by raising of

the [V+affIx] complex at LF is a V-trace (under the assumption that this

process involves substitution, not adjunction), which cannot be deleted. Thus,

ooly in affIrmative sentences can such LF-raising apply in order ta correct ECP

violations; in a negative sentence, the trace in Agr would be blocked from

antecedent govemment by Neg.

Chomsky's analysis is driven by a principle of the grammar which forces

the least 'costly' derivation ta talœ place, wherever two derivations are possible.

In the case of negative sentences in English, the less costly derivation (using an

LF raising and traee-deletion strategy) is impossible, 50 a language-specific rule

of do-insertion is required.8

1.3.3. RJzzI (1990): NEGAnON AS AN A'-SPEOFIER

Pollock (1989) and Chomsky (1989) focus on the head status of

sentential negation in French and English in arder ta derive an apparent

blocking effect of negation on head-movement. Rizzi, on the other band,

17

• working within a revised analysis of minimality barriers, examines data which

lead him to postulate that negation in these languages is an A'-specifier. 9

1.3.3.1. RELATIVlZED M1NIIIfAUTY

•

•

Rizzi proposes a conjunctive formulation of the ECP, such that a trace

must simuitaneously fulfiII a formai licensing requirement and an identification

requirement.

[W]e may think of the theory of each type of nuII elements asconsisting of two components: A principle of formai Iicensing,which characterizes the formai environment in which the elementmay be found, and a principle of identification [...] whichrecovers sorne contentive property of the nuII element on thebasis of its immediate structurai environment. [Rizzi 1990:32]

The identification requirement may be fulfùled by government from an

actual antecedent or from a theta-marking head.

(33) CONJOINED ECP:A nonpronominal empty category must be(i) properly head-governed (formallicensing)(ii) antecedent governed or theta-governed (identification)

Proper head government is government by a head within its immediate

projection. Theta-government is government by a theta-assigner (e.g.

government of an object by a theta-assigning verb). Antecedent government is .

government by an antecedent (an eIement that govems and binds the governee).

Rizzi further proposes that the concept of minimality described in

Chomsky (1986a) shouid be reIativized, such that only a lypicaI potential

governor creates a barrier for government. A few definitions are required here.

(Rizzi 1990: 6-7)

18

• (34) RELATIVlZED MINIMALITY:X CL-govems Y only if there is no Z such that(i) Z is a typical potential CL-govemor for Y(ü) Z c-eommands Y and does not c-eommand X

(35) TYPICAL POTENTIAL GOVERNOR:

1. Head-governmenr.Z is a typical potential head govemor for Y, Z is a headm-eommanding Y

2. Antecedent governmenr.Z is a typical anteeedent govemor (TAG) for Y, Y in :(i) A-chain: = Z is an A-specifier c-eommanding Y(ù) A'-ehain:= Z is an A'-specifier c-eommanding Y(üi) XO-ehain: = Z is a head c-eommanding Y

(36) HEAD GoVERNMENT:X head govems Y iff

(i) X e{A,N,P,V,Agr,T}(ù) X m-eommands Y(m) no barrier intervenes(iv) re1ativized minimality is respected

(37) ANTECEDENT GoVERNMENT:X anteeedent govems Y iff

(i) X and Y are co-indexed(ù) X c-eommands Y(m) no barrier intervenes(iv) relativized minimality is respected

Rizzi proposes that sentential negation in English is a potential

A'-anteeedent (thus a potential anteeedent governor which creates a barrier only

for A'-ehains). He uses this 10 account for the observation that negation appears

10 affect adjunet extraction capabilities: Ross's Inner Island Constraints.10 The

19

•

•

•

effect of negation is illustrated in (38)-(39), where the affinnative sentence has

two possible readings, while the negative sentence has only one.

(38) It is for this reason that 1 believe John was fired.[1 believe irfor this reason, or he wasjiredfor this reason)

(39) It is for this reason that 1 don't believe John was fired.[Only: 1don'r believe irfor mis reason)

In the first sentence, the adjunct can be construed with either the main or

the embedded clause. In the second sentence, the adjunct can only be construed

with the main clause. This is accounted for by the relativized notion of

antecedent govemment as follows: Adjunct traces are not theta-govemed; hence,

they rely on meeting the antecedent govemment requirement in order to satisfy

the identification requirement of the conjoined ECP. Negation, an A'-specifier,

counts as a potential antecedent govemor which blocks antecedent govemment

of the adjunct trace in the lower clause by creating a minimality barrier. Since

the presence of negation ruIes out a co-indexed trace of the adjunct in the lower

clause, the negated sentence can only be interpreted as having the adjunct trace

in the higher clause, above the negative operator.

Rizzi's analysis of negation as a specifier is compatible with the proposaI

that the negative market heads an a'Jtonomous projection, NegP, only ifpas and

nor are analyzed as the specifier, not the head, of NegP. The facts described by

Pollock (1989) and Chomsky (1989), which they analyze in part as resulting

from a barrier created by the head ofNegP, receive a different treatment.

Under the assumption that non-lexical verbs can raise to 1 in English, but

that for lexical verbs affix-Iowering is required, (a pattern claimed by Pollock to

20

• result from the opacity of English agreement, which restricts V-raising to non­

theta-assigning verbs), Rizzi takes the acceptability of sentences like those in

(40)-(41) be10w 10 support the fuct that negation, conttary 10 the claims of

Pollock and Chomsky, does not create a barrier for antecedent government by a

head.Il

(40) They should [not have left]

(41) They have [not t left][Rizzi 1990:22]

Rizzi notes what he refers 10 as a residual blocking effect of negation on

the association of the inflectional morpheme with a lexical verb in English.

• (42) abcd

John smokes.*John smokes not.*John not smokesJohn does not smoke.

[pol1ock 1989:23]

•

Accepting the widespread assumption that the we11-formedness of (42a)

involves affix-hopping, the ill-formedness of (42c) seems to indicate that an

intervening negation rules out a derivation involving affix-Iowering, forcing the

operation of the Ianguage-specific rule of do-insertion. He accounts for the

observation that negation blocks affix-Iowering but not regular V 10 1 movement

as follows. Essentially, Rizzi concurs with Chomsky and Pollock that negation

creates a barrier forli raising (the repair strategy for constructions that include

affix-Iowering), but not for V-raising prior 10 LF.

21

• Under the assumption that [tense] is 'operator like'. a c1aim alse made

by Pollock (1989), Rizzi accounts for this by arguing that the verb inflected for

tense adjoins 10 IP at LF, an instantiation of operator movement. This

adjunction creates an A'-chain at LF, which is necessarily affected by an

interVening negation (a typical potential A'-anteeedent). The verbal trace cannot

be anteeedent govemed, and the structure violates the ECP.

1.3.4. OUHALLA (1990): THE NEG PARAMETER

Ouhalla's paper proposes that, cross-linguistically, the sentential negation

marker can be either a head or a specifier of NegP. He alse c1aims that the

position of NegP is pararneterized such that it can appear above TP or above

VP. These possibilities are instantiated cross-linguistically as below. The two

components of sentential negation in French, ne and pas, are the head and

specifier of NegP, respective1y.

OverTP:SpeclNegP: Swedish, FrenchHeadlNegP: Berber, French

OverVP:GermanEnglish, Turkish

According te Ouhal\a, the morphological nature of negation alse plays a

role in the syntax. In Berber and Turkish, negation is affixal, forcing an XO

(e.g. verb) 10 taise 10 it.12

In a language where the negative marker is a head, he proposes that there

is an empty operator generated in the specifier of NegP. If the negative marker

is a specifier, an abstraet morpheme is generated in the head position. Ouhal\a's

analysis requires the assumption that an empty operator in specifier position

22

• induces a violation of relativized minimality, but an abstraet morpheme in head

of NegP does not.

1.3.5. BELLETII (1990): NEGATION AS A CLlTlC IN ITALIAN

BelIetti (1990) analyzes the sentential negation marker non in Itallan as

the head of NegP. Based on the order of inflectional morphoiogy, Belletti

reverses the order of TP and AgrP proposed by Pollock, but follows Pollock in

generating NegP between the twO. 13 Also in agreement with Pollock, she

derives the position of negatior: with respect to the verb in ltalian in part by

attributing eUtie status 10 non. As sueh, by assumption, it must raise, via head­

to-head movement, and adjoin to the head of the highest funetional projection,

AgrP. Verb-raising, following Rizzi & Roberts (1989), is taken to be an

instance of substitution, not adjunetion, where inflectional morphology within

the verb derives from substituting the verbal roct for a slot subcategorized for

by the morphology.

The verb can raise from T 10 Agr over Neg without violating the ECP

because of the nature of Agr. BelIetti adopts Moritz's (1989) proposaI that

antecedent government holds between two ehains whieh share the sarne hcad.

The antecedent-government relation whieh is required 10 hold between any two

members of a chain is defined in terms of non-distinetness from the indexation

of the head of the chain. The Agr node created by head-to-head movement

carries the index of bath the raised Neg and the raised V. Thus, in example

(43), antecedent government h01ds between the trace [toi) and the trace [t-k).

23

• (43)

NP Agr-i-k

AAgr-i Agr-k

1 1

i !

toiGianni non·j parla-k

t-k

v

t·k

Belletti also extends the role of NegP by allowing adverbs of the type

piu, mai, ancora (anymore, never, yet) to raise to the specifier position created

by the negative projection.14 Note, however, that these negative adverbial

elements differ from pas in French in their optionality. These adverbs can

otherwise appear in the specifier of VP, resulting in the order illustrated by

sentence (45), below.

• 1.3.5.1. NEGAlIVE ADVERBS

(44) Gianni non ha piu parlatoGianni Neg has more talkedGiQ1l1Ù bas Mt talked arrymore.

(45) Gianni non ha parlato piuGianni Neg has ta1ked moreGiQ1l1Ù bas Mt talked arrymore.

[Bel1etti 1990:33]

24

•

•

Given that the past participle mises to AgrO, a sentence with the

negative adverb following the past participle bas the negative adverb in

speclVP, as shown in (46).

(46)

NP

Gianni

non

present

avere

-tlo) V

piu parla

According ta Belletti, the V, and the clitic head of NegP (non), raise ta

the head of the highest functional projection which makes up Infl. Since NegP is

immediately below the highest Infl projection, and negative adverbs are in

speclNegP, the word order exhibited in these sentences is as expected.

2S

• 1.3.6. LAKA (1990): NEC AS THE HEAD OF SIGMA-l'HRASE

•

Laka, working primarily on English and Basque, proposes thal the

invenlory of funetional categories available eross-linguistically ineludes a Sigma

Phrase, whieh can be headed by either Neg or an emphatie marker.15 SigmaP is

generated above TP in English. The relative order of the negative marker and

the tensee! verb at S-strueture is derived via a proposee! universal condition, the

Tense C-Command Condition (TCC), whieh forces T to e-command all

propositional operalors of the clause al S-strueture. This forces the [V+Tl

complex to adjoin 10 SigmaP, and precede the negative marker.

Her arguments for this proposal inelude the observation thal negative and

emphatie markers are in complementary distribution. Thus, as noted by Klima

(1964), they cannot co-occur in English, and b"th require either an auxiliary

verb, a modal, or do•

(47) 1 didn'l, as Bill had thought, go to the store.

(48) 1 DID, as Bill had thought, go to the store.

(49) *1 DID not, as Bill had thought, go to the store.

When both EMPH and Neg are present, negation is analyzed as

constituent negation. This is supported by the faet that it does not allow wide

scope over a universal quantifier in subject position, as shown be1ow.

(50) AlI of them did not go to the store.[For aIl X. X did not•.• OR For no X. X did••• ]

26

• (51) AlI of them DID not go to the store.[only For ail X, X did ooc... ]

Furthermore, Laka demonstrates that EMPH and Neg trigger identical

syntaetic effects, reflected in do-support in English, and AUX-raising in

i3asque. She derives the required do-support found in negative sentences from

the combined effect of the Tee and the base-positions of NegP and TP. She

maintains that affix-lowering is not permitted in negative sentences, because it

would result in Tense not c-commanding Neg, a propositional operator. Instead,

the dummy verb do is inserted under T.

1.3.7. ZANUITINI (1991): NEGA110N AND 1'ENsE

Restricting her study 10 Romance languages, Zanuttini (1991) proposes

that sentential negation in a given language can be distinguished along two Unes.

First, a language may have either the functional projection NegPl or NegP2,

which differ in their selectional requirements. NegPl is restricted 10 sentences

that instantiate a TP. This allows her to account for an interesting range of facts

involving constraints on the type of constructions in which a given type of

sentential negation can occur. Second, NegP in a given language is generated

either above or below TP. She further allows the negation marker 10 be

generated in either the head or the specifier of NegP.

1.3.7.1. THE CIme STAnJS OFNEGA110N

Zanuttini analyzes a given negative marker as a head if it interferes with

clitic-climbing, and as a specifier if it does nol Although she concedes that

there may be good arguments for analyzing ne as a clitic in French, she claims

(contra Belletti 1990) that the negative clements in other Romance languages

27

• (Spanish no, ltalian non) do not pattern with clitics with respect to syntactic

behaviour. She bases the claim that the negative particles in Spanish and ltalian

are not clitics on the following arguments.

First, pronominal clitics in these languages cannot normally bear stress,

unIess in a context of 'repair' .

(52) Non GU, ma LE ho parlato.not to-him, but to-her have-lsg spoken] have spoken not ta hùn, but ta her.

The negative particles can, however, bear stress, without the context of

repair.

(53) Preferirei NON farlo.prefer-lsg-cond Neg do-it]'d rather not do it.

Second, as shown in (54) below, the order V-ADV-clitic is

ungrammatical in ltalian, but the order V-ADV-negative marker illustrated in

(55) is marginally acceptable.

(54) *Essendo di certo vi un dottore...being for sure CL a doctorThere cenainly being a doclor•••

(55) ?Essendo di certo DOD un dottore.••being for sure Neg a doctorThere cenainly not being a doclor•••

28

•

•

•

Third, pronominal clitics precede a finite verb and follow a non-finite

verb, but the negative marker always precedes the verb, whether it is finite or

not.

(56) E meglio non parlarleis better Neg talk-to-herIr's better Mr ro ralk ro her.

(57) E meglio che non le parliis better that Neg to-her-talklr's better that 1don 'r ralk ro her.

Fourth, she makes the argument that, whereas the respective ordering of

pronominal clitics varies from language to language in Romance, the negative

marker consistently appears before the pronominal clitics. For example, in

ltalian the order 3rd DAT clitic preceding ACC clitic is exhibited, but French

exhibits the reverse order. ln both languages, however, Neg (Mn and ne)

precedes the pronominal clitics.

These facts, though interesting, do not rule out the possibility that the

negative marker is also a clitic. Her first argument, that the negative marker cao

be stressed even outside of a context of repair, is weakened by the fact that a

context of repair seems to be present in these cases as well, the difference being

that the context is pragmatic. Negative sentences cao be seen as denials of their

affirmative counterparts. As such, a pragmatic context of repair is always

possible for these utterances.16 Generally, stressing the negative marker is

appropriate ooly in a situation where the speaker is contradicting a contextual

affirmative presupposition.

29

• 2anuttini's second argument, regarding the relative order of adverbs and

clitics vs. adverbs and the negative marker, is weakened by the fact that the

negative marker in these sentences is conceivably a constituent negator, adjoined

10 the NP un doaore, in (55). The pronominal clitic vi (there) cannot be

associated with an NP in this way. If this is the case, the ungrammatica1ity of a

structure where vi appears between an adverb and an NP is not surprising.

The other arguments put forth by 2anuttini with respect to word order

are also conceivably compatible with an analysis of the negative ma-ker as a

clitic, if it is base-generated in a position distinct from tlle argument clitics. In

this way, if the negative marker were generated above the landing site of both

finite and non finite verbs, one would expect it to precede both. Argument

clibes, on the other hand, would be generated within VP, and their final

position might vary with respect 10 finite and non-finite verbs.

Although 2anuttini's proposal that the negative markers in Romance are

not clities may weIl be correct, these arguments do not appear 10 constitute a

decisive argument against the clitic analysis adopted by Pollock (1989) and

Belletti (1990).

1.4. SUMMARy

This review of recent proposals regarding NegP ilIustrates the various

claims researchers have made, the justification for them, and the data they are

claimed 10 account for. Fïrst, evidence for the status of negation consistently

involves its blocking effect on movement. Thus, if negation interferes with head

movement (either verb-raising or clitic-climbing) it may be analyzed as a head.

If it does not interfere with head movement, or if (under Rizzi's relativized

30

• minimality) it interferes primarily with non-argument XPs, it may be analyzed

as a specifier.

Consistently, the negative marker in the languages studied, which

include ltalian, Spanish, French, English, Basque, Turkish, Berber, Swedish,

and German, appears 10 bear a strong affinity with the verb. This relationship is

derived in a number of ways. ProposaIs include (i) its affixal nature (forcing the

verb 10 raise 10 support it), (ù) its clitic nature (forcing it 10 raise 10 the highest

functional head dominating it, which may be the position that the verb also

raises 10 for independent reasons), (m) the requirements of the Tense C­

Command condition (which force T 10 :aise to Neg after the V raises to 1'), or

(iv) its base-generated position, immediately dominating TP, the S-structure

position of the verb.

Clearly, these proposaIs regarding the syntactic nature of negation are

inseparabIe from analyses of the constraints on syntactic movement in generaI.

In this thesis, the interaction of sentential negation with syntactic movement will

play an important roIe, but evidence as to its syntaetic status as a maximal

projection will be derived primarily from its interaction with agreement

processes and Case-assignment. Evidence for this position will be derived from

changes in agreement patterns in relative clauses and from agreement marking

surfacing on the head of NegP in Literary WeIsh; from changes in Case­

marking, agreement, and word order in Russian; and from variations in the need

for a:1 additional Case-assigner in ColloquiaI WeIsh.

31

•

•

•

1.5. ENDNOTES TO CHAPTER 1

IThc panidc yr surfaces as y whcn prcccding a conson.~nl·inilial word.

20u.: 10 a lack of the appropriale 'J'mbols on my wordproce."",r. certain palataliz.edRussian consonanlS may be repre.<én!ed as coronals. Thus. [z] repre.<énl' both [z] and [tl. [siboth [sI and ["s'J. and [cl both [cl and [:fI.

3The Mirror Princip/e: Baker (1985).

4Note that the ph~omcnon of a ~"Upportive verb n:quired in negativc scntt:nC~ i~ notlimiled 10 English. In Chukchee, for ""ample, the verb 10 he is added 10 negative dalL"'-', and inother languagéS "ith"r he or ha,'t! funclions as a support v"rb (Payn" 1985).

5 Th" proposai thal V-raising is d"P"Rdenl on wh"th"r Agr is Iransparenl or opaquefor theta-assignméOl will nol play a roI" in this théSis, and a more detailed d=riplion ofPollock's analysis will tak" us far from the SlIbj,,"1 al hand. Th" inl"resled reader is ref"rred 10Pollock's articl".

&rh" n"g'dliv" mark"r ne is OftéO dropped in conl"mporary s'JlOken French.

7Th" diff"renC" beIWeen V raising to INFL and INFL low"ring 10 V was inlroducedby Emonds (1976) 10 accounl for th""" diff"rencéS belWeen French and English.

8No!" thal this only follows on th" assumplion that any langu..i;~ '1"'Cific rul" ofinsertion is inœrenUy more costly !han a d"rivation which involvéS 10WérÎng followed by LF­raising, s.,., C.L. Bak~r (1991) for an all"mativ" ViéW, in which d"riving th" faclS ofdo-supportvia languag"-spéCific ruléS is argued to ""lIlt in a simpl"r grammar. takéO as a whol". Oea1ingOnly with facls of English, Baker propoSéS a languag~ific rul" which disallows v"rbs fromraising over the negative adv..m nor. Do-support is required to support!lon.<è morphology.

9Although the anaIysis in Rizzi's monograph dOèS not includ" a cal"gory N"8P, hedOèS allow for the possibility that the negalive markers in English and French (nol and pas) arein fact in spéCifier of NegP (Rizzi 1990:115; tn 15). What is crucial for his trealméOt is not theexistenC" ofa proj""tion N"8P. but that the negative marker be a poléntial A'-anl..cedéOl.

10Ross (1983). Noté that th""" eff,,"lS are not restricled to SéOléntial negation.Compare the following IWO senlénces. the latter of which is limited 10 a reading where theadjunct is construed with the higher clallSlO: Il was for lhis /'t!QSon lhot everyone heUeves lharJohn was fired, Il was for lhis reason lhar no one heUeves John was fired. Rizzi suggèSlS thationer island eff,,"lS are d~ned by 'affective' operators (Klima 1964. Batwise and Coop.".1981). defined as eleméOlS that cao (a) license negative polarity itéms and (h) trigger subjo:,­AUX inversion in English (Liberman 1974): Under no circumslances will 1 go vs. ·Under nocirCumslances 1 will go. s.,., Rizzi (1990: 19 ft) for discussion, and Rizzi (1991) for a moredeveloped treatment which relatéS inversion in th""" caséS to inversion in wh'iJUcstiODS. Notéalso that negative adverbs like seldom and never. which give rise to ioner island constrainls,apparenUy allow affix-hopping derivations. Rizzi arguéS that, unIike nol. these adverbs cao

32

•

•

surface in pre-INFL position. They therefore need not inlervene between the raise<! verb and itsl.'aCe al LF (Rizzi 1990: 116, fil. 18).

lIAccording ID Pollock (1989) ana Chomsky (1989), the barrierhood created by thebead of NegP is voided in these examples via L-marking of NegP by the verh in T. The negativemarker nol is, nevertbeless, a potential barrier for bead-movement for these authors, and thisdistinguishes their anal:, ses froID that of Rizzi (1990).

120uhalla's analysis explicitly disallows syntaetic lowering.

13Zanuttini (1990:38) suggests that an unfortunale result of Belletti's analysis is thatthe AgrP node is required in arder ID derive the correct word order in French (the negativeclement ne preceding the negative clement pas) even in infinitivals, which do not exhibitagreement features.

lolne suggestion that certain clements must he in Spec/Negp, either al S-structure oral LF, is also maintained by Zanuttini (1991), Haegeman (1991), Haegeman & Zanuttini (1991),and Rizzi (1991). These proposais will he discussed in greater detail in the Appendix, withrespect ID secondary negation markers in Colloquial We1sh.

ISI will mention only the English data bere. In Basque, Laka's evidence comes fromAUX-fronting triggered by both EMPH and NEG, and by their complementary distribution.

16ne view that a negated sentence can he analyzed as the denial of an affirmativesentences bas a long hislory. Sec Hom (1989: Chapter 1 and elsewbere) for a review ofarguments that negated clauses are subordinate, in this sense, ID their affirmative counterparls, asweil as arguments against this view.

33

• 2. CHAPTER 2: NEGATION AND RELATIVIZATION IN LITERARYWELSH

2.1. INTRODUCTION

In this chapter, 1 introduce a paradigm where sentential negation appears

10 trigger changes in relative clause fonnation in Literary Welsh (LW). Negated

relative clauses differ from their affinnative counterparts with respect 10

complementizer selection and agreement. The fucts that will be accounted for

are presented, and three previous analyses are reviewed. It will be shown that

certain problems associated with these analyses can be eliminated under an

analysis that includes the functional projection NegP. Evidence for the role both

of the head and the specifier of NegP will be presented.

2.2. lNmAL PARADIGM

2.2.1. TwO RELAnvJZAnON STRATEGIES

2.2.1.1. Tm: D1RECI' STRATEGY

Welsh is traditionally described as having two distinct relativization

strategies. The tirst or direCl relativization strategy is required when the

relativized argument is a subject or the direct object of a verb inflected for

tense. The pre-sentential particle is a, and agreement cannot mark the site of the

relativized argument. This strategy is illustrated in (1).

(1) Y dynion a ddarllenodd [el y llyfr.lthe men PT read-past the bookthe men who read the book

34

• (2) *y dynion a ddarllenasant [el y Ilyfrthe men PT read-3pl-past the book,he men who read the book

However, it is not possible te capture the class of arguments for which

the direct strategy is available purely by referring te the grammatical functions

subject and direct object. A relativized subject is not accessible te the direct

strategy if a clause intervenes between the head of the relative clause and the

extraction site. Furthermore, the direct object in a periphrastic sentence, !bat is,

a sentence of the form AUX-S-V-O, is also not accessible to the direct strategy.

The periphrastic direct object immediately follows an untensed,

([-finite]) verb, which itseif follows the subject. The untensed verb is

traditionally referred to as a verbal noun (VN). It appears as the final verbal

element in periphrastic sentences where an auxiliary verb carries the tense

marking, as weil as in [-finiœ] clauses. VNs differ crucially from tensed verbs

in the realization of agreement. While VNs surface with object but not subject

agreement, tensed verbs surface with subject but not object agreement. 1 will

refer te the arguments of tensed verbs and VNs as VSO direct objects and

periphrastic direct objects, respectively, for ease of exposition. An example of a

periphrastic sentence is given in (2a).

(2a) Mae Sion yn ddarllen y llyfr.is-3sg John PROG read[-finite] the bookJohn is reading the book.

Subjects relativized across an intervening clause and periphrastic direct

objects require the indirect relativization strategy, described below.

3S

• 2.2.1.2. THE INDIRECT STRATEGY

•

The indirect strategy is required when the relaùvized argument is the

complement of a preposition, a noun, or a [-finite] verb. In this strategy. the

pre-sentential particle is y(r), and there is obligatory agreement marking the site

of the relativized argument. This strategy is illustrated in (3).

(3) Yllong y gwnaeth Sion ei werthu [elthe boat PI' did-3sg John 3sg selllhe boat Ùlal John soM

(4) *y llong y gwnaeth Sion gwerthu [elthe boat PI' did-3sg John selIthe boat Ùlal John soM

Although subjects usually require the direct pattern, the indirect pattern

is required if a clause intervenes between the head of the relative clause and the

clause containing the relativized position, even if the relativized position is a

subject or the direct object of a tensed verb. This is illustrated in (5)-(6).

(5) Ydynion y dywedodd Sion y darlIenasant [el y lIyfrthe men PI' said-3sg John PI' read-3pl the bookthe men who John said read lhe book

(6) y dyn Ygwn Ygwel Wyn eCthe man PI' know-lsg PI' will-see-3s Wyn himthe man Ùlal 1 laww HjIn will see

36

• 2.2.2. NEGATED RELATIVE CLAUSES

•

Sentential negation in LW takes the fonn of a pre-verbal negative

particle. Of particular interest 10 this investigation, negating a relative clause

appears 10 force the use of the indirect relativization strategy in Welsh. Thus, in

the case of subject relativization, agreement with the relativized argument is

obIiga1Ory if the relative clause is negated.

(l) y dynion na ddarllenasant [el y llyfrthe men Neg read-3pl [el the bookthe men who didn 't read the book

(8) *y dynion na ddarllenodd [el y llyfrthe men Neg read-3sg [el the bookthe men who didn't read the book

In the case of direct objects, the negative relative differs from the

affirmative in that an independent pronominal can, and in fact must, appear in

the relativization site. In this way, an asymmetry Iacking in affirmative relative

clauses is present in negative relative clauses. Whereas the subject exhibits the

so-caIIed indirect strategy, the object has the sarne charaeteristics as an

unreIativized direct object: an obligatory in situ pronominal. This contrast is

shown in (9)-{lO).

(9) *y llyfr na ddarllenais i [elthe book Neg read-lsg 1the book which 1didn't read

37

•

•

(10) Yllyfr na ddarllenais i efthe book Neg read-1sg 1 itthe book which 1didn 't read

[Sadler 1988: 128)

However, negated direct object relatives allow a second possibility. In

these sentences, an agreement marker surfaces on the negative marker na itse1f.

If such an agreement marker occurs, then the pronominal may drop.

(11) Y llyfr nas ddarllenais i efthe book Neg-3rd read-1s 1 (it)the book which 1didn 't read

[Sad1er 1988: 113)

Before reviewing the ways in which previous researchers have dealt with

these data, 1 will present a brief description of some aspects of the syntax of

Welsh which are relevant for the analyses of this paradigm.

2.3. SOME lNmAL FACTS ABour WELSII SYNTAX

2.3.1. WORD ORDER AND CONS'ITlUENT STRUC1lJRE IN WELSII

Welsh is a head-initial language; within a phrase, the head normal1y

precedes its complements and modifiers. Thus, the head precedes the possessor

in a possessive construction (PC), prepositions precede their objects in a

prepositional phrase (PP), anà ~e head of a relative clause (Re) precedes the

relative clause (fallerman 1991:311).

(12) NP [[ci) [bach])dog smallasmaIldog

38

• (13) PC [(ci] [Gwyn]]dog GwynGlryn'sdog

(14) PC [[[ci] [bach]] [y meddyg]]dog small the doctorthe doctor's small dog

(15) PP r[yn] [yr eglwys]]in the churchin the church

(16) RC y dyn [[y] [mae Wyn yn ci weld [el]the man [C was-3sg Wyn PROG 3sg see]the man thalll)ln was seeing

Verbs take a full NP direct object in post-head position.

(17) VP Roedd hi'n [[gweld] [y dyn]].was-3sg she-PROG [see the man]She saw the man.

As mentioned above, Welsh sentences display two basic sentence

patterns, VSO in simple sentences and AUX-S-V-O in periphrastic sentences. In

the latter, the second verbal element appears without tense marlâng. .

(18) Darllenodd Sion y llyfr.read-3sg John the bookJohn retJd the book.

39

•

•

•

(19) Mae Sion yn ddarllen y llyfr.is-3sg John PROG read the bookJohn is reading the book.

One of the problems raised by VSO languages for GB theory is that the

elements associated with the VP appear to be discontinuous. This led to claims

that these languages do not instantiate a VP constituent. Following proposaIs by

McCloskey (1983) for Irish and Harlow (1981), Sproat (1985), and Sad1er

(1988) for Welsh, 1 will assume that Welsh is underlyingly SVO, and that the

surface V-;nitial order is caused by raising of the leftmost verbal element in

order to support tense morphology.2 The [-finite] verbal elements remain in

their base-position in VP. A [+finite] S-structure is illustrated below,

abstracting away from complexities introduced by additional functional

categories.

(20)

[verb[tensell

t NP

,,

1._._._._.

As predicted under such an analysis, we find VSO order in tensed

clauses, but SVO order in untensed clauses.

40

•

•

(21) Bwriadai 'r athro i ['r plant ddarllen llyfr ara1].intended the teacher to [the children read book other]The teacher imended the children to read another book.

[Rouveret 1990:59]

(22) Darllenodd y plant y llyfr.read the children the bookThe children read the book.

[Rouveret 1990:48]

As a variant of the VSO sentence, the verb gwneud ('do') may carry the

tense features, while an uninflected lex;cal verb remains within VP.3

(23) Gwnaeth John ennill.did John winJohn won.

Note that, unlike English do-support sentences, such sentences in We1sh

are not emphatic. They appear to be synonymous with the simple VSO

structures. There is, however, a slight tendency in Colloquial Welsh ta use VSO

ta refer to future tense, and the periphrastic structure to refer to present tense (J.

Williams, p.c.).

2.3.2. AGREEMENT AND Nuu. ARGUMEN'IS lN WELSH

2.3.2.1. Nuu. SUB.JEcrs

•

In common with the other Ce1tic languages, We1sh is a null-subject, or

pro-drop language, with a rich system of agreement marking. A null-subject

language is one in which the subject of a tensed clause may be non-overt. In GB

theory, lexical requirements (subcategorization and theta-assignment) of

predicates must be met at each syntaetic leve1. This requirement, which provides

41

•

•

a fundamentaI part of the motivation for the theory of empty elements, is known

as the Projection Principle. The Extended Projection Principle (EPP) combines

this with the requirement that all clauses have subject!.. This requirement

captures a basic asymmetry between NP and S: subjects may be freely omitted

in NP, but are required in S even when they are not assigned a theta-role by the

predicate.

(24) NP John's claim that the police are responsiblethe claim that the police are re.<;ponsible

(25) S It is believed that the police are responsible.*Is believed that the police are responsible.

The ability of a language to drop subjects in tensed clauses is a locus of

parametric variation. Taking the perspective of a functional definition of empty

categories, we will follow Chomsky (1982) in assuming that the empty element

in null subject constructions has all the properties of a pure pronominal, that is,

an element with the feature specifl:-->tior.. [+pronominal,-anaphoric], which, like

overt pronouns, occurs in governed positions. A pronoun may be non-overt

under certain specific conditions, i.e. if governed by ricn agreement.

In Welsh, verbs that are [+tense] necessarily exhibit person/number

agreement with a pronominal subject This agreement permits but does not force

the pronominal ta be null.4

(26) Maent (hwy) yn canu.is-3pl (they) PROG singThey are singing.

42

• A tensed verb does not agree with a non-pronominal subject. Instead, we

find default 3sg agreement.

(27) Mae Yplant yn canu.is-3sg the chiIdren PROG singThe children are singing.

(28) *Maent y plant yn canu.is-3pl the chiIdren PROG singThe children are singing.

Thus, personlnumber agreement is in complementary distribution with

full lexical NP subjects, but is obligatory with pronominaI subjects.

2.3.2.2. 01lŒR NUIL ARGUMENTS

Not only tensed verbs, but prepositions, nouns, and [-finite] verbs aIIow

null arguments in Welsh. Just like tensed verbs and subjects, these heads must

exhibit agreement morphology if their complement is pronominaI (null or

overt), and cannot exhibit agreement if their complement is a non-pronominaI

lexical NP. Agreement on the head permits a pronominal complement 10 be

null.

The form agreement takes depends on properties of the lexical head

itseIf. In a possessive construction, formed in Welsh by placing the head noun

before the possessor NP, a clitic agreeing with the personlnumber features of

the (pronominaI) possessor appears 10 the immediate left of the possessum.

(29) Brynais i *(ei) dy (el).bought-lsg 13sg house (him)1boughl ms house.

43

•

•

Object agreement on a verb uninflected for tense features (VN) also

surfaces as the same form of the clitic, which occurs to the left of the VN.

(30) Mae Wyn wedi ·(ei) weld (ef).is-3s Wyn PERF 3sg see (him)~n Juzs seen mm.

Many prepositions are inflected for the personlnumber features of their

pronominal complement. As with all instances of clitics and agreement

inflection, the preposition must be inflected when it can be inflected; otherwise,

the sentence is ungrammatical.

(31) Soniais i amdano (ef).spoke-Isg 1 about-3sg (him)1spoke about him.

(32) ·Soniais i am (ef).spoke-lsg 1 about (him)1spoke about mm.

In these constructions, just as with tensed verbs, neither a c1itic nor

inflection can occur with a non-pronominal lexical NP.

(33) Mae Rhys yn (·ei) gweld Megan.is-3sg Rhys PROG 3sg see MeganRhys is seeing Megan.

(34) Brynais i (·ei) dy Sion.bought-lsg 1 3sg house Sion1bought Sion's house•

44

• (35) Soniais i am (*amdano) Sion.spoke-lsg 1about (about-3sg) Sion1spoke about Sion.5

Not all prepositions in Welsh have inflected (synthetic) forms. The non­

inflecting prepositions take neither a pre-head clitic nor agreement morphology;

as expected, a pronominal argument of one of these prepositions cannot be

dropped. This is shown in (36)-(38).

(36) Euthem i a Mair i'r sinema.went-ls 1 with Mary 10 the cinema1 wenr wirh Mary to the movies.

(37) Euthem i a hi i'r sinema.went-ls 1 with her to the cinema1 wenr with her to the movies.

(38) *Euthem i a [el i'r sinema.1 wenr with her to the movies.

The parallels between the clities that appear on VNs and Ns and the

agreement inflection that surfaces on tensed verbs and inflecting prepositions are

clear. However the phenomenon of null subjects in Welsh is 10 be analyzed, the

analysis should be extendable 10 other null arguments.

2.3.3. WS:-QUES110NS AND CLEFT CONSTRUcnONS

•Wh-questions and clefts in Welsh pattern with relative clauses with

respect 10 the pre.sentential particle (a or y(r) and the pattern of agreement.

Thus, when the extraction site is in local subject or VSO direct object position

4S

• in these constructions, there is an empty category in the extraction site and the

pre-sentential partic1e is a.

(39) SUBJECTA. RELATIVE CLAUSE

Ydyn a ddarllenodd [el y llyfr ar y trenthe man PT read-3sg the book on the trainthe man who read the book on the train

B. WH-QUESTION

Pa ddyn a ddarllenodd [el y lIyfr ar y tren?which man PT read-3sg the book on the trainWhich man read the book on the train?

•C. CLEF!"

Y dyn a ddarllenodd [el y lIyfr ar y tren.the man PT read-3sg the book on the train(It was) the man who read the book on the train.

•

(40) DIRECT OBJECTA. RELATIVE CLAUSE

Yllyfr a ddarllenodd y dyn [el ar y trenthe book PT read-3sg the man on the trainthe book t1uJl the man read on the train

B. WH-QUESTION

Pa llyfr a ddarllenodd y dyn [el ar y tren?which book PT read-3sg the man on the trainWhich book did the man read on the train?

46

• C. CLEFT

y llyfr a ddarllenodd y dyn [el ar y tren.the book PT read-3sg the man on the train(It was) the book tJuu the man read on the train.

[Sadler 1988: 163-166]

•

If the extraction site is a periphrastic direct object, the object of a

preposition or in a possessive NP, relatives, clefts, and wh-questio!lS aIl exhibit

properties of indirect relatives: the pre-sentential particle is y(r), and agreement

marks the ext.-action site.6

(41) PREPosmONALOBlECTA. RELATIVE CLAUSE

Ytren Ydarllenodd y dyn Yllyfr amo [elthe train PT read the man the book on-3sgthe train tJuu the man read the book on

B. WH-QUESTIONPa tren Ydarllenodd y dyn Yllyfr amo [el?which train PT read the man the book on-3sgWhich train did the man read the book on?

-.

C. CLEFTy tren Ydarllenodd y dyn Yllyfr amo [el.the train PT read the man the book on-3sg(It was) the train thal the man read the book on.

[Sadler 1988:165]

(42) POSSESSlVENPA. RELATIVE CLAUSE

Ydyn Ygwelais i ei fab [elthe man YI' saw-ls 1 3sgm sonthe man whose son 1 saw

47

• B. WH-QUESTION

Pwy y gwelais i ei fab [e]?who PT saw-1s 1 3sgm sonlWwse son did 1see?

•

c. CLEFT

y dyn Ygwelais i ei fab [el.the man PT saw-1s 1 3sgm son(Il was) the man wJwse son 1sGW.

[Sadler 1988: 164]

2.3.4. PlED-PlPING AND COMPLEMENTIZER SELECTION

As shown above, questioning and clefting an NP in Welsh patterns with

relativization structures with respect 10 agreement and type of pre-sentential

partic1e. However, while a relative clause cannot be headed by a PP, a PP can

be c1efted or questioned; !hat is 10 say, pied-piping is pennitted. If a pp is

questioned or c1efted we find the y(r) (indirect pattern) comp1ementizer

occurring with the direct agreement pattern. Neither agreement nor a

pronominal marks the questioned or clefted site.

(43) PREPosmONAL PHRASESA. WH-QUESTION

[Ar ba tren] y darllenodd y dyn Yllyfr [el?on which train PT read the man the bookOn which train did the man read the book?

B. CLEFT

[Ar y tren] y darllenodd y dyn Yllyfr [el.on the train PT read the man the book(Il was) on the train thal the man read the book.

48

• These structures deviate from the usuaI pattern found with respect to

complementizer selection and agreement.

2.3.5. "l'RUE" IsLAND CONSTRAINTS

•

•

We have seen three agreement patterns in Welsh. First, in an

unrelativized structure, we find obligatory agreement with a pronominal

argument, the latter surfacing optionally. Second, in the direct relativization

strategy, we find that neither agreement-marking nor a full pronominal can

surface. Third, in cases of indirect relativization, agreement with the relativized

argument is present, but a pronominal cannot surface in the relativization site.

In this way, the indirect pattern differs from unrelativized arguments, whose

pronominal argument optionally co-occurs with agreement.

As noted by Tallerman (1983), the pattern associated with non­

relativized contexts (obligatory agreement and optional surfacing of a

pronominal) is also found in some reIativized structures; specifically, if

relativization is out of a complex or conjoined NP. Compare the (a) sentences

below where an in situ pronominal is strongly preferred, with the indirect

strategy illustrated by the (b) sentences, where a pronominal is completely

ungrammatical.7

(44)

a. Dyma'r eneth garedig y diolchais i [iddi hi a'i thad].here-is-the girl kind that thanked-lsg 1 to-3sgf her and her fatherHere is the kind girl that 1 thanked (her) and herfather.

rrallerman 1983:202]

49

• b.

(45)

Dyma'r eneth garedig y diolchais i iddi (*hi).here-is-the girl kind that thanked-I sg 1 to-3sgf (ber)Here is the lcind girl tha! 1 thanked.

a. y dyn Yseniais [arndano ef ae Ann]the man comp spoke-ls about-3sgm him and An••the man tha! 1 spoke about him and Ann

[rallerman 1983:201]

b. Ydyn y seniais arndano (*ef)the man that spoke-lsg about-3sgm (bim)the man tha! 1 spoke abOut

2.3.6. ~y

These data should be addressed by an adequate analysis of relativization

in We1sh. The faets, whieh are quite complex, are summarized below. Although

the issues raised by sentential negation are critical to this thesis, the analysis,

incorporating as it does agreement and relativization faets, must handle the other

issues as we11.

(1) THE NEGATION IssUE: Negating a relative clause appears to forcethe pattern of agreement associateci with the indirect strategy forthose arguments that exhibit the direct pattern in affirmativerelative clauses.

y dynion a ddarllenodd [el y llyfrthe men PT read-3sg the bookthe men who reatf (he book

so

• y dynion na ddarllenasant [el y llyfrthe men Neg read-3pl the booklhe men wluJ didn '1 read lhe book

(D) THE Co-OCCURRENCE ISSUE: Whereas a pronominal argumentmay co-occur with agreement or a clitic in a simple sentence, itcannot co-occur with agreement or a clitic if the argument inquestion is relativized, clefted, or questioned.

Soniais i amdano (el).spoke-ls 1 about-3sgm (him)1spoke aboU/ him.

y dyn Y soniais amdano (.el)the man PT spoke-lsg 1 about-3sgm (*him)/he man lha! 1spoke aboU/

(ID) THE LACK OF AsYMMETRY Is..<;UE: With respect to agreementpatterns in affirmative relative clauses, subjects pattern with vsodirect objects.

y dynion a ddarllenodd le] y llyfrthe men PT read-3sg the bookthe men wluJ read the book

y llyfr a ddarllenodd y dynion le]the book PT read-3sg the men [elthe book t1wJ the men read

51

•

•

(IV) THE DIRECT OBJECT ISSUE: VSO direct objects require the directstrategy, but periphrastic direct objects require the indirectstrategy.

y 11yfr a ddar11enodd y dynion [e)the book PT read-3sg the men [elthe book tha! the men read

y l10ng y gwnaeth Sion ei werthuthe boat PT did-3sg John 3sg sel!the boat tha! John soM

(v) THE ASYMMETRY ISSUE: Although subjects and VSO directobjects pattern tegether in affirmative relative clauses, theyexhibit distinct agreement patt"'..rns in unrelativized contexts and innegated relative clauses. Subject agreement, '>ut never objectagreement, surfaces on a [+finite) vero. Thus, in a VSOstructure, a subject pronominal, but not an object pronominal,can be dropped.

Agoron (hwy) *(et).opened-3pI they itThey opened it.

In negated relatives, while the subject exhibits the agreementpattern associated with the indirect strategy, the VSO directobject patterns with unre1ativized structures in allowing apronominal te optionally double agreement.

y llyfr nas ddarllenais i (et)the book Neg-3sg read-lsg 1 (it)the book which 1didn 't read

52

•

•

(VI) THE NON-ARGUMENT IssUE: Questions and cleft constructions inWelsh pattern with relativization structures with respect toagreement and complementizer selection. However, if a pp isquestioned or clefted instead of an NP, then we find the yrcomplementizer (like the inclirect strategy) occurring with a lackof agreement (like the direct strategy).

PREPOSITIONAL PHRA.cES

Wh-question[Ar ba tren] y darllenodd y dyn Yllyfr [el?on which train Pr read the man the bookOn which train did the man read the book?

Cleft[Ar y tren] y darllenodd y dyn Yllyfr [el.on the train Pr read the man the book(It was) on the train tJuu the man read the book.

•

(vn) THE IsLAND CONSTRAINT IssUE: Indirect relatives must bedistinguished both from direct relatives and from cases ofrelativization out of complex and coordinate NP structures. Thelatter, which 1 will refer to as true islands, differ from both directand indirect relativization in permitting (and, in fact, preferring)an in situ pronominal to co-occur with agreement.

Dyma'r eneth garedig y diolchais i [iddi hi a'i thad].here-is-the girl kind that thanked-lsg 1 to-3sgf her and her fatherHere is the kind girl tJuu 1 thanked (her) and herjarher.

S3

• 2.4. PREvIouS ANALYSES oFRELATIVIZAnoN IN WELSH: THE EMYIT

CATEGORY PRINCIPLE

2.4.1. INTRODUcnON

In many previous analyses of relativization in Welsh, the differences

between the distinct pre-sentential particles and agreement patterns in direct and

indirect relativization are derived by proposing distinct syntactic processes.

Based on the intuitive observation that the indirect strategy seems to be required

when the relativization site is 'farther away', the required agreement marking is

taken to indicate that no movement has taken place.8

Awbery (1976), for example, captures the distinction by claiming that

only the direct strategy involves a movement transformation. Under her

approach, negative relative clauses formed on subject and direct object positions

are taken ta be non-movement cases, unlike their affirmative counterparts, since

the indirect pattern of agreement is employed.

Commonly, in previous analyses, the apparent correlation between the

pre-sentential particles and the distinct agreerr.el;t patterns is treated as causal.

The particle a, associated with the direct pattern, is treated as in sorne way

forcing or licensing a gap in the relativization site. The particle y(r), associated

with the indirect pattern, occurs when a gap is not permitted in the relativization

site.

Recent analyses of relativization in Welsh carried out in the Govemment

and Binding ftamework share certain features. They generally invoke sorne

version of the Empty Category Princïple (ECP) ta derive the complementarity

between the direct and indirect patterns of relativization. The definition of the

ECP bas evolved in certain essential respects since its inception, but consistently

• involves an asymmetry between subjects and objects. We will see that the faet

that subjects appear 10 pattern with direct objects in Welsh relativization raises

problems for analyses that derive the agreement patterns by a condition on

empty categories.9

1t has also been claimed that the structurai properties of Welsh are

significantly differer.t from those that have been proposed for other languages;

i.e. that heads do not project specifiers in Welsh (Rouveret 1990) or that Welsh

instantiates a structure-pruning mechanism at S-structure (Harlow 1981). These

revisions, proposed within an ECP-driven account of the relativization

strategies, are driven by the need 10 have subjects and VSO direct objects

pattern 10gether with respect 10 government, an unexpected fact if the ECP is

the crucial principle involved. The analysis that will be proposed in this thesis

will not require Welsh 10 be treated as exceptional with respect 10 any of the

major components of the theory.

Below, 1 outline three previous analyses of relativization in Welsh:

Harlow (1981), Sadler (1988) and Rouveret (1990). 1 will argue that certain

problems common to aIl three analyses follow from this assumption, and can be

eliminated by analyzing both the direct and indirect pattern as resulting from

wh-movement, leaving a trace in the relativized position. 1 will concentrate on

how these analyses handie the issues listed in (I)-(VII) above.

2.4.2. HARLow (1981): 'l'BE EM1'TY CATEGORY PRlNOPLE

Harlow assumes a fairly standard initiai phrase structure for Welsh,

given in (46). To derive the surface VSO order, Harlow maintains that the

leftmost verbal element moves up 10 [+tense] Infl.

55

• (46) a.b.c.d.e.f.

S"-> X" S' (where X"=~l', PP, VP, AP)S'-> CSS-> Infl NP VPVP- > V (NP) (PP)NP->NP (S')Infl-> {TENSE, Agr}

•

•

Harlow adopts Chomsky's (1981) formulation of the ECP and definition

of Govemment, given below. 10

(47) ECP: [el must be govemed.NP

cr. govems ~ iff cr. minimally c-eommands ~ and:a. cr. = N, V, A, P (=Iexical govemment)

or b. cr. is co-indexed with ~.

Where minimally c-eommands = cr. c-eommands ~ and there is noy such that cr. c-eommands y and y c-eommands ~ and X does notc-eommand cr.

Harlow mair.tains that languages which support empty subjects have an

indexed Agreement (Agr). Agr selects pro, a [+pronominal -anaphoric]

element, which is then deleted, leaving a co-indexed empty category. The rule

of pro-deletion is generalized from McCloskey's rule of relative deletion

proposed for Irish (McCloskey 1978:50).

Following pro-deletion, the resulting empty category is govemed by the

co-indexed Agr, satisfying the ECP. Harlow follows Chomsky (1981) in

extending this theory of agreement 10 object clitics, arguing that these are of the

form

(48) [Agr V] (Agr=clitic)V

•

•

•

The missing pronominal object is the consequence of strict

subcategorization for pro, indexing, and pro deletion. The [e] is properly

governed by the co-indexed Agr, in accordance with the ECP.

To alIow for the co-occurrence of agreement and a pronominal NP,

Harlow maintains that pro-deletion in WeIsh, unlike McCloskey's rule of

relative deletion in Irish, operates optionally. VSO direct objects cannot undergo

pro-deletion because they are not governed by Agr, 50 the resulting [e] would

violate the ECP.

(49) Darllenais [e] y llyfr.read-ls the book1read lhe book.

(50) Darllenais i ef.read-ls 1 it1 read il.

(51) *Darllenais i [e].1 read il.

Harlow's mie accounts for the possible co-occurrence of agreement and

pronominals in non-reIativized contexts, but fails te explain why, in the indirect

relativization cases, an overt pronominal cannot occur in the reIativized

position, doubling the agreement or clitic which he treats as pronominal (the Co­

Occurrence Issue). Compare this with a nonrelativized context, in which

agreement on the preposition may co-occur with an overt pronominal NP.

57

•

•

(52) Soniais i arndano (et).talked-ls 1 about-3ms him1 taIked about him.

Harlow (1981:228) accounts for this by proposing that pro-deletion is

obligatory in relative contexts, provided that its output does not violate the ECP.

In this way, Harlow's rule of Relative Deletion is subject to exactly the sarne

conditions as pro-deletion, but it is obligatory, a theoretically problematic

notion.

Having argued for a rule in order to account for the obligatory missing

arguments in indirect re1ativization, Harlow proposes an analysis of the direct

strategy. Recall that the direct strategy is used for subjects and for direct objects

of finite verbs, and requires the pre-sentential particle a. Neither agreement nor

a pronominal can mark the relativization site.

According te his treatment of agreement, pro-deletion cannot apply to

non-relativized VSO direct objects because the resulting em;>ty category would

not be governed by Agr, in violation of the ECP. In direct relativization,

however, although no agreement elements are present to license [el, the

re1ativized position is obligatorily empty. Furtherrnore, the use of agreement to

license pro-deletion in subject position results in an ungrarnmatical sentence.

(53) *y ferched {y/a} aethant [el i Gaerdyddthe girls PT went-3pl [el to Cardiffthe girls thal wer.r to Cardiff

Harlow deals with this issue by claiming that it is the pre-sentential

element a itse1f that provides a governor for the ernpty category in subject and

VSO direct object positions. His proposai thus relies on treating the e!ement a as

fundarnentally distinct from y; whereas y is a complementizer, a is a pronominal

58

• element generated under C. Since the same pattern is exhibited for subjects and

for VSO direct objects, he must c1aim that a governs both of these positions in

Welsh. l1 He does this by proposing a derived structure where S immediately

dominates V, Infl, NPI (subject) and NP2 (object), a fairly radical claim about

Welsh phrase structure. The structure, given in 54), is derived by structure­

pruning following V-raising. 12

(541

V NP1 NP2

Harlow's account encounters difficulties motivating the choice of

relativization strategy. In order to block the occurrence of a in indirect relatives,

he daims that the agreement affixes are in fact pronominaIs, and that

pronominaIs in Welsh are translated as bound variables at LF. The restriction is

then related to a constraint on variables at LF. If the agreement affix on the

head of, for example, a preposition is co-indexed with [el, then [el is properly

govemed, but then the pronominal a is a free variable at LF. If a but not the

(agreement or clitic) pronominal is co-indexeè with [el, then [el is not properly

govemed. If they are both co-indexed with the [el, the structure violates a one­

to-one requirement between lexical NPs and argument positions.

ln this way, the analysis blocks the use of a with the indirect cases, but

fails te explain why the grammar bas this second strategy at all. Additional

modifications are required te rule out the following sentence, with Agr

governing [el.

59

• (55) * y dynion y darllenasant [el y llyfrthe men C read-3pl the bookthe men that read the book

Harlow maintains that an NP in subject or direct object position

potentially has two governors: Agr and a co-indexed element in C. However,

only one govemor is permitted. Essentially, he stipulates that the highest C must

bind an NP in S, effectively dictating the choice of govemor, namely a.13 A co­

indexed null element in C satisfies this requirement in the indirect cases.

To sum up this aspect of his proposal, sentences like (a) below are

ungrarnmatical because [el is doubly govemed, whereas (b) is ungrarn:natical

because the head of the relative must bind an NP in C. Only (c) satisfies both

binding and government.

• (56) a. *y dynion [ai [ddarllenasanti [~]

•

b. *y dynion [y [darllenasanti [~] ...

c. y dynion [ai [ddarllenodd [~].

To account for the patteming together of subjects and VSO direct objects

with respect to relativization (the ùu:k of asymmetry issue), he proposes a

pruning transformation which follows verb-raising to I. This transformation

leaves both subjects and direct objects in non-periphrastic sentences governed by

C. This pruning also serves to distinguish direct objects of tensed verbs in

simple sentences from direct objects in periphrastic sentences (the direct object

issue); pruning can only occur if the verb bas raised out of VP. In a periphrastic

construction, the verb remains in VP, and the VP blocks government from C. In

60

• this way, Harlow accounts for the patterning together of complements of [­

finite] verbs, complements of prepositions (proteeted by PP) and noun phrases

in genitive constructions (proteeted by NP).

2.4.2.1. NEGATED RELATIVE CLAUSES

e

Harlow's proposed constraint that the head of a relative must bind an NP

in the adjacent C holds in negated relative clauses as weil. This means that there

must be a nuIl eIement co-indexed with the head of the relative clause in negated

relatives, just as in indirect relativization. Since bis analysis of the direct pattern

of relativization requires that the pronominal element a fulfill both the co­

indexation requirement and the govemment requirement for subjects and VSO

direct objects, and since govemment is defined in terms of c-command, he

suggests that the negative eIement na is generated under C, creating a branching

C. Thus, C in negated relatives dominates two nodes, one containing the

negative complementizer na and the other containing a null NP eIement co­

indexed with the head of the relative. The co-indexed NP in C no longer fuIfills

the c-command condition, 50 govemment of an empty category in subject or

direct object position from an eIement in C is blocked, and local agreement is

needed.

•

(57) S'

V INFL NP' NP2

61

• 2.4.2.2. DISCUSSION OF HARLOW'S ANALYSIS

•

Harlow's analysis relies on the c1aim that a a pronominal in C and y(r) is

a complementizer. The c1aim that these are distinct elements receives some

support from Lie fact that they trigger distinct consonant mutations. Whereas a

triggers soft mutation of a following consonant, y does not trigger any

mutation.14 However, the mutation effect is not very slrong evidence for

claiming that a is pronominal. As noted by Sadler, the question partic1e a also

triggers soft mutation, shown in (58), but one would not want to c1aim that this

particle too is an NP.

(58) A welais ti ef?PT saw-2s you himDid you see him?

Furthermore, certain problems fol1ow from the c1aim that the head of the

relative must be co-indexed with (bind) an element in the adjacent C. There

does not seem to be a requirement that this coindexed element in C be lexica1ly

ft1led by a. In fact, in indirect relativization, it cannot be fil1ed.

(59) "y dynion a y dywedodd Dafydd Ydarl1enasant le) y lIyfrthe men NP C said-3sg David C read-3pl the bookthe men thal David said read the book

The structure of an indirect reiative must be as shown in (60).

(60) HEADi [ [e]i y [••.[y [ Agr.•.]))

But if we allow this e1ement te optionally be lexically fi11ed, we do not

have any explanation for the ungrammaticality of (61).

62

• (61) .y dynioni [ [e]i y [darllenasant [el y llyfrthe men NP C read the bookthe men wJw read the book

In order to account for the fact that a pronominal cannot co-occur with

agreement in indirect relativization, but may if non-relativized, even though

neither is assumed to involve movement, Harlow must resort to the c1aim that

the same de1etion ruIe applies in both cases, but it applies obligatorily only in

the latter. Rule-based analyses of this kind are now genera1ly considered

inadequate. We shall now examine two more recent analyses of these facts:

Sadler (1988) and Rouveret (1990).

2.4.3. SADLER (1988): A CONJOINED ECP

With respect to agreement in We1sh, Sadler (1988) interprets the pre­

head agreement markers found in possessive NPs and on [-tense] verbs as clitics

indicating an agreement relationship holding between the heads of these phrases

and their pronominal complements, just as agreement inflection on tensed verbs

and on sorne prepositions is indicative of such a relationship. Their presence,

like that of inflection, allows the personal pronoun to be dropped.

Like Harlow (1981), Sadler proposes two distinct structures for the

direct and indirect patterns of relative clause formation. Her analysis alse

exploits the difference between an empty category with the status of a trace, and

pro, an ernpty category co-indexed with agreement and base-generated in

argument position.

Her analysis rests on the initial assumption that traces in We1sh are in

complementary distribution with the presence of agreement morphology. Thus,

for Sadler, only in direct relativization is the empty NP in the ~:elativized

63

•

•

argument position a non-pronominal [el, subject to the ECP. In indirect

relativization, the empty element is pro, licensed by agreement on the governing

head.

Sadler proposes the following structure for direct relative clauses.

(62) rOi a[...[e]i· ..]]S' S

As a non-pronominal empty NP, [el is subject to the ECP and must be

properly governed. She proposes a conjunctive formulation of the ECP, the

CECP, given in (63).

(63) CECP: NP[e]i must be lexicalIy governed and locally co-indexed.[Sadler 1988:149]

Consider the schematized S-structures of subject and VSO direct object

reIativization, given in (64) and (65), respectiveIy.

(64) [XiS·

(65) [xkS'

[Infl+VS [aagr]

[Infl+VS [aagr]

[~] [VP]]NP

NP [... t [~]]]

[aagr] VP NP

•

Sadler maintains that NP [~ and NP [~ are lexically governed by Infl

and the trace of V respectiveIy. The chain involving the fronted V and the trace

of V provides the lexical governor for the direct object ernpty eategory. Subject

and VSO direct object ernpty categories are aIso co-indexed with an antecedent

64

• governor-- the operator in C. This co-indexation is sufficiently local to satisfy

the CECP.

In the indirect pattern, pro is doubled either by person inflection or a

clitic on its governing head (x), as iIlustrated below. When the head does not

have agreement forrns (i.e. with a non-inflecting preposition), we find a

resumptive pronoun in base position as in (66b).

(66) a. y[... [x... [pro]i...]]+AgTj

•

•

b. y[... [x... [pronoun]i...]]-AgTj

Given that the structures proposed for both the direct and indirect

strategies satisfy the requirement of lexical government (P in PP, VN for objects

of [-finite] Vs in periphrastics and infinitivaIs), the indirect strategy must

involve a failure of antecedent government. Sadler concludes that the empty

NPs above are not co-indexed with a sufficiently local antecedent, defined in

terrns of an element's Agreement Domain (AD). ADs are specified as S' for

subjects and VSO direct objects, NP for the possessed NP in a possessive

construction, pp for prepositionai complements, and VP if the head M is not a

trace (i.e. in periphrastic VPs). Thus, VP is an agreement domain only if the

verb bas not raised to 1. This allows Sadler 10 account for the fact that VSO

direct objects pattern with subjects and not with periphrastic direct objects; VSO

direct objects and subjects share an AD.

In principle, Sadler's anaiysis allows subjects and VSO direct objects 10

exhibit the indirect pattern. To account for the fact that this option would result

in an unacceptable sentence (the gap strategy must occur where it can), Sadler

6S

• makes the plausible assumption that in a language in which both a gap (direct)

and a resumptive pronoun (indirect) strategy are available for relativization. we

would expect the latter 10 be more or less restricted to cases in which a gap is

unacceptable.

2.4.3.1. NEGATED RELATIVES

•

•

Reca11 that the direct strategy is unavailable for subject and VSO direct

object positions in negative relative clauses.

(67) Ydynion a ddarllenodd [el y llyfrthe men PT read-3sg the bookthe men who read the book

(68) Ydynion na ddarllenasant [el y llyfrthe men Neg read-3pl the bookthe men who didn 't read the book

As we have seen, within Sadler's set of assumptions, there is no failure

of lexical govemment in the negated relatives, so the unacceptability of a gap

strategy must be related 10 a failure in anteeedent govemment. She concludes

that the negative element interferes in sorne way with the C-indexation

mechanism, thus blocking anteeedent govemment of the empty category in

subject or direct object position by the operator in C, and forcing a resumptive

pronoun strategy. Accordingly, like Harlow, Sadler analyzes na as generated

under C.

66

• 2.4.3.2. DISCUSSION OF SADLER'S ANALYSIS

•

The lack of asymmetry between subjects and VSO direct objects is

handled by the fact that the agreement domain is defined as identical for subjects

and VSO direct objects. She derives the extension of the AD for direct objects

via verb raising.

Sadler' s proposai discusses but does not provide a principled reason for

the appearance of the y complementizer with an empty category when ? pp is

c1efted or questioned (the Non-Argument Issue). She is led to stipulate that PPs

do not have an AD. Given that these facts arise with all non-argument clefts and

questions, not merely with PPs, her analysis will have to specify that no adjunct

has such a domain. It would be preferable to capture these facts in a more

principled manner.

With respect to the Co-Occurrence Issue, Sadler (1988:138) mentions the

unacceptability of a pronominal doubling agreement in indirect relativization,

but does not propose an explanation. Accordingly, her analysis fails to account

for a very striking difference between agreement in indirect relatives and

agreement in non-relativized structures.

2.4.4. ROUVERET (1990): A BARRIERS ApPROACH

•

Rouveret (1990) analyzes both clitics and inflectional agreement as

clitics: XO-level elements, base-generated in argument positions. Rouveret's

proposai captures the differences between direct and indirect relativization by

analyzing the pre-sentential particle a as a clitic which (under his assumptions)

cannot raise out of a maximal projection in Welsh.

This enables him to maintain that the direct and indirect relativization

structures are rather similar. In bath types, a c1itic element is affixed onto the

67

• minimal head governing the relativization site. He bases this on the observation

that the same Iocality conditions appear to hold bctween the clitic and the

position it duplicates.

Neither the effect of negative complementizers nor the impossibility of

an in situ pronominal doubling agreement in the indirect relativization pattern

are directly addressed in his paper, but his proposaI will be evaluated in the light

of these and other facts.

Like Sadler, Rouveret takes the obligatory occurrence of a clitic in the

indirect cases to suggest that wh-movement and NP-movcment are not available

options in those structures. Thus, the position governed by a lexical head N,

VN, or Pis not an appropriate site for NP-trace or wh-trace in Welsh.

His analysis rests on the claims listed in (69).

• (69) a. Welsh phrase structure includes both a VP and a VNP (verb-nounphrase).

•

b. Welsh X-bar differs erucially from other familiar languages in notprojecting the level XP. Accordingly, there are no Specifierpositions in Welsh. Elements that would otherwise be analyzed asoccurring in a Specifier are treated as adjunction structures. Thus,subjects are generated adjoined to VP, and remain there at S­structure.

c. Agr is not instantiated in Welsh; 1 is exclusively spccified for thefeature [tense).

68

• d. Adjunction 10 argument categories is not permitted. VNP isincluded as an argument category. Thus, only VP and IF can beadjoined to.

•

e. Clitics are heads, and traces of clitics are subject to the ECP.

f. A version of minimality as proposed in Chomsky (1986) isasssumed. Thus, unlike Rizzi (1990), the definition of minimalityis not relativized

If Welsh clitics are indeed to be analyzed as heads, as Kayne (1991)

analyzes French clitics, then an explanation is needed for the fuct that there is

no clitic-elimbing in Welsh. Rouveret's explanation for this forms part of a

wider proposai that movement is generally restricted in Welsh. His proposai is

outlined below.

2.4.4.1. CImes IN WElSH

•

Clitic-elimbing rtSults in a clitic attached 10 a lexical item of wliich it is

not an argument. This can be illustrated with en-constructions in French. The

genitive clitic en is J'l'alized not on the nominal head that governs the empty

position, but on the verbal head that immediately governs the NP domain.

(70) J'en ai vu le frere, de Marie.1 of-her have seen the brother, of Marie1have seen Marie's brocher.

According 10 Rouveret, c1itics in Welsh always occur on the minimally

goveming head. The sentences given below, where the complement clitic

surfaces on the pre-sentential emphatic partic1e Je, rather than on the minimal

governing head, are ungrammatical.

69

• (71) *Fe'i brynais i [dy [e]].EMPH-3sg bought-ls 1 house1bought his house.

(72) Fe brynais i [ei dy [e]].EMPH bought-lsg 1 3sg house1bought his house.

(73) *Fe'i wnaeth y plant [ddarllen [e)).EMPH-3sg did-3sg the children readThe children read il.

(74) Fe wnaeth y plant [ei ddarllen [e)).EMPH did-3sg the children 3sg readThe children read il.

In other cases, however, these pre-sentential particles can support a

clitic; specifically, if the clitic is a VSO direct object clitic.

(75) Fe'i ddarlleno<!d y plant [el.EMPH-3sg read-3sg the childrenThe children read il.

This leads Rouveret 10 claim that in these cases a VSO direct object clitic

can surface on the pre-sentential particle because that particle is the minimal

goveming head for the direct object, due 10 verb-raising. We will retum 10 this

point later.

2.4.4.2. Two RELA'llVIZA'lION PATl'ERNS

Rouveret captures the distinctions between direct and indirect

relativization structures as follows: In the former, the goveming head is a tensed

70

•

•

V, not a VN or a P, and there is only one maximal projection separating the

empty ca~gory in relativized position from its antecedent.

He argues that the unacceptability of pre-sentential a in the indirect cases

is related te the constraints on movement in Welsh. He proposes that a be

analyzed as a relative pronominal clitic, base-generated in argument position,

which moves up te the minimal head goveming the relativization site: in the

case of subjects and VSO direct objects, this is [V+tense]. Since the tensed verb

is always sentence-initial, except in cases where a pre-sentential particIe is

found, this relative clitic appears to be in the same position as yr.

In defense of the cIaim that a is substantially different in kind from yr (a

position taken earlier by Harlow (1981», he notes that a has certain clitic-like

properties, incIuding the fact that it cannot be separated from the verbal head te

which it is affixed except by an intervening pronominal clitic. lS The relative

clitic, being a head, cannot raise out of an opaque domain, and is trapped if

generated inside of a maximal projection (pP, NP, VNP). Rouveret's analysis

ofXPs as opaque domains in Welsh is descnbed below.

2.4.4.3. No SPECIFIERS, ND XP MOVEMENT

Based on the fact that questions and cIefts pattern with relative clauses,

Rouveret must restrict wh-movement in WeIsh, as weIl as clitic-movement. He

does this by claiming that Welsh phrase structure does not include specifier

positions, which makes al! maximal projections in Welsh barriers te

govemment.

Within a Barriers-type fiamework (Chomsky (1986a», the specifier and

the head of a domain can be govemed from the outside, but its complements

71

•

•

•

cannot. Thus, in the structure gi\'cn bclow. if a L-marks X·, thcn y is go\'crncd

by .:x, but Pis not.

(76) ...a [y [ ...X ...p...]]X" X'

The presence of a singlc barrier bctwecn a and P prc\'cnts a from

goveming p. The presence of two or more barricrs bctwecn a and P pre\'cnts

syntactic movement from Pto a. Thc dcfinitions of thc ECP and Subjaccncy are

given below.

(77) PROPER GOVERNM:!ôNT:

a propcrly govems Piff a 6-govems or antcccden' govcms p.

(78) SUBJACENCY:

If (ai, ai+ r) is a link in a chain, then ai+ 1 is subjacent 10 ai.ai+ 1 is subjaccnt to ai iff thcre is at most onc banicr for ai +1that excludes ai.

The combination of restrictions on adjunction and the proposed lack of

spccifiers in Welsh blocks wh-movement out of maximal projections thal arc

barriers for govemment. He therefore assumes thal in questions and c1efts, no

movement occurs except in the case of local subjects and VSO direct objects,

where movement is restricted to raising of the relative c1itic a 10 the minimal

goveming head (the verb which has raised to 1).

Recall that a c1itic generated within a possessive NP cannot raise to a

pre-sentential particle (example (71». However, as shown in (79), VSO direct

object clitics can surface on these particles.

72

• (79) Fe' i ddarllenodd y plant.EMPH-3sg read-3sg the childrenThe children read il.

Rouveret maintains that the clitic fust rai= te V, and then V-to-I

movemolt neutralizes the barrierhood of VP, allowing an object clitic te raise

further te a pre-sentential particle.

2.4.4.4. NEGATED RELATIVES

•

Consider now how Rouveret's analysis might handle negative relatives in

Welsh. His daim is that the relativf.; paniclea can occur on the tensed V if the

relativized position is, in sorne sense, clCise enough. In non-periphrastic

sentences, both the subject and direct C'lIject positions meet this locality

condition. However, if the relative clause is negated, then the relativized

argument is too far away. .The indirect strategy found with negated relative clauses suggests that

sentential negation introduces a barrier. This barrier could be related te the

presence of an additional maximal projection, Negation Phrase (NegP). Note,

however, that adjUllClion te NegP would have to be blocked, and NegP would

have te be generated without a specifier, in oroer te capture the fuct that

interrogative and cleft constructions pattern with relative clauses with respect te

negation. In principle, Rouveret disallows adjunction only te argument-type

maximal projections. He would therefore have te claim that NegP is an

argument-type maximal projection, an unintuitive result.

2.4.4.5. DISCOSSION OFROUVERET'S ANALYSIS

Other problems are associated with Rouveret's treatment of WeIsh.

Rouveret analyzes agreement markers as clitics, base-generated in argument

73

position and then raised 10 their minimal goveming head. Accordingly. his

analysis leads one to expect the lack of clitic doubling in relativized structures,

but fails 10 exp1ain why doul>ling is in fact permitted in non-relativization

structures and in cases of relativization out of complex NPs ('true islands',

discussed earlier).16

Rouveret points out certain inadequacies in his analysis. First, it cannot

account for the dependence of object clitic climbing in VSO on the presence of a

pre-sentential partic\e. Also, it does not capture the lexical class of elements !bat

cannot support c\itics in Welsh (adjectives, certain prepositions, etc.).

Finally, it is worth emphasizing that his analysis requires a phrase

structure for Welsh which is radically different from other languages, a proposal

that 1will argue is not necessary.

•

• 2.4.5. SUMMARY

e

The analyses discussed above, which appeal 10 constraints on empty

categories to derive the different agreement patterns in Welsh relativization, are

not entirely satisfactory.

First, because a strong causal relationship is posited between the choice

of complementizer and the agreement patterns, there is no natural explanation

for why pied-piping in clefts and wh-questions requires the yr complementizer,

associated with the indirect strategy, but neither an agreement clitic nor a

pronominal marks the extraction site.

Second, while these analyses account for the aceeptability of an empty

category in the relativized position in direct relatives, they do not provide a

unified explanation for the fact that, whereas a pronominal can double

agreement in a non-relativized position, it cannot in indirect relativization. This

74

•

•

is a robust fact that should fall out naturally from an adequate analysis of

relativization in Welsh.

In addition, by treating the indirect strategy as essentially a resumptive

pronoun strategy, there is no explanation for certain relativization structures

(i.e. relativization out of complex noun phrases) which, unlike bath the direct

and indirect patterns, do aIlow a pronominal ta co-occur with agreement or a

clitic marking the relativization site.

Finally, relativized subjects and VSO direct objects are treated

identically in these analyses; neither a pronominal nor an agreement marker can

mark the relativization site. By removing the subject-object asymmetry in a

radical manner, i.e. by strucrure-pruning (Harlow 1981) or by voiding

barrierhood, there is no immediate explanation for the fact that a subject-object

asymmetry does appear in negated relative clauses. Whereas negated subject

relatives have only one option (obligatory agreement on the tensed verb, and an

obligatory [e] in argument position, the pattern associated with the indirect

strategy), it is not clear that the indirect strategy is found with negated VSO

direct objects. 11Iese exhibit a pattern associated with unrelativized structures,

an optional pronominal in the extraction site, as long as an agreement clitic

appears on the negative marker.

1 will argue for a revised analysis of these facts, one which suggests a

different account of relativization, agreement, and the role played by sentential

negation.

7S

• 2.5. A NEW PROPOSAI..

2.5.1. INTRODUCIlON17

•

•

In this section, l propose an analysis for the paradigm involving

agreement and negation in Welsh relative clauses. Two of the syntactic roles

played by negation that will be relevant for this thesis will be introduced: its

role as a potential barrier for Xo raising, and its role in Case-licensing.

l will present arguments for a revision of the treatment of relativization

in Welsh, eliminating the traditional distinction between direct and indirect

relativization. l will analyze sentential negation in Literary Welsh as projecting

a Negation Phrase, generated above TP. The phenomenon of agreement

surfacing on the negative marker in Welsh will be analyzed, and the restriction

on agreement-marking on the negative marker to direct objects in simple tensed

clauses will be accounted for.

The analysis will exploit recent proposals concerning th.: analysis of

agreement and X-Bar theory. l will claim that both the direct and indirect

patterns reflect wh-movement. Agreement will be treated as the reflex of an S­

structure Case-checking operation, and the different pre-sentential particles and

agreement patterns associated with the distinct relativization strategies will be

analyzed as resulting from the configurational relationship between the A'-chain

formed by operator movement to the specifier of CP and a Case-licensing head.

Like the analyses described above, l will argue that a resumptive strategy

is found in Welsh only when wh-movement is not available, but l will not

analyze the so-ca1led indirect strategy as being forced by the ECP. A resumptive

strategy will be proposed only for those cases of relativization where an in situ

pronominal may be found, as is the case with relativization out of complex NP

76

• islands, out of PPs headed by non-inflecting prepositions, and of VSO direct

objects in negated relative clauses.

The discussion is organized as follows. An introduction 10 the structural

assumptions and the treatment of a.,areement and Case assumed is presented

below. In section 2.6 1 propose a unified analysis of Welsh relative clauses

which accounts for the pre-sentential particle and the agreement patterns in the

direct and indirect strategies. In section 2.7 the analysis is extended 10 negated

relative clauses and we address the phenomenon of agreement surfacing on the

sentential negation marker.

2.5.2. TBEoRETICAL FRAMEwORK

This analysis is based on the following initial assumptions about phrase

structure. Structures obey a binary-branching requirement.. AU categories obey

X-bar theory and may project a specifier. 1 also assume a two-tiered VP (Larson

(1987», the verb being generated in the head of the lower VP, and the head of

the higher VP being empty. The head of the higher VP will be referred ta as

Della, and 1 will occasionally refer 10 this phrase as a Delta-phrase in order 10

distinguish it from the lower VP. Subjects will be generated in the specifier of

the higher VP (Kitagawa (1986); Koopman & Sportiche (1988». 1 will follow

Pollock (1989) in positing a TP, but 1 will not inc1ude AgrP in this analysis.18

Somewhat more radically, 1 will assume that a TP is generated only in

[+tense] sentences in WeIsh. Given the assumption that subjects are generated

intemal 10 VP, in the specifier of the higher VP in a Larsonian she11, the surface

•

•

2.5.2.1. STRUcnJRAL ASSUMP110NS

77

• SVO order found in infinitivals marks the order of these constituents in their

base positions within VP.

A language may include the category Negation Phrase (NegP) in its

phrase structure. 1 will maintain that NegP in Welsh can only he generated in a

sentence that instantiates TP.19

2.5.2.1.1. THE POSITION OF NEOP

•

A negative sentence in fonnal Welsh takes the simple fonn of

introducing a negative particle initially before a positive sentence. There are five

such negative pre-sentential particles, depending on whether the sentence to be

negated is a statement, a tag-question, an imperative, or a relative clause. In this

section we will be concerned only with the first and the last kinds of sentences,

which require the negative e1ements ni(d) and nard), respectively.20 They are

illustrated below.

(80) Mae John yn arcs.is-3s John in stayJohn is sraying.

(81) Nid yw John yn arcs.Neg is-3s John in stayJohn isn 'r sraying.

[Jones & Thomas 1977:317-318]

(82) y dynion a ddarllenodd [el y llyfrthe men Fr read-3sg the bookthe men w.1o read the book

78

•

•

1

(83) Ydynion na ddarllenasant [el y llyfrthe men Neg read-3pl the bookthe men who didn't read the book

[Harlow 1981:237]

Consider the simple case of a negated statement. Welsh is a surface VSO

language. The negative element always occurs in S-initial position, immediate1y

preceding the tensed verb. Nothing can intervene hetween the negative marker

and the verb.

Proposals as 10 the base position of NegP cross-linguistically include

generating it immediate1y below TP, immediately above TP, AgrP, and VP.

Analyses which do not allow movement of the negative head itself ascertain th~

base position of NegP by its S-sttucture position relative 10 other constituents.

However, once we allow for the possibility of treating the negative marker as a

head which is itself capable of movement, the ground becomes very unstable

indeed.

Given that the negative marker is the leftmost lexical item in a simple

sentence, if the negative marker itself has not undergone movement then the

most straightforward assumption would be that NegP is the highest functionaI

category in S.21 However, if we analyze the negative marker as a head, it could

be argued that it arrives at the S-initial position in Welsh from any of a number

of lower potential base-positions. This movement could occur independently

(i.e. forœd by its clitic nature, as argued by Pollock (1989) and Belletti (1990»,

or it could he driven by the requirements of the Head Movement Constraint,

whereby the verb would he forœd 10 move through Neg, necessarily taking the

negative marker along with it as it moves up the structure (Ouhalla (1990».

79

• However, l will adopt the more sttaightfowarC: initial assumption that NegP in

LW is generated above TP.

Given a treatment of sentential negation as a hcad, we can appeal to its

status as a potential barrier to support this base position for NegP. If negation is

generated below TP, we might expect it 10 block V-ta-T raising, forcing a verb

10 surface without [tense] marking in a negated clause as in do-support sentences

in English. This is not the case. Negation in Welsh will be shown 10 interfere

only with V-raising to C.

2.5.2.2. AGREEMENT AND S-STRUcnJRE CASE-CBECKING

•In non-relativized structures, the agreement paradigm is sttaightforward.

As described earlier, agreement cannot surface if the argument in question is a

non-pronominal NP but is obligarory if the argument is pronominal. A

pronominal that triggers agreement can then optionally drop.

Throughout, l take the proclitics that surface on nouns and untensed

verbs to be derived under the same conditions as inflectional agreement.

Whether these agreement features are prefixed or sufflxed 10 the head is

determined by the lexical specifications of the head. The basic agreement

paradigm is repeated below for convenience.

(84) SUBJECT AGREEMENT

a. Gwelodd (*gwelsant) y dynion y ci.saw (saw-3pl) the men the dogThe men saw the dog.

80

• b. Gwelsant (*gwelodd) (hwy) y ci.saw-3pl (saw) (they) the dogThey saw the dog.

(85) OBJECT AGREEMENT

a. Mae Wyn yn son arndano (ef).is-3sg Wyn PROG speak about-3sgm (him)~n is spea/dng about him.

b. Mae Wyn yn prynu ci dy (ef).is-3sg Wyn PROG buy 3sgm house (him)~n is buying his Muse.

•c. Mae Wyn yn ci ddarllen (el).

is-3sg Wyn PROG 3sgm read (it)~n is reading it.

[Sadler 1988:123]

Object agreement never surfaces on a tensed verb. Thus, in VSO

structures, the direct object pronominal is obligatory. This is illustrated in (86).

(86) Gwelsant (hwy) *(ef).saw-3pl they itThey sawil.

The position taken here is that agreement in Welsh reflects the operation

of Case-checking. In much =nt work within GB theory, inc1uding Chomsky

(1992), Sportiche (1990), Mahajan (1990), Johnson (1990), and Noonan (1992),

the proposai that structural Case must be checked in a SpecifierlHead

configuration, either at S-Strueture or at LF, bas been adopted. This is

formulated here as the Licensing Condition on Chaïns (LCC), given in (87).

81

• (81) LCC: A chain must be Case-checked at S-structure or at LF.

1 will adopt the view, advocated by Sportiche (1990) and Mahajan

(1990), that agreement reflects Case-checking in a Spec/Head configuration

established at S-structure.22 The definition of a licit Spec/Head configuration

will be formulated in terms of chains. Given a maximal projection XP, a licit

Spec/Head Case configuration holds between an element in the head of XP and

an element in the specifier of XP if the following conditions hold.

•

(88) Llcrr SPEc/HEAD CONFIGURATION:

(i) The element in the head of XP must be a morphologicallycomplete member of the chain of a head specified [+Case).and(ii) The element in either the head or the specifier of XP must bethe head of a chain {i.e. not a trace).

This can be illustrated by the structures given in (89), where H is a Case­

licensing head. The structure in (ii), where both the element in the head of XP

and the element in the specifier of XP are traces, is not a licit Spec/Head Case

configuration.

(89)

(i) (ii) XP

~t

•Under these assumptions, the obligatoriness of agreement with

pronominal NPs and the unacceptability of agreement with non-pronominal NPs

82

•

•

has a possible explanation, one that appeals 10 the distinction between S­

structure and LF Case-ehecking. Non-pronominal NPs are not Case-ehecked

until LF; accordingly, they do not trigger agreement. Pronominal NPs, on the

other hand, have 10 he Case-ehecked in a specifier position at S-structure.23

Recal1, however, the surface word order in Welsh, where a head

precedes its complement even when the complement is a [+pronominal] NP.

This entai1s that pronominal NPs are Case-ehecked at S-structure but do Dot

mave 10 the specifier position until LF. We will maintain that at S-structure the

Lee is satisfied by pronominals via the formation of an A-ehain by indexation

between the in situ pronominal and a null pleonastic in a licit SpeclHead Case

configuration.

The chain that is formed can be compared 10 the chain between there and

the argument NP in existential sentences like (90), which allows agreement 10

be triggered on t'Je verb at S-structure, although the NP three men does Dot raise

to replace the expletive in subject position until LF.

(90) There were three men in the room.*There was three men in the room.

The pronominal NP can be dropped, under identification by agreement

morphology on the governing head (Rizzi 1986). A pronominal which has Dot

triggered agreement cannot he dropped. The S-structure configuration of a

pronominal prepositional object and a pronomina1 subject of a [+finite] verb are

illustrated in (91).24

83

• (91 ) pp

~~! P +agr-i NPi! +pronominal!,1 A1 :, '! !

::xplctivc-argumcnt cbain

NPs! + pronominal!A

...* ..._ •.__ .• __ t

c.~lctive-argurncnt chain

2.5.2.2.1. OBJEer-OBJEer AsYMMETRIES

•One of the more problematic facts for previous analyses of Welsh to

account for is the inability of pronominal VSO direct objects to trigger

agreement on a tensed verb. This distinguishes VSO direct "bjects from

periphrastic direct objects, which trigger agreement on the lower [-tense] verb

(VN). Because it does not trigger agreement, a pronominal VSO direct object,

unlike a pronominal periphrastic direct object, does not have the OptiOll of being

dropped.

(92) Welais (i) ·(et).Saw-lsg (1) himlsawhim.

(93) Mae Wyn wedi ei weld (et).be-3sg Wyn PERF 3sgm seel\-)II Ms seen him.

Under the analysis proposed here, this distinction is expected. The verb

in a tensed sentence is not morphologically complete until it bas incorporated

with the tense morphology generated under T. The verb in T is in a SpeclHead

84

• configu;ation with SpeclTP, a position available for the null pleonastic co­

indexed with the subject, not with the object.~ Accordingly, a direct object

pronominal will never be in a licit SpeclHead configuration with its Case­

assigner if that Case-licenser is a vero in T. The situation is iIIustrated in (94).

t-v NP-k

1 :..: t...__.._.-._.__." :

V+ Agr-i NP·; [+ pronominall

[eli

Î1

i

(7"P

V'

NP-k ...

t-v

e-i

(94)

On the other band, the lower verb in a periphrastic construction such as

(93) does not have to raise to T; it is morphologicaIly complete without [tense]

features. Thus, the null pleonastic co-indexed with the object in the specifier of

the lower VP is in a licit SpecfHead Case configuration, and agreement surfaces

on the untensed verb. This is shown in (95).

(95)

•

•

85

•

•

Note that, while this analysis rules out Case-licensing the pronominal

direct object via SpeclHead coindexation at S-structure, the manner in which

pronominal VSO direct objects satisfy the Case-licensing requirement has not

been specified. There are two possibilities. First, it could be aIgued that the

pronominal can be Case-licensed at LF, where the requirement of morphological

completeness is relaxed.26 In this way, the lack of surface agreement with these

pronominals would follow from the fact that they are are not Case-ehecked at S­

structure.

Altematively, if we maintain that a SpecfHead configuration which is not

licit at S-structure is not licit at LF either, the VSO direct objects may be Case­

!icensed via govemment by the vero. 1 will maintain that the latter is the correct

analysis. Sorne support for postulating a Case-licensing mechanism via

govemment cornes from non-inflecting prepositions. Prepositions can he divided

into two classes: those that can be inflected with agreement marking, and those

that cannot be 50 inflected. The latter cannot support either inflectional

agreement or the agreeing pre-head c!itic found on VNs and NPs in possessive

constructions. Instead, an in situ full pronominal must occur.

There are good arguments for making the claim that these non-inflecting

prepositions are best analyzed as not projecting a specifier.27 As such, they do

not provide a position for Case-ehecking in a SpecfHead configuration either at

S-strueture or at LF; in other words, neither [+pronominal] nor [-pronominal]

NPs find a !icit SpecfHead configuration in these phrases. The grammar of

Welsh must therefore allow Case licensing via govemment by the head of pp at

S-structure. Given that Welsh requires a mechanism whereby Case is assigned

other than via SpecfHead coindexation in any case, we will appeal to this

86

• mechanism to account for the Case-licensing of VSO direct objects rathcr than

introducing a relaxation of the definition of licit Spec/Hcad configuration at

LF.28 The notion of Case-licensing under govemment is refincd in the following

section.

2.5.2.2.2. CASE-ASSIGNMENT TO VSO DIRECT OBJECTS

•

•

l will suggest that an argumen: can be Case-licensed by the verb undcr

govemment rather than by Spcc/Hcad coindexation if the following conditions

hold. First, the capacity is sensitive to a position in the structure. Il is movement

through this position that activates the Case-licensing mechanism. In Welsh, the

position at which the V may assign Case under govemment is Delta, the hcad of

the higher VP in a Larsonian shell.29 The member of the verb-ehain in Delta

can assign Case to an XP in the specifier of the projection that it immediately

dominates. Second, if the specifier of the projection that Delta' immediately

dominates is a potential licit Spec/Head Case configuration, the ability to assign

Case to that position under govemment is blocked.30 Third, Case-assignment

under govemment by the verb-ehain takes place at LF, following raising of an

NP to the appropriate specifier position; agreement, the reflex of S-structure

Case-licensing in a Spec/Head configuration, is not triggered.

The LF representation of a VSO sentence in Welsh with a non­

pronominal direct object is given in (96).

87

• (96) TP

t-v

4NPobj

CASE t-v t-obj

•The member of the V-chain in Delta Case-licenses the direct object. In

the next section, this treatment of Case and agreement will be shown 10

eliminate the traditional notion of two distinct re1ativization strategies in Welsh.

2.6. AGREEMENT JN RELA11VE CLAUSES

2.6.1. BACKGROUND

•

Previous analyses have treated the relativized position in an indirect

relative as a resumptive pronoun.31 In the analysis which follows, 1 argue that

both strategies involve syntactic movement.

1 will maintaÎn that a resumptive strategy is available in Welsh, but it is

not refiected in the so-eal1ed indirect strategy. Rather, it is found only when

movement is impoSSlble, i.e. in cases of relativization out of a complex or

coordinate NP island.32 In these cases, an example of which is given in (97), the

c1aim that movement is blocked is consistent with the acceptability of a full

88

•

•

•

pronominal in the relativized position co-occurring with agreement marking.

The resumptive pronominal in (97) is in a conjoined NP.

(97) Dyma'r eneth garedig y diolchais i [iddi hi a'i thad).here-is-the girl kind that thanked-lsg 1 to-3sgf her and her fatherHere is the lànd girl tJuu 1 thanked (her) and herjarher.

Compare this with an instance of the indirect strategy, given in (98),

where a pronominal argument is complete1y ungrammatical.

(98) Dyma'r eneth garedig y diolchais i iddi (*hi).here-is-the girl kind that thanked-lsg 1 to-3sgf (ber)Here is the lànd girl that 1 thanked.

A natural inclination is to argue that these contructions do not involve

wh-movement, and that the optional surfacing of the pronominai indicates that

the pronominal NP remains in its base position and can be dropped under

identification by agreement on the governing head, just as in non-relativized

structures. This raises problems for analyses which maintain that the indirect

strategy, charaeterized by the inability of an in situ pronominal to surface, does

not involve movement.

1 will propose an alternative explanation for the agreement patterns

associated with the two relativization strategies. Both will be analyzed as

involving wh-movement to spec/CP. The agreement marking the relativization

site in the indirect strategy will be argued to reflect C3se-checking of the A'­

chain created by wh-movement; not the presence of a resumptive pronoun.

Since 1 am denying the claim that two distinct strategies are involved, 1

will re1"er to the direct and indirect patterns of relativization. The pre-sentential

partïcles a and yr are analyzed as complementïzers. The patterns of agreement

89

•

•

associated with the two strategies are characterized for convenience below,

where H represents a head.

(99) Direct Pattern:

NP...[CP ÜJli a [TI' ... [HP H(*agr)..til ..l

(100) Indirect Pattern:

NP... [CP Opi yr [TP...[HP H*(agr) ..ti l···l

Since 1 am proposing a movement analysis for both the direct and

indirect patterns, the challenge is not to explain the lack of an in-situ pronominal

in these structures (a constraint which is common 10 both patterns and which is

expected if there has been wh-movement) but 10 motivate the different

agreement patterns exhibited in the direct and indirect cases.

Consider first the fact that subjects and direct objects of tensed verbs

pattern 10gether with respect 10 relativization. There is one feature which

subjects and VSO direct objects in Welsh share that di~tinguishes them from an

other arguments: the heads that are responsible for Case-licensing these NPs, V

and T, are combined at S-structure. 1 will claim that the differences between the

direct and indirect relativization patterns follow from the different positions in

which Case-checking takes place.

1 will assume that the [V+1] complex raises 10 C in relative clauses.33

Both the indirect and direct strategies will be analysed as wh-movement 10

SpeclCP, and the distinct agreement patterns and complementizer selection will

be argued 10 result from differences in head movement. It will then follow that

only subjects and direct objects of tensed verbs exluoit the agreement pattern

90

•

•

and complementizer associated with the direct strategy because it i~ only their

Case-checking heads which raise to C. Other Case-licensing heads (i.e.

prepositions, [-finite) verbs, and noun heads of possessive NPs) remain within

their projection.

In this way, the LCC hCllds for A- a.'1d A'·chains in Welsh. In

accordance with the LCC, the A'-ehain created by this movement must he Case­

checked at S-sttucture. In cases which involve wh-movement, the A'-ehain may

create more than one !icit SpeclHead configuration. 1 will maintain that a

constraint operates such that agreement is always realized at the lowest licit

SpeclHead configuration.34 This is formulated as the Condition on Agreement

ReaIization in (101):

(101) CONDrnONON AGREEMENTREA1JZAnON (CAR):Agreement is triggered once, at the lowest licit SpeclHeadconfiguration.

2.6.2" DERlVING THE DISllNcr COMPLEMEN'I1ZERS

The distinct complementizers, y(r) in the indirect pattern and a in the

direct pattern, will he derived as follows. The head of CP is taken to he y(r).

Since it is the default C, this claim is consistent with the fact that it is the form

found in regular embedded clauses, illustrated in (102).

(102) Rwfyn gobeithio [y daeth John ddoe).be-1sg in hope [C came John yesterday)1hope thal John came yesterday.

If C is in an agreement configuration at S-structure (that is, if the A'­

chain is case-checked in ~CP) then C surfaces as a.35 The particle a is thus

91

CP

/

analyzed as the realization of (C+agr]. The fonTi agreement takes is highly

impoverished, and does not vary according to number, person or gender.

However, it is sufficient to satisfy the requirements of the LCC.36 The

configuration triggerÎilg an agreeing C is illustrated in (103).

(103)

Op-k C

1 7' ~C+agr-k ~

! V+T...._-_!

•

•This analysis predicts that if the complementizer is a, there will be no

agreement with the relativized argument elsewhere in S. Since Case is checlœd

only once on a given chain, agreement on C precludes agreement with the same

argument elsewhere. In this way, the complementarity between the pre­

sentential particle a and agreement with the relativized NP elsewhere is

explained without resorting to the claim that a is a pronominal (Harlow 1981) or

a clitic (Rouveret 1990).

While the complementizer a will never co-occur with agreement realized

on a head lower in the sentence, this analysis does not make the reciprocal

prediction. Thus, a lack of agreement marking the extraction site does not force

the a complementizer. If agreement, the realization of Case-licensing, is not

triggered, then we should find neither agreement marking the relativization site

nor an agreeing complementizer. This prediction is discussed in section 2.6.5.

92

• 2.6.3. DERIVING THE INDIRECT PATfER1"

•

•

As an illustration of relativization resulting in the indirect strategy.

consider relativization out of a prepositional phrase. Under the assumptions

specified above, the object of the preposition undergoes wh-movemerlt to

spec/CP. Movement must proceed through the argument's A-position, which in

the case of pp complements is the specifier of PP. The A'-chain headed by the

operator in specJCP includes a trace in spec/PP. Since the head of PP is

morphologicaIly complete and does not raise, its D-structure position determines

the lowest potential SpecJHead configuration for agreement. Thus, in

accordance with CAR, agreement is realized on P. Since agreement is realized

only once for a chain, we find the non-agreeing C. This is illustrated in (104).

(104) Ydyn y canodd efamdano (*ef)the man comp spoke-3sg he about-3sm (*him)the man rhar he spoke about

The structure of (104) is given in (105).

93

amdano

t·k<----/0-! P+Agr·k t·k1,! l ,_____._..L--_...~..,.;

11,

et

Np·jt Ai'__J:

11,111111111,11

canodd

~ V+[T+Agr·iJ111111,,,,

NP

y dyn y

(105)•

•Note that, even though the [V+1] complex has raised to C, agreement

with the subject pronominal is triggered in specfTP. This is predicted by the

CAR: the configuration at TP is the lowest licit SpeclHead configuration for the

pronominal subject at S-structure. The specifier contains the pleonastic head of

the chain, and the head of T dominates a morphologically complete member of

the chain of the raised V.

A question should be raised at this point. This account incorporates V­

movement to C that is not forced by morphological requirements. What drives

this movement? 1 will propose that the driving force is the wh-operator in

speclCP. In a structure where the speclCP does not contain an operator, i.e. in

non-relativized embedded clauses, V-to-C movement does not take pIace. This

94

•

•

daim will be addressed further in section 2.6.6, when long-distance

relativization is ciiscllssed.

This analysis accounts straightforwardly for the fact that VSO direct

objects exhibit the direct strategy but direct objects in periphrastic constructions

exhibit the indirect strategy. Recall that in periphrastic constructions,

relativization of the direct object results in agreement on the untensed verb, and

the complementizer is y(r). This is illustrated in (106).

(106) Ydyn Y mae Sion wedi ei weld (*ef)the man comp is-3s John PERF 3sgm see (him)the man thaJ John has seen

In these cases the Case-licensing head (the VN) does not raise out of its

projection; hence its D-structure position determines the 10west potential

SpeclHead configuration for Case-checking. Since the chain of wh-movement

includes a trace Ïil spec of the 10wer VP, the configuration is licit and agreement

surfaces on the verb, in the form of the clitic ei. The S-structure of an instance

of relativization of a periphrastic direct object is given in (107).37

95

• (107)

•y dyn y mae Son wedi i ë\gr-k

: ",__.•.•.' l

11111

ei weld

t·k

The realization of agreement, the form of the complementîzer, and the

unacceptability of an in situ pronominal that characterize the indirect pattern of

relativization are thus accounted for. Wh-movement to spec/CP does occur, and

agreement is reali:ml at the lowest licit SpeclHead configuratior:: either on the

preposition, if a prepositional object is relativi:ml, or on the [-tense] verb, if a

periphrastic direct object is rclativi:ml. The complementizer is therefore the

non-agreeing form of C, whicb surfaces as yr. An in situ pronominal cannot co­

occur with agreement in relativization contexts because it bas undergone wh­

movement to spec/CP.

However, wben the direct object of a tensed verb (a VSO direct object)

is relativi:ml, agreement does not surface on the verb. Instead, we find the

96

• direct pa~ern; the complemenùzer is a. and neither a pronominal nor agreement

marks the extracùon site.

2.6.4. DERIVING THE DIREcr PATI'ERN

•

Consider fust subject relativizaùon. In a non-relaùvized sentence. T

remains in its D-structure position, and if the subject NP is pronomillal. the

coindexed null pleonastic in speclTP triggers agreement on T. This agreement is

realized as personlnumber agreement. However, in a relaùvized structure, the

verb moves through T, where it picks up tcnse morphology, and thcn raises to

C. The subject operator is in spec/CP. Case cannot be checked at TP, a lower

potential SpeclHead configuration, because neither the head nor the specifier of

TP contains the head of a chain, as required by part (ii) of the definiùon of a

licit SpeclHead configuration given in (88). Since the Case-checking head is in

C and the subject operator is in spec/CP, the lowest potential Spec/Head

configuration for the A' chain created by wh-movement is between C and

spec/CP. Accordingly, we find agreement on C (the head of the projecùon that

[V+T] has adjoined to) instead of the rich person/number agreement associated

with specITP. The agreeing form of the complementizer is a.

The S-structure representation of subject relaùvization is shown in (109).

(108) Ydyn a ddarllenodd [el y llyfrthe man C+agr read the bookthe man tJwl read the book

97

• (109)

t-V NP

L:::::..

Now, consider the case when a VSO direct object is relativized. When

the relativized argument is the direct object of a finite clause, the direct object

operator is in spec/CP. The Case-licensing head for the direct object is V. As

with all cases of affirmative relative clauses, the verb raises to T, and then

[V+T] raises ta C. Since the Case-licensing head for the direct object is in C,

the lowest potential Spec/Head configuration is again between the complex head

dominated by C and its spec position. Agreement is realiwl on C, the head of

the complex head, which then surfaces as Q. The S-structure representation of

VSO direct object relativization is shown in (111).

•y dyn a ddarllenodd y lIyfr

(110) Ydyn a welodd Sion le]the man C+agr Sion sawthe man thal Sion saw

• (111 )

t-V t-ky dyn a welodd Sion

•

•

In example (110), there is no subject agreement because the subject NP

is [-pronominal]. However, given the same structure with a pronominal subject,

agreement is established at TP even though the [V+T] complex has raised

further to C.

(112) Ydyn a welsant (hwy) [elthe man C+agr saw-3pl (they)the man rhal they saw

The head of TP dominates the trace of a morphologically complete item

([V+T]), and the spec of TP dominates the pleonastic head of the chain created

by the pronominal subject. 'This satisfies the requirements of a licit SpeclHead

Case configuration. Accordingly, personlnumber agreement triggered by Case­

licensing of a pronominal subject in specITP is found in affirmative relative

clauses as long as it is not the subject itse1f that is relativized. If the subject is

99

•

•

relativized, then both the head and specifier of TP dominate a trace, and TP is

not a licit Spec/Head Case configuration.

In this way, the agreement patterns and complementizer selection

associated with the direct and indirect strategies reflect the S-structure position

where Case-licensing takes place. In the following section, l present data from

wh-que:;~ons and clefts which provide additional support for a Case-based

treatment of agreement and complementizer selection.

2.6.5. WB-QUESTIONS AND CLEFfS: THE "1NDIRECX'"COMPLEMENllZER WlTB THE "DIRECX'" AGREEMENT PATI'ERN

With respect to the choice of complementizer and the agreement pattern,

wh-clefts and questions in Welsh pattern with relative clauses; the direct pattern

is found if the subject or the direct object of an inflected V is questioned or

clefted. Otherwise, the indirect pattern is required. These data were given in

(39)-(42).

In Welsh, relative clauses may not be headed by a PP. However, in

clefts and questions, there are two options: stranding the preposition or pied­

piping the whole PP. If the preposition is stranded, we find the typical indirect

pattern: the complementizer is y(r), and agreement surfaces on the stranded P.

This is illustrated in (113).

(113) [Pa tren] Y darllenodd y dyn Y llyfr [amo le]]?which train C read-3s the man the book on-3sgmWhich train did the man read the book on?

However, ifpied-piping occurs, the y(r) complementizer is still required,

but there is neither an agreement marker nor a pronoun marking the extraction

site.

100

• (114) [Ar ba tren] y darllenodd y dyn y llyfr [el?on which train C read-3s the man the bookOn which train did the man read the book?

These facts provide support for the account proposed herc, which derives

agreement marking and complementizer selection by Case-licensing. Given that

a is the rea1ization of [C+agr], it should surface only with A'-ehains which are

Case-ehecked. Since a pp chain does not require Case, we predict that

agreement will not be triggered on C.

In this way, questioning, clefting or relativizing elements that are not

assigned Case should always exhibit the y(r) complementizer, in spite of the

apparent gap in the relativized position. This prediction is correct, as illustrated

in (115).

(115) [Pa bryd] Ydaw ef [el?what time C will-eome-3s heWhen will he come?

[Sadler 1988:166]

2.6.6. LoNG-DISTANCE RELA1IVIZAll0N

As mentioned earlier, the so-ca11ed indirect strategy of relativization is

found when relativization takes place out of an embedded clause. Thus, even

subject relativization, which exhibits the direct agreement pattern in both

periphrastic and non-periphrastic sentences, confonn to the indirect pattern

when in an embedded context.

(116) NON-EMBEDDED SUBJECT:

y dynion a ddarllenodd y llyfrthe men C+Agr read the bookthe men thor read the book

101

•

•

(117) EMBEDDED SUBJECT:

y plant y dywedodd Wyn y darllenasant [el y llyfrthe children C said Wyn C read-3pl the bookthe children thot vryn said had read the book

These cases reflect a wh-movement strategy, as shown by the

unacceptability of a pronominal in the relativization site, a fact attested to by

Sadler (1988: 112). The proposai that these structures involve wh-movement is

also supported by crossover effects, cited in Hendrick (1988:185ff).

(117b) Ybachgen ddywedodd e oedd Mair wedi sôn arndano [elthe boy C say-past he be-past Mary PT talk about-3sgmthe boy thot he said Mary had been taIking about

[Hendrick 1988:192]

The sentence is ungrammatical on the reading where [el is bound by the

pronominal e(j) ('he'). This would be surprising if [el is a resumptive pronoun.

According to the analysis proposed here, there must be a licit A'-chain

between the operator in the higher C and the reJativization site in the embedded

clause. However, instead of an agreeing form of C, we find person/number

agreement associated with subject agreement established at TP. Thus, agreement

with the reJativized subject is triggered not in the spec of the lower C, as it is in

direct reJativization, but in the spec of TP.

This fact is expected under the analysis presented here. The S-structure

of 10ng-distance reJativization is derived as follows. Wh-movement proceeds

through the spec of the lower C 10 the higher C. However, the V+T complex

raises 10 the lower C only if there is an operator in spec/CP. In these

constructions, the e1ement in spec of the lower C is a trace.

102

•

•

(118)

NP

y plant y dywedodd Wyn y darllenasant

t-v NP

y lIyfr

With a trace of the A'-chain in the specifier and a morphologically

complete member of the chain of a Case-licensing head in T, we expect the

lower TP 10 be the lowest licit SpeclHead configuration for agreement with the

A'-chain of the relativized subject.

It is worth noting that in Breton, a closely related language which has

similar agreement patterns in affirmative and negated relative clauses, long­

distance relativization exhibits a distinct pattern. The embedded clause does not

exhibit subject agreement on the tensed verb; this can be contrasted with the

same sentences where the embedded clause is negated, in which case agreement

on the tensed verb is obliga1Ory, as in LW. This paradigm suggests that in long-

103

•

•

•

distance relativization in Brelon, V-to-C raising takes place regardless of the

presence of an Operalor in spec/CP.

(119) Ar baotred a soiij din a lenne al levriou a zo arnaii.the boys PT think to-lsg PT read-3sg the books PCL is hereThe boys tha! 1 think read the books are here.

(120) Pelore paotred a soiij deoc'h a lenne al levriou?the boys is PT think to-2sg PT read-3sg the booksWhich boys do you think read the books?

[Borsley & Slephens 1989:420]

(121) Ar baotred a soiij din ne lennent (*Ienne) kel al levriou a zoarnaii.the boys PT think to-lsg Neg read-3pl (*read-3sg) not the booksis hereThe boys tha! 1 think did net read the books are here.

(122) Petore paotred a soiij deoc'h ne lennent (*Ienne) ket al levriou?which boys PT think to-2sg Neg read-3pl (*lenne) not the booksl'tÏlich boys do you think did net read the books?

[Borsley & Stephens 1989:425]

In the following section, 1 retum to the role played by sentential negation in

relative clauses, and propose that the specifier ofNegP provides an A-position for

VSO direct objects in both matrix and relative clauses in LW.

104

• 2.7. CASE-UCENSING IN SPEclNEGP IN MAnux AND RELATlYE CLAUSES

2.7.1. NEGATED SUBJEcr RELATIYES

Let us first consider negated subject relatives, which show a fairly

straightforward pattern. Compare subject relativization out of an affirmative

sentence, as illustrated in (123), where agreement is not permitted, with the

negated case in (124), where personlnumber agreement is obligatory, and a

pronominal cannot surface in the relativization site.

(123) Ydynion a ddarllenodd (*ddarllenasant) y llyfrthe men C+Agr read (*read-3pl) the bookthe men who reat! the book

(124) Ydynion na ddarllenasant (*ddarllenodd) y llyfrthe men Neg read-3pl (*read) the bookthe men who didn't reat! the book

ln the preceding section, 1 argued that the difference between the direct

and indirect strategies derives from a difference in the S-structure position of the

Case-licensing head: whereas the 10west Case-licensing head for complements of

Ps, Ns and VNs remains in its projection, the Case-licensing head for the

subject and direct object raises to C. This results in distinct agreement marking

patterns. If the same reasoning is carried over 10 negated relatives, the pattern

we find suggests a paralle1 explanation: in negated relatives, the [V+'1]

complex does not raise 10 C, but rather remains in T.

1 have suggested that negation heads its own projection and is generated

above TP. 1 will further maintain that NegP intervenes between TP and CP,

lOS

•

•

blocking verb movement into C via the Head Movement Constraint (HMC;

Travis 1984).38

The HMC can be reduced ta the operation of the ECP as defined in

tenns of barriers (Baker (1988); Chomsky (1986a». The trace of a moved Xo

must be properly govemed and can only be licensed by the raised XO itse1f

through antecedent-government. This means that the antecedent-trace link of the

chain must not cross a barrier. An intervening head Y creates such a barrier.

This blocks XO-raising from skipping over an intervening head; such a chain

would include a trace that is not properly govemed.

If the claim that the Case-licensing head for the relativized subject (T)

remains in T is correct, then the lowest Case SpeclHead configuration for the

subject is specITP. As above, agreement realized on T surfaces as

morphological person/number agreement. The structure is given in (125).

(125)

NP,! Op-k

Y dynion C

yr

•welsant

t-V NP

1

y lIyfr

106

•

•

•

The presence of negation does not block wh-movement of the subject,

which proceeds as usual. This accounts for the unacceptability of a pronominal

doubling subject agreement in negated subject relatives.

(126) Ydynion na ddarllenasant (*hwy) y Ilyfrthe men Neg reaad-3pl (they) the bookthe men tJuu didn 't read the book

The analysis accounts straightforwardly for subject agreement on the

tensed verb and the ungrammaticality of a pronominal in subject position.

Unexplained, however, is the Jack of overt complementizer in the negated

relatives. Given the derivation shown above, we would expect the

complementizer yr to precede the negative marker ni(d). Instead, we find no

complementizer, and the negative marker surfaces as TUl(d).

There are !wo possible explanations for this facto The fact that the

negative marker never co-occurs with an overt complementizer could be

construed as support for the cIaim that the negative marker is itself generated

Jnder C; a position taken by Harlow (1981) and Sadler (1988). However, on the

assumption that negation heads a maximal projection itself, a more appropriate

explanation is that the negative marker found in negated relatives is a synthetic

fonn, created by the head of NegP raising te C.

For Neg-raising te C te occur, we must determine whether this raising

follows raising of the [V+1] complex te Neg. Such a derivation would be

permitted according te Rizzi & Roberts (1989), who propose a relaxation of

Baker's (1988) constraint on the structure of derived XOs. According te Baker,

an XO created by syntaetic adjunction cannot dominate a trace and a lexical item.

Thus, the derived structures on the right in (127) are both ill-fonned.

107

• (127)

x y X t

t Y

•

•

Rizzi & Roberts suggest that the lower structure is acceptable, as it is the

head of the newly created Xo that bas raised further, leaving behind a trace. This

possibility is restricted to the head itse1f; foIIowing Baker, they assume that the

y element cannot raise further, leaving X behind.

If we accept that these two possibilities are available in Welsh, and that

Neg-raising to a higher functional head is obligatory when such a projection is

generated, it is consistent with thlo fact that the form of the sentential negation

marker is different in main clauses and embedded contexts. Thus, nard) can be

viewed as the morphologicaI merger of the negative marker and the

complementizer yr.39 This is shown in (128).

(128)

Neg

"'i'-l,,,

1,,1

1V T

In this way, there are two potential derivations. In one, the [V+'I1

complex adjoins te Neg, and Neg then raises te C. In the other, the [V+'I1

lOS

•

•

complex remains in T, and Neg raises independently to e. In either derivation,

the merger of e and Neg results in the form nard). However. as is supported by

the presence of agreement triggered on T, in either derivation. the tensed verb

cannot raise al! the way 10 e.

However, there are certain problems associated with the proposai that

Neg-to-e raising follows [V+ T]-to-Neg raising. First, if [V + T] adjoins to

Neg, then we must determine what blocks further raising of the [Neg-V+T]

complex to e.

Second, if the [V+ T] complex raises 10 Neg, then under subject.:

relativization both the head and specifier of TP will dominate a trace; an illicit

SpeclHead Case configuration as defined by the Lee. Accordingly, we would

not expect 10 find personlnumber agreement associated with the subject on the

tensed verb in these cases. These issues are discussed in detail below. First, we

will present a potential derivation based on the claim common to both

derivations; that the [V+ T] complex does not raise to e in negated clauses.

2.7.2. NEGATED OBJEeI' RELA1lVES: NO AGREEMENT AND AREstlMP11VE PRONOUN STRATEGY

While the negated subject relatives are accounted for in a straightforward

manner under this analysis, the negated abject relatives show more interesting

behaviour. Here we find two different patterns. In the fust pattern, illustrated in

(129), the direct object pronominal is obliga1Ory.

•(129) a. y llyfr na ddarllenais i *(el)

the book Neg read-ls 1 (it)the book which 1didn 'r Tead

109

•

•

•

1 have suggested that the intervening negation prevents the verb from

moving into C. While this presents no problem for the relativized subject, for

which agreement is simply established at TP as opposee! to CP, it does present a

problem for the object: the same problem confronted by a VSO direct object in

non-relativized structures. As discussed earlier, the object of a finite verb in a

simple unrelativized structure has no licit agreement configuration; the only

position where the verb is morphologically complete is in T, whose specifier is

required for the subject. Since Case is claimed to be assigned under govemment

in this instance, agreement is not triggered at S-structure, and an in situ

pronominal is obligatory.

The situation for the direct object is similar in negative relatives, which

block V-raising to C. The V in T provides no licit Spec/Head configuration for

the direct object, because speclTP is reserved for the subject. This inability of a

relativized direct object to satisfy the LCC in negated relative clauses forces a

true resumptive strategy, the strategy found in relativization out of complex

NPs. In these cases, there is no wh-movement; a base-generated operator in

specfCP A' binds the object pronoun in its base-position.40 In this way, the

asymmetry between subjects and VSO direct objects in negated relatives is

captured: The former trigger agreement at TP and still undergo wh-movement,

but the latter trigger no agreement and cannot undergo movement.

This analysis also suggests an account of cases of extraction out of non­

inflecting PPs. Such prepositions do not agree with a pronominal object.

(130) Siaradasoch chwi ag ef.talked-2pl you with bimYou taIked with him.

110

•

•

Not surprisingly. we find that if the object of such a preposition IS

relativized, the pronoun occurs obligatorily in base position.

(131) Ydyn y siaradasoch chwi ag efthe man C talked-2pl you with himlhe man lha! you talked with

1 maintain that these prepositions do not licence a specifier position.

Since the pronoun cannot enter into a chain with a pleonastic in spec/PP, it does

not have the option of bcing Case-checked in a Spec/Head configuration at

S-structure. The pronominal cannot drop, as pro is not identified by agreement

morphology on the governing head. Further, the lack of specifier means that a

complement of P has no way of moving out of PP. This forces a true

resumptive strategy.41

2.7.3. NEGATED OBJEcr RELATIVES: AGREEMENT ON NEGATION ANDARESUMl'TIVE PRONOUN STRATEGY

ln the alternative option of negated object relatives, we find an abject

agreement marker on the Neg element itself. Thus, (132) exists as a variant of

(129), with the 3sg agreement marker s surfacing on Neg. In these cases, the

object pronoun is optional.

(132) Yllyfr nas ddarllenais i (ef)the book Neg-3sg read-1s 1 (it)the book which J didn '1 read

The agreement paradigm for negation is given in (133) (adapted from

Sadler (1988:72».

111

• (133) lsg -m2sg -th3sg -sIpl -n2pl -ch3pl -s

Agreement is taken here to be a reflex of a licit SpeclHead configuration

at S-structure. Agreement on the Neg element leads us 10 posit that a licit

SpeclHead configuration can be established between the head of NegP and an

element in its specifier position; Le. NegP introduces a potential Case-liœnsing

configuration for the direct object.

•

•

Note, however, that this construction is unlike the indirect pattern of

relativization found with relativized subjects in negated clauses, in that a

pronominal is free to occur in direct object position, doubling agreement. This

is only possible where movement bas not occurred, e.g. in unrelativized

structures. The fuct that the pronominal can co-occur with the agreement marker

suggests that these reflect agreement with an A-chain, rather than with an

A'-chain formed by wh-movement.

By deriving object agreement on Neg by formation of a A-chain, we

expect the agreement that surfaces on Neg to appear in unrelativized structures

as well. This is illustrated in (134)-(135).

(134) N"1S gweIodd Wyn (el).Neg-3sg saw Wyn himll)'n did not see hïm.

112

• (135) Gwn nas gwel Wyn (el).know-lsg Neg-3sg see Wyn him1know rha! ltYn will ner see him.

[Awbery 1976:189]

Furthennore, as shown in (136), object agreement markers may appear

on Neg in relativized (or more genera11y A'-moved contexts) when the

relativized argument is not the object.

(136) Ydyn nas gwelodd (el)the man Neg-3sg saw-3sg (him)rhe man who didn't see him

In this way, the issue of agreement on Neg is quite separate from the

issue of relativization. The possibility of a VSO direct object triggering

agreement on Neg is not dependent on wnether or not it enters into an A'

relation. It is simply a Case-licensing option made available by the projection of

sentential negation. Accordingly, our analysis of agreement surfacing on Neg

will paralle1 our analyses of agreement surfaeing on prepositions and [-tense]

verbs in non-relativized contexts: triggered by a chain created at S-strueture and

headed by a null pleonastie in specifier position.

However, this paradigm, specifie to VSO direct objects, differs from

agreement in non-relativized contexts in one important respect. In non­

relativized contexts, agreement is genera11y required wherever it is possible;

otherwise, an in situ pronominal is obligatory. In negated VSO direct object

relatives, however, agreement may, but need not, surface on the negative

partïc1e. Naturally, if there is no agreement, then the pronominal is obligatory,

just as in the non-relativized contexts.

113

•

•

There are two reasons to consider an alternative to this proposaI. First,

the claim that the head of NegP provides an A-position in which an NP can be

Case-ehecked constitutes a fairly radical departure from previous claims about

the properties of this projection. Second, agreement on Neg is optional, unlike

agreement on other heads in Weish. l will first consider an account which will

derive both the optionality of agreement and the Case-ehecking ability of Neg

by optional V-raising to Neg. However, this will be rejected in favour of an

analysis that incorporates an optional chain formation operation with a

pleonastic in speclNegP that is independent of V-raising.

The apparent optionality could in principle be related to the two

proposed S-sttucture positions of the [V+ 11 complex in negated clauses

out1ined above: T and Neg. It is not implausible that agreement on Neg should

depend on whether or not the tensed verb adjoins to the head of NegP in

negative clauses, and that if V-to-Neg movement does occur, the spec of NegP

can (and therefore does) provide a Case-position for the direct object. If the

verb remains in T, no such position wouId be available for the direct object, and

a resumptive pronoun would be required. In this way, it would not be agreement

but rather V-raising which encodes an optional aspect. The claim, then, wouId

be that the spec of NegP can provide pronominal objects in non-periphrastic

sentences with a way of satisfying the S-sttucture Case-Iicensing requirement,

but that this ability is dependent on V-raising to Neg.

One way of implementing this technically wouId be 10 cIaim that the

[Case] feature of a verb is inherited by the head of a functional projection that it

raises into. Thus, if a [+Case] verb raises into Neg, the specifier of NegP

becomes a licit SpeclHead Case configuration.

114

•

•

However, this proposaI is incompatible with the proposaI that agreement

is triggered at the lowest !icit Spec/Head configuration (CAR).'fhe crucial case

involves subject relativization in negated clauses. Specifica1ly, if we allow the

[V+'1'] complex to raise to Neg but block it from further raising to C, then

there will be no !icit Spec/Head configuration for the subject NP. Both the head

and specifier of TP would dominate a trace, and spec/CP could not provide a

!icit Case position because the head of C wouid not contain the Case-licensing

head, T. Thus, Case-licensing of the subject in specITP would be expected only

in those cases where the [V+'I'] complex does not raise to Neg. Since we have

associated V-raising to Neg with the appearance of object agreement on Neg,

we would not expect any su!lject agreement on the raised verb when object

agreement is triggered on Neg. This prediction is incorrect.

Since subject agreement is found in ail negated relatives, whether the

direct option agreement option is taken or not, we cannot say that the optionality

derives from optiOnal V-raising to Neg. This is discussed further below.

2.7.4. ON THE AuroNOMY OF NEG AS A C~LICENSING HEAD

This analysis finds itself confronted with !WO possible explanations for

the appearance of agreement on Neg. In one, the ability to Case-license an NP

is dependent on V-raising. This approach bas the benefit of relating the Case­

licensing potentia1 in speclNegP to the verb, which is a somewhat less

controversial c1aim than attributing Case-licensing capability to the head of

NegP itself. It also bas the benefit of capturing the optionality of object

agreement on Neg by maintaining that agreement is obligatory whenever it is

possible. However, it fails to prediet the appearance of subject agreement on a

ilS

•

•

•

tensed verb in all negated clauses, whether object agreement on Neg surfaces or

not.

In the second option, where [V+Tl does not raise te Neg, the specifier

of NegP provides an A-position in which the chain of the direct object can be

Case-ehecked, independently of V-raising. This has the benefit of correctly

deriving the agreement facts in negated clauses, and removes the need for

invoking language-specifie morphological constraints te block further raising te

C, but requires optional Case-licensing in speclNegP, apparently a unique

optionality in the agreement system of Literary Welsh.

The fact that, like objects of prepositions and [-finite) verbs, relativized

subject NPs never trigger agreement on Neg in negated relatives provides sorne

additional support for the claim that [V+Tl does not raise te Neg. If we allowed

such movement, then speclNegp would be a lower potential Case-position for a

relativized subject in negated relative clauses, and we would expect subject

agreement on Neg. An earlier analysis of these structures (de Freitas & Noonan

1993) required that speclNegP be specifical1y reserved for direct objects, a bald

stipulation, in order te block this possibility. However, if [V+Tl never raises te

Neg, then the agreement triggered at TP is predicted in negated clauses, and

there is no need te specifical1y reserve speclNegP for the direct object. This

restriction follows from our definition of a !icit SpeclHead configuration and the

operation of the CAR.

Whether or not the head of NegP depends on V-raising in order te Case­

license an NP in its specifier could conceivably be determined by finding data

where a lexical item intervenes between the negative marker and the tensed

verb, and object agreement still surfaces on the negative head. This would

116

•

•

demonstrate that the head of NegP can independently Case-license the chain of

the direct object. However, 1 have found no data where lexical material (other

than object agreement) intervenes between sentential negation and the tensed V.

Another possible test could come from [-finite] clauses, which exhibit

SVO order. Accordingly, if sentential negation is S-initial, the subject would

intervene between negation and the tensed V. If agreement surfaces on Neg it

cannot be due to V-raising. Again, though, this potential disambiguating data is

not avai1able, because sentential negation is not generated in [-finite] clauses in

We1sh. Instead, the negative verb peidio ('cease') is used (see Chapter 4, section

4.3.1).

Accordingly, we can neither definitively reject nor confirm the

hypothesis that [V+1] never raises to Neg by appealing to word-order facts.

However, this claim (and the accompanying claim that the head of NegP can

Case-license a chain in its specifier independently of V-raising) follows from the

proposais made in this thesis conceming the definition of a licit SpeclHead Case

configuration and the CAR (the Condition on Agreement Rea1ization).

Accordingly, we will conclude that the [V+1] complex remains in T in negated

clauses, and that the head of NegP left-adjoins to C independently in negated

relative clauses, resulting in the fused form 1Ul(d) [ni(d) + y(r)] at PF. We will

maintain that the possibility of triggering agreement on Neg is optional, unlike

other agreement paradigms in We1sh.42,43

2.7.4.1. REsnuCIlONS ON AGREEMENT ON NEG

The only NP in a sentence which can trigger agreement on Neg is the

direct object of a [+tense] verbe Direct objects in periphrastic sentences do not

have this option; agreement must surface on the [-tense] verb.

117

•

•

•

(137) *Nis wnaeth y plant darllen [el.Neg-3sg did the children readThe chi/dren did net read it.

(138) Nid wnaeth y plant ei ddarllen [el.Neg did the children 3sg readThe chi/dren did net read it.

Furthermore, the object of an inflecting preposition cannot trigger

agreement on Neg.

(139) *y dyn nas canodd Sion amthe man Neg-3sg spoke Sion aboutthe man tha! Sion didn 't speak about

(140) Ydyn na canodd Sion amdanothe man Neg spoke Sion about-3sgthe man tha! Sion didn't speak about

The restriction of agreement on Neg to VSO direct objects follows from

the CAR; speclNegP will be the lowest licit Spec/Head configuration only for

these arguments. Complements of tensed verbs are exceptional in that their

Case-licensing head, V, is morphologically incomplete until it raises to T. In a

sentence without a pre-sentential NegP, the complement of the V bas no licit

agreement position. However, if a NegP is generated, this creates a licit

configuration for the complement of the tensed V which is also the lowest such

configuration.

The lowest licit Spec/Head configuration for a periphrastic direct object

or an inflecting preposition, however, is defined by the S-structure position of

Ils

• the untensed V or P. Accordingly. agreement is established at the lower VP or

PP.

These facts are illustrated in the structure in (141). The lowest licit

SpecJHead configuration for the prepositional complement is speclPP; the

lowest such configuration for \he subject is specITP. SpecfNegP is the lowest

licit Case configuration only for a VSO direct object.

•

(141)

P NP3

i it... _._.•._ __ __...i

2.7.4.2. A-CBAJNs: CREAnON BY INDEXAnoN vs. MOVEMENT

•

As noted above, although there is an agreement position (speclNegP) for

the direct object, the direct object still cannot undergo operator movement 10

speclCP in negated relative clauses, as is the case with the normal indirect

strategy (i.e. periphrastic direct objects). This is evident from the optional

surfacing of the pronominal, an option unavai1able in relativized contexts

119

•

•

•

derived by operator movement. Thus, the agreement pattern for VSO direct

objects in negated relatives is triggered by an A-chain and is unaffected by

relativization; relativizing a direct object still requires a non-movement strategy

with a base-generated operator.

This raises the following question: Why is a resumptive strategy forced

in such cases? Since agreement on Neg would in principle satisfy the LCC (a

condition on A' chains), why is operator movement te spec/CP not pennitted?

1 will offer the following explanation, which invokes a distinction

between A- and A'-movement. It has been maintained here that an element

moving te spec/CP is Case-licensed as low as possible in the structure. In this

way, a relativized argument moves through its lowest Case position before

raising to spec/CP. If this is taken to imply that it must A-move to its lowest A­

position, and then proceed by A' movement to spec/CP, then we see why such

movement is not pennitted for VSO direct objects. They are the only arguments

whose A-position (in specfNegP) is higher than the A-position of the subject.

Given the restrictive theory of movement proposed by Sportïche (1988,1990),

where all A-movement must proceed through each intervening specifier, in

order to A-move to spec/NegP the object would have te move through speclTP,

which is occupied by the subject.44 As shown in (142), skipping over this

position would create an ill-fonned chain.

120

•

•

(142)

Op-k

/,'-,,,,: Neg + C t-k: i .If', , ,. , ,• J 1,

A' movement

: ~k

l-----X---------------J ! t-V t·k,

A-movement : ', '... J

We are thus 100 to the conclusion that, although the creation of an A­

chain by indexation with a pleonastic in spec of NegP is possible across the

subject, A-movement to that position, which necessari1y includes A-movement to

speclNegP as an initial step, is not permittOO; at least, not in the syntax. This

condition would have to be reIaxOO at LF, assuming LF expletive replacement in

line with Chomsky (1986a).

In a sense, what is proposed for A-ehains is implicit in sorne of the

literature on resumptive strategies in A' chains: A'-binding of a pronoun in situ

by a base-generated operator is subject to less severe locality conditions than

syntaetic A'-movement. (McCloskey (1990), Shlonsky (1992» Here the same

claim is made for A-ehains: chain formation by indexation is less restricted !han

chain formation by movement.

121

•

•

To conclude, under this analysis there are only two possibilities for

direct objects of verbs to undergo operator movement: (i) if the verb does not

raise out of VP (in infinitivaIlperiphrastic constructions), and (ii) if the verb

raises aIl the way ta C (in non-negated object relatives). This restriction on wh­

movement for direct objects in Welsh follows from the fuct that they can be

Case-licensed higher in the tree than subjects, but cannot A-move to that

position.

In spite of the technicaI questions raised by proposing that the vsa

direct object agreement found on Neg results from the creation of an extended

A-ehain, 1 would claim that this is plausible. This kind of agreement is already

remarkable in that it is the only apparent case of optionality in the agreement

system in LW, a characteristic that raises problems for any treatment.

Furthermore, the potential doubling occurrence of a pronominal in argument

position is consistent with the idea of a base-generated chain rather than with a

chain formed by movement.4S

abject agreement on the sentential negation marker is an instance of a

more generaI phenomenon in Welsh. Agreement with vsa direct objects is

found on other pre-sentential particles as weIl, and just as in the case of

agreement on Neg, it is optional. Consider (143).

(143) Fe'i ddarIlenodd y plant.EMPH-3sg read the children'Ihe children read il.

[Rouveret 1990:51]

This raises the possibility that sententiaI negation may be one

instantiation of a more generaI functionaI category (Laka (1990). A crucial

122

• question, then, is whether there is a functional projection NegP which is distinct

from the projection generated by these other sentential markers.

2.7.5. OmER PRE-SENTENllAL MARKERs IN WELSH

Jones & Thomas (1977:357ff) describe the situation pre-theoretically as

follows: there are three distinct pre-sentential markers which occur on matrix

sentences, marking the sentence as declarative, interrogative, or imperative.46

Each of these markers can be positive or negative. There are additional markers

which appear in 'fronting' contexts, which they define as questions, clefts, and

relative clauses.

•(144)

DECLARATIVE

positivenegative

INTERROGATIVE

positivenegative

IMPERATIVE

positivenegative

FRONTING

positivenegative

MATRIX

milje/i47

nï(d)

aonï(d)

na(c)

EMBEDDED

y(r)nard)

aIy(r)nard)

Below, 1 will propose a treatment for the pre-sentential markers in Welsh

which distinguishes NegP and CP from what will be termed IllP, or

lliocutionary Phrase.48

123

• 2.7.5.1. NEGPvs. CP

•

•

In the analysis presentee! above, 1 have analyzed the pre-sentential

particle in what Jones & Thomas refer to as fronting contexts as the head of CP

(a being the agreeing form, y(r) being the non-agreeing counterpart of Cl. The

particle nard) is analyzed as resulting from the morphological merger of the

head of NegP (ni(d)) with a non-agreeing form of C, y(r).

The ni(d)/na(d) altemation has received sorne attention in the lite:ature.

Awbery (1977) concludes that the: are in fact the same lexical item, and that a

rule gives the correct vowel in each case.49 Another possibility, adoptee! by

Harlow (1981) and Sadler (1988), is that nard) is a negative complementizer,

generated under C. The fact that this marker never coincides with an c','ert

complementizer of the form y(r) or a could be considered good support for this

hypothesis. Laka (1990) uses similar complementary âistribution facts from

English and Basque to argue for a single functional category, Sigma-phrase,

which can be headee! by elements conflating a similar collection of illocutionary

features to those seen here.

Problematically, if we adopt Laka's argumentation for a unified account

of the pre-sentential elements in negative relative clauses in Welsh, and claim

that nard) is a negative complementizer, we lose our account for the indirect

pattern of agreement required for relativized subjects and VSO direct objects in

negative relative clauses. If nard) is simply a negative complementizer, we

would expect agreement to be triggered at CP when the subject of a negated

relative clause is relativized, in the same way that it is triggered on the

affirmative complementizers. Instead, we find that subject relativization exhibits

the indirect pattern (obligatory subject agreement on the tensed verb, and an

obligatory empty category in the subject argument position).

124

• The daim that nard) is a base-generated negative complementizcr could

be salvaged if we stipulate that the [V +T] complex can only raise to C if the

head of C is not negative; the lowest Spec/Head configuration for the subject

would be TP, and the person/number agreement on the verb would be predicted.

Alternatively, it could be claimed that the lack of subject agreement on nard)

results from an idiosyncratic morphological gap in Welsh, such that there is no

[+agr] form of the negative complementizer. However, such a daim would

encounter problems with direct objects. As discussed at length above, direct

object agreement can surface on nard). However, the kind of agreement that

optionally surfaces on nard) does not pattern with agreement on C. First of all,

there is its optionality. Second, there is the acceptability of a pronoun doubling

the agreement.

(145) Y dyn nas welais i (ef).the man Neg-3sg saw-lsg 1 (him)The man rhal 1didn 'r see

(146) Y dyn a welais i (*ef).the man C+Agr saw-lsg 1 (*him)The man rhal 1saw

If we claim that nard) is a negative complementizer that can only surface

with direct object agreement, then we would expect wh-movement of the direct

object to spec/CP, leaving a trace in its base position, just as is the case in an

affirmative relative clause. In such sentences, we would expect an in situ

pronominal ta be ungrammatical, and this is not the case.

Accordingly, instead of daiming that negation interferes with V-raising

and having the differences between subjects and objects faU out from more

125

• general principles, one would have to propose that negation interferes in sorne

way with wh-movement of direct objects only. It will therefore be maintained

that NegP has a syntactic status independent of the CP projection, and that both

NegP and CP are generated in negated relative clauses. Below, 1 will propose

that NegP is also distinct from the projection headed by the other pre-sentential

particles found in LW.

2.7.5.2. NEGP vs. ILLP

•

•

First, note that the agreement paradigm on the negative marker differs

from that on the other pre-sentential markers. Specifically, agreement with 3rd

person singular and plural direct objects on the negative particle differs from the

3sg agreement that surfaces on the other pre-sentential particles, and the number

distinction found on the other particles with 3rd persan agreement is lost on

Neg.

(147)

Neg other pre-sentential particlesIs m m2s th rh3s s iIpl n n2pl ch ch3pl s u

Assuming that Case-checking is sensitive ta the properties of the head

defining the SpeclHead configuration, it is plausible that this differing

agreement pattern reflects the distinctive properties of the head cf NegP.

Another potential argument for a distinct NegP follows from co­

occurrence restrictions on the other pre-sentential markers. These markers are in

126

• complementary distribution. Thus, there are no forms combining interrogative,

declarativelemphatic or imperative. The impossibility of combining these forms

is consistent with a treatment where they are distinct potential heads of a non­

iterating llIocutionary Phrase. However, forms combining negation with these

other markers do occur. The examples below illustrate the negative-interrogative

combination.

(148) Ni ddarllenodd Sion y Ilyfr.Neg read-3sg John the bookJohn didn't read the book.

(149) Oni ddarlenodd Sion y Ilyfr.Q-Neg read-3sg John the bookDidn't John read the book?

• [Sadler 1988:6]

(150) Nid oedd John yn chwerthin.Neg was John PROG laughJohn was not Iaughing.

(151) Onid oedd John yn chwerthin'?Q-Neg was John PROG laughWasn't John Iaughing?

[Jones & Thomas 1977:358]

This is consistent with the claim that, while the markers indicating

Declarative, Interrogative, and Imperative may be alternative functional heads

of a single phrase, TIlP, NegP is generated independently.So

A further restriction supports our hypothesis that the pre-sentential

markers indicating illocutionary force are distinct from NegP and CP. Urlike

127

•

•

the particle y(r) that 1 have analyzed as a complementizer, which appears only in

embedded clauses (its surface form varying according te whether it combines

with Neg and whether it is in an agreement configuration at S-structure), and the

negative particle ni(d), which occurs in both embedded and matrix clauses (its

surface form also varying depending on other functional heads with which it

combines and whether it is in an agreement configuration at S-structure), the

pre-sentential interrogative, imperative and declarative markers are restricted to

matrix sentences, whether negative or affirmative. 51 This suggests that not only

can NegP be distinguished from the pre-sentential particles that indicate

iIlocutionary force, but these in tum can be distinguished from

complementizers.

This line of reasoning leaves us with a three-way distinction, between a

CP headed by y(r), a NegP headed by ni(d), and an D1P headed by the

DECL(arative), INT(errogative), or IMP(perative) marker. This is iIlustrated in

(152).

(152) D1PDECL: Tnilfe/iINT: aIMP: B

CP y(r)

NegP ni(d)

The other surface forms are derived as in (153).

128

• (153) Neg+CNeg+INTNeg+IMPNeg+DECL

= nard)= oni(d)= na(c)=ni(d)

What the projections NegP and lllP have in common is that an A-ehain

can optionally be created by indexation between a pronominal direct object and

a null pleonastic in their specifier, triggering agreement. However, only the

spec/CP is a landing site for A'-movement (where agreement is incompatible

with an in situ pronominal); the specifiers of lllP and NegP are only available to

Case-license the head of an A-ehain created by indexation.

2.7.6. PROBLEMS WlTB LoNG A-CRAINS

•This analysis includes a potential A-position for the direct object in

simple VSO sentences (spec/NegP or spec/lllP) that is higher than the A­

position for the subject. A way in which this can be implemented technica1ly is

by appealing to a notion of complete .fùnctional complex. in the sense of Aoun

(1981) and Chomsky (19800): if the verb (the theta assigner) taises over the

subject, then it extends the A-domain for the object 50 that it is able to form an

A-ehain across the subject.52

Note that this dilemma is not unique to the analysis presented here.

Given a combination of the VP internaI subject hypothesis and the hypothesis of

Case-ehecking in spec/AgrO for objects (where AgrOP dominates the maximal

projection of V), the definition of acceptable NP-movement will have to be

adjusted in any case, sinee the object NP will have to A-move to its Case

position across the D-strueture position of the subject.'3

Among the predictions that might follow from such a daim is that

binding facts in LW would parallel those exhibited in Ergative languages.

129

•

•

•

However, it has been maintained that A-movemeru of the direct object to a

position higher than the A-position of the subject is prohibited. Accordingly,

these effects might not be instantiated in the same way as if the object actually

moved te an A-position higher than the subject

2.7.7. CONCLUSION

This anaiysis of sententiai negation and agreement patterns in matrix and

relative clauses in WeIsh rests on the following assumptions about Case:

(i) Structural Case is checked in a SpeclHead configuration at S­structure or at LF.

(ii) Agreement is a reflex of structural Case-checking at S-structure(either by movement to spec or by chain-formation).

(iii) In WeIsh, non-pronominai NPs are Case-checked at LF, andpronominai NPs are Case-checked at S-structure by chain­formation (when possible).

1 have argued that the difference between the direct and indirect

strategies does not reflect a movement versus non-movement strategy, but rather

variations in the S-strueture position of Case-licensing heads. Under this

anaiysis, there are only three cases where a resumptive strategy is forced: when

the reIativized argument is (i) the VSO direct object in a negated clause, (ri) the

complement of an uninflecting preposition, or (Iii) in a complex or coordinate

NP island.

This anaiysis provides a natura! account for ways in which subjects and

direct objects pattern tegether with respect te reIativization in WeIsh, as weIl as

130

•

•

for those ways in which they differ. Furthermore, the definition of licit

SpeclHead configuration accounts for the lack of direct object agreement on

tensed verbs.

With respect ta negation in Welsh, 1 have maintained that NegP

influences Case (and hence agreement) in two ways. First, it is generated

between CP and TP and blacks movement of the [V+T] complex to C. Thus,

Case-checking of a relativized subject takes place in spec/CP in affirmative

clauses but spec/TP in negated clauses.

Furthermore, 1 have proposed that NegP provides a licit Spec/Head

configuration for Case-licensing VSO pronominal direct objects. The

pronominal object is thus provided with a licit SpeclHead configuration and can

be Case-checked at S-structure. This procedure is reflected in agreement on

negation itse1f.

1 have argued that Case-checking of either the subject or the direct object

in We1sh takes place in spec of CP if the [V+T] raises to C. In other words, an

NP can be Case-licensed in that position. However, it is not an A-specifier,

because A' binding takes place from that position. The analysis thus allows for a

subject or direct object NP to be Case-licensed in an A' position.54

2.8. ENDNOTES TO CHAPTER 2

IThe endiDg -odd in this seotence does Dot mark agrœmeDt; the relativized NP isplural but the verb is iDflected for 3sg-past. The 3sg endiDg is used as a default wheo teDSC butDot agrœmeDt is triggered, as there is DO morpheme that marks teDse &IODe. The 3sg f?rmappears OD the teosed verb with DOD-proDOminal NP subjecls. See Harlow (1981) for discUSSl°D•

2 See Sadler (1988:19f1) for arguments in Cavour of &DaIyzing Welsh as having a VPcoDSlitueDt.

3Wbether this verb is inserted dinlctly UDder T Or raised from a lower positiOD (i.e.AGR, as in the Pollock (1989) &DaIysis) will he left open.

131

•

•

•

4Welsh differs crucially from Irish in this respect. In Irish, neither a pronominaJ nora non-pronominaJ subject cao C<HlCCur with inllection on the verb. Furthermore, whereas Irishinllected paradigms are incomplete, Welsh bas complele paradigms for those ilems which doinllect for agreement features.

SNole that these inflected prepositions contain a consonant (-<i) inlervening betweenthe preposition (am) and the agreement marking (ano). Rouveret (1991) proposes an interestinganalysis of inflected prepositions as motphologically complex clements, consisting of apreposition, an intervening functional head (DEl), and agreemenL

6In eolloquial Welsh (eN), it is becoming increasingly acceptable ta have an in situpronominaJ when the head of a possessive NP is relativized, questioned, or clefted. 1 do DOthave an explanation for this fact, other !ban ta suggest that perilaps wh-movement in general islosing out ta a resumptive strategy in CYl. Nole that, even in dialects which permit doubling inthese cases, the constraint against doubling agreement holds when the extraeled clement is aproper name: Emrys y gwekzis i ei llyfr ("ej) ('It's Emrys whose book 1 saw') (Awhery1976:79fl).

7Nole that in WeIsh, as in Irish, agreement is with the first member of a conjoinedNP. This bas been argued ta he related ta an asymmetrical structure for these phrases in Irish(Noonan 1992). We will DOt address this issue hore.

8Sec also ShIonsky (1992) for an analysis of relativizatiOll strategies in Irish, Hebrewand Palestinian Arabic !hat argues for a movement vs. non-movement analysis.

9Nole that an apparent subject-<lbject asymmetry does appear in lIegaled relativeclauses, but direct object relativization out of negated clauses does not pattern with the indirectstrategy; rather, it pattems with non-relativized structures. These constructions are discussed indelai! in sectiOll 2.7.

IllHarlow (1981:217). Sadler (1988:250, fil. 4) notes that Harlow must mean'properly govemed' bore.

llSec Sells (1984) for an alternative way ofavoiding the predicted asymmetry.

12Harlow cIaims that since insertion of the dummy verb gwneud blacks pruning, theuse of the y comple:lCtlti= is correctly predicled for these structures. This is similar taRouveret's (1990) cIaim that the barrierbood of a verb-noUll phrase (VNP) is neutralized by V­raising; a process DOt instantiated in periphrastic sentences.

13Harlow (1981:242ft).

14 Regarding the collleSted status of a, Morris·Jones (1931) considers it a pronOUll,but McCloskey (1978) argues that the Irish equivalent (aL) is not a pronOUll, because it Iacksnumber and gonder features. However, as Harlow notes, these features are oftell neutralized inrelative prollOUllS, cf. English who vs. he, she, they. For many speakers ofEnglish, Case is aIsoDeutralized iD these proDOUDS, and occasiona11y even the p"jmpte'jnanjmate distinction (tM manwhich 1 WQS talking to, the book which 1 WQS Tt:Dding). Sec Sadler (1988) for a critique ofHarlow's cIaim regarding the status of a. Also sec Borsley (1984), Fife (1986) and Borsley(1987) for an exchange ofviews on this matter.

132

•

•

•

lSNotc:. how~vc::r. that this is c:qually truc:: of the:: r;:h:m~nt yr. which Rouycrct doc::.... notanalyu: as a cHtic. This wc:.akens his argument.

16n,e fact that an overt pronoun cannot occupy lhe relalivi2Cd position in an indiroctrelative, which mak~ il strikingly difft:n:nt from a non-rdativiZ&:d ditic-eonfigul"dtion (a... ootooby Awhery (1976, 1977), Harlow (1981), Sadler (1988), and others), is mcntioned but notaddresse<! in Rouveret (1990: fn. 58).

17This analysis is a revi~ and c:xpandc:d version of a pap:r givcn with MalteNoonan at the 27th R.gional Meeting of th. Chicago Linguislics Society.

IgTh.re are sever.1 points wh.re additional functional calegories might facilitate theanalysis. Howevor, given th. vi.w of agreement as the surface reflex of a relationship helWeon aCase-Iiœnsing head and an eh:mc:nt in its ~-pc:cifit:r position. positing agreement phl1L~'\ issomc::what redundant. particularly sinet: th~ agreement phl".L~ ~m to require the pn:scnce ofhoads which have beon t.rmod ea....-a.o;sih'll.rs in th. pa.'l: [+Ten....] and verbs.

19Zanullini's (1991) ciaim that the lowor h""d T det.rmin.s the availability of thehighor phrase N.gP might he consid.re<! somewhat uninlUitive. Evidenc. for the d.pcndencyrelation helWeon N.gP and TP will he presente<! in th. chaptor on Colloquial W.bh. Th. facl'of LW are consistent with this claim.

2On. final (d) appcars only if th. following word is vowel-initial.

21If Wc:: do not assume that sc:ntential negation generatt:S a NegP. then an alternativepossibility for th. hase position of N.g would he in Comp. This is discus.~ in section 2.7.

22s.. Rouv.ret (1991) and Tarald.= (1992) for analyses of Celtic agreement a.'involving synlactic incorporation of an agreem.nt morph.mc. Tara1dsen argues againS! analyzingsubjcct v.rh agreement (and polentially all agreement) as Spcc/H""d agreement.

23A distinction helWcoo pronominals and non-pronominal NPs with rc';pcct tomov.ment into spec position is proposed in Johnson (1990) for English and Noonan (1992) forIrish.

24n,é claim that a full NP cannot creat. '"\Ich a chain is porhaps ,...rprising, andcertainly stipulativ•. 1 will tentatively suggest that only pronominals are define<! soldy in termsof specifications for th. f""lUres persan. numher. and g.nder. and are ,...fficienlly 'light' in thisrespect to form a chain with a nuU pleonastic. If this claim is correct, it could he rclatcd te th.arguments made by Noonan (1992) to account for th. fact that only pronominals cao undergo S­structurc raising te specifior position in Irish. Th. chains iIIustratcd in (91) are essontially th.mirror imagé of th. chains shé dcscribcs for Irish pronominals. An alternative treatmont wouldappcal to Chomsky's (1992) di,'tinction helWcoo checking operations that must accur at S­structure and those that tak. place at LF. However, further research is ""luire<! te adcquatelyaddrcss Ibis question.

133

•

•

•

25Note that this analysis does not "''luire a stipulation 10 the effect that the specrrp isin any way uservt!d for the subject. As will he clarified in the following sections, an NPgenelllled helow Delta (i.e. a direct object) bas an additional Case-liccnsing cption: Case undergovemment by the verb. In this way, a structure where the specrrp is used by the subject NPcao avoid a violalion of the Case filte<. If, on the other band, the specrrp were used by thedirect object. the subjecl NP (which is genelllted in spec/DeltaP) would not he in a position to heCase-licensed under govemment by the verb, and the structure would he ruled out.

26n,is could he achieved (a1heit with some not insignificant technica! difficulties) bya1lowing the IV +TI compl"" to lower at LF to a position where it could Case-liccnse the directobject. Altematively, it could he maintained that the notion of morphologica! completencss issimply irrelevanl at LF.

27One ""ample is the inability of such a complement 10 raise ouI of ilS base positionin wh-question, relative clauses, and clefts. To accounl for these fuelS while maintaining that ailheads project a sl""'ifier, one would have 10 propose that these prepositions do not project aspecifier until LF, a possibility that 1 prefer not to allow the grammar without more directt:vid~ce.

28This c1aim makes an interesting prediction. The pronominai vs. non-pronominalCase distinction is lost in this position. If the mechanism by which Case is assigned 10 VSOdirect objeclS is inlcrfcred with in some way. bath pronominal and non-pronominai NPs sbouldhe affected. 1 will retum to !his point in Chaple< 4.

29nùs is similar 10 the proposai made by Sportiche (1990, 1992) that movementthrough a pal1icular position allows a V to Case-license a direct object. Thanks to Daniel Valoisfor s"Uggesting how this could apply to the Welsb paradigm.

30nis additionai restriction will come into play in the anaIysis of Russian, wben: theprojection immediately dominaled by delta is not VP but NegP.

3lPreviolL< analyses include Sells (1984), wbo (working within a very different set ofassumptions from those adopted here) relates the pre-sentential particle a 10 leftward Case­assignment. In this paper also, a is closely tied to the presencc of a gap in the relativizedposition. Sells (1984:137ft) mentio:lS the \ack of parasitic gap constructions in Welsb as apossible argument against a movement anaIysis for direct and indirect relatives.

32The optionai surfacing of an in situ pronominal in constructions is also mentionedby Jones & Thomas (19n:181) in the con~t of a discussion of copy pronouns. They includethe sentence given below. wben: relativization is not from inside a compI"" NP. This suggeslSthat wh·movement is governed by additionai restrictions in Welsb, whicb we will not addressben:.

Dyma 'r dyn yr oeddwn i'n dadlace am y gem efa fo.bere the man PT was I-PROG argue about the game with-3sgm bimHt!:1'f! is the man that 1was arguing about the game wilh.

331 adopt the position that raising occurs wben: it must. The movemcnt of the tensedVIOC is driven by the presence of an Operator in SpeciCP. ta facilitate Casc-Iicensing of theA'-clIain of the re1ativized argument.

134

•

•

•

34nüs can be considen:<! in the spirit of Pesetsky's Earliness Prindpk. wbereby anoperation takes place as early as possible in the derivation, This ana\ysis diffe", from a previousversion pn:sentec1 al the CLS. wben: the CAR was formulatec1 sucb !bat agreement was realizodal the highesl licit S~ead configusation. That version requin:<! Iwo additional stipulationsnot required ben:; (1) !bat NP complements of prepositions and [-tense] verbs in sorne sense'know' wbat their Case-liccnsing head is. and (2) !bat the specifier of NegP is rcservcd for Case­liccnsing of a direct object pronominal. These stipulations are not requin:<! in the anaIysisproposcd here.

3S Sec SbJonsky (1992) for a proposai !bat sorne Cs licensc: an A-specifier. Under hisanaIysis. the operator in spocICP adjoins to CP al LF in order to be in an A-bar position. Welshdoes not lend support to this proposai. It wiU be shown !bat an object is not able to A-moveacross the subject in Welsh (negated object relatives). If Spec/CP were an A-position. then theobject would not be able to move there. and we would always have a forecd resumptive strategyfor non-negated object relativization.

36Note !bat agreement on functional categories is generally impoverishcd in Wc1sh.These facts are discusscd in section 2.7. If wc view agreement not rnerely as a formai liCCDSingrequirement but as having a functional role to play. the process clearly brea1cs down withagreement on C. Thus. the sentence y dyn a welodd ~ [the man C saw-3sg he] is ambiguousbetween the reading the man that M saw (where the fina1 pronominal is understood to be thesubject) and the man that saw him (where the fina1 pronominal is understood to be the directobject).

371 have includcd an Aspect phrase between the VP tiers. as proposcd by Travis(1991). For our purposes. however, the labeling of these markers is not crucial.

38This enlai1s !bat the inlervening functional head Neg croates a minimality barrierbetween the [V+Tl complex and its trace. Baker & Hale (1990) propose !bat the notion ofminimaljty as it pertains to the ECP be made sensitive to the distinction between lexical andfunctional heads, such !bat onIy a functional head is a minima1ity barrier for the chain of afunctional head, and onIy a lexical head is a minimaUty barrier for the chain of a lexical head.The ana\ysis proposcd here is consistent with this claim, since V-movement into T is asubstitution process, and the trace in T is the trace of the functional head, T. Thus. thefunctional head (Neg) creates a minimaljty barrier for the chain of the functional head T.

39An interesting aside here is !bat in data from Middle We1sh, when the negativemarker co-occuns wilh an OVer! complementizer, the complementizer is alwaya yr (Evans 1964).This is as expecled under our ana1ysîs, where an agreeing C would be impossible in a neptcdrelative clause.

40r leave the question ofwltether Ibis binding talces place al 5-Structure or al LF open(cf. McC10skey 1990 and Sblonsky 1992).

135

•

•

41Noonan (1992) points out lhat this paradigm closely resembles preposltlOn­stranding facts in Dutch: Ibose prepositions lhat license a specifier position appear as apostposition if Ibeir complement is a pronoun, because Ibe pronoun moves 10 Spec position.Since Ibey provide a specifier position for Ibeir complement 10 move througb, Ibese are Ibeprepositions lhat cao he stranded in wb·movement constructions. Sec van Riem..<dij!:: (1978).Note also lhat in lhese instances, our analysis panillels lhat of RoUVerel (1990), in Ibe sense lhatIbe lack of specifier blocks movemenL

42There appears 10 he an optionality involving Ibe appearance of direct objectagreement on Ibe inserIed Case-licensor ""Iui:::J in negalcd lt3lIsÏtive clauses in PembrokesbireWclsh (sec Awbery (1991) and Cbapter 4 of this Ibesis), 50 this problem may he less critical!ban suggestcd bore. Furl1lermore, agreement on Neg ""IUÎres Ibe formation of an A-cllain !batcrosses Ibe position of Ibe subject, whicb may come al 50me oost 10 Ibe grammar.

43A question arises al this point. Anolber set of NPs whicb are unable 10 satisfy IbeS-structure Case-licensing ""Iuirement are objects of non-inflecting prepositions. Theseprepositions are analyzed bere as beads whicb do not project a specifier position. The specifierof NegP migbt provide a Case position for pronominal objects of non-inflecting prepositions asweil, triggering agreement and Iicensing pro-drop. However, what we find is !bat Ibecomplements of non-inflecting prepositions cannot trigger agreement on Neg, and mw>t surfacein silu. This is as expected, according 10 Ibe c\aim lhat Ibese prepositions do not generate aspecifier position. The lack of specifier means lhat Ibey cannot form an A-cllain wilb an clementoutside of Ibeir projection cilber by movement or by co-indexation, wilbout violating Ibe ECP.

44ne conclusion !bat A·movement would he blocked in this case al50 followsaccording 10 Rizzi's (1990) notion of relativized minimality, as Ibe clement in Specifier of TPwould induce minimality for Ibe moved objecL

4Snere is possible support for Ibe claim lhat Ibe A-cllain created by indexation overIbe subject is exceptional in olber respects as welI. Sadler (1988:76) states !bat Ibe process ofoptional VSO direct object agreement on negation is blocked if Ibe subject is overL Sbe givesIbe following sentence as an illustration: - ni'm gwelodd Sion (Neg·lsg saw Sion: 'Sion c1idn'tsec me'). Sec Campana (1992) for an account ofobject pronoun binding in an Ergative languagelhat migbt he applicable 10 Ibese structures in We1sh. However, olber data provided by Sadler(1988: 113) appear 10 contnulict this c\aim, and Ibe possibility of finding adclitional support forthis distinction must he left for future researcb.

46ne pre-sententia1 lIllUker descrlbed as Ibe declarative form for matrix affirmativeclauses is termed emphatic by my informants, who speak a North Wales clialecL For Ibem, pre­sententia1 mi marlts emphasis, and ail olber pre-sententia1 lIIBIkers are cIropped (sec Cbapter 4).This may reflect ~onaI variation. This interesting question must await future researcb.

47The forms mi,Je. and i reflect regional variations.

4Snumks 10 M. Baker for suggesting this tenD.

49n.e form of !he negalcd complementi= nard) migbt he talten 10 inclieate !batnegation bas merged not wilb y(r) but rather wilb !he agreeing form of Ibe complementi=, a.However, Awbery (1977:189) mentions lhat in Middle We1sh Ibe form usee! in negalcd relativeswas ny(l) and bas evolved inlO nard) only in modern Welsh.

136

• 5~f the claim that thcre is a distinct NegP is corrdet. theu pnxiictions LOi ta theirrebtive position in the struclUre might be deriv<d from the type of agreement triggered un the.""fuse<! fonns. Specifica1ly. the CAR would force agreement te be eslllblish<d al the lowe.<l licilSpeclHead configuration. If we fi-.d s on these lIl:U'kers (the 3rd person agreement IIl:U'king thalsurfaces on Neg). it would constilUle an argument for generation of NegP below IllP.

51lt is plausible that only one UlP is permitted for every sentence, in the pragmnticsense of spuch oct. Thus, one cannotlll:U'k the higher clause as interrogative and the embedd<dclause as imperative; the illocutionary force necessarily bas scope over both clauses.

52The possibility of exlending the A-<1omain by movement of the thela as.<igner isdiscuss<d but not adopt<d in Sportiche (1990).

53 See Chomsky (1992) for a discussion of!his problem.

54rhis cc.nclusion is drawn for A-bar mov<d subjects in French and English inNoonan (1989) and investigat<d in grealer del:Ùl in Noonan (1992). It is alse suggest<d byKayne (1983:5fl). where an element piclcs up Case in COMP.

137

•

•

•

3. CHAPTER 3: SUBJECT-OBJECT ASYMMETRIES ANDDEFINITENESS EFFECTS IN THE GENITIVE OF NEGATION INRUSSIAN

3.1. INTRODUCfION

In this chapter, a correlation found in Russian between sentential

negation and changes in Case-marking will be exarnined, and the analysis of the

interactions between negation and Case developed in the preceding chapter will

be extended in two directions. We derive the apparent sensitivity of thlS feature

to the thematic roles of NPs in a sentence, in part by proposing that the position

of NegP in Russian is lower than in Welsh, and in part by localizing Genitive

Case-checking in the specifier of the functional category NegP. Two accounts of

the relationship between negation, Case and definiteness in Russian,

corresponding to two documented paradigms found among native speakers, will

be proposed and evaluated.

Determining whether an NP in a negated sentence has the option of

surfacing with Genitive Case is established by judgements concerning the

grammaticaIness of a given negative sentence. These judgements appear to be

consistent across dialects and levels of formality; an analysis which accounts for

them is given in Part 1. The intriguing but slightly rnurkier faets involving an

apparent definiteness/indefiniteness effect (specifically, the correIation of the

Genitive of Negation option with the interpretation of that NP as definite,

indefinite, or ambiguous) will be discussed in Part 2.

1 will rnaintain that NegP in Russian is generated as a funetional

projection between the two VP tiers. Thus, NegP is generated above the lower

VP tier which is present in all sentences (transitive and intransitive), not above

138

•

•

•

the higher VP tier, which is generaled only if 3 verb selects an exlemal

argument. The basic paradigm is outlined below.

3.2. INmALPARADIGM

In a negated transitive sentence, an NP thal would otherwise be assigned

Accusative Case can surface with Genitive.

(1) ja vizu kniguI-NOM see book-ACC1see a book.

(2) ja ne vizu kniguI-NOM Neg see book-ACC

(3) ja ne vizu knigiI-NOM Neg see book-GEN1don 't see a book.

[Neidle 1988:34]

The subject of a transitive sentence can never surface with the Geniùve

of negation; it is a1ways marked with Nominative, whether in an affirmative or

negated sentence.

(4) ni odna gazeta ne pecataet takuji erundunot one newspaper-Cem-NOM-sg Neg prints such nonsenseNo newspaperprintS sud! nonsense.

(5) *ni odnoj gazety ne pecataet takuju erundunot one newspaper-Cem-GEN-sg Neg prints such nonsense

[Pesetsky 1982:46]

139

•

•

The initial generalization is that direct objects in negated sentences have

an additional Case-licensing option not available in affirmative sentences, and

that this option is unavailable to subjects. However, two principal facts

complicate the paradigm. First, a.~ pointed out by Pesetsky (1982) and Neidle

(1988), the ability ta surface with the Genitive of Negation cannot be accounted

for by referring to S-structure grammatical relations like subject and object;

rather, the phenomenon is sensitive ta the thematic raIes of NPs.

Second, the Genitive of Negation is not obligatory where it is permined

in Russian. Thus, two negated sentences may be identical except for the Case­

mar1àng on an NP, and this distinction appears ta carry information bearing on

the interpretation of the sentences.! These facts are summarized below.

3.2.1. INmANSITIVE SENTENCES: D-STRUCTURE GRAMMATICALRELATIONS AND THE GENlTIVE OF NEGATION

The term intransitive here refers to the praperty of a verb which

subcategorizes for only one argument; it does not specify whether that argument

is internal or external. Unlike the Nominative NPs in transitive sentences like

the one given in (5), Nominative NPs in sentences with certain intransitive verbs

do exhibit Genitive Case as an alternative in negated sentences. The

generalization is the following: while the Genitive of Negation cannot appear on

the sole argument of an intransitive verb if that argument is an AGENT, it is

permitted on the intemal argument of a verb in a negative intransitive sentence,

i.e. a PATIEN!' or THEME. It is also permitted on the internai argument of a

passive verb, which surfaces with Nominative Case in affinnative clauses. The

data in (6)-(9) are taken from Pesetsky (1982:42-62).

140

• (6) *takix sobak ne kusaetsjasuch dogs-fem-GEN-pl Neg bite-3sgSuch diJgs don 'r bite.

•

(7) ni odnogo goroda ne bylo vzjato (vragom)not one city-masc-GEN-sg Neg was taken-neut-sg (enemy­INSTR)No ciry was raken (by rhe enemy).

(8) gribov zdes' ne rastetmushrooms-GEN-pl here Neg grow-3sgMushrooms diJn 'r grow here.

Under the assumption that D-structure encodes the basic thematic

relations in the sentence as determined by the argument structure of the

predicate, and S-structure reflects the position of arguments following the

application of move-a., the class of NPs that can surface with the Genitive of

Negation can be captured by appealing to the distinction between the D-structure

and S-structure positions of arguments. Thus, whiIe the NP dogs in (6) and the

NP mushrooms in (8) appear in the same S-structure position with Nominative

Case, they do not originate in the same position at D-structure.

With respect ta the Russian data, this hypothesis allows us to distinguish

between the NPs in intransitive sentences that can surface with the Genitive of

Negation from those that cannot, as weil as capture the common feature of the

NPs in transitive and intransitive sentences that can surface with the Genitive of

Negation, by making reference ta their D-structure position. Those that can

surface with the Genitive of Negation, like the direct objects of transitive verbs,

are internai arguments of the verb.

141

• 3.2.2. OPTIONAUTY AND (IN)DEFINlTENESS EFFEcrs

•

•

The second complicating factor involves the semantics of these

constructions. While the Genitive of Negation used to be obligatory in Russian,

negated sentences now exhibit an apparent optionality, illustrated in (9)-(10).

The direct object in a negated transitive sentence can appear either with

Accusative Case (as it would in the affirmative counterpart) or with Genitive

Case.

(9) ja ne vizu kniguI-nom Neg see book-ACC

(10) ja ne vizu knigiI-nom Neg see book-GEN1do not see a book.

It is well documented that these options have repercussions on the

interpretation of these sentences; specjfically, the choice of Case-marking

appears to be related to whether the NP in question is definite or indefinite. In

English, definite and indefinite NPs can be distinguished by the choice of

determiner, i.e. a book (mdefinite) vs. the book (definite). This contrast is not

overt1y Wllized in Russian. The resu1ting ambiguity with respect te the

deftniteness of a direct object NP assigned accusative Case is illustrated in (11).

(11) on cita kniguhe-NOM read book-ACCHe read a/the book.

Negated sentences in Russian can potentially be disambigœted with

respect te the definiteness of the direct object by the appearance of either

142

•

•

Genitive or Accusative Case. However, there is a lack of consensus on the exact

pattern of interpretations. Reformatskij (1967), cited in Neidle (1988:34), gives

the following contrast.

(12) ja ne vizu ImiguI-nom Neg see book-ACC1do nol see lhe book.

(13) ja ne vizu ImigiI-nom Neg see book-GEN1 do nol see a book.

Pesetsky (1982) states that an NP with the Genitive of Negation tends to

be interpreted as indefinite, and the non-Genitive option is ambiguous.

However, Neidle (1988) maintains that it is the Genilive option that is

ambiguous in this respect, and that the Accusative option forces a dejinile

reading of the NP. These apparently contradictory facts will be addressed in

Part 2 of this chapter.

In the following sections, 1 will concentrate on capturing the constraints

on which NPs have the option of surfacing with the Genitive of Negation. Only

after accounting for the structural constraints on Genitive marking will the

described definiteness effects be incorporated inta the anaIysis.

3.2.3. SUMMARY

•

The basic paradigm can therefore be captured as follows. There are two

kinds of conditions on the occurrence of the Genitive of Negation, beyond the

initial restriction ta negated sentences. The fust condition can be defined

structurally, in terms of the D-structure position of the NP in a negated

143

• sentence. The second condition is defined semantically, in tenns of the ultimate

interpretation of these sentences. In Part 2, this second restriction will also be

derived by referring to characteristics of the NP itself; its specification for the

feature [defmite].

Before proposing an analysis, 1 will briefly outline Pesetsky's (1982)

treatment, which will demonstrate how these facts have previously been

accounted for within the Government & Binding Framework, as weIl as

introduce sorne additional features of the paradigm that require explanation.

3.3. A 1'REVIous ANALYSIS: 1'ESETSKY (1982)

Pesetsky proposes an analysis of the Genitive of Negation in Russian that

is driven by the observed parallels between this and two other constructions in

Russian: the Genitive Case assigned in numeral phrases and the Dative (DA'l)

Case assigned in distributive po-phrases. These phrases are argued to pattern in

certain central respects with phrases in the Genitive of negation.

Phrases that surface with Genitive under negation and distributive po­

phrases that surface with DAT (i) are restricted to D-structure objects, (ii) are

overridden by the requirements of oblique Case-assigning verbs, and (m) cio not

trigger subject a..oreement on the verb. Numeral phrases are somewhat less

straightforward, in that unlliœ phrases in the Genitive of Negation they may or

may not trigger verbal agreement.

Pesetsky claims that these three types of phrases in Russian are base­

generated Quantifier Phrases (QPs) rather !han NPs. He deals with the

apparently optional triggering of agreement by numeral phrases by maintaining a

•

•

3.3.1. C-SELEcnON OF NPs VS. QPs

144

•

•

distinction at sorne level of the grammar between agreement and no-agreement

numeral phrases. He cIaims that the latter, like Genitive phrases under negation

and po-phrases, are Q(uantifier) phrases rather than NPs.

(14) Distributive po phrase:[ [PO] [jabloku]]QP Q Ndist apple

(15) No-agreement numeral phrase:[ [sest'] [studentov))QP Q Nsix students

(16) Genitive phrase under negation:[ [el [pisem))QP Q N/elters

[pesetsky 1982:88]

According to Pesetsky, Genitive phrases under negation differ from the

others in that the quantifier is nul!. These phrases can only occur in negated

sentences because they require negation ta license this nuIl quantifer. This is

essentially the only role played by negation in his analysis.

The distinction between agreement and T/()-agreement numeral phrases

lies in their choice of head. If Q is the head, the phrase is a QP; if N is the

head, the phrase is an NP. An agreement numeral phrase is iIlustrated in (17)•

145

•

•

•

(17) Agreement numeral phrase:[ . [sest'] [studentov]]NP Q Nsix studems

[pesetsky 1982:89]

Pesetsky appeals to the difference between NPs and QPs to derive the

syntactic restrictions on and interpretation of phrases in the Genitive of

Negation. The essential components of his analysis are given below.

3.3.2. CASE-ASSIGNMEI\'T

With respect to structural Case assignment, Pesetsky maintains that

Nominative can be assigned to any NP governed by Agr (working within a set

of assumptions where Agr is part of J, the head of IP) and Accusative can be

assigned to any NP governed by a vero, regardless of theta raIe [pesetsky

1982:92].2

Pesetsky daims that QPs are like semences in that they do not require

Case, but like NPs in that they bear theta-rales. The potential problem for

subcategorization associated with a QP occurring in a position in which a vero

subcategorizes for an NP is handled by claiming that C(ategory)-selection is not

checked until LF, at which point there is oniy the trace of the raised QP ii:! the

position subcategorized for by the vero. The trace of a moved QP is an NP­

trace, which satisfies C-se1ection.

Depending on where one initiates this interconnected set of daims, one

could in principle derive a different fact each time. Pesetsky's analysis stipulates

(1) !hat these phrases are QPs, (2) !hat the trace of a QP is an NP-trace, and (3)

that C-se1ection is checked al LF. TIùs forces the QP to adjoin to IP at LF, in

order for C-se1ection to be satisfied. TIùs in tum disallows a QP in subject

146

•

•

position at S-structure (where it would surface with Nominative Case) because

its trace would violate the ECP following LF adjunction; the moved QP does

not agree in categorial features with the trace, and therefore cannot

antecedent-govem it. The inability of a QP to be in spec/lP at S-structure

accounts for the lack of subject agreement with phrases in the Genitive of

Negation.

The restriction to D-structure objects is thus accounted for: LF raising

must take place from a properly govemed position (i.e. a position governed by

V, not by Agr), or the trace would violate the ECP.

Pesetsky notes that certain expressions of duration of rime in Russian,

though not theta-marked by the verb, can bear Genitive in negated sentences.

This is illustrated in (18)-(19).

(18) ja spal odnu minutu1 slept one-ACC minute-ACC1sleprfor one minute.

(19) ja ne spal ni odnoj minuty1 Neg slept NOT one-GEN minute-GEN1didn 'r sleepfor one minute.

These facts raise problems for Pesetsky's analysis, which posits the

theta-assigning properties of the verb (specifically the C-selection requirement

checked at LF) as the driving force behind the patterns associated with the

Genitive of Negation. Duration of time expressions are not assigned a theta role,

but can neverthe1ess occur in the Genitive of Negation, in which case they are

subject to the same restrictions as other phrases which he analyzes as QPs.

147

• Pesetsky cannot resort to the C-selection properties of the verb to force

these phrases to raise at LF, sc he proposes that such expressions are

'inherently' C-selected (although not theta-marked) NPs. This forces them to

raise at LF in order 10 leave an NP trace and satisfy C-selection requirements.

He leaves open the question of how inherent C-selection works.3

3.3.3. OPTIONALITY AND INrERPRETAnON OF GENITIVE NPs

•

•

Pesetsky maintains that, in sentences exhibiting a Nominative-Genitive

alternation, as well as in those exhibiting an Accusative-Genitive alternation, the

interpretation of Genitive phrases is more restricted than that of their non­

Genitive counterparts. He captures this difference in interpretation by their

configuration at LF; whereas the sentences containing a Genitive NP (e.g. (20»

are necessarily interpreted as quantifier-variable structures, those containing a

Nominative or Accusative NP (e.g. (21» are ambiguous.

(20) ne pojavilos' studentovNeg showed up-neut-sg students-GENNo studell1S showed up. [i.e. Studell1S didn 'r show upl.

(21) studenty ne pojavilis'students-NOM Neg showed up-pl1he studell1S didn 'r show up. or No studenls showed up.

In the LF representation of (20), negation bas scope over an existential

operator.

(22) ...,3x, (S)x 1\ x showed up

148

• This forces an interpretation where the existence of the NP is negated.

The second sentence allows this interpretation, but also allows an interpretation

with the existential operator having wide scope, as illustrated in (23).

(23) 3x, (S)x ....(x showed up)

This is interpreted as positing the existence of the NP, but negating sorne

property attributed to it by the predicate. Following May (1977), Pesets\....y treats

the LF structure shown in (22) as derived by adjunction at LF. He derives the

obligatory quantifier-variable reading of Genitive phrases under negation by the

fuct that Quantifier Raising (QR) of the Genitive phrase is oLligatory, driven by

the C-selection requirements holding at LF.

While this treatment predicts an obligatory indefinite interpretation of an

NP when the Genitive option is used, the fuct that two readings are available if

the Nominative option is taken is more problematic.

Regarding agreement vs. no-agreement numeral phrases in Russian,

Pesetsky claims that if a numeral phrase is a QP, it must raise at LF. It cannot,

therefore, be in subject position at S-structure, and neither surfaces with

Nominative Case nor triggers agreement on the verb. If generated as an NP, it

is not subject to this restriction. It can (in fuct must) he in subject position al S­

structure. The appearance or non-appearance of agreement marking on the verb

was introduced as evidence for whether or not a phrase is a QP, and,

accordingly, whether or not it undergoes QR.

This strict correlation between agreement-marking and lack of QR, and

no-agreement marking and obligatory QR, is problematic for an account of the

• 3.3.4. PROBLEMS wrm THE TREATMENT

149

•

•

•

optional reading associated with Nominative NPs in negated sentences. Beth the

reading associated with QR and that associated with no QR are available to these

arguments.

If we loosen the correlation by allowing NPs in subject position (bearing

Nominative Case) te undergo optional QR, as suggested by these data, then the

optional reading associated with Nominative NPs in negated sentences is

accounted for. However, the account of agreement vs. no-agreement numeral

phrases becomes somewhat uninteresting. Essentially, the argument is that a

numeral phrase that is generated as an NP must raise to subject position at S­

structure. From this position it can optionally undergo QR, because it can

antecedent govern its trace. A numeral phrase generated as a QP cannot raise to

subject position at S-structure, because it will leave behind an NP trace when it

undergoes obligatory QR. Given that there are no means of deciding whether a

numeral phrase is a QP or an NP other than by whether it triggers agreement on

the verb, and agreement reflects its position in spec/IP, all that this in fuet

amounts te is a claim that agreement and no-agreement numeral pllral.:es are in

different positions at S-structure. This is not related te any difference in

interpretation. The question raised by this argumentation is why certain numeral

phrases are generated as QPs at all.

An additional problem is !hat, while the Nominative option discussed

above is ambiguous between the !wo readings, an NP marked with Accusative in

a negative sentence is not optionally interpreted as a quantifier-variable

structure; in fact. it appears te be unambiguously interpreted as a detinite NP

outside the scope of negation. Pesetsky (1982:66 and 214f, fn. 20) notes this as

a problem, but does not address il. It is difficult te see how bis analysis could

ISO

•

•

account for such a restriction, given that he must allow NPs to undergo opùonal

QR.

Pesetsky's analysis of these constructions is impressively thorough, and

although it relies quite heavily on stipulations (as pointed out by the author

himself) and there are problems with the predictions it makes, it provides an

interesting explanation for many of the restrictions on the Genitive of Negation.

Furthermore, it relates the restrictions on phrases in the Genitive of Negation to

no-agreement numeraI phrases and the DAT-Case of distributive po-phrases- a

correlation which will not be directly addressed in this thesis.4

In Pesetsky's analysis, the role played by sentential negation is relatively

minor; essentially, negation is required te license the null quantifier which

makes Genitive phrases QPs instead of NPs. However, as argued above,

positing a distinction between QPs and NPs runs into difficulties with the

asymmetries found between phrases that otherwise bear Accusative case (objects

of transitive verbs) and phrases that otherwise bear Nominative case (objects of

unaccusative or passive verbs).

Below, 1 propose an analysis which gives significantly greater weight to

the role of sentential negation in aIlowing the Genitive of Negation. 1 will argue

that the Genitive Case found in negated sentences is checked in the specifier of

NegP.

3.4. PART 1: STRUC'roRAL REs'1'RICIlONS ON THE GENITIVE OF NEGATION

3.4.1. INrRODUCIlON

•1 will assume a version of the Uniformity of Thematic Assignment

Hypothesis (UTAH: Baker (1988», such that AGENTS are always generated in a

151

•

•

position structurally higher than the position where PATIENTS/THEMES are

generated.s Agents are generated in spec/DeltaP, PATIENTS/THEMES in t.he

complement of V position.

In this analysis, as in the preceding chapter, agreement will be seen as

the reflex of S-structure Case-checlâng. Since only the NP marked with

Nominative Case triggers agreement, 1 will assume that only Nominative Case­

checlâng is an S-structure phenomenon; other Cases in Russian are checked at

LF.6

Unlike in Welsh, S-structure Case-checking is not satisfied by the

creation of an A-chain with a pleonastic anteeedent. It requires movement of the

NP into specifier position. Thus, basic Russian ward arder is SY~, with the

Nominative subject (in specrrP) preceding the verb in T. A further difference

between the Russian and Welsh Case-systems is that in Russian there is no

distinction between the levels of the grammar at which Case-!icensing of

pronominal and non-pronominal arguments takes place.

As in Welsh, speclVP in a [+tense] sentence is not a !icit SpeclHead

configuration for Case-checking of the direct abject, because the head of VP

does not contain a morphologically complete member of the verb-chain.

Accordingly, in a [+tense] sentence, Case is assigned ta the direct abject at S­

structure via govemment by the member of the verb chain in Delta.

In the preceding chapter, it was argued that speclNegP can Case-!icense

an NP independently of V-raising. In this chapter, we will extend this claim ta

Russian. This will account for the ability of an NP ta surface with the Genitive

of Negation even when the argument of a [-Case] (i.e. unaccusative or passive)

verb.

152

• 3.4.2. RUSSIAN PHRAsE STRUCI1JRE

•

The basic Russian word order is SVO. Phrases can be scrambled fairly

freely, obscuring the underlying order. However, the scrambled sentences are

typicaIly focussed or topieaJ;zed constructions, marked by obligatory stress.

Phrases are head-initial. The minimal initial hypothesis, in keeping with the

assumptions adopted in this thesis, is that Russian phrase structure consists of

head-initiaI phrases, with specifiers occurring to the left (i.e. preceding the head

and its complements). This is illustrated in (24).

(24)

v

3.4.3. PosmON OF NEGP

Since we base our analysis of the Genitive of Negation on Case-licensing

options in negative contexts, we must establish the base position of NegP in

Russïan. Sentential negation in Russian, represented by the element ne, foIIows

the Nominative NP and immediate1y precedes the first verbal eIement in a

clause. In an anaIysis which incIudes an AgrSP above TP, this might be taken to

suggest that NegP in Russian is generated between AgrP and TP, and

153

•

•

•

Nominative NP raises 10 spec/AgrSP. However. 1 will propose a lower base­

position for NegP in Russian, immediately above VP.

The D-structure of a negated sentence with an unaccusative verb in

Russian is given in (25).

(25)

V NP

On the assumption that a higher VP is generated in all sentences whose

verb projects an externaI theta role, the cIaim that NegP selects VP becomes

ambiguous; it couId conceivably be generated immediately above the higher VP

(DeltaP), or it could he generated immediateIy above the lower VP. 1 will

maintain that NegP is always generated immediately above the lower VP. In this

way, its base-position is above the base-position of the V, and intervenes

between this position and Delta, as shown in (26).

154

• (26)

•

•

V NP-imernal

The S-structure position of Neg preceding the verb can be accounted for

in one of two ways. First, the head of NegP in Russian may be a clitic, in which

case it may raise to the head of a higher functional category, i.e. T., at S­

structure. Clitic status is invoked to determine the S-s!IUcture position of the

pre-verbal negative marker in French (pollock 1989) and in Italian (Belletti

1991), both analyses positing a NegP be10w TP.

A second possibility for deriving the surface word order makes use of

syntactic affixation; that is, affixation which takes place via head-to-head

movement in the syntax. This idea is most fully worked out in Baker (1988).

Under such an analysis, the negative marker precedes the verb at S-structure

because the verb bas picked up the negative marker as it raises from its base­

generated position to T. The verb does no! have the option of •skipping over'

the head of NegP because the interVening negative marker would create a

lSS

•

•

minimality barrier between the V and its trace in VP, resulting in an ECP

violation.

In the second option, the raised verb can antecedent govern its trace over

the NegP by virtue of the Government Transparency Corollary (GTC, Baker

1988). If the raised head governs the matimaJ projection out of which it bas

raised, that maximal projection is no longer a barrier 10 government by that

head.

The fust option requires sorne further stipulations concerning the

def1nition of barrierhood. In principle, the verb should not be able to raise over

the head of NegP and still antecedent govern its trace (since the head of NegP

creates an interVening barrier). Bel1etti (following work by Moritz (1989) and

Rizzi & Roberts (1989» avoids this problem by a further assumption- that the

Agr node into which the verb raises carries the index of the raised negative clitic

and of the raised verb. By defining minimality in terms of chains, Bel1etti can

argue that the head (Agr) is the head ofboth the Neg-chain and the V-chain, so

the potentially problematic trace of the verb can, in essence, be antecedent

governed by the Neg-trace in NegP.

In the absence of independent empiricaJ motivation from Russian for

altering the def1nition of barrier in terms of chains and allowing raising of the

head of NegP independently of V-raising, the relative order of negation and

verb at S-structure is derived as folIows. The head of NegP in Russian is a

morphologically dependent item, like Tense morphology. It is lexically specif1ed

as dominating an empty sIot 10 its right, into which the V must raise.

(27) Neg: [Neg _]V

156

• (28) T: [_11v

•

This head-to-head movement, being an instance of subsùtuùon, resullS in

a Neg head that dominates both a Neg and a V element.7 Following V-raising to

Neg, further verb-raising (i.e. to Delta or to T to support tense morphology)

necessarily carries the negative marker a10ng with the verb. This is consistent

with the fact that no lexical material may intervene between the sentenùal

negation marker and the verb in Russian.

The cIaim that V a1ways combines with Neg is consistent with the fact

that in spite of the fairly free word order found in Russian due to scrambling,

the order Neg+Vis not disrupted. If a verb in a [-tense] sentence had the option

of not raising to Neg, then sentences might be expected where only the lower

VF had been scrambled and the negative marker left behind in VP, but this is

not the case. Scrambling the negated verb necessarily carries the negaùve

marker a10ng with it.

The S-structure of a negated [+tense] transitive sentence is given in

(29).

(29)

T

nNeg T

nNeg V

t NPo

157

•

•

•

In summary, the morphemc order and word order found in Russian

negative sentences are compatible with the treatment of the sentential negative

marker as the head of a functional projection generated between the two tiers in

a Larsonian VP, under the assumption that the negative marker is

morphologicaIIy specified as dominating ai: empty slot to its right, which the

verb moves into.

In the following section, 1 propose an analysis of the assignment of

Genitive Case in negated sentences. This analysis is an extension of the proposaI

that the specifier of NegP is a position in which NP-chains can be Case­

Iicensed.

3.4.4. CASE-ASSIGNMENl"UNDER GoVERNMENT IN NEGATED

SENTENCES

In a [+tense] sentence, Accusative Case is assigned under govemment,

by the member of the verb chai'1 in Delta to an NP in the specifier position of

the functional projection it immediately dominates. The specNP in a [+tense]

sentence is not a licit Case position for Spec/Head coindexation because the V is

not morphoIogicaIIy complete until it has combined with T. Hence, in an

affirmative sente.lce, the PATIENT/THEME NP raises 10 specNP at LF and is

Case-Iicensed under govemment.

Now, consider how this system is affected by the presence of NegP

which intervenes between Delta and the specNP. The V raises through

Negation, then into Delta, then on 10 T. The member of the V-chain in Delta is

a trace of the [Neg+V] complex (technicaIIy, the trace of Neg). We have said

that from this position Case can be assigned 10 an NP in the specifier of the

lSS

•

•

•

functional projection it immediately dominales, which in this case is spec/NegP.

This is illustraled by the S-structure given in (30).

(30)

Neg+V-t

4CASE

V-t NPo

However, what we find in such sentences is Genitive Case, not

Accusative Case. TIùs is compatible with IWO possible explanations: First, it is

conceivable that a V which has combined with Neg assigns Genitive instead of

Accusative Case under government, perhaps by virtue of the fact that the

member of the verb-ehain in Delta is not of the category V but of the category

Neg.

Second, it is possible that the NP in speclNegP is Case-!icensed through

Spec/Head coindexation; that while specIVP is not a !icit SpeclHead

confi!:uration in a [+tense) sentence (the verb at that point in the derivation not

meeting the morphological completeness requirement in the definition of a !icit

Spec/Head configuration), Neg carries a [+Case) feature which makes

speclNegP a !icit Case configuration.

159

•

•

Based on data from unaccusative and passive verbs, 1 will argue for the

latter analysis. As 1 will show in the following section, these verbs are not

specified as [+Case]. This motivates NP-raising to speclTP in affirmative

sentences. However, the head of NegP provides the needed [+Case] features for

Case-ehecking of an NP in negated sentences.

3.4.4.1. ON THE R.E:LAnONSHIP BETWEEN GENl'IlVE ANDACCUSATIVE

First, consider the ramifications of the fust proposaI: anaIyzing the

Genitive of Negation as an aItered version of the verb's ability to assign

Accusative Case under govemment. We would therefore expect that only

sentences whose verbs carry the feature specification [+Case] should exhibit the

Genitive of Negation. Such a proposaI is immediately contradicted by the

occurrence of Genitive of Negation on the D-structure objects of passive and

unaccusative verbs- verbs which, according to Burzio's Generalization (BG:

Burzio 1986), neither project an extemaI argument nor assign Accusative Case.

BeIow, 1 will show that Russian conforms to BG, and that an anaIysis which

treats the Genitive of Negatio:l as independent of the Case specifications of the

verb is preferable.

3.4.4.1.1. BURZIO'S GENERALIZATION

The relation between a verb's internaI argument structure and its ability

to assign Accusative Case is described as a kind of mutuaI dependency in BG,

given in (31).8

160

•

•

(31) T<->AWhere:T = Theta marking of an extemal argument,and A = Accusative Case.

[Burzio, 1986: 185]

The effect of BG can be ilIustrated by the behaviour of passive verbs.

The intemal argument (THEME/ PATIENT) of the verb surfaces with Nominative

Case and, in English, appears in subject position. Under standard GB analyses,

it is claimed that this follows from the passive verb's failure to assign

Accusative Case, which forces the argument generated in complement position

to raise to the position where Nominative Case can be assigned.

Given the assumption that theta roles are assigned to theta-positions,

movement from D-structure object position to S-structure position is potentially

a violation of the theta-criterion, given in (32)•

(32) THETA-CRlTERION:A lhela ,ole musl be assigned 10 one and only one argument. andevery argument must be assigned one and on/y one theta raIe.

Such movement is permitted in these cases because the subject position is

not assigned a theta role by passive verbs. The raised NP forms a chain with its

trace in complement of V position; the chain is rendered visible for theta­

marking by Nominative Case assignment. The D-structures and S-structures of

the sentences 11ùs slory was believed by the villagers (passive) and These men

arrived (unaccusative) are given below.

(33) DS [eIP

[was [believed [this story) ]by the vilIagers]l' VP

161

(wasl'• (34) SS [This storyi

IP1 _

NOMINATIVE

[believed (ej] ]by the villagers]]VP

1__-a-role

(35) DS (eIP

( (arrived (these men] ]VP

NOMINATIVE

(these meniIP

(36) SS (arrivedk1'1

1 1

( ... tk (ej] ]VP ,--

a-role

•

•

The assimilation of verbs like arrive 10 the c1ass of passive verbs is

motivated in part by the possibilities of ne-eliticization in Italian, which is

possible only if the NP is dominated by V'. It is argued that the fact that ne can

diticize in the first two sentences given below, but not in the third, supports the

daim that the NP in (37) and (38) is in object position (the position dominated

by V').

(37) ne furono arrestati moItiof-them were arrested manyMany ofthem were arrested.

(38) ne arrivano moItiof-them arrived manyMany ofrhem arrived.

(39) ·ne telefonano moItiof-them telephoned manyMany ofrhem telephoned.

162

•

•

Now, BG identifies just those intransitive verbs in Russian which permit

the Cenitive of Negation on their arguments. Problematically for the claim that

Genitive of Negation is dependent on the Case-assigning potential of the verb,

BG also identifies those verbs as lacking Accusative Case-assigning potential.

Thus, in (40), the NP gribov shows a Nominative-Genitive altemation in

negated sentences.

(40) gribov zeles' ne rastetmushrooms-GEN-pl here Neg grow-3sgMushrooms don 'l grow here.

griby zeles' rastutmushrooms-NOM-pl here grow-3plMushrooms grow here.

[Pesetsky 1982:43]

Before concluding that the Genitive of Negation is independent of the

verb's Case-licensing potential, we will consider some apparent violations of

BG, on the reasoning that if BG does not hold in Russian, then it should not be

invoked as an argument against analyzing the Genitive of Negation as a marked

variant of Accusative Case.

3.4.4.1.2. APPARENT VIOLATIONS OF BURZIO'S GENERALIZATION

•

Certain verbs (discussed in Babby (1975) and Pesetsky (1982»,

including verbs of uncontrollable physical sensation, like losnil ('nauseate') or

znobit ('chil1'), are followed by NPs bearing Accusative Case. What is

problematic for BG is that there is apparently no extemal argument in these

constructions. This is illustrated in (41).

163

•

•

•

(41) tosnilo Masunauseated-neut-sg Masu-ACCMasafelr nauseared.

[pesetsky 1982:211 fn.15]

Also included in the class of verbs that apparently contradict BG are

verbs that "denote a physica1 action that can occur spontaneously, without

initiation or intervention of an animate agent" (Babby 1975: note 4). Thus, in

example (42), the instrument must be sorne non-volitional force such as by lhe

wind or by fare.

(42) veter unes lodkuwind-masc-NOM-sg carried away-masc-sg boat-fem-ACCWind carried away rhe boar.

(43) uneslo lodku vetromcarried-away-neut-sg boat-fem-ACC-sg wind-masc-INSTR-sgThe boar was carried away (by wind).

(44) ·uneslo lodku lvanomcarried-away-neut-sg boat-fem-ACC-sg lvan-masc-INST-sgThe boar was carried away by Ivan.

[pesetsky 1982:211-212, fn. 15]

Rather than conclude that BG does not hold in Russian, we will adopt

Pesetsky's (1982) ana1ysis of these verbs as extema1 theta-role assigners. The

theta role is assigned to a quasi-argument, Le. "natura! cause", similarly to the

extema1 theta role assigned 10 the ù-subject in weather constructions in English

(e.g. Il Tained yeslerday). 11ùs is consistent with the restricted interpretation

accorded such sentences; there is no volitional agent implied. Assuming that the

impersonal subject is not specified for these features, it is also consistent with

164

•

•

the 3rd person neuter agreement found on the verb when the overt argument

surfaces with Accusative Case.

Crucially, this restriction is not found with passive verbs which surface

with the Genitive of Negation; the understood agent can be animate and the

action can be volitional.

(45) ni odnogo goroda ne bylo vzjato vragomnot one-GEN city-GEN Neg was-neut-sg taken enemy-INSTNor one city was taken by the enemy.

(Chvany 1975, cited in Pesetsl.:y 1982:42]]

Furthermore, although the Genitive of Negation is permitted on the sole

arguments of passive and unaccusative verbs, Accusative Case on these

arguments is not an option in either affirmative or negative sentences. Thus,

whatever analysis is proposed to account for constructions where the sole overt

argument bears Accusative Case, it is distinct from occurrences of the Genitive

of Negation on arguments of passive verbs.

The apparent exceptions to BG can be put aside, and we will assume that

in Russian, as in other languages, passive and unaccusative verbs do not assign

Case to their internal arguments, and the Genitive of Negation on arguments of

these verbs cannot be analyzed as an altered version of the verb's Accusative

Case-assigning ability. The alternative proposaI, which treats the Genitive of

Negation as independent of the Case assigning potential of the verb, is

deve10ped be1ow.

165

• As argued earlier, any member of a chain can Case-license an element in

its specifier if it is specified [+Case]. 1 will propose that the head of NegP is

inherently specified as [+Case], and that Genitive Case can be checked in its

specifier. In this way, the J)-structure object of an unaccusative or passive verb

can raise te spec/NegP for Case-checking. This is shown in (46).

•

(46) TP

NP-t1 ! j~'---'--'I

1 V-t NP-t,i, ,L-- .,

Below, 1 will show how this proposaI accounts for restrictions on the

Genitive of Negation.

3.4.5. STRUCl1JRAL CONSI'RAINTS ON GENnIVE OF NEGATION

3.4.5.1. REsnucnON 1'0 NON-AGENllVE NPs

•Consider first a negated transitive sentence. The NP that would surface

with Accusative Case in the affirmative equivalent can surface with Genitive

Case, but the NP that would otherwise surface with Nominative Case cannot.

166

• Given that NegP is generated below the D-structure position of the

agentive NP, this constraint is expected. The agent is generated in spec/DeltaP,

higher than the base position of NegP. As such, its only potential Case position

is speclTP, where it surfaces with Nominative Case. Genitive Case on such an

argument would require lowering to spec/NegP, leaving an antecedentless trace

in spec/DeltaP in violation of the ECP. Only NPs generated below NegP can be

Case-licensed in speclNegP. The situation is illustrated in (47).

•

(47)

i1

1: NP-agentl ~1 .:

Vot

3.4.5.2. COMPLEMEN'IS OF OBUQUE CASE-ASSIGNING VERBS

•

A class of Russian verbs, termed Oblique Case-Assigning Verbs by

Babby (1980), do not allow the Genitive of Negation on their complements. To

address these data, we will initially refer to a distinction that bas been made in

Russian grammar between structural Case and semanric case.

This distinction is motivated in Russian primarily by constraints on

passivization. Passivization genera1ly results in the D-structure object of the

167

•

•

•

verb raising te subject position, where it surfaces with Nominative case. There

is a class of verbs in Russian which do not allow this construction; instead, they

require an existential construction with a null pleonastic subject, and the Case of

the object remains unchanged. An example is the transitive verb vredil ('harm').

In a non-passivized sentence, the object of vredil surfaces with Dative case. If

the sentence is negated, this NP does not have the Genitive Case option. This is

illustrated in (48)-(50).

(48) on vredilljudjamhe-NOM harmed people-DATHe hanned people.

(49) on ne vredilljudjamhe-NOM Neg harmed people-DATHe didn '( hann people.

(50) *on ne vredilljudejhe-NOM Neg harmed people-GEN

[Neidle 1988:37]

1 will fust investigate the possibility that these verbs assign inherent Case

te their complements at D-structure, and then propose an analysis which invokes

a null preposition in these constructions.

The inability of these arguments te raise te Nominative position under

passivization could in principle be derived by appealing te a disti.'1et Case­

marking operation which takes place at D-strueture: inherent Case-marking. If

arguments of these verbs are assigned Case at D-structure, raising te a position

at S-structure where Nominative is assigned would result in an NP chain with

168

two Cases, creating a Case-eonflict which has been argued to be a violation of

the Case-Filter.

Since this sarne class of verbs does not pennit the Genitive of Negation

on their complements, this analysis would be consistent with the claim that the

Genitive of Negation, like Nominative, is a structural Case, and cannot be

assigned ta an NP that already bears inherently assigned Case.9 Below, 1 will

argue lhat, based on the behaviour of prepositional objects, the objects of sa­

called oblique Case-assigning verbs can be analyzed as PPs headed by null

prepositions. In this way, parallels between the behaviour of these phrases with

respect to the Genitive of Negation and passivization can be captured without

appeaJing to two distinct Cas .-mechanisms.

According to Babby (1980), objects ~f prepositions are also assigned

semantic Case. Support for the claim that prepositions assign semantic Case

comes from prepositions that express motion; whereas prepositions that express

motion roward require Accusative case, prepositions that express motion away

from generaJIy requite Genitive Case. Il has been suggested (Neidle 1988:12)

that prepositions assign a partially specified feature bundle, and that the second

feature value is detennined by the directionality of motion.

For our analysis, accounting for the absence of the Genitive of Negatior:

on prepositional complements does not require reference ta a structurallsemantic

Case distinction. The lack of the Genitive of Negation is expected because the

complement of P is Case-licensed by the preposition itself, within pp.Jo In this

way, we do not rnd the Genitive of Negation on the object of a preposition for

the sarne reason that we do not find the Case associated with the verb on the

•

•3.4.5.3. PREPosmONAL COl\fPI..EME!'<"I'S

169

• object of a preposition in other languages: the P determines the lowest possible

Case-eonfiguration for its complement, and Case is checked only once for a

given chain. This is illustrated in the LF structure given in (51).

NP

/i'Li P1,, ,, ,, ;

NPo-t i :_._---_.---.__••,1

toi

NPo·GEN

Lt_i

NPo-t

NPs·NOM

L [Neg+V+n·j

(51)

•These fucts suggest that there may be a structural explanation for the lack

of Genitive of Negation on the arguments of oblique Case-assigning verbs as

weIl. We can extend this account to oblique Case assigning verbs by analyzing

their complements as PPs, with a null prepositionaI head which is responsible

for Case-assignment (Libert 1992).

•(52) on

he-NOMnevrediINegharmed

Ijudjam[0 [people-DAT]]pp

170

•

•

The prepositional ("oblique") object is Case-licensed by the null

preposition P. Like objects of lexically reaIized prepositions, slJch arguments do

not undergoing raising to speclTP under passivization of the verb; passive

morphology on the verb does not affect the ability of these arguments to be

Case-licensed by P.

Support for this analysis of prepositional objects and oblique-Case

assigned objects cornes from double-object constructions. The Dative object in a

double object constructions does not surface with Nominative Case in passives;

nor can it surface with the Genitive of Negation. On the assumption that double

object constructions are structurally asymmetrical (Larsen (1987», the Dative

can be seen as generated within a pp headed by a null preposition (Baker 1988),

and these three paradigms where Genitive of Negation is not found have a

common explanation. The structure of a negated double object construction in

Russian is given in (53).

•

(53)

spec

TP

171

• 3.4.5A. A CLAUSAL CONSTRAINT

•

•

The possibility of the Genitive of Negation being triggered in a 10wer

clause than the clause containing sentential negation, as described in Pesetsky

(1982), is constrained by the following conditions: Genitive cannot appear on an

NP object in the lower clause "unless the clause containing the NP and aIl

intervening clauses are infinitives" (pesetsky 1982:149). Thus, in (54a), neither

the subject nor the object of the lower tensed clause can occur in the Genitive of

Negation. However, in (54b), the Genitive of Negaêcn is permitted on the

object of the [-fmite] V.

(54a) *ja ne skazal [cto [ty pises stixov]1 Neg said that you write [indic] verses-masc-GEN-pl1didn'r say rhatyou wrire verses•

(54b) ja ne xocu [pisat' stixov]1 Neg want write [-finite] verses-masc-GEN-pl1don 'r wanr ro wrire verses.

[pesetsky 1982:149]

Under our account it is not surprising that the Genitive of Negation

would be blocked in the tensed embedded contexts. What does require an

explanation is the claim that the Genitive of Negation con appear in lower

[-finite] clauses.

1 argue that the descriptive generalization given by Pesetsky is somewhat

misieading, as it suggests a far greater flexibility !han is found. 1 will maintain

that these cases repre:;ent a small class of Russian verbs that can take a VP

complement. The sentences cited by Pesetsky as evidence that the Genitive of

172

• Negation can be assigned to an NP in a lower clause will be analyzed as

monociausaI.

3.4.5.5. GENI'I1VE OF NEGATION IN AN E!lmEDDED (-FINITE)CLAUSE?

Only a small class of veros in Russian allow a (-finite) vero to follow

them. The class of veros which allow a (-finite) complement express modai-like

properties, such as desire, abiliry, or inrenr. 1 will propose that these sentences

are inherently monoclausaI, as illustrated by the D-structure given in (55).11

(55)

V NPo

1 1ja ne xocu pisat' stixov

The higher V raises through Negation and Delta 10 T, and the subject

raises 10 spectrP for Nominative Case-checking at S-structure. As there are no

barriers between the speclNegP and the direct object, Genitive Case can be

checked in speclNegP at LF. The derived structure is given in (56).

173

• (56) TP

•

Il

(Neg+V+TI

ja ne xocu

stixov

v t-o1

1

pisat'

These cases can be distinguished from object-control verbs like

'persuade', illustrated in (57), which take a regular complement. Negating these

verbs is incompatible with the Genitive of Negation on the lower object, even

though the lower verb is [-finite].

(57) *ja ne ugovorila Natasha citat' knigI-NOM Neg persuaded Natasha read-INF books-GEN1 didn't persuade Natasha to read books.

[Neidle 1988:157]

The Genitive of Negation on the lower object is ooly pennitted if

negation intervenes between the two verbs.

174

•

•

(58) ja ugovorila Natasha ne citat' knigI-NOM persuaded Natasha Neg read books-GEN1persuaded Nalasha not to read books.

This is as expected. In cases like persuade, the lower constituent is a TP,

with an empty subject controlled by the matrix object NP, Nalasha. For the

lower object to be Case-licensed in speclNegP in the higher clause it would have

te A-move over the embedded subject. Accordingly, the object of the lower

verb appears only with Accusative Case. If, as in (58), NegP is generated in the

lower clause, then the Genitive of Negation is acceptable. Nothing blocks

movement from the base position of the direct object to speclNegP.

In impersonal constructions as weil, negation can intervene between the

[+finite] and [-finite] verbs, negating the infmitival.

(59) stalo ne suscestvovat' takix ljudejstarted-3-neut-sg Neg exist-inf such-GEN people-GENThere staned not tO exist such people.

[Neidle 1988:82]

Again, the Genitive of Negation is acceptable on the lower object in

these cases, as expected under our analysis. Tne Genitive of Negation is checked

in speclNegP at LF, following V-raising into Neg (which we claim te be

obligatery).

It is interesting in this regard that the sentence given in (60), with a

subject control verb trJ. is ungrammatica1 with the Genitive of Negation.

(60) takie 1judi staralis' ne prixoclit'such-NOM peop1e-NOM tried Neg come-infSuch people tried not tO come.

17S

• (61) ·staralos' ne prixodit' takix Ijudejtried-3-neut-sg Neg come-inf such-GEN people-GEN.

The NP takix ljudej ('such people') is agentive, generated in

spec/De1taP, and cannot lower te spec/NegP te be Case-licensed.

In sum, we have seen that the clausaI restriction on the Genitive of

Negation is not related te whether or not a lower clause is [-finite]. Rather, the

possibility is conditioned by the ability of an NP te raise te spec/NegP at LF for

Case-checking.12

3.4.6. SUPPORTING EVIDENCE

A major syntactic difference between sentential and what has been

termed constituent negation assumed here is that only the former generates a

NegP. Constituent negation is not the head of a phrase NegP, but a lexical item

adjoined te a phrase. By positing such a close relationship between the presence

of a NegP and the possibility of the Genitive of Negation, this analysis predicts

that the Genitive of Negation should not be available under constituent negation.

This is uphe1d by the data. The incompatibility of the Genitive of Negation with

constituent negation has been noted by many researchers.13

The distinction between sententiaI and constituent negation may not be

evident from surface word order aIone. Rence, pragmaticaIIy, one can choose te

negate onIy the vero in a sentence, rather than the clause, by contrastive1y

stressing the vero. This can be seen in the English sentence, 1 dü!n't SEE the

•3.4.6.1. SENTENIlAL NEGATION VS. CONS11TUENT NEGATION

1

film.

176

•

•

Sentences of this type can be distinguished from senlcntial negation in

that they allow violations of the downward entailment requirement associated

with sentential negation. In this way, while the truth of the sentence 1 didn't ear

a green vegetable logically entails the truth of the sentence 1 didn 't ear a

zucchini, il is permissible to state: 1 didn't ear a GREEN VEGEl'ABLE, 1 are a

ZUCCHINI.

These cases are therefore instances of constituent negation, not senlential

negation, and our analysis predicls that they should not permit the Genitive of

Negation. This prediction is substantiated, and the constraint is observed even if

the word order is identical to the unstressed (sentential negation) version, where

Genitive is permitted. The appearance of the Genitive of Negation forces a

reading where negation has sentential scope.

(62) on ne prosmatrivaet sls'ju a citaethe-NOM Neg looks over article-ACC (*GEN] but readsHe does not look over the article. but reads it.

[Neidle 1988:40]

Neidle supports this distinction between sentential and constituent

negation and the possibility of the Genitive of Negation with reference ta the

Academy of Sciences of the USSR Grammar (1980, VoI.2:417) where it is

stated that an Accusative object is required when one verb is negated in contrast

to another (Neidle 1988:40). This supports the dependency of the Genitive of

Negation on the functional projection NegP.

In sentences exhibiting the Ct'.nitive of Negation, these phrases most

naturally occur post-verbally, even if they ::l"~ the negated equivalents of•3.4.6.2. WORD ORDF.& DIFFERENCES

177

•

•

•

sentences where the same argument would occur in S-initial position, bearing

Nominative case. 14 Thus, the exarnples given below show the unmarked order

for a phrase bearing Nominative Case vs. a phrase bearing Genitive Case.

(63) ni odna devuska ne prislanot one-NOM girl-NOM Neg came-fem-sgNOl one girl came.

(64) ne prislo ni odnoj devuskiNeg came-3rs-neut.sg. not one-GEN girl-GENNOl one girl came.

[Neidle 1988:77]

This is exactly what is expected under this analysis. We have maintained

that Nominative Case is checked in speclTP at S-structure (triggering

agreement). Since we have maintained that Genitive Case reflects LF Case­

licensing in spec/NegP, a position lower than the S-structure position of the

verb in T, the Genitive NP cao only be S-initial if preposed by scrambling,

creating a marked structure which requires sorne kind of emphatic Stress. 15 As

shown in (65), the Extended Projection Principle is satisfied by a null pleonastic

subject NP.

178

•

•

(65)

t NP2

LF·raising

3.4.6.3. LACK OF AGREEMENT wrm GENl'IlVE PHRAsEs

•

In the intransitive sentences we have discussed, arguments of

unaccusative and passive verbs may surface with either Nominative Case or

Genitive Case. Subject agreement is analyzed as the reflex of Case-licensing in

specrrP. Given that the Genitive of Negation is checked in speclNegP, Genitive

NPs should not trigger agreement on the vero, even if they are superficially in

"subject" position, preceding the vero. These facts are borne out; Genitive NPs,

even when they are the sole NP in an intransitive S and preceding a tensed vero,

do not trigger agreement on the vero. This is shown in (67).

179

•

•

(66) griby zdes' rastutmushroom-NOM-PL here grow-3PLMushrooms grow here.

(67) gribov zdes' ne rastetmushroom-GEN-PL here Neg grow-3SGMushrooms don'( grow here.

[Pesetsky 1982:43)

3.4.7. CONCLUSIONTO PART 1

This analysis rests on the following c1aims about the syntax of Russian.

First, l have proposed that sentential negation in Russian (un1ike constituent

negation) generates a NegP. l have proposed that Case-checking of the Genitive

of Negation takes place in spec/NegP, which is made possible by the [+Case)

feature of the head of NegP. This accounts for the fact that the Genitive of

Negation is not possible on complements of prepositions or so-called oblique­

Case assigning verbs. Finally, we have accounted for the appearance of the

Genitive of Negation on a 10wer verb when a higher verb is negated, by

proposing that the higher verbs in these constructions take VP complements.

With these daims, we have developed an analysis that accounts for the

facts Iisted in (68) without requiring additional stipulations:

•

(68) (a) The Genitive Case found on an NP otherwise assigned Accusativein negated transitive sentences;

180

• (b) The Genitive Case found on an NP otherwise assignedNominative in (certain) negated intransitive sentences;

(c) The ungrammaticality of the Genitive of Negation on an NPotherwise assigned Nominative Case in a transitive sentence;

(d) The lack of subject agreement on the verb when an NP otherwiseassigned Nominative Case surfaces with the Genitive ofNegation;

(e) The unmarked position of an NP with the Genitive of Negation(following the verb);

•(f) The ungrammaticality of the Genitive of Negation on

complements of prepositions;

(g) The ungrammaticality of the Genitive of Negation oncomplements of oblique-Case assigning verbs;

(h) The ungrammaticality of the Genitive of Negation on NPs inlower [+finite] clauses when the higher clause is negated;

(i) The acceptability of Genitive Case on the direct object of certainlower [-finite] verbs when the higher verb is negated.

In the following section, 1 will propose an analysis for the apparent

definiteness effects exhibited by these constructions. The initial analysis will be

driven by the faets described by Neidle (1988), which are supported by my

informant work. 1 will, however, also address Pesetsky's findings.

181

• 3.5. PART 2: DEFThTIENESS EFFEcrs ASSOCIATED wrru TIIE GEI\'ITIVE OF

NECATION

3.5.1. INI'RODUcnON

Phrases that can appear with Genitive Case in negated sentences also

have the option of surfacing with Accusative Case or Nominative Case.

depending on the verb. However, these variations have repercussions on the

interpretation of the sentences; repercussions which differ dependir.g on whether

it is an otherwise Accusative-marked NP or an otherwise Nominative-marked

NP that surfaces with Genitive Case.

We have aiready proposed an account for how the D-structure objects of

transitive and intransitive verbs pattern together in their ability to bear the

Genitive of Negation. The greatest challenge to the (in)definitencss aspect of the

analysis of the Genitive of Negation is to account for the fact that there is an

asymmetry between these arguments when it comes to interpretation. Whereas

with the D-structure objects of intransitive verbs the non-Genitive option

(Nominative) leads to an interpretation that is ambiguous with respect to

definiteness, with the D-structure objects of transitive verbs the non-Genitive

option (Accusative) forces the definite reading. We will address these facts

bel(\w.

3.5.2. DEALING \\TIll THE CONCEPT OF (IN)DEFINITENESS

3.5.2.1. ACAVEAT

A caveat regarding the notion of (in)definiteness is appropriate here. In

much of the work on the Genitive of Negation in Russian, the NPs that surface

with this Case are described as indefinite NPs, or as NPs which can be

182

• interpreted as within the scope of negation (henee indefinite). In fact, the

tendency to use the Genitive of Negation is sensitive to severa! factors, and is

perhaps best described as a series of oppositions, with one extreme tending to

use the Genitive, and the other tending not to use it. The oppositions described

by Timberlake (1975), are summarized in (69).

(69) Genitive of Negation~COMMON->MORECOMMON

proper :.ounsconcretecountanimatesingulardefinitetepicalizedmodified

- > common nouns->abstract->mass- > inanimate->plural- > indefinite->neutra!- > unmodified

These features, Timberlake argues, can be reduced te the general

property of individuation.

Furthermore, the use of Accusative under negation is described as

gaining in frequency. The diachronic progression seems te be the following: the

Genitive of Negation used to be obligatory on the direct object NP in negative

sentences. Later, the use of Accusative Case becarne an option, but an option

that carried with it a restrictive ([+definite]) interpretation, a restriction from

which the use of Genitive was free. Neidle (1988) discusses this effect in the

dialect she describes.

183

•

•

•

To the extent that automaùc Geniùve marking of an object(regardless of scope relaùons) rep.esents an older grammaùcalsystem [sùll retained in Polish]... , an Accusaùve object wouldprovide more informaùon about scope relaùons (namely, that theobject is outside the scope of negaùon) than a Geniùve object(which may bear Geniùve for purposes of scope marking, or a~

an automaùc consequence of being contained in a negaùvephrase).

[Neidle 1988:37].

However, the data reported in Pesetsky (1982) suggest that their roles

are becoming reversed, with GEN now restricted to the indefinite, or

interpretation within the scope of Neg, and ACC relaùvely freely used without

any commitment to particular scope relaùons. Contemporary Russian appears to

be in transition between an initial state of affairs like that in present day Polish,

where the Genitive of Negation surfaces on any NP in the appropriate structural

configuration, regardless of its semantic features, and that in present-day Czech,

which has practically 10st the construction altogether. Thus, as noted in a study

by Ward (1959:213-214), the use of Accusative in Russian negated sentences

increased from 21.7% in 1918-23 to 38.3% in 1959.

The analysis presented here does not attempt to address the semantic,

pragmatic, and stylistic factors that are clearly invoived in delimiting the class

of NPs which (within the appropriate syntactic configuraùons) surface with the

Genitive of Negation. What is important here is that certain arguments in

Russian are appropriate candidates for the Genitive of Negation and others are

not; given this basic division, the task is ta capture the role pIayed by sentential

negation and the syntactic constraints on the assignment of Genitive of

Negation. As the construction evoives, there appear ta be new semantic

184

• restrictions on which NPs surface with this Case, but the syntactic restrictions

have not changed.

3.5.2.2. Do WE HAVE A PARADIGM?

While the pattern discussed (and anaIyzed) by Pesetsky (1982) can be

illustrated as a uni-directionaI implicature, with Genitive imp1ying [-definite]

and the non-Genitive option being arnbiguous, the data discussed in NeidIe

(1988) and supported by my own informant work indicate that the use of

Genitive permits the NP to be interpreted as either [+definite] or [-definite],

while the Accusative option is necessari1y interpreted as [+definite]. This is

ilIustrated in (70).

•(70) Genitive

Accusative-> [-definite] or [+definite]-> [+definitc]

•

Strong support for Neid1e's c1aim that the Genitive option is not

restricted to indefinite NPs comes from the acceptability of Genitive Case on

pronouns, demonstrative NPs, and proper narnes, aIl of which are incompatible

with any traditionaI interpretations of the concept indefinite.

(71) on ne vidaI etoj stranyhe-NOM Neg saw this-GEN country-GENHe did Mt see this country.

[NeidIe 1988:76)

(72) ja ne vidaI masi1 Neg saw Masa-GEN1didn't see Masa.

[Neid1e 1988:47)

185

• In the analysis 1 will propose here, 1 will assume that NPs are lexically

marked with a value for the feature [definite]. The actual name that we give this

feature is not crucial te the analysis; 1 will refer to it as [definite] for case of

exposition and because it appears to approximate the common property of the

NPs in question. In what follows, 1 will argue that the distinction between

definite and indefinite N?s drives the pattern of interpretation and Case­

assignment exhibited in negated sentences in Russian.

3.5.3. DEFINlTENESS AND THE ACCUSA11VE-GENrnVE ALTERNATION

Given that it will be the specification for definiteness carried by the NP

itself that will limit the Case-marking options of NPs in negated sentences,

rather than the Case-marking that limits interpretation, the paradigm can be

clarified by reversing the direction of entailment of Neidle's observations. This

results in the pattern given in (73).

(73) [+definite][-definite]

- > Accusative or Genitive-> Genitive

•

According to the analysis developeci up to this point, in an affirmative

[+tense] sentence Accusative Case is assig;led to a VSO direct object in

spec/VP, under government by the trace of the verb-chain in Delta. We have

alse maintained that the presence of the functional category NegP immediately

dominated by Delta introduces an additional Case-checking position. We have

maintained further that where Case-checking in a SpeclHead configuration is an

option, Case-licensing under government, a marked option, is not triggered.

Without further refinement, this predicts that the Genitive of Negation would be

obligatory on the direct object in a negated sentence. Indeed, this captures the

186

• situation historically in Russian, where any NP that met the structural

requirements for the Genitive of Negation was obligatorily Genitive. In

contemporary Russian, however, we see that Accusative Case on the direct

object is an option.

Given that the position that would otherwise be Case-licensed under

govemment in an affirmative sentence, the specifier of the functionaI projection

immediately dominated by Delta, is a licit SpeclHead configuration for Case­

licensing in a negated sentence (due to the [+Case] feature of the negative

head), the Accusative marking on NPs in negated sentences in Russian must

have a source other than govemment by the verb. 1 will argue that there is an

additionaI SpeclHead configuration where Accusative Case can be checked.

Such an additional Case position has been proposed in much of the recent

literature on Case-licensing of direct objects (Chomsky (1989), Mahajan (1990),

Johnson (1990), Noonan (1992».

1 will propose that this projection, which 1 will refer to as the functional

projection (PP), is generated below Neg; the position otherwise taken by VP.16

The D-structure position of FP in a negated [+tense] transitive sentence is given

in (74).l7

187

At S-structure, the verb raises through the head of FP, NegP, and

DeltaP, to T. There <>re two potential Case-positions for the direct object at LF:

speclFP and speclNegP. The properties of FP are considered in the following

section.

•

•

(74)

spee

NOM eheeked(S-structure)

spee

Neg

spee

:,

GEN eheekedILF)

ACC eheekcdILF) v NP

3.5.4. Tm: PROPERTlES OF FP

3.5.4.1. TBESPECIFlEKOFFP: LnwTEDTO [+DEFINITE] NPs

•

1 will maintain that Case-checking in spec/FP is restricted to [+definite]

NPS.IB As this is the only means for Accusative Case-checking in negated

sentences in Russian, the corresponding interpretations of these sentences are as

expected. Accusative NPs that are D-strueture objects of negated transitive verbs

must be [+definite]; if they were [-definite] they would not have the option of

188

•

•

•

being Case-checked in this position. Further, since no such restrictions in terms

of definiteness are attributed to speclNegP, the fact that a phrase in the Genitive

of Negation is not disarnbiguated with respect to defmiteness is also expected.

This proposai accounts quite naturally for the observed asymmetrj

between the internaI arguments of intransitive and transitive verbs with respect

to interpretation; the observation that only an Accusative marked internaI

argument in a negated sentence is unambiguously interpreted as [+definite]. In

Pesetsky's analysis, the fact that the non-Genitive option is interpreted

differently depending on whether it is marked Accusative or Nominative is

problematic.

This also provides an account for an additionaI asymmetry between D­

structure objects of transitive and intransitive verbs, noted by both Pesetsky

(1982:214-215, fn. 20) and Neidle (1988:77). D-structure objects of transitive

verbs, which may be Case-checked either at FP (Accusative) or NegP

(Genitive), appear to require the Genitive of Negation option when preceded by

the negative intensifying particle ni ('no', 'not any').

(75) ??ja ne poluca1 nikakie pis'maI-NDM Neg received no-ACC Ietters-ACC1didn 'c receive arry lerrers.

[Neidle 1988:77]

This can be contrasted with D-structure objects of intransitive verbs,

which allow Case-checking at NegP (Genitive) or TP (Nominative). These

arguments maintain optionality of Case-checking even when preceded by the

negative particle.

189

•

•

•

(76) ni odna devuska ne prislanot one girl-NOM Neg came-fem-sgNot one girl came.

[Neidle 1988:77]

This asymmetry is predicted by the anaIysis we have developed here.

The addition of the negative partic1e ni preceding the NP renders it [-definite].

As such, the option of Case-checking in the spec/FP option (Accusative) is ruled

out for these NPs, which forces Case-checking in spec/NegP (Genitive) instead.

The internaI arguments of intransitive verbs, on the other hand, can be Case­

licensed in specITP (Nominative) even if [-definite).

3.5.4.2. THE HEAD OF FP: ACCOUNTING FOR THE LACK OF FPWITH UNACCUSAllVE AND PASSIVE VERBS.

The fust fact to note about the FP option is that, unlike the Genitive of

Negation, it is unavailable for the arguments of passive and unaccusative verbs,

which according1y show only a Nominative-Genitive aIternation in negative

sentences. This is illustrated in (77)-(78).

(77) *ne suscestvuet takuju stranuNeg exists-3rd sg such-ACC country-ACCThere does not exist such a country.

(78) ne suscestvuet takoj stranyNeg exists-3sg such-GEN country-GENThere does not exist such a country.

[NeidIe 1988:76]

We have suggested that only an element with the feature [+Case) can

license an NP in its specifier. Above, we argued that the head of NegP is

[+Case), based in part on its ability 10 license an NP in its specifier even in

190

• passive sentences, and therefore cannot rely on inheriting this feature from the

verb. The Jack of Case-licensing in ~pec/FP in sentences with unaccusative or

passive verbs suggests that the head of FP is not specified for the feature [Case].

As such, it is entirely dependent on inheriting this feature from a lexical item

that raises into it. 19 In an affinnative sentence with an unaccusative or passive

verb, this forces the internai argument to raise to specfTP to be licensed with

Nominative Case.

3.5.5. A l'ROBLEM: OBUGATORY GENlTIVE OF NEGATION

Russian existential sentences take Nominative Case on their arguments.

Howevcr, negated e:tistentiaIs which use the form byt' ('be') do not show a

NominativelGenitive alternation. Instead, the sole argument of the negated

existentiaI verb is necessarily Genitive. The negative form of existential byt' is

net (present), ne bylo (past), and ne budet (future).

(79) knigi est'books-NOM (there) are1here are books.

(80) net knigNeg (there) are books-GEN1here are not books.

DisaIlowing the Nominative option in these sentences is mOre complex

than disaIlowing the Accusative option for unaccusative and passive verbs or for

NPs preceded by the negative particle ni. Below, we will explore means of

deriving the obligatoriness of the Genitive of Negation on these existential

sentences.

191

•

•

•

It is weIl documented across human languages that existential sentences

with expletive elements in the subject position do not permit a [+definite]

argument in object position.

(81) There ensued a/*lhe riO( on Moss. Ave.

This is Safir's (1982) 'definiteness effect', and Milsark's (1974)

'quantification restriction'. As noted by Reuland and ter Meulen (1989:2ff),

constructions of this type are found in Chinese, Chamorro, Dutch, Finnish,

French, German, Hungarian, Italian, and Spanish.

Given the posited relationship between Case-checking in spec/FP and the

feature [+definite], it is tempting to suggest that since the objects are

necessarily indefinite, these sentences require Genitive as opposed to

Accusative, which is restricted to definite NPs. However, two immediate

problems arise; the first is that it is not Accusative but Nominative that we have

to rule out, and the second is that any such argument would predict that ail

existential verbs in Russian should show this behaviour, which is not the case.

Other existential sentences (e.g. those using the verb form 10 exisl) differ from

bYl' in allowing the expected Nominative-Genitive altemation. This suggests

that sorne form of lexical specification is at the root of the observed constraint.20

One possibility is that the negated existential byl' is lexically specified as

a morphologically negative verb, which has the special property of checking

Genitive Case in its specifier. This alone will not produce the correct results, as

r-..ising 10 specfTP is still permitted. Accordingly, we must analyze these verbs

as inherent Genitive Case-assigners. The D-structure object of byl' cannot raise

10 specfTP and surface with Nominative, because it already bears Case.

192

• While this account amounts to an ad hoc solution, it is conceivable that a

basic existential use of the verb co be should become a lexicalized negative verb;

many of the world's languages have a negative form which has evolved from the

negated form of the verb co be. Moreover, in languages where negation is

usually an independent morpheme, it is common to find a restricted subset of

verbs subject to morphological negation, and :he most frequent is undoubtedly

be.21 Accordingly, these forms may carry special features including features

relating to Case-assignment.

INDEFINITE NPs AS VARJABLES, NEGATION AS AN OPERATOR•

(82) [nist] =Neg-be[laysa] =Neg-be[eino] =Neg-be

3.5.6.

PersianArabicModern Hebrew

[payne 1985:228]

3.5.6.1. INTRODUcnON

•

In the following section, 1 present an analysis which is driven by the data

presented by Pesetsky, wherein the Genitive of Negation unambiguously denotes

indefiniteness. This analysis is presented with certain reservations. On a

somewhat meta-theoretic nO\e, the fact that Genitive Case was historically

obligatory for alI NP D-structure objects in negated sentences, and Accusative

or Nominative Case is a more recent option, suggests that if either option should

have associated restrictions on its availability it is the Accusative or Nominative

option.

Further, 1 was unable to duplicate these facts among my own informants,

for whom the Genitive of Negation does not entai! an indefinite interpretation.

193

Nevertheless. 1 present the analysis of this paradigm as an alternativ.: to the

preceding analysis. As Russian evolves. the use of the Ge.1itive mav he

becoming increasingly marked (as noted by Timberlake (1975) and others):

perhaps what follows can serve to suggest an approach to the language of

younger speakers who use the Genitive of Negation in a more restricted way.

In what follows. it is assumed that Genitive Case on an NP in a negated

sentence results in an unambiguously [-detinite] rcading. As above. the basic

daim on which the analysis rests is that Genitive is Case-checked in spec/NegP.

The additional daim will be that only [-definite] NPs have this option. We will

derive this by an additional restriction on Case-licensing in spcc/NegP in this

dialect.

In her (1982) dissertation, Heim proposes that ail NPs, induding

indefinite NPs, are assigned a referenrial index. This index is dcscribed as a

numerical subscript, unrelated to traditional notions of reference (Heim

1982: 132). Indefinite NPs are treated as variables, elements that require a c·

commanding operator. An obligatory rule termed Quantifier indexing operatcs

to copy the referential index of every indefinite NP as a selection index onto the

lowest c-commanding operator. Selection indices are also numerical subscripts,

differing only in that they appear on Quantifiers rather than NPs.

Selection indices are assigned in (Wo ways. First, there is the Quantifier­

indexing operation. Second, when a Q is moved out of an NP it takes the

referential index of !hat NP with it as a selection index.

•

•

•

3.5.6.2. HEIM (1982): INDEFlNITE NPs AS VARIABLE.."

194

•

•

QUANTIFIER INDEXlNG: Copy the referential index of every indefinite NPas a selection index onto the lowest c-commanding quantifier.[Heim 1982: 146]

Quantifier indexing occurs at the level of the grammar where the

interpretive component operates. Crucially, only indefinite NPs copy their

indices onto an operator.

In what follows, 1 will suggest that quantifier indexing is a necessary

(though not sufficient) condition for LF Case-checking in spec/NegP in this

dialect. It is the requirements of this condition that rule out the Genitive of

Negation on definite NPs, and force an indefinite reading of NPs in the Genitive

of Negation.

Under the assumption that the negative operator adjoins to TP at LF, C­

commanding al! NPs in its binding domain (TF), the referential index of any

indefinite NP in a negated clause will be copied as a selection index onto the

negative operator. However, index-copying is not a sufficienr condition for the

Genitive of Negation. As we have seen in this chapter, not all indefinite NPs

can surface with Genitive of Negation. This fcllows from the fact that Case­

checking is subject to stricter structural constraints than co-indexation; an NP

bearing the Genitive of Negation must raise to speclNegP at LF.

To clarify the distinction between the index-copying mechanism and the

ability to be Case-checked in speclNegP, consider sentences where the Genitive

of Negation is ruIed out for other reasons. Thus, the object of a preposition

(Case-checked in speclPP) in a negated sentence can be interpreted as an

indefinite NP in the scope of negation, even though it cannot surface with the

Genitive of Negation.

195

•

•

Clearly, Quantifier indexing is not thc problematic factor for the

Genitive of Negation within PPs; it is the more restrictive constraints on Case­

checking that are not met. This leads us to the following conclusion. The

Genitive of Negation and co-indexation between negation and an indefinite NP

are not mutually entailing processes. While the Genitive of Negation requires

index-copying, index-copying does not guarantee the Genitive of Negation.

3.5.6.3. SAMPLEDERJVATION: A NEGATEDTRANSlTIVESENTENCE

Anned with these assumptions about the relationship between indefillite

NPs, Quantifier indexing, and Case-checking in spec/NegP, the restriction of

the Genitive of Negation to indefmite NPs are accounted for as follows.

Qoly an indefinite NP copies its index onto a c-commanding negative

operator. As in the preceding analysis, an AGENT NP, whether definite or

indefinite, cannot be Case-checked in specfNegP because it would have undergo

LF-Iowering to do 50. Instead, it raises to the specifier of TP and is Case­

checked there at S-sttucture, reflected in personlnumber agreement on the

tensed verb.

An indefinite D-sttucture object has two options for LF Case-checking:

specIFP or speclNegP. Given that specIFP is limited 10 [+defmite] NPs (the

unavailability of an indefinite reading for the Accusative option is al50 a feature

of the dialect described by Pesetsky), [-definite] NPs of transitive verbs have

ooly one option, which corresponds to Genitive Case; Case-checking in

speclNegP. If an NP is [+definite], no index-copying takes place, Case­

checking in speclNegP is not permitted and the NP must be Case-checked in

specIFP.

196

•

•

•

To re-phrase this in interpretive terms, since Genitive Case-assignment in

spec/NegP is only possible if index-eopying between the operator Neg and an

NP has taken place, and since this co-indexation is restricted to [-definite] NPs,

a Genitive object is unambiguously interpreted as indefinite, as desired.

Likewise, since spec/FP is restricted to [+definite] NPs, an Accusative NP is

unambiguously interpreted as definite.

3.5.6.4. SAMPLEDERIVATION: A NEGATED!NIRANSITIVESENTENCE

The derivation in passive and unaccusative sentences proceeds exactly as

in transitive sentences, except that the option of Case-checIàng in speclFP is

ruled out (the verbs are not specified [+Case)), and an additional option, S­

structure Case-checIàng in specITP, is available. An indefinite D-structure

object copies its referential index onto the c-eommanding negative operator.

This permits Genitive Case-checIàng in spec/NegP. If the D-structure object is

[+definite], index copying does not take place and the NP must raise to specITP

at S-structure, where it triggers agreement on the tensed verb. An agenrive

argument of an intransitive verb, whether definite or indefinite, cannot lower to

spec/NegP. Accordingly, it raises to speclTP, where Nominative Case-checIàng

takes place.

Given that specITP is not restricted to [+definite] NPs, this alIows two

Case positions for an indefinite D-structure object of an unaccusative or passive

v~; Nominative Case-checIàng in specITP at S-structure and Genitive Case

checIàng in spec/NegP at LF. This prediets that the Nominative option alIows

both the definite and indefinite interpretations, while the Genitive option is

unambiguously interpreted as indefinite for arguments of both transitive and

197

•

•

•

intransitive verbs, which is consistent with Pesets1:y's data. Further. we accounl

for the unambiguously definile reading associaled with the Accusative option.

These facts are summarized in (84)-(85).

(84) UNACCUSATIVE AND PASSIVE VERBS

GEN -> [-definite]NOM -> [+definite] or [-definile]

(85) TRANSITIVE VERBS

GEN -> [-definite]ACC - > [+definite]

3.6. S~y

This analysis, which posits indefiniteness as a necessary initial condition

for the Genitive of Negation, is plausible and perhaps captures best the language

of younger speakers for whom the Genitive of Negation is a more marked

option. A wealmess in the analysis is that the main supporting c1aim, that Case­

checking in speclNegP at LF requires Quantifier indexing, is otherwise

unmotivated. This can potentially be related to an additional feature-matching

restriction on case-checking in NegP; the co-indexation requirement for Case­

checking in speclNegP is met by Quantifier indexing.

3.7. ENDNOTESTOCHAP'rER3

l1bis is an option that was nol available historically in Russiao. when the Genitive ofnegation surfaced whenever struetural conditions allowed (Neidle 1988:30).

2This is required ID account for the Accusative Case that surfaces OD expressions ofduration of lime. a fact that is troublesome for Case theory in general bUI appears in otberlanguages as we11. inc1uding Latin.

198

•

•

3Pesetsky (1982: fn. 57). Note that these expressions raise severa! additiolllÙproblems for Case theory in genera!, including the fact that they surface with ACC Casefollowing passive verbs and verbs that have been termed unaccusative. Note further that theability of these expression la bear the Genitive of Negation is dependent on the presence of thenegative intensifying particle ni. In !his way they are distinguished from the internai argumentsof the verb, whicb do not require the particle.

4We do not rule out a related explanation. Anticipating the aIllÙysis of Russian !hat isproposed in !his cbapter somewhat, the restriction la D-structure objects for distributive po­phrases may also he related la a funCtiOIllÙ projection whicb is generated helow the base-positionof agentive NPs. Regarding agreement and no-agreement numera1 phrases, these might reflectwhether the phrase is Case-Iicensed in the funCtiOIllÙ projection al LF or in TP al S-structure(reflected by agreement on the verb). Sa essentia1ly what these phrases have in common is anadditional Case-licensing possibility whicb is restricted la D-structure objects, rather !han aprinciple whicb requires LF-raising and therefore restricts their distribution la properly governedpositions.

Snis con he seen as a Re!ativized UTAH (1.arson 1990:601): Idenrica1 r1lemalic,..,/alionships are represenred by idenrica! ,..,/alive hierarchical ,..,/alionships berween items al D­strucrUrt!. This relaxation of the hypothesis is required for the distinct base positions of D­structure objects in CW, whicb neverthe!ess maintain the same hierarchica1 relationship withAgent NPs•• Sec Larsen (1990) for further discussion of!his proposai.

6Thjs is not necessarily universal. However, sec Chomsky (1992) for a proposai !hatit is the presence of agreement !hat forces S-structure Case-checking; otherwise, NPs raise al LFfor Case-checking, on the understanding !hat raising al !his leve! is "cbeaper".

7Note !hat, for the structure in (29) la he consistent with the morphologica1specification given in (28), the requiremenl must he salisfied by linear adjacency, salisfied by T,rather !han by the calegory label of its sister, which is Neg.

8BW2Ïo 1986:178ff

9The conclusion !hat the Genitive of Negation is a structural Case is also arrived al byLasnik (1990).

IONole that the ward order within PPS entails that the preposilional complement doesnot raise lO speclPP until LF. This is consistent with the proposai that only NOM is Case­checked al s-structure (hencc triggers agreement) and ail other Cases are checked al LF,including GEN Case in speclNegP.

11The concept of a class of Russian verbs which is distinguished in this way is notnow lO the literature. Neidle (1988) refers lO a distinction between open and closedcomplements, the former of which aIlow the Genilive of Negation and negative polarity item(NP!) Iicensing by a higber negative operalOr. Likewise, Progovaç (1991 and eisewhcre) arguesfor the transparency of certain clauses in Serbo-Croatian (subjunctive clauses) hased on NP!licensing by a negative operator in a higber clause. The proposai made here that thesecomplements are VPs is one way of implementing this idea stIuctura1ly•

199

•

•

12Note that lbe examples which permit lbe Genitive of Negation an: "reSlnlcturing"­1ike environmenls (Rochette 1988). According1y, lbey an: amenable to an ultimatelymonoclausal structure analysis even if base-generated as bi-elausal.

13Chvany (1975:156), Babby (1980:105fl), Neidle (1988:53fl).

14Neidle (1988:86 fn. 14) mentions Peskovskij (1956:366-367), Peselsky (1981:8),Babby (1980:14fl), Karcevskij (1927:125-126), and Lobanova (1975:202-203).

15Note that even lbough lbe NP does not raise to specINegP unli1 LF, il must hebelow spec!NegP al S-strueture if it is to raise to specINegP al LF.

16A possibility which 1 will nol explore here is that FP con he related to a DP­analysis of argumenls as in Abney (1987). We cou1d maintain !hal [+definite] NPs generate afunetional projection FP (indefinites are haro NPs). Accordingly, Ibis FP would he optionally,getierated in any [+definite] argument position. This would have lbe positive resu\l of relalingthe [definiteness] restriction to olber proposals about lbe internai structure of definite andindefinite NPs. However, it would he difficult to avoid overgenerating Ibis projection. Giventhat it is only generatcd in negative sentences (a fact that is more evident in Colloquial Wdsh,where lbe head of FP is not null), lbe proposai that it is optionally sdected by NegP allows US to

constrain ils occurrence. However, if lbe prohlem ofovergeneration is dealt wilb by appealing 10

some notion of economy, as in Chon:.s1cy (1989) (FP being a language-specifie feature of lbegrammar), lben Ibis alternative opti':ln is viable. The analysis itself would not change, except!hat lbe relative order of FP and VP would he reversed, and F would have to raise to V to makespecNP a licit SpeclHead configuration in tensed sentences. However, given !hat we coneludebelow!hat lbe head ofFP is not specifiee! [+Case], according to U (1990) lbe combination of Fand V in the syntax will only he [+Case] if V raises into P, not the reverse. U proposes !hat agiven lexical item con be specifiee! as [+Case] or [-Case] or he unspecifiee! for the feature. Ifunspecifiee!, it con inherit lbe feature [+Case] from a head that moves into it, via fCllturepercolation à la Di Sciullo and Williams (1987). Thus, while the causative affix in somecompounds may be unspecifiee! for lbe feature [Case] it con become [+Case] if the head of lbecompound is [+Case]. See U (1990: 409-10 and m.9) for discussion.

17A1lbough Ibis structure posils a radically different position for NegP !han !hatproposee! by Pollock (198!!), it could be arguee! !hat it is in the spirit of that analysis, where hisproposee! AgrP (like our FP) is generatcd below NegP if NegP is present. The differences derivefrom our claim regarding the base position of NegP, not the relative hierarehy between NegPand a Case-position for the direct objecL

18Attributing SOUle restriction to Ibis projection is not wilbout precedent in lbeliterature. The extra Case position for direct objects proposee! by Mahajan (1990) for Hindi isidentifiee! wilb specifie NPs, and Johnson's (1990) and Noonan's (1992) projections areassociatee! wilb pronomina1 NP&.

19RecalI that the specifier of the VP dominating a velb-trace is not a Case-licensingposition in a [+tense] sentence, even if lbe verl> is specifiee! [+Case], due to the morphologicalcompleteness requirement of the definition of licit SpeclHead configuration. This forcesAeeusative Case licensing under government by the member of the verl> chain in Della inaffirmative sentences. However, lbe verl>'s [+Case] feature is inheritcd by the head of FP viahead-to-head raising, enabling the head ofFP to Case-license an NP in ils specifier.

200

•

•

2llNeidle (1988) aIso concludes that Gcnitive Case assignmcot must he a lexicalproper:y of the negatcd existential 11)1'.

210001 vcrbs that arc commonly morphologically negative arc typical modals, i.e.won 'f, can 'f, musfn 'f, wouldn 'f, etc., in English. $ce Payne 1985 for discussion.

201

•

•

4. CHAPTER 4: CASE-LlCENSING, 'EMBRACING' NEGATION, ANDDEFINITENESS IN COLLOQUIAL WELSH

4.1. INTRODUcnON

In Chapters 2 and 3 of this thesis, analyses were proposed 10 account for

the syntactic role of sentential negation in Case-licensing in Literary Welsh

(LW) and in Russian. It was suggested that while the head of NegP creates a

barrier for verb-raising in LW, it does not constitute a barrier in Russian. This

was related to the claim that Neg is a morphologically dependent lexical item in

Russian, but an autonomous lexical item in LW. This chapter, which deals with

data from Colloquial Welsh (CW), provides a kind of bridge between these two

languages. l

NegP in CW, as in LW, will be analyzed as generated immediately

above TP. It will be maintained that the two negative markers found in CW

(which can be loosely referred to as the pre-verbal and clause-medial negative

markers) are generated in the head of NegP and adjoined to VP, respectively.

The head of NegP in CW will be analyzed as a morphologically dependent

lexical item, as in Russian, which cannot surface independently of another

lexical head.

The relationship between NegP and TP proposed by Zanuttini (1991)

will be addressed, and 1 will argue that NegP is dependent on TP in CW. The

differences between the agreement patterns in negated relative clauses in LW

and CW will be derived from the morphologically dependent nature of the head

of NegP in CW.

202

•

•

1 will then propose an account for the fact that the negated counterparts

of non-periphrastic transitive ser.tences in CW require a lexical item 10

immediately precede the direct object; a requirement not found in LW. This will

be driven by the presence of the secondary negation marker intervening between

Delta ;,.nd the base-position of the direct object.

The fact that direct objects in negated periphrastic constructions do not

require this lexical item will provide support for the claim made in Chapter 2

that they do not reIy on the [+finite] V for Case; rather, they are Case-Iicensed

in the specifier of the phrase headed by the [-finite] verb.

1 will then extend the account of definiteness effects in negated sentences

in Russian 10 account for the fact that in at least one documented dialect of CW

(pembrokeshire Welsh), this lexical item is not found when the direct object is

[-definite], and is optional when the direct object is [+definite].2 With respect to

the Genitive of Negation construction in Russian, it was proposed that the

Specifier of NegP provides a potential Case-position for NPs; 1 will suggest that

this analysis can be extended to account for the paradigm in CW negated

clauses. The final section contains a paradigm involving variable word order

facts in negated sentences in Pembrokeshire Welsh.

Colloquial Welsh is sufficiently different from Literary Welsh in the

areas addressed in this thesis 10 warrant the brief description that follows.

4.1.1. WORDORDERlNCOUOQUIAL WELSB

The unmarked sentence type in CW is periphrastic. Periphrastic

constructions are made up of an inflected auxiliary (i.e. bod ('be'», an aspect

marker (progressive yn or perfective wedi) and an untensed verb (VN). Thus,

203

•

•

the typical way of conveying the senœnce John sings is not with VSO order, as

in (1) below, but rather with AUX-S-V-O order, as in (2).

(1) Canith Sion.sing-3sg SionSion sings.

(2) Mae Sion yn canu.is-3sg Sion PROG singSion sings.

The preponderance of the periphrastic construction in CW is noted by

Jones & Thomas (1977), and is supported by my work with native speakers. In

fact, certain verbs are restricted such that they may appear only in periphrastic

constructions.3

The clements within the VP in periphrastic constructions appear in the

fixcd order given in (3), and no other lexical items may intervene. Following

Harlow (1981), Sproat (1985) and others, we will assume that, as in LW,

surface VSO order results from the raising of the leftmost verbal element te

Tense (T).

(3)

PERFECTIVE[boc! wedi]

PROGRESSIVE VERB[boc! yn]

•An example of a periphrastic sentence which incorporates ail of these

clements is given in (4).

204

• (4) Mi dylai John [fod wedi bod yn edrych ar y teledu].PT should-3sg John be PERF be PROG watch at the televisionJohn sJwuId Jwve been looking al the television.4

[Jones & Thomas 1977:25]

V-raising characterizes regular VSO sentences as weil as periphrastic

sentences. The structure given in (5) illustrates the S-structure that will be

assumed for periphrastic sentences in CW, whicl. features an Aspect Phrase

(AspP) selected by V ar.d headed by the aspect marker (progressive yn or

perfective wedl).5

(5) TP

(6) Mae Sion yn canu.is John PROG singJohn is singing.

205

• 4.2. SEl'.'l'ErIo'TIAL NEGATION IN COLLOQUlAL WELSH

4.2.1. TwO NEGATIVE MARKERS

•

Unlike LW, in which the pre-verbal negative panicle ni(d) is sufficient

to negate a clause, negation in CW requires an additiomù, sentence medial

marker, dim.6 The pre-ve,bal negative marker found in LW is generally

dropped, (in fuct almost all of the sentential-initial markers discussed in Chapter

2 are dropped in CW), but should not be considered optional. While

considerably weakened, its presence can be established by its effect on the

following word. When preceding a vowel-initial verb, the pre-verbal negation

marker appears in reduced form as a prefix (d-). When preceding a consonant­

initial verb, it tends to he dropped altogether, but its presence can be noted in

the form of obligatory consonant mutation on the verb.7 As illustrated in (7)­

(10), the verb form palodd obligatorily appears as phalodd in a negated sentence

in LW, under the influence of the preceding negative panicle, ni(d). This initial

consonant mutation is obligatory in negated sentences in CW, even though the

negative marker itself is no longer overt.

(7) UCW:Palodd John yr ardd.dug-3sg John the gardenJohn dug the gartien.

(8) LW:Ni phalodd (*pa1odd) John yr ardd.Neg dug-3sg John the gardenJohn did not dig the garden.

206

•

•

(9) CW:Phalodd (*Palodd) John ddim 0 'r ardd.8

Neg-dug-3sg John Neg of the gardenJohn didn 't dig the garden.

(10) CW*Palodd John ddim o'r ardd.dug-3sg John Neg the gardenJohn didn't dig the garden.

At S-structure, the secondary negative marker dim follows the tensed

verb and the subject, but precedes the untensed verb and ail other elements in

the VP, including the aspectual markers bod wedi and bod yn. The example

given in (12) illustrates the position of dim with respect to the elements in VP.

Note the influence of the pre-sentential negative mark-er on the form of the

inflected auxiliary.9

(11) Mae Sion wedi bod yn canu.is John PERF be PROG singJohn /ws been singing.

(12) 'Dyw Sion ddim we<ii bod yn canu.Neg-is John Neg PERF be PROG singJohn /ws not been singing.

4.2.2. SENTEN11AL NEGATION VS. CONS'll1UENT NEGATION

We maintain here that ooly sentential negation generates a NegP. Given

that the secondary negation marker in CW is homophonous with the constituent

negator, it is important te be able to distinguish these occurrences. First, while

the secondary sentential negation marker is dependent on the presence of the

207

•

•

pre-verbal negation marker (as evidenced by consonant mutation on the tensed

verb), the constituent negator can appear in affirmative clauses, adjoined to any

constituent. Further, under certain conditions, the initial consonant of the

clause-medial negation marker is subject to the phenomenon of soft mutation,

widespread in the Celtic languages. The effect of soft mutation is represented

orthographicaIIy as ddim. This process distinguishes the sentential negation

marker from the constituent negator dim, which never undergoes soft mutation.

They are further distinguished by the effect of multiple negation markers

in a sentence. Whereas the sentential negative marker dim combines with the

pre-verbal negative marker to CMry a single negative force, constituent negation

adds a second negative quantity which, in combination with sentential negation,

results in canceling out the negative force. A sentence including both the

sentential and constituent negators is given in (13).10 The initial dim is adjoined

to the fronted constituent John.

(13) [NidIDim John] oedd ddim yn helpu.Neg John Neg-was-3sg Neg PROG helpIt wasn't John that wasn't helping.

[Jones & Thomas 19TI:325]

Distinguishing sentential from constituent negation in CW is therefore

relatively straightforward.

4.2.3. PREPosmON INSERTION IN NEGATED SENTENCES

•In CW, we find that the negative counterparts of simple transitive

sentences require that a particie be inserted before the direct object11 This is

illustrated in (14), as compared with the affirmative counterpart in CW, given in

(15).

208

•

•

•

(14) Phalodd John dc!im *(0) 'r ardd.Neg-clug-3sg John Neg PT the gardenJohn didn't dig the garden.

(15) Palodd John Or ardd.dug-3sg John the gardenJohn dug the garden.

Compare this with the LW forms of the same sentences, given below.

While the affirmative sentences are identical, sentences (14) and (16) differ with

respect to both the presence of the secondary negation marker and the presence

of a lexical item preceding the direct object.

(16) Ni phalodd John yr ardd.Neg dug-3sg John the gardenJohn didn't dig the garden.

(17) Palodd John yr ardd.dug-3sg John the gardenJohn dug the gartfen.

This particle is found preceding direct objects in VSO structures; in other

words, direct objects of [+tense] verbs. In periphrastic sente:lces, however,

negaton does not appear 10 affect the realization of the direct object. This is

illustrated in (18)-(19).

(18) Oedd John yn palu 'r ardd.was-3sg John PROG dig the gardenJohn was digging the gartien.

209

•

•

(19) 'Doedd John ddim yn palu 'r ardd.Neg-was John Neg PROG dig the gardenJohn wasn't digging the garden.

Accordingly, the occurrence of this particle in negated sentences

disûnguishes direct objects in periphrastic constructions from direct objects in

simple VSO constructions. Recall that these argument:; are distinguished with

respect to Case in LW; direct objects in simple constructions are exceptionally

Case-licensed under govemment, but periphrastic direct objects are

Case-checked in a Spec/Head configuration.

4.2.4. THE STATUS OF THE PRE- Al'lD PoST-VERBAL NEGATIVEMARKERs

These fucts raise severa! questions regarding the syntactic status of the

negative markers in CW. The first question hinges on their co-occurrence.

Given that the negative marker !hat surfaces on the tensed verb in CW cannot

occur without a c1ause-medial negative marker, it is plausible that their co­

dependence derives from being generated as obligatary constituents of the same

maximal projection, a possibility explored in the Appendix to this dissertation.

However, we will maintain that the secondary negation marker is base-generated

adjoined ta a maximal verbal projection be10w NegP (VP or AspP).12

As in LW, 1 will assume !hat NegP is generated above TP. In LW, it

was argued that the head of NegP creates a barrier for V-to-C raising. In CW,

the head of NegP is not an independent lexical item; it will he analyzed as

affixal, and we will maintain that V-movement into Neg constitutes an

obligatory substitution inta an empty sIot. The secondary negative e1ement, dim,

does not undergo raising, and its linear order relative ta other components of the

210

• sentence reflects its base position in the structure. The S-structure of the CW

sentence Dyw Sion ddim yn CarlU ('John is not singing') is given in (20).

(20) NegP

A,

;,~_._,_._.- ..-----

'dyw Siont

yn

v

canu

It has been proposed (Z2nuttini 1990, 1991) that NegP may have an

additional property: that of requiring TP in order for it to be generated. In the

following section, 1 argue that this requirement holds in CW as well.13

4.3. SENTENTIAL NEGATION AND '1'ENSE IN COU,OQUIAL WELSH

In her work on negation in Romance languages, Z2nuttini (1991) claims

that sentential negation may involve two distinct maximal projections: one

(NegP1) that cannot occur independently of T, and another (NegP2) that bas no

such requirement. In a sentence without TP, NegP1 will not be generated,

rendering this kind of negation unavaiIable. She maintains further that

211

parametric variation allows NegP to be generated immediately above TP or VP,

and that a sentential negation marker can be either the head or the specifier of

NegP. A language may instantiate NegPl, NegP2, or both.

According to Zanuttini, given that NegPl requires TP, NegPl-type

sentential negation should not be available for negating past participles, absolute

constructions, or lrue (as opposed to suppletive) imperatives. In Chapter 2 the

fuct that NegP is not used ta negate infinitivals was mentioned in the context of

determining the base position of NegP. Since this analysis assumes that [-finite)

clauses do not generate TP in Welsh, this fuct suggests that Welsh may

instantiate Zanuttini's NegPl. We discuss these predictions beIow, and show

that in WeIsh, too, the generation of NegP is dependent on the generation of

•

•TP.

4.3.1. INFINrnvALS

•

We adopt Borsley's (1984, 1987) proposaI that complements of control

verbs in WeIsh are VPs, not TPS.14 Accordingly, NegP should not be able to

negate these constituents. Consistent with the dependency observed, NegP does

not appear in [-finite) constructions in CW. Instead, these constituents are

negated by using the negative verb peidio (a) ('to cease').15 This is illustrated in

(21)-(23). (Jones & Thomas 1977:333)

(21) Mae Mair wedi trio gweId John.is Mary PERF try see JohnMary Ms tried 10 see John.

(22) Mae Mair wedi trio peidio â gweId John.is Mary PERF try stop with see JohnMary Ms lried MIlO see John.

212

•

•

•

(23) 'Dydy Mair ddim wedi trio peidio â gweld John.Neg-is Mary Neg PERF try stop with see JohnMary hasn 'rrried Mr to see John.

This can be compared with the use of the negative verb to negate [-finite]

verbs in modal constructions. (Jones & Thomas 1977:335)

(24) Ddylet ti ddim aros.Neg-should you Neg stayYou shouIdn 'r stay.

(25) Ddylet ti ddim peidio ag aros.Neg-should you Neg stop with stayYou shouldn't Mt stay.

(26) Mi ddylet ti beidio ag aras.PT should-you stop with stayYou shouId Mr sray.

4.3.2. PAST PARllCIPLES

Consider the sentences given be1ow. Sentence (27) has a negated

auxiliary, and (28) a negated past participle, and these result in distinct scopal

relations between the negative operator and the adverb.

(27) Mary has not aIways paid taxes.[Ir is Mt the case thor Mary has aIways paid taxes]

(28) Mary has aIways not paid taxes.[It is the case thor Mary has aIways Mt paid taxes]

213

• zanuttini claims that these options are available in English because it

instantiates both NegPI and NegP2. In (28), NegP2 is available even though

NegPI is blocked.16 A language that has only NegPI will not be able to express

these two propositions by negating either the auxiliary or the past participle.

Thus, in Italian, zanuttini demonstrates that sentential negation can be invoked

to express only the first proposition. This is iIIustrated in (29)-(30).

(29) Maria non ha sempre pagato le tasse.Maria Neg AUX a1ways paid-past part. DET taxesMary hasn't always paid taxes.

(30) *Maria ha sempre non pagato le tasse.Maria AUX a1ways Neg pay-past part. DET taxesMary has always not paid taxes.

If sentential negation in Welsh is of type NegPI, dependent on the

generation of TP, we would expect it to pattern with Italian with respect to these

constructions. Significantly, what we find is that in the interpretation of

sentences like (31) the negative operator necessarily has scope over the

adverbial.

(31) 'Dyw Mair ddim wedi talu trethi yn rheloaidd.Neg-is M. Neg PERF pay taxes in ruleMary hasn't always paid taxes (i.e. has Mt paid taxes regularly)

To express the proposition that Mary has a1ways Mt paid taxes, CW

rather naturally uses erioed ('never') in place ofdim.

214

•

•

(32) 'Dyw Mair erioed wedi talu trethi.Neg-is M. never after pay taxesMary 1uJs never paid laxes.

This is not te say that Welsh is incapable of capturing different scopes of

negation. To negate the lower [-finite] VP in a perlphrastic sentence, the verb

peidio is used. Thus, the required contrast is captured by using this negative

verb. This is illustrated in (33)-(34).

(33) Wneith Sion ddim arcs.Neg-will John Neg stayJohn won '1 slay (= unwilling 10 Slay).

(34) Gwneith Sion peidio aros.will John cease stayJohn will nol slay (= willing not to slay)

What is critical to this investigation is the observation that NegP cannot

be invoked to negate a constituent that is [-tense] in Welsh.

4.3.3. ABsOLUΠCONSTRucnONS

zanuttini adopts Belletti's (1989) and Kayne's (1989) proposai that

absolutives are AgrPs, not TPs. Accordingly, she predicts that NegPl will be

unavailable with these constructions. Again, sentential negation in CW shows

the predicted constraints. Instead of NegP, a negative fonn of the perfective

marker is used. 17 The data are given in (35)-(36).

(35) Wedi arcl cyrraedd yn hwyr, ynddi heurodd Mary.PERF on arrive late, 3fsg apologîzed MaryHaving arrived lare, Mary apologized.

21S

• (36) Heb iddi cyrraedd yn hwyr, ...without to-her arrive lateNor having arrived lare... (lit. Wirhour arriving lare)

4.3.4. IMPERATIVES

•

•

zanuttini's discussion of imperatives invokes a critical distinction

between rrue and suppletive forms, only the latter of which generate a TP.

Verbal inflection distinguishes the two forms. While true imperative forms lack

tense features, suppletive forms use an inflected form borrowed from another

paradigm, e.g. present indicative. She proposes that NegPI should occur only

with suppletive forms of the imperative.

This hypothesis finds strong support in Italian, which has both forms of

the imperative. The pre-verbal marker non, associated with NegP l, is found

only with the suppletive forms, used for 2pl and 2sg.18

The imperative in Welsh uses the 2sg and 2pl forms of the present

indicative paradigm, thus satisfying the definition of a suppletive imperative,

and, according to zanuttini, should permit NegPl. This is consistent with what

we find in LW, where the imperative pre-sentential particle results from the

merger of NegP and IllP.

(37) Cenwch!sing-plSing!

(38) Na chenwch!Neg sing-plDon'( sing!

216

•

•

•

CW, however, displays a strong preference ta negate an imperative not

with NegP. but rather with the negative vero peidio, aIso found in the negation

of infmitivais and past-participles. t9 This is illustrated in (39).

(39) Paid a ehenweh!Cease with singDon't sing! (Stop singing!)

Given that the imperative in CW, as in LW, is a suppletive form, the

constraint is unexpected. There are severa! possible explanations for this. Ether

imperatives are subject ta a more genera! constraint, in addition to the TP

requirement, or imperatives do not generate TPs in CW, even though they use a

suppletive form, or sententiai negation in Weish is not in faet a NegPI, and

hence does not require TP.

Basee! on the consisteney of the other tests, 1 will adopt the first

explanation. SpecificaIly, it seems plausible that a special form of negation is

required for imperatives, but this is not driven by the presence or absence of

TP. Note that sueh a position is compatible with Zanuttini's anaiysis. According

to her proposai, it is not the case that suppletive imperative forms (as found in

CW) are required ta be negated by NegPI; rather, true imperatives are required

not be negated by NegPl. Accordingly, these facts do not constitute

counterexarnples; rather, they indicate the operation of an additionai restriction.

Note, in this regard, that while the use of the negative verb ta express

negated imperatives is strongly preferred, sentences like those in (40)-(41) are

alternative means of expressing prolu"bition with regular sententiaI negation.

217

• (40) 'Dach chi ddim i ganu.Neg-be you Neg PROG singYou are not ta sing.

(41) 'Dach chi ddim i fod i ganu.Neg-be you Neg PROG be PROG singYou are not ta be singing.

Thus, NegP is available for sentences expressing prohibition as long as

the imperative is not used.

l will briefly remark here that the reIationship between a unique form of

sentential negation and imperatives has long been noted. Hom (1989) mentions

that, in a survey of syntactic realization types for speech act distinctions by

Sadock and Zwicky (1985:175ff), approximately half of the languages surveyed

displayed a negative imperative sentence type, the prohibitive, which differed

significantly from other negative and/or imperative types. "In three-quarters of

the languages surveyed there are no straightforward negative imperatives, the

functional gap being filled by special negative markers, non-imperative verb

forms, or both". (Hom 1989:447). It is likely that there is a more general

constraint operating here, which is alse instantiated in Welsh.20

We will leave this as an unexplained but not contradictory facto As a

whole, the restrictions on sentential negation of the type ni(d)•••dim in CW

support the hypothesis tbat NegP patterns with Zanuttini's NegP1, and cannot

be generated in a sentence that does not instantiate TP.

4.3.5. CONCLUSION

•In this section, we have proposed tbat NegP in CW bas the following

properties.

218

• (i) NegP is generated above TP, and is only generated in sentencesthat include TP;

(ii) NegP is headed by a reduced (morphologically dependent) fonnof the LW negation marker (like Russian, but unlike LW, wherethe head of NegP is an independent lexical item);

(iii) The c1ause-medial negation marker dim is generated adjoined to averbal projection beicw DeltaP.

Be1ow, we will show how these characteristics attributed to NegP in CW

provide an explanation for the interaction between sentential negation and Case­

Iicensing manifested in agreement marlàng, and account for sorne strilàng

paraIIels between CW and the LW and Russian paradigms discussed in Chapters

2 and 3.

4.4. SENTEN'I1AL NEGATION AND CASE lN CW

4.4.1. NEGA'IED RELAllVE CLAUSES lN CW

ln Chapter 2, we developed an analysis of relativization out of negated

clauses in Literary Welsh driven in large measure by the blocking effect of the

head of NegP on XO-raising. RecaI1 that in LW, a negated relative clause is

introduced by the element na, ana1yzed as the reflex of Neg raising 10 C. When

the argument base-generated in speclDeltaP is relativized out of a negated

relative clause, subject agreement must surface on the verb and a pronominal

NP in specIDe1taP is not permitted. If a VSO direct object is relativized, a

resumptive pronoun stIategy is required, and the in situ pronomina1 may be

dropped ifagreement is triggered on Neg. This is illusttated in (42)-(44).

219

•

•

•

(42) Ydynioni na welsant (*welodd) [e]i Mairthe men Neg+C saw-3pl (*saw-3sg) [el Maryrhe men rhar didn'r see Mary

(43) Ydyn na welodd Mair *(eC)the man Neg+C saw-3sg Mary himthe man rhar Mary didn 'r see

(44) Ydyn nas welodd Mair (eC)the man [[Neg+Agr]C] saw Mary (him)rhe man rhar Mary didn'r see (him)

This was derived by the blocking effect of Neg on [V+ T]-raising 10 C.

Thus, while wh-movement to spec/CP is permitted, agreement with the

A'-chain headed by the subjecl operalor is established at T. There is no licit way

for the direct object to raise to spec/CP (which would entail A-movement over

the subject ta spec/NegP, then A'-movement to spec/CP). This forces a

resumptive strategy for relativization of VSO direct objects in negated clauses.

We have taken the position that V-movement into Neg is obligatory in

CW. This is driven by the morphologically reduced status of Neg (unlike in

LW, it never occurs as an independent lexical item). This line of argument leads

us to assume !hat there is no barrier interfering with V-to-C raising in CW, and

we predict that subject relativization in negated clauses shoul':! pattern with

subject relativization in affirmative clauses. Agreement should be established al

CP.

The agreement patterns found in negated relative clauses in CW

substantiate this prediction. Unlike in LW, in CW there is no subject agreement

220

•

•

•

on the tensed verb when a subject is relaùvized out of a negated clause. This is

iIlustrated in (45).

(45) Ybechgyn welodd mo'r eneth21

the boys Neg-saw-3sg Neg-P-the girllhe boys who didn '1 see lhe girl

[rallerman 1990b:298, fn. 5]

The pattern of agreement under subject re1ativization in negated clauses

in CW is identical to the pattern found in afftrmative relative clauses. While

there are!Wo pctenùal Case conftgurations below CP; TP and NegP, neither of

these conftgurations saùsftes the deftnition of a licit SpeclHead Case

conftguration, because the head and the specifter of these maximal projecùons

are traces. Accordingly, the lowest licit Case conftguration for the subject is at

CP, just as in afftrmative relative clauses in CW and in LW.22 This is illustrated

in the structure given in (46).

221

•

•

(461

NP-;

y bechgyn welodd

t

This provides a straightforward account for the lack of subject agreement

on the tensed verb; such personlnumber inflection is the reflex of Case-checking

at TP.

Note, however, that the pre-sentential particle analyzed in Chapter 2 as

an agreeing complementizer, a, does not surface. Thus, although the [Neg+V)

complex raises to C, providing a licit Case-configuration for an A'-chain, the

effect of Case-checking in speclCP is not overt.

As stated earlier, the pre-sentential particles do not surface in CW. Just

as the presence of negation is reflected in the mutation effects on the first

consonant of the tensed verb (as weIl as the presence of dim), the underlying

222

• presence of the head of CP can generaIly be established by its mutation effects

on the initial consonant of the tensed verb. The agreeing and non-agreeing forms

of the complementizer are distinguishable by their mutation effects: a triggers

soft mutation, and y(r) does not trigger any mutation.23 However, the mutation

effect on the tensed verb in negated clauses is controlled by negation, not by the

complementizer. This makes it impossible te determine by mutation effects

whether the form of the complementizer is the agreeing form a or the non­

agreeing form y(r). As a result, Case-checking at CP in these structures is

reflected only in the lack of agreement with the chain elsewhere in the clause.

4.4.1.1. VSO DIREeT OBJEeT RELATIVIZA110N

•

•

When we tum to relativization of VSO direct objects, we find that an

asymmetry exists in CW that is not found in LW. Thus, while subjects in CW

pattern with what has been termed the 'direct' pattern of agreement, even in

negated relative clauses, VSO direct objects continue to exhibit a resumptive

strategy. This asymmetry is consistent with the account given so far. In Chapter

2 it was argued that the in situ pronominal found in these constructions in LW is

due to the fact that, although the direct object can create an A-ehain by

indexation with a null pleonastic in specJNegP, it cannot mise te specfCP, as

this would entail A-movemelll over the subject in speclDeltaP. A-movement

must proceed through eaeh interVening specifier. The lowest A-position for the

VSO direct object is in specJNegP, whieh is higher than the A-position for the

subject. This forces a resumptive pronoun strategy.

In CW the situation for the direct object bas not ehanged, despite the faet

that NegP does not introduce a barrier for V-mising. We find that, in CW too, a

resumptive strategy is required for VSO direct objects.

223

•

•

(47) Ybechgyn welodd yr eneth monyn nhwthe boys Neg-saw-3sg the girl Neg-P-3pl themthe boys that the girl didn 't see.

[Tallerman 1990b:298]

In this way, the proposai that the head of NegP in CW does not

constitute a barrier for verb-raising to C provides an explanation for a

subject-object asymmetry in negated clauses in CW that is not found in LW.

While the chain created by subject relativization is Case-checked in spec/CP, a

direct object is still blocked from moving te that position.

Note that the third option in LW, where a pronominal NP can trigger

object agreement on the head of NegP, is not found in CW. Case-checking in

spec/NegP will not be ruled out in principle (in fact, LF Case-checking of

[-definite] NPs will be an option in CW); the lack of agreement on Neg is

derivable from the CAR. As will be discussed in the following section,

pronominal VSO direct objects in negated sentences are preceded by the lexical

item o. Agreement with a pronominal NP is triggered, as expected, on this

head, which provides a lower licit SpeclHead configuration for the VSO direct

object. This will be clarified in the following section.

4.4.2. SUBJEcr-OBJEcr ASYMMETRIES AND CASE IN NEGATEDCLAUSES IN CW

In Chapters 2 and 3 of this thesis, a Case-licensing mechanism was

invoked which allows the trace of a raised V in Delta te assign Case under

government te a VSO direct object. Case via govemment is not blocked by the

presence of NegP in LW. Thus, while a pronominal VSO direct object may

forro a chain with a pleonastic in speclNegP in LW (triggering agreement on

224

•

•

•

Neg), this is not required. A pronominal in this position is acceptable even if it

does not uigger agreement. This is illustrated in (48)-(49).

(48) Palodd Sion ef.dug-3sg John itJohn dug it.

(49) Ni phalodd Sion ef.Neg dug-3sg John itJohn didn't dig it.

Similarly, as shown in (50)-(51), negation does not interfere with the

ability of a non-pronominal NP to be Case-licensed under government.

(50) Palodd Sion yr ardd.dug-3sg John the gardenJohn dug the garden.

(51) Ni phalodd Sion yr ardd.Neg dug-3sg John the gardenJohn didn't dig the garden.

This is as expected, given that NegP in LW, unlike in Russian, does not

intervene between the position from which Case is assigned under government

and the VSO direct object. What we find in CW is quite different. If the clause

is negated, the lexical item 0 must occur immediateIy preceding the VSO direct

object. This item is homophonous with a preposition meaning 'of that surfaces

in partitive constructions, and is often glossed as such.

22S

•

•

(52) Phalodd Sion ddim ohono fo.Neg-dug-3sg John Neg of-3sg itJohn didn't dig it.

[Jones & Thomas 1977:323]

Since we maintain that agreement reflects S-structure Case-checking in a

Spec/Head configuration, the agreement found on this element is analyzed as

resulting from such an operation.

Given that this analysis obviates the distinction between Case-licensing

of pronominal and non-pronomi.,al NPs in VSO direct object position (they are

both Case-licensed under government), we predict that whatever interferes with

the Case-licensing of pronominal VSO direct objects should interfere equally

with that of non-pronominal NPs. This is borne out; non-pronominal VSO direct

objects alse require the preposition.24 This is shown in (53)-(54).

(53) Palodd Sion yr ardd.dug-3sg John the gardenJohn dug the garden.

(54) Phalodd Sion ddim o'r ardd.Neg-dug-3sg John Neg of-the gardenJohn didn'l dig the garden.

First, we will address the issue of what blacks Case-as~ignment under

government te these NPs in negated sentences. Based on the anaiysis of Russian

in Chapter 3, we will then propose an account of how Case-licensing is

accomplished in these structures.

226

• 4.4.3. MEDIAL NEGATION AS A BARRIER 1'0 CASE-ASSlGNMENT

Reca1l how this Case-licensing mechanism functions. The chain of the

raised V includes a trace in Delta. From this position, the V-chain can assign

Case to an NP that it governs. However, the ability of the V-chain to assign

Case is constrained such that Case can only be assigned into the specifier of the

maximal projection it immediately dominates. The constraint on this kind of

Case-licensing will now be refined further, such that an intervening potential

Case-assignee will block Case-assignment to a lower argument.

In a paper on stylistic inversion in French, Valois & Dupuis (1989)

argue that an intervening subject blocks Case-assignment under government to a

lower NP, even though the subject does not receive this Case itself.25 They

consider this proposai to be in the spirit of Rizzi's (1990) reIativized minimality:

"[A]lthough the subject NP does not receive case from the verb, it is

nonetheless a potential case "receiver", just as a wh-phrase in an intermediate

COMP is a potential antecedent for a trace it c-commands." (Valois & Dupuis

1989:20, fn. 8) Thus, an intervening potential Case-assignee blocks Case­

assignment to a lower potential Case-assignee.26

In the negative structures in CW, the clause-medial marker dim

intervenes between the position from which Case can be assigned under

government (Delta) and the position into wlùch Case can be assigned under

government (specIVP). 1 will maintain that it is the presence of dim that

interferes with the ability of the chain of the raised V to Case-license a VSO

direct object; that dim in CW counts as a cIoser potentiaI Case-assignee.27

This cIaim is consistent with the fact that dim was diachronicalIy an NP

meaning 'thing', and is still homophonous with it.2S Thus, aside from the

227

•

•

question of whether dim actually requires Case, it is not inconceivable that il has

the status of a potential Case-assignee.

Nevertheless, support for the hypothesis that dim is in fact assigned Case

might come from a phenomenon referred to by Harlow (1981) and BalI &

Muller (1992) as direct object mutation (DOM). DOM is so-càJled soft

mutation; under its influence an initial d surfaces as dd. Zwicky (1984) proposes

that the trigger for soft mutation is syntactic as opposed to lexical, and that

Accusative Case is the trigger for DOM. While the discontinuity of the verb and

the direct object raised problems for Zwicky's analysis, our hypothesis that Case

is assigned by a member of the verb-chain in Delta avoids this dilemma. Given

that the sentential negation marker dim (as opposed to the homophonous

constituent negation marker) is subject to this Iànd of mutation (Sad1er

1988:230), surfacing as ddim, we might argue that it bears Accusative Case.

This situation is illustrated by the structure given in (55), in which the

direct object is not Case-licensed.

(55)

t t

yr ardd

228

• (56) *Phalodd Sion ddim yr ardd.Neg-dug-3sg John Neg the gardenJohn didn'( dig the garden.

If this daim is correct, then o-insertion is required for an NP that would

otherwise be Case-licensed under government; not for an NP that is Case­

checked in a SpeclHead configuration. Thus, we have an explanation for why

the preposition is not required under negation for the direct object in a

periphrastic sentence. Since these complements are Case-licensed in the specifier

of the [-finite] verb, and do not rely on the raised V for Case, it is not

surprising that an additional Case marker is not required. The S-structure of a

negated periphrastic sentence is given in (57).

•

•

(57)

'dyw

Sion

t

~ddimCASE

.1

: V+Agr NP-i

l---r----11yn ei weld ef

229

• (58) 'Dyw Sion ddim yn ei weld (ef).Neg-is-3sg Sion Neg PROG 3sgm see (him)John didn'[ see him.

4.4.4. CASE-ÙCENSING OF VSO DIRECT OBJECTS

•

Thus, as in Russian, VSO direct objects in negated clauses in CW are

unable to be Case-licensed under govemment. It was proposed in the preceding

cho.pter that these arguments have the option of being Case-checked in spec/FP,

a functional projection generated within DeltaP. The introduction of this

p,ujection within our analysis of Russian was driven largely by theoretical

considerations, the presence of I-P being inferred from Accusative Case-marking

which, according to our analysis, could not come from the verb. Motivation for

the introduction of this projection was derived from a property unique to it: its

restriction 10 [+definite] NPs. These data from CW suggest an analysis where

the inability of the verb 10 assign Case under govemment to VSO direct objects

forces Case-licensing within FP, as in Russian.

Let us assume that the Case-licensing head for VSO direct objects in

negated constructions in CW is FP, the head of which is 0.29 A pronominal, if

oven, follows the head of FP. On the assumption that neither pronominal nor

non-pronominal NPs in Welsh raise to their Case positions until LF, these NP~

must be generated inside FP.30 The S-structure of a negated transitive sentence

with a pronominal direct object is given in (59).

230

• 1591

F+Agr·k NP·k [ +pronominal)

: ~ ,, ,: ;l. .__ ....•__.•......_.....~

'dyw Sion ddim ohono el

A non-pronominal NP raises ta specIFP for Case-checking at LF.

4.4.5. (IN)DEFINI'ŒNESS REvlSITED: CASE-UCENSING IN SPEClNEGP

It has been noted that the lexical item we have analyzed as the head of

FP is not found preceding [-definite] direct abjects (Awbery (1988, 1990),

Hendrick (1988:261 fn. 27». This is ilIustrated in (60)-(61).

(60) Yfais i ddim te.drink-past-lsg 1 Neg tea1 didn 'r drink any rea.

231

•

•

(61) Ddarllenais 1 ddim llyfrau.read-past-1sg 1 Neg books1didn 'r read any books.

[Hendrick 1988:262)

According to Awbery (1990), a three-way distinction is found in

Pembrokeshire Wclsh regarding the appearance of 0 (which she refers to as a

preposition) in negated clauses. Her conclusions, based on archival recordings

of spontaneous spoken speech by a number of individual native speakers of

Pembrokeshire Wclsh, are that pronouns a/ways occur with the preposition,

indefinite NPs never appear with the preposition, and definite NPs may or may

not appear with the preposition.

It is worth noting that Awbery's reference to the oprionaliry of the

preposition is based on its appearance in sorne but not all sentences where the

direct object is [+definite]. It does not necessari1y follow from this that any one

speaker uses the preposition optionally in such structures; instead, the exhibited

variation may he due to certain individuals consistently using the preposition and

others consistently dropping it, under the same conditions. Thus, optionality

may not exist within any individual grammar. Accordingly, our analysis will

primarily be concerned with addressing those structures where the preposition is

always found (pronominal direct objects) and those where it is never found

(mdefinite direct objects). We will, however, aIlow for optionality of the

preposition with non-pronominal [+definite] direct objects.

Awbery summarizes the situation for object NPs as shown in (62).

PRONOUN

•(62)

ddim+objectddim+o+object +

lNDEFINITE NP

+

DEFINlTENP

++

232

• We have analyzed the presence of 0 as reflecting Case-ehecking in

spec/FP (where 0 is the head of FP). Let us now further extend the analysis

developed for Russian and analyze its absence as indicating that an NP is

Case-ehecked in speclNegP at LF. If this is correct, the paradigm parallels to a

remarkable degree the Russian data described by Neidle (1988). There, the

Genitive of Negation (Case-ehecking in spec/NegP) is available to both

[-definite] and [+definite] NPs, and the Accusative option (Case-ehecking in

spec/FP) is restricted to [+definite] NPs. The LF structure of a negated VSO

structure with a [-defmite] NP is given in (63). Both the subject and object NPs

have raised to their LF Case-ehecking positions.

t-o

•

•

(63)

NPo~

1!

ii1

iL__I

1 t,

l -'

233

•

•

Note, however, that these data suggest an additional restriction: that

pronominal NPs cannor be Case-licensed in spec/NegP. One met.'lod of

implementing this teehnically would be by positing a restriction associated with

Case-licensing in spec/NegP in this dialect to [-pronominal] NPs. However, a

less ad hoc alternative is available. In Chapter 2 we proposed that pronominal

NPs must be Case-checked at S-structure in Welsh, whenever possible. This

requirement is reflected in agreement, the reflex of S-structure Case-ehecking;

only pronominal NPs trigger agreement in Welsh. The obligatory presence of

FP with pronominal direct objects in negated clauses arguably foIIows from the

same language-specific feature of Welsh. Pronominal NPs must make use of the

S-structure Case-ehecking option, in spec/FP, rather than the LF option of

raising to spec/Negp.31

Consider next the optionality of 0 with non-pronominal [+definite] NPs.

Given the FP restriction to [+definite] NPs and the S-structure Case-checking

requirement for pronominal NPs, non-pronominal [+definite] direct object NPs

are unique in CW in having IWO Case-licensing options in negated clauses;

Case-licensing in speclNegP or spec/FP. The exhibited optionality of 0 with

[+definite] NPs follows.32

LF Case-checking in speclNegP requites the kind of movement over an

intervening subject that was ruIed out at S-structure for relative clauses ir.

WeIsh. Note, however, that this constraint on movement bas to be relaxed at LF

in any case. The VSO direct object that creates a chain by indexation with a null

pleonastic in speclNegP in LW must raise te that position at LF fer expletive

replacement. Accordingly, the fact that the non-pronominal VSO direct object

234

• raises to spec/NegP at LF for Case-ehecking does not introduce any new

complexity to our account.

4.4.6. VARIABLE WORD-ORDER IN PUmROIŒSHIRE WELSH

As discussed at length above, the main negator in CW is dim. According

to Awbery (1990), its exact position in Pembrokeshire Welsh (PW) varies

systematically according to the nature of the subject NP. We do not consider the

grammatical functions subject and object to be primitives of the theory; rather,

they are defined structurally. For our purposes, a fundamental distinction exists

between arguments that are base-generated in spec/De1taP (AGENTS) and

arguments that are base-generated inside the lower VP (PATIENTS, THEMEs).

Viewed from this perspective, the generalization appears te he that D-structure

objects can oceur in two distinct S-structure positions in PW, and that this is

sensitive to the property of definiteness.33

First, the data show that an indefinite NP which is aD-structure object

always follows the negative marker, as in (64).

(64) Ond ddath ddim pethe i ben fel odd hi wedi fwriadu.but carne-3sg Neg things to end as was she after intendBur things didn 'r work our as she Md intended.34

However, when such an NP is a definite noun, as in (55)-(56), it may

precede or follow the negative marker.

(65) A fywodd ddim 'r 'en grwban bach.and lived-3sg Neg the old tortoise littleand the little oTd ronoise didn 'r survive.

235

• (66) Ath 'y nhad ddim i mas i ddryehidwent-3sg my father Neg to outside to lookMy falher didn 'r go owside lO look.

The caveat reg:trc!ing the optionaiity of definite NPs. initiaily raised with

respect 10 the presence of the particie 0, applies here as weil. Nevertheless. until

these faets can be ehecked with a native speaker of Pembrokeshire Welsh. our

"Ulaiysis will allow for definite NPs to have (WO options in PW, as in the

preceding section.

When the NP in question follows dim, the appearanee of 0 is related to

whether it is [+definite] or [-definite]. Thus, as in (67), a definite NP following

the negative marker dim is optionaily preceded by the preposition 0, but as in

(68), an indefmite NP following dim is never preceded by this preposition.3s

(67) A ddath ddim (0) )' gyfreth i rym nesbod hi 'n y Ionawr.and came-3sg Neg the Iaw into force till be it in the JanuaryAnd rhe law didn 'r come inro force rill JanU4ry.

(68) Alle ddim (*0) mashin i ddwad lawr 'na.could-3sg Neg machine to come down thereA machin~ couldn'r come down rhere.

1 will tentative1y suggest the following explanation for these variable

word order facts. In Chapter 3, an anaiysis was proposed for the Russian diaiect

described by Pesetsky, where the Genitive of Negation is restricted 10 [-definite]

NPs. 1 proposee! that Case-checking in speclNegP was restricted 10 NPs that

have copied their referential index onto the negative operator, and that index­

copying affects only [-definite] NPs. We can potentially extend this proposai

further to account for the Pembrokeshire We1sh paradigm by claiming that

236

•

•

index-copying does not merely allow but forces an NP ta be Case-licensed in

speclNegP. Accordingly, indefinite NPs must remain in their base-position at S­

structure, within the lower VP, following dim. Definite NPs, on the other hand,

have [wo options: the option of being Case-Iicensed in spec/FP at LF (thus

appearing following dim 0), or, alternatively, the option of raising ta

speclDeltaP at S-structure (preceding dim) and being Case-Iicensed in speclTP

at LF.36

4.5. S~~y

In this chapter, we have proposed that sentential negation in CW differs

from LW in the following respects. First, the head of NegP in CW is a

morphologicalIy dependent lexical item. Second, sentential negation in CW

requires an additional negation marker, generated in an adjoined position

between DeltaP and VP.

We have argued that NegP in WeIsh is dependent on the generation of

TP. The lack of NegP in [-finite] clauses, absolutive constructions and past

participles were all derived by this constraint.

The analysis also accounts for the fact that relativized subjects in negated

clauses in CW differ from the equivalent sentences in LW (exhibiting the direct

pattern of agreement), while relativized direct objects continue 10 exhibit a

resumptive pronoun strategy. In this way, an asymmetry that does not exist in

LW has been accounted for in CW.

Furthermore, we have accounted for the lexical item found preceding

VSO direct objects in CW negated clauses, by proposing that the option of

Case-licensing under govemment is blocked by the secondary negation marker,

dim, which counts as a closer potential Case-assignee. We have proposed that

237

•

•

the additional Case-licensing option required (reflected in the presence of the

preposition 0) is related to the presence of the functional projection FP,

proposed in the preceding chapter for Russian Case-licensing of direct objects in

negated sentences.

ln the final section we argued that speclNegP provides an alternative Case

position for VSO direct objects, and data was introduced that supports the daim

that FP is restricted to [+definite] NPs, as proposed in Chapter 3.

4.6. ENDNOTES TO CHAPTER 4

IThe distinction between LW and CW is nol hetw.... a written fonn and a spokenfonn of the same language; LW and CW con bath he written and spoken. LW is rescrvc:d formore fonnal situations. for example speeches. officiBJ letters, ete. Il is closest 10 the fonn thalWe\sh lOOk when the Bible was original1y lnlIIS1ated inlO We\sh in 1563. and bas beonesscntially fro:œn in that state. (This exp\ains why certain sentences suggested 10 my informantswete tetmed 'biblical sounding'.) Nevetthe1ess, bath LW, in the sense of the fonnal or srandardlanguage. and CW are subjecl 10 regional variation. Sec PUch (1971) for discussion. The CWdata in !his section are taken from a variety of so=. principally Jones &. Thomas (19TI) andAwbety (1988. 1990). The data in the discussion of NegP and TP. excepl where otherwisenoted, were providc:d by my We1sh informants. John Williams and Ewen Edwards.

:zone Jack of 0 preccding indcfinile direct abjects in CW is also nolcd bY Hcndrick(1988).

3These include nabod ('10 know'), gaIujo ('10 rain'). gobeijo ('10 hope'). kan ('10love').~ ('10 he accustomed 10'). ete. (pUch 1971:139). Many of these vetbs con heclassified as statives. Sec NoolI&II (1992) for an intriguing analysis ofa similar l'acts in Irish.

4-rhe fi."St instance ofbod bas undergone initial consonanl mutation. surfacing as/ad.

Sne IWo VPs surrounding AspP provide base-positions for bod and the [-tcnsel vcrb,respective1y. The base position for bod must dominate the base-position of the aspecl markerSÏDce il precedes the 1atter even when [-tensel. when. byassumption. il bas nol raised out ofVP.

6ne marker dim is usee! for regular sentence negalion colTeSpOnding 10 Eng1ish not.However, dim is replaced by the negalive markers byth or erioed 10 negate a sentencecorresponding 10 an Eng1ish~on. This is discussed in greater detail in theAppendix.

710nes &. Thomas (19n:320f). Payne (1985:224). There are IWo consonants whichdo not undergo lenition; IsI and /hl. Thus, in a sentence like Synnwn i ddim ('1 would not hesurprised'). ddim is the sole evidence of negalive contenL

238

• 8The initial consonant of the =ondary negative marker bas undergone soft mutation.The conditions whieh triggcr consocact mutation in Wclsh are a matter for spirite<! dchate in theliteraturc, although the argumccts for the triggcr being SCDSitive te syntactie facters as opposcdte word order alone appear te be convineing. Sec TaIlcrmac (19908) for discussion and reviewof previous proposais. The triggcricg of soft mutation on the negative marl:cr is discusscd insection 4.4.3.

9nere is tremeedous regiocaJ divcrsity with respect te negation of the verb bod ('tebe'), whethcr usee! as copula or auxiliary. Standard LW: nid yw Pam yn y gwailh (neg-is Pamprog worlc: Pam is nol worldng) vs. Northem (tydi Pam ddim yn y gwairh) vs. Southwest(smolsimo Pam yn y gwairh) vs. Westcre (Mg yw Pam yn y gwairh) vs. Southem (dyw Pamddim yn y gwailh).

I0Compare Ibis Ilegate<! cleft wilb the equivalect in LW, discusscd in Chapter 2,which requires the negative partiele nard): Nid John 1IIld oedd yn help". The remnact of nard) (­d) may show up in CW, allbough Jones and Thomas (1977:325) state that the more typicalpattern is lbe occurrence of medial dim alone Ilegating lbe clause.

l1ln this section data are from Jones & Thomas (1977: 322ft). Lexical insertion ofthis item in Ilegate<! sentences is a rohust feature of CW, as IlOte<! by Sadler (1988), Williams(1980:115), and Awbcry (1990, 1991), and Hecdricl: (1988).

12Although Rouveret (1991) mentions the possibility ofgeccrating dim in speclNegP,he acaIyzes Ibis marlcer as gecerated adjoined te the higher VP projection. There are severa!lbeory-intercaJ reasons for rejecting the Spec/NegP acaIysis, lbe first bcicg that givec thesimilarities betweec the Wclsh data and the Russian facts, there is evidecce that the specifier ofNegP is available for SpeclHead Case-licecsicg. Sec the Appecdix te this Dissertation forfurther discussioll. Note that givec that NegP is IlOt gecerated in [-tiIlite] clauses, it is Ilotpossible te derive lbe base positiOIl of dim by comparing its positiOIl relative te [+tecse] and [­tecse] verbs, as Pollock (1989) proposes for bis acaIysis ofFrecch llegatiOIl.

13nese arguments aIso bold for LW, discusscd in Chapter 2. While democstratinglbe depecdecce of NegP 011 TP, 1 malee 110 claims as te what drives this depecdecce. The claimmade by Zacuttini that T, in a sense, subcategorlzes for NegP is somewhat bizarre in the sensethat Ibis cac involve a lower bead subcategorizicg for a higber bead. Zacuttini aIso notes in berdissertation (1991) that the notion of subcategorization becomes somewhat obscure wbec appliedte functiocaJ beads.

14Data in Ibis section are takec from Borsley (1984:293).

ISNote that in the data in the sectiOIl covcricg negation of past participles theprepositiOIl a does IlOt appear. That data is takec from informant worlc, and the particle wasdropped.

16ln a taII: givec at Université de Québec l Montréal in 1991, Zacuttini suggestedthat while the higher negatiOll found in the first sentence is a funetional projection NegP, the10wer negatiOll is lexical; an adverb. This differecce is reflected in COIltraction possibilities; thehigber negative clement cac be COIltraeted but the lower CBIIIlOt: Mary hasn't always paid taus,*Mary has alwaysn 'r paid taus. However, in Zacuttini (1992), she acaIyzes bolb the higher andlower negative marlcers as NegPs.

239

• 17The form ni(d) occurs on abselutives, bul these are in.<tancos of conslituenlnegation. nol sentential negation. Thus, if the pre-sentenlial negalive marker is use<! (i.e. nidarol cyrraedd yn h"Y"•••. ). il means Il wasn'l afta arrhing lau rhal... (i.e. il was some othertime).

18See Zanuttini (1990:6-7. 1991) for a discussion ofthose fiu:l<.

19ne use of!his vern ta negate imperatives in CW is aise noted by Jones & Thomas(1977:348).

2~orn suggests that this constrainl may he pragmatic in nature. hasing his argumenlon possible dangers associated with post-vetbal negation in negative imperatives. An examplewould he ·Jump•..••norl·. While perbaps appea1ing al an intuitive leve!. il is bard ta """,ncile!his suggestion with the ltalian data. where the constraint operates only on true (uninflected)imperative forms.

21Ta1Ierman glosses mo'r as NEG-the. but mentions!hat 'mo is a contaction of dim 0

(NEG-P). se 1 have glossed it as such.

22Hence. subject agreement is found in negated relative clauses in CW only if thesubject itself is not relativized.

23See Harlow (1981) for discussion.

24ne not unexpected difference between pronominal and non-pronominal NPs innegated sentences in CW is that only the former trigger agreement on the preposition.Case-checlcing of pronomina1s in the specifier of the projection headed by 0 is an S-structurephenomenon; Case-checlcing of non-pronominals in !his position occurs at LF. Nole thal thedaim that these arguments are generated as sisters ta F rather !han V in negated structures mayhave non-trivial consequences for the Uniformity of Thematic Assignment Hypothesis (UTAH:Baker 1988). and may require revision. See fn. S. Chapler 3.

2SVaIois & Dupuis (1989) argue against the Casc-assigning ability of vetb-traees. aposition incompatible with the approach taken bere. They do. bowcvcr. suggcst a possiblercvision to the constraint in vicw of VSO languages, such that only the lowcst vcro-uacc isincapable ofassigning Case. This vicw is compabole with my elaim that the vcro-uacc in Deltabas this capability.

26NOle that, for the CW data. the constraint could altemativcly he formulated as anadjaceney condition (expressed in structural terms). as dim is the only lexical item thatpotentia1ly intervencs between Delta and specNP. Au!hier (1991:726-7) proposes that theadjacency condition on Case assignment in Eng1ish he captured structurally as follows: '[•••) Vassigns ACC Case only ta a sister with which it is in a stri~t (first branching node) C-commandrelation.' Sadlcr (1988:245 fn. 6) also suggeslS that preposition insertion in negated sentences inCW should he analyzed in terms of Case thenry. as docs Hendrick (1988).

27As an alternative ta the treatment of dim wc propose hcre. one could maintain thatit is its ro1e as a doser potential antecedent goveroor that is the cruc:ial property affectingCasc-licensing in CW ncgated clauscs. Anticipating our discussion of Pemhrokcshire Wclsh. itcould he claimed that the presence of dim b10cks LF movemcnt of a VSO direct object ta

240

•

•

•

SJ"'C/NegP for Case-chocking. The lrac<: in VP would nol b<: anloc<:denl govem<:d. forcing theappearmce of the dummy Case a.........ih"l1Cr o. However, this would rt:Sult in incorrect p~ictions.

SJ"'Cifically. LF movem<:nl 10 SJ"'C/NegP is p<:rmitl<:d for NPs (in facto it is obligatory for[-definit<:) NPs which cannol b<: ea.",·li=s<:d in SJ"'C/FP). It is bard to """ how an analysis ofdim as a pott:ntial barric:r to antt:Ceddlt govemmc::nt could he:: made sensitive to the specificationfor the f""lure [definilei. An analysis whereby dim inlerferes only with ea.",-assil,'llm<:nt undergovemment is then::fore preferJ.ble.

28The particJe dim which acts as a negative intc:nsifier in LW is pennitted in personalpassives (Chafodd mach ddim o'u /lad yma [NEG-got-3sg pigs NEG of-3pl kill here] Pigsweren't kill<:d here) but not imp<:rsonal passive structures (Ni /addwyd mach yma [NEG kilI<:d­imp<:rsonal pigs here] Pigs wer<:n't kill<:d here. (Sadler 1988:230). a fact that is otherwiseanomalous. but might b<: amonable 10 a ea.'i<:-driv<:n <:xplanation.

29While th= facts are suggestive of English of-insertion. an analysis along the linesof Chomsky (l986a) is problematic. given the framcwork WC assume here. Chomsky analyzesthe p=c<: of of in sentencos like Kim's ~nry of Jean (cf. "Kim 's ~nry J~an) as a ref1<:x ofinher<:nt Case-assignment by the noun. Given that a in CW always prce<:des the direct objoct ofthe verb, the CW paradigm is consistent with the claim that a l<:xical it<:m cao ooly assigninher<:nt ea..., if it thela-marks an NP. but encounlers problems in other ""'l'OCts. If we retreatfrom the position that the verh combines with Neg in the synlall. and claim instead that verbs arcgeneral<:d in the l<:xicon as morphologically negative. we could in principle claim that a negatedverh assigns inherent Genitive Case to its objoct. We could maintain further that Ibis is ref10ctedin Genitive Case marking on the objoct in Russian. and in the presence of a preceding the objoctin CW. The first problem encountercd is accounting for Accusative ea..., assignment in negaledsenl<:nces in Russian and negative sentencos without a in CW (discussed in the next soction).How are the direct objects assign<:d Case in these scntencos? If wc app<:al to a functionalprojoction aldn 10 the propos<:d FP for Case-licensing when inh=t Case-assignment does nottake plac<:, then we arc left with the problem an apparent sensitivity to definileness and thelriggering of agreement on O. Giv<:n that agreement ref10cts Case-cbocking in a SpocIHeadconfiguration. agreement on a is incompatible with Case assignment under govemment.

30rhe fact that FP is base-generaled in distinct positions in Russian and CW is anundesirable feature of this llIll!lysis. A possible resolution would involve maintaining that FP inRlLo;sian is aise generaled in complement of V position, and forœ F-to-V raising ta facilitateCase-licensing, but certain problems arc associated with sucb a claim. Seo Cbapler 3. fn. 18.

31While this analysis accounts for the lack of LF Case-cbecking of pronominals in.l""'fNegP. still unexplained is why pronominal NPs in CW do not bave the option available inLW of crcating an A-<:hain at S-structure, headed by a pl""nastic in spocINegP. and triggeringagreem<:nt on the head of NegP. While this inability is not surprising. given the rcduced staNsof the negative head in combination with the availability of a Case-cbocking position lower thanthe ."Ubjoct NP. this analysis as it stands does not provide a principJ<:d means ofruling it out.

32Nolc !hat if the CAR constrains not only agreemcnt rca1ization but wherc in a ehainCasc-chccking accu"," [+dctinite] NPs arc prcdietcd to rcquirc o. as Ibis rcprcscnts a lowcrSpcclHcad configuration for thc direct abject !han spcc/NcgP. Howcvcr. maintaining CAR forCasc-chocking is impossiblc givcn thc claimcd optionality ofo.

241

• 33Awhery CllÛms lhat pronominal 'subject' NP. differ from non-pronominal NPs.whether definite or indefinite. in lhat they must precede dim. However. since the pronominalNPs in the examples sbe CileS are aIl ageotive. !his word order is expected. since they would hebase-generated in speclDeItaP. above dim. Accordingly. we will not discuss them here.

34Note that Awbery glosses wedi as 'afler'. as traditiona1 grammars do. We willmaintain her glosses hore. but we follow Sadler (1988) in treating !his element as a perfectivemarker.

35The descriptive statement that indefinite NPs following dim cannot he preceded bythe preposition is made by Awbery; however. as her data is based on spontaneous speech rather!han grammaticaIity judgements. her data include only possible sentences. not ungrammatica1ones. Accordingly. 1 have inserted the asterisks te mark the unacceptability of 0 te clarify theexamples.

36What is particularly surprising about these data is lhat they suggestlhat NP-raisingOCClUS in the syntax. The claim that an NP may raise in the synlaX but not te its ultimate Case­position is problematic for the principle of economy. One stage in the movement from within VPte SpecfIl' is completed al S-structure. the next stage at LF. Th.... is no apparent reason for !hismovement heing broken up inte IWO distinct levels of the grammar. as opposed 10 taking placeentirely al LF. CIearly. !his paradigm requires more extensive investigation. Furtber. whileDoting these facts for the interested reader in the context of an NP-movement ana1ysis. it is notinconceivable lhat the data are in fact demonstrating IWo possible positions for dim placement. Ifthis approach is taken. the difficulty would he in making !his optiona1ity sensitive 10 the notionof (in)definiteness. Sec Rouveret (1991) for an alternative ana1ysis of these facts.

242

•

•

•

S. CONCLUSION

The objective of this dissertation is to broaden and sharpen our

understanding of the syntactic nature of sentential negation. To this end, r

undertook a close examination of changes in Case-marking, agreement and word

order that appear to be triggered by negation in Literary Welsh, Colloquial Welsh

and Russian. The proposai that the inventory of functional projections available

cross-linguistically includes a NegP was adopted as a starting point for the

investigation.

In Literary We1sh, the initial paradigm investigated was the interaction

between negation and changes in relative clause formation. Previous researchers

had noted that subject and direct object relativization out of negated relative

clauses appeared to force a different choice of complementizer and distinct

agreement patterns. By positing that sentential negation generates a NegP whose

head interferes with V-raising and whose specifier provides a Case coilfiguration

for A-chains, we were able to account for a wide array of data, induding

agreement patterns, the possibility of an in situ pronominal in relativization

constructions, and VSO direct object agreement surfacing on the head of NegP

itself. In that section, the framework of c1aims and assumptions led us to propose

that the head ofNegP can Case-Iïcense an NP in its specifier independently of V­

movement into Neg.

In the context of these daims, modifications to the analysis of relative

clause formation in Welsh have been proposed. By modifYing the treatment of

agreement in Welsh, we were able to do away with the traditional analysis that

distinguishes relativization of subjects and VSO direct objects from ail other cases

of relativization as a movement strategy as opposed to a resumptive pronoun

243

•

•

•

strategy. In trus way, we were able to propose a non-movement analysis for

relativization out of complex NPs, out of PPs headed by non-inflecting

prepositions, and for direct object relativization in negated clauses, ail of wruch

differ from both direct and indirect cases in permitting an in situ pronominal.

The proposai that the head of NegP can Case-license an NP in its specifier

found additional support in our investigation of the Genitive of Negation

construction in Russian. We saw that the Genitive ofNegation, wruch we analyzed

as subject to Case-checking in specINegP, appears in sentences traditionally

analyzed as lacking Accusative Case assignment: unaccusative and passive

constructions. T1ùs supported the c1aim that the Case-licensing ability ofNeg is not

sirnply an a1tered version of the verb's Case-assigning ability, but is rather a

property ofNeg itselfin these languages.

We argued further that the presence of NegP in Russian interferes with

Case assignment under govemment to VSO direct objects. This was related to the

lower base-position posited for Russian: irnmediately below DeltaP rather than

irnmediately above TP. The base position ofNegP proposed for Russian accounted

for the apparent sensitivity of the Genitive of Negation to thematic roles,

specifica1Iy the inability of agentive NPs to surface with Genitive Case in negated

clauses.

The lack of interference with V-raising found in Russian negated clauses

was related to the morphologica\ly dependent status of the head of NegP. Vero­

movement into Neg was analyzed as an obligatory instance of substitution, wruch

does not block further raising to T.

The chapter on Colloquial Welsh offered a bridge between the analyses of

LW and Russian. Thus, as proposed for LW, we maintained that NegP in CW is

244

•

•

•

gcncratcd above TP, but like Russian. the head of NegP in CW is a

morphologically dcpendent item which does not interfere with V-raising. In this

way, an account was provided for both the similarities and the differences between

relative clause formation in negated clauses in LW and CW.

However, as argued for Russian, the presence of sentential negation in CW

intcrferes with the verb's ability to assign Case under govemment. This was related

to the prescnce of the secondary negative marker required in CW. It was proposed

that this secondary negative marker is a potential Case-assignee, that blocks Case

assignment to a direct object NP lower in the structure.

Furthermore, we showed that the presence ofNegP in CW is dependent on

the presence ofTP, as claimed by Zanuttini (1990) for Italian.

The definiteness effects found in both Russian and CW negative clauses

were relatcd to the additional functional projection (FP) required to replace Case

assignment under govemment, which is blocked in negated sentences in these

languages. The potential Case-checking position provided by the specifier of FP

was characterized as available only to [+definite] NPs. Based on the interpretation

ofGenitive NPs vs. Nominative or Accusative-marked NPs in negated sentences in

Russian. as weil as the option of dropping the lexical item 0 which otherwise is

required as a Case-assigner for VSO direct obejcts in CW, we proposed that in

these languages [-definite] direct object NPs are forced to be Case-checked higher

in the structure, in specINegP.

Our analysis of Russian aise provided an explanation for the asymmetric

interpretation ofNominative and Accusative NPs in negated clauses; an asymmctry

that is not adressed by Pesetsk..y (1982). Only Accusative NPs have a restricted

[+definite] reading, because only spec/FP is restricted to [+definite] NPs.

245

•

•

•

In the analysis of the dialect of Russian discussed by PesetsJ,.-y (1982), il

was proposed that Case-checking in specINegP is subjecl 10 an addilional

restriction; only an NP that has copied its referenlial index as a selection index onto

the negative operator can be Case-checked in specINegP. Following Heim (1982),

we maintained that only [-definite] NPs are subject to Quantifier indexing, thus

accounting for the necessarily [-definite] re::ding of these NPs in the dialect

discussed by Pesetsky.

This investigation into the syntactic nature of sentential negation in LW,

Russian, and CW supports sorne daims that have been made about the properties

of NegP in other languages. As argued by Pollock (1989) and Choms\...-y (1989),

we found that the head of NegP can interfere with XC-raisins. by crealing a

rninimality barrier. Our analysis also supports the daim made by Ouhalla (1990)

and by Zanuttini (1990), and assumed in severa! other papers, that the base­

position ofNegP is subject to cross-Iinguistic variation. Our work on Russian and

Welsh supports their proposais that NegP can be generated either above TP or

aboveVP.

246

• APPENDIX: NEGATION AS HEAD AND SPECIFIER OF NEGP

In this section, we explore a treatment of sentential negation in CW that

posits a lower base-position for NegP and generates dim in spec/NegP. We will

see that while this treatment of NegP provides an interesting explanation for the

required secondary negation marker in CW, the interactions between negation,

Case, and definiteness are problematic.

If, following Pollock's (1989) treatment of sentential negation in French,

we generate dim in the specifier and ni(d) in the head of NegP, we must account

for surface order of the negative constituents. This could be achieved by

appealing to the reduced status of the head 0= NegP in CW. If it is analyzed as a

c1itic, it is arguably subject to obligatory raising to a higher functional head at

S-structure. Given that the element in specifier position, dim, does not undergo

raising, we can appeal to its S-structure position to determine a base position for

NegP which is be10w De1taP. This is illustrated in (i).

lil

Neg

'"!t!

I--I-·--~

•dyw Sion ddim yn

V NP1 1

canu "God Save the Queen"

247

•

•

(ii) Dyw Sion ddim yn canu "Gad Save the Queen".Neg-is John Neg PROG sing "Gad Save the Queen"John isn 'c singing "God save che Queen".

CW ne6ation patterns with French in requiring the presence of a

secondary negation marker. 1 Just as in French, a negative adverb (bych or

erioed in Welsh) can appear in the place of the standard marker dim, satisfying

this requirement.:

(iii) FRENCH:

Je n'ai {pas/jamais} vu Marie.I Neg-have-lsg Neg seen Mary1have {noc/never)seen Mary.

*Je n'ai vu Marie.

I Neg-have-lsg seen Marie

(iv) COLLOQUIAL WELSH:

'Dyw Sion {ddimlbyth} yn palu'r ardd.Neg-is John Neg/never PROG dig-the gardenJohn did noc dig/never dug lhe garden.

*'dyw Sion yn palu'r arddNeg-is John PROG dig the garden

When these elements do not appear in the position otherwise taken by

dim, they appear in S-final position. When they appear in a negated sentence

when dim is present, they appear S-finally, and contribute no additional negative

force to the sentence. This is illustrated in (v).

(v) 'Dydy John ddim wedi nofio yn y mor erioed.Neg-is John Neg PERF swim in the sea everJohn has never swwn in lhe sea. [Lit: John has noc swwn in lhesea ever]

248

•

•

•

To account for these facts, we can appeaJ to sorne recent work by

Haegeman (1991) on sentential negation in West Flemish. Haegeman maintains

that if the secondary negative marker found in West Flemish, which she terms

suppon negarion, is not generated in spec/NegP, it must move there, either at

S-structure or at LF. This movement is driven by the Neg-Crîrerîon, which

Haegeman bases on Rizzi's (1991) wh-Crîrerîon.

(vi) Neg-Criterion: (Haegeman 1991:9)A Neg operator must be in a Spec/Head configuration with anX°[Neg).An X°[Neg) must be in a Spec/Head configuration with a Negoperator.

Let us maintain that the Neg-criterion is operative at S-structure in

Colloquial Welsh. This will account for the obligatory nature of secondary

negation. The Neg-criterion requires that the head of NegP be in a SpeclHead

configuration with a negative operator at S-structure. This requirement is

satisfied by the presence of dim, base-generated in speclNegP. If dim is not

base-generated, a negative adverb must move 10 speclNegP at S-structure in

order to satisfy the requirement.

(vii) ·'Dydy John wedi nofio yn y mor erioed.Neg-is John PERF swim in the sea everJohn has never swum in the sea.

However, given that this thesis is primari1y concerned with secondary

sentence modifications triggered by sentential negation, specifically those

relating to Case and agreement, we must determine whether such a treatment of

sentential negation in CW can account for these facts as well.

Consider first those ways in which CW negated relative clauses differ

from the LW constructions. Recall that subject relativization in CW exhibits the

so-called 'direct pattern' of agreement. Given that the 10wer NegP will not

intervene between T and C, the direct pattern found in negated clauses is

successfully derived for subject relativization.

249

•

•

•

The analysis must also account for the resumptive strategy found with

direct object relativization out of a negated clause. If we analyze the secondary

negation marker in Welsh as a potcntial A'-antecedent (in conjunction with a

c1aim that the verb-trace cannot satisfy the ECP by theta-government), we can

block wh-movement from direct object position in negated clauses by appealing

to the ECP; its trace in VP would not be properly governed.

However, by appealing to the barrier status of dim to block wh­

movement of VSO direct objects in CW, we would lose a parallel account of the

resumptive pronoun strategy exhibited in VSO direct object relativization in

LW. In the analysis proposed in this thesis. the resumptive strategy is forced in

both LW and CW by the A-position of the object provided by NegP bcing

higher than the A-position of the subject.

Regarding the definiteness effects discussed in this chapter, an analysis

whp.re dim is generated in spec/NegP also encounters difficulties. It could still

be argued that the dummy Case-marker in CW is driven by the presence of dim

in spec/NegP (a position no longer available for Case-Iicensing). In order to

account for sentences where 0 does not appear, one could maintain (following

Belletti 1988) that verbs have an inherent Case to assign (which she terms

panitive Case), but this Case is constrained to [-definite) NPs. Accordingly, we

could conclude that the dummy Case assigner is not found with [-definite) NPs

because they cao receive inherent Case from the verb.

Such an analysis, extended to the Russian facts, encounters problems.

The Genitive Case option exists for all NPs regardless of their specification for

the feature [definite) (as evidenced by its appearance on pronominal and

demonstrative NPs, as well as on proper names). As it is the Accusative option

which is restricted in terms of definiteness; and the restriction is to [+definite),

not [-definite) NPs), we would be forced to c1aim that inherent Case à la Belletti

(reflected in Genitive Case marking) is unrestricted. Such an extension of

Belletti's analysis would result in a criticalloss of explanlltory power.

Similarly, the CW facts do not fit neatly into Belletti's analysis. The

dummy Case marker should be obligalOry for [+definite) NPs, given the

unavailability of Partitive Case. As we have secn, this is not correct; [+definite)

NPs in CW, as in Russian, exhibit an optionality not in keeping with the

restrictions associated with inherent Partitive Case assignment.

250

•

•

Thus, given the focus of this investigation and the results obtained in the

investigations of LW and Russian, the proposa! that dim is generated in

speclNegP in CW does not appear to be correct.

tThis leaves aside the issues raised by sentences in French where only the initialnegative ma:ker occurs, as in j" Il<! douz" qu'il vi"ndra. Tbese sentences cac he argued tainclude a negative force elsewhere, as in the doubt expressed by the main verb.

2The adverbs byth and aio<!d bave the same meaning, but are in complementarydistribution. Byth is restricted ta [·PERF] and aio<!d ta [+PERF) clauses. See Jones & Thomas(1977:322-331) for discussion. Nole lbat since these items are restricted ta perfect've andprogressive (bence periphrastic) sentences, the issue of preposition insertion does not arise.Such COnstruCtiODS bave a [-finile) verb which remains in ilS base position and Case-licenses anNP in ilS specifier. Thus, direct objeclS in these CODSIrUCtiODS do Dol rely on Case bygovemmenl, which is blocked by the presence of secondary negation•

251

•

•

•

REFERENCES

Abney, S. 1987. The English Noun Phrase in its Senrential Aspects. Ph.D.Dissertation, MIT.

Academy of Sciences ofthe USSR, 1980. Russkaja Grammatika Tom II:Sinraksis. Nauka: Moscow.

Aoun, J. 1981. ECP, Move-a, and Subjacency. Linguistic Inquiry 12.4.

Autlùer, J.M. 1991. V-Govemed Expletives, Case-Theory, and the ProjectionPrinciple. Linguistic Inquiry 22, 721-740.

Awbery, G.W. 1976. A Transformationai View ofWelsh Relative Clauses. TheBulletin ofthe Board ofCeltic Studies 27, 155-206.

__....:. 1977. The Syntax ofWelsh: A Transformational Study ofthe Passive.Cambridge University Press: Cambridge.

__....:. 1988. Pembrokeslùre Negatives. The Bulletin ofthe Board ofCelticSrudies 35, 37-49.

__....:. 1990. Diaiect Syntax: A Neglected Resource for Welsh. in R. Hendrick(ed.) The Syntax ofthe Modem Celtic Languages. Syntax and SemallticsVol. 23. Academie Press: New York.

Babby, L.H. 1975. Impersonai Verbs and their Lexical Specification. Slavic andEast European Jouma119, 182-187.

__....:. 1980. Existenrial Sentences andNegation in Russian. Karoma: AnnArbor.

Baker, C.L. 1991. The Syntax ofEngiish Not: The Limits ofCore Grammar.Linguistic Inquiry 22, 387-429.

Baker, M. 1985. The Mirror Principle and Morphosyntactic Explanation.Linguistic Inquiry 16, 373-415.

___. 1988. Incorporation: A Theory ofGrammatical Function Changing.University (IfClùcago Press: Clùcago.

Baker, M. and K Hale 1990. Relativized Minimality and Pronoun Incorporation.Linguistic Inquiry 21, 289-297.

BalI, M.J.(ed.) 1988. The Use ofWelsh: A Contribution to Sociolinguistics,Multilingual Matters Ltd.:Clevedon.

BaIl, M.J. and N. Müller 1992. Mutation in Welsh. Routiedge: New York.

252

•

•

•

Barwise, 1. and R. Cooper 1981. Generalized Quantifiers and Natural Language.Linguistics andPhilosophy 4, 159-219.

Belletti, A. 1988. The Case ofUnaccusatives. Linguistic Inquiry 19, 1-35.

___. 1990. Generali;ed Verb Movement: Aspects ofVerb Syntax. Rosenberg& Sellier: Torino.

Borsley, R. 1984. VP Complements: Evidence from Welsh. Journal ofLinguistics20, 277-302.

___" 1987. A Note on Traditional Treatments ofWeIsh. Journal ofLinguistics23, 185-190.

Borsley, R. and 1. Stephens 1989. Agreement and the Position ofSubjeets inBreton. Natural Language andLinguistic Theory 7,407-427.

Burzio, L. 1986. Italian Syntax. Reidel: Dordrecht.

Campana, M. 1992. A Movement Theory ofErgativity. Ph.D. Dissertation, McGilIUlÙversity.

Chomsk.")', N. 1981. Lectures on Government andBinding. Foris: Dordrecht.

. 1982. Sorne Concepts and Consequences ofthe Theory ofGovemment--- and Binding. Linguistie Inquiry Monograph 6. MIT Press: Cambridge,

Mass.

___. 1986a. Barriers. MIT Press: Cambridge, Mass.

__. 1986b. Knowledge ofLanguage: Ils Nature, Origin, and Use. Praeger:New York.

___. 1989. Sorne Notes on Economy ofDerivation and Representation. MITWorking Papers in Linguistics 10, 43-75.

__. 1992. A Minimalist Program for Linguistic Theory. MIT OccasionalPapers in Linguistics 1.

Chvany, C.V. 1975. On the Syntax ofBe-Sentences in Russian. Siavica Publishers:Cambridge.

. 1983. On 'Defirùteness' in Bulgarian, English, and Russian. American--- Contributions to the IXInternational Congress ofSlavists. Slavica

Publishers: Cambridge.

de Freitas, L. to appear. Case-Licensing in Specifier ofNegP: The Case ofRussian. Proceedings ofthe SecondAnnual Workshop on Lexical­Semantie Relations, February 1992, McGilI UlÙversity, Montréal.

253

•

•

. in press. Case-Licensing and (In)definiteness in Russian and Welsh.--- Proceedings of the Colloque Internationale sur La Négation, November

1992, Université de Paris X, Nanterre.

de Freitas, 1. and M. Noonan 1993. Head Movement, Agreement and Negation inWelsh Relatives. In L. Dobrin (ed.) Proceedings ofthe 27th RegionalMeeting ofthe Chicago Linguistics Society: Parasession on Negation.University ofChicago Press: Chicago.

de Freitas, 1., M. Noonan and B. Shaer. 1991. Ignore-a as a Principle ofEconomy. Ms., McGil1 University.

Di Sciul1o, AM. and E. Williams 1987. On the Definition ofWard, MIT Press:Cambridge, Mass.

Enç, M. 1991. The Semantics ofSpecificity. Linguistic Inquiry 22, 1-25.

Evans, O.S. 1964. A GrammarofMiddle Welsh. Institute for Advanced Studies:Dublin.

Fife, J. 1986. Additional Faets about Welsh VPs. Journal ofLinguistics 22, 179­186.

Haegeman, L. 1991. Negative Concord, Negative Heads. Ms. Université deGenève.

Haegeman, L. and R Zanuttini 1991. Negative Heads and the Neg Criterion. TheLinguistie Review 8, 233-251.

Haïk, 1. 1981. On Clitic en in French. Journal ofLinguistic Research 2.

Harlow, S. 1981. Government and Relativization in Celtic. In F. Heny (ed.)Binding andFiItering. Croom Helm: London.

__.. 1983. CelticRelatives. YorkPapersin Linguistics 10, 72-121.

__.. 1989. The Syntax ofWelsh Soft Mutation. Natural Language andLinguistie Theory7, 289-316.

Heim, 1. 1982. The Semantics ofDefinite andIndefinite Noun Phrases. Ph.D.Dissertation, University ofMassachusetts, Amherst.

Hendrick, R 1988. Anaphora in Cellie and Universal Grammar. KluwerAcademie Publishers: Dordrecht.

Hirschbüh1er, P. and M. Labelle 1992. L'évolution des propositions infinitivesnégatives en 1hmçais. Ms., Université d'Ottawa et Université du Québec àMontréal.

Horn, LoR 1989. A NaturaI History afNegalion. The University ofChicago Press:Chicago.

2S4

•

•

Iatridou, S. 1990. About Agr(P). LinguisticInquiry 21, 551-577.

Jespersen, O. 1917. Negation in English and Other Languages. AF. Host:Copenhagen.

Johnson, K. 1990. Object Positions. Ms., University ofWisconsin, Madison.

Jones, D.G. 1988. literaI)' WeIsh. In M.J. BalI (ed.) The Use ofWelsh: AContribution to Sociolinguistics. Muitilingual Matters Ltd: CIevedon.

Jones, M. and AR Thomas 1977. nIe Welsh Language: Studies in its Syntax andSemantics. University ofWales Press: Cardiff.

Kayne, R 1984. Connectedness andBinary Branching. Foris: Dordrecht.

___" 1991. Romance Clitics, Verb-Movement, and PRO. Linguistic Inquiry 22,647-686.

Kitagawa, Y. 1986. Subjects in Japanese andEnglish. Ph.D. Dissertation,University ofMassachusetts, Amherst.

KIima, E. 1964. Negation in English. In J. Fodor and J. Katz (eds.) The StructureofLanguage. Prentice-Hall: Engiewood CIiffs, N.J.

Koopman, H. and D. Sportiche. 1988. Subjects. Ms., UCLA

Laka, 1. 1989. Constraints on Sentential Negation: The Case ofBasque. MITWorking Papers in Linguistics 10, 199-216.

___.1990. Negation in Syntax: On the Nature ofFunctionaI Categories andProjections. Ph.D. Dissertation, MIT.

Larson, R 1987. On the Double-Object Construction. Linguistic Inquiry 19.3,335-391.

__.1990. Double Objects Revisited: Reply to Jackendoff. Linguistic Inquiry21,589-632

Lasnik, H. 1990. Case and Expletives: Notes Towards a Parametric Account. Ms.University ofConnecticut.

Li, Y. 1990. XO-Binding and Verb-Incorporation. Linguistic Inquiry 21,399-426.

Liberman, M. 1974. On Conditioning the Rule ofSubject AUX Inversion. NELS5,77-91.

Libert, A 1992. On the Distinction between Syntactic and Sernantic Case. Ph.D.Dissertation, McGil! University.

Macana, P. 1975. Notes on the Affixed Pronouns in Welsh. Studio CelticaX, 318­325.

2SS

•

•

•

___.. 1978. Notes on the Abnormal Sentence. Studia CeItiea XIll, 174-187.

Mahajan, A. 1990. The A/A'Distinction andMO~'emelltTheory. Ph.D.Dissertation, MIT.

May, R 1977. The Grammar ofQuantification. Ph.D. Dissertation, MIT.

McCloskey, J. 1978. A Fragment ofa Grammar ofModem Irish. Texas LinguistieForum 12. University ofTexas, Austin.

. 1983. A VP in a VSO Language. In Gazdar, G. E Klein l1Ild G.K. Pullum---' (eds.) Order, Concord & Constituency. Foris Publications: Dordrecht.

___. 1990. Resumptive Pronouns, A-Bar Binding and Levels ofRepresentationin Irish. In R Hendrick (ed.) The Syntax and Semantics ofthe ModemCeltie Languages. Academie Press: New York.

McCloskey, J. and K. Hale, 1984. On the Syntax ofPerson-Number Intlection inModem Irish. Natural Language and Linguistie Theory 1, 487-535.

Milsark, G. 1974. Existential Sentences in English. Ph.D. Dissertation, MIT.

Moritz, L. 1989. Aperçu de la syntaxe de la négation enfrançais. Mémoire deLicense, Université de Genève.

Morris-Jones, J. 1931. A Welsh Grammar, Historieal and Comparative. ClarendonPress: Oxford.

Neidle, C. 1988. The Role ofCase in Russian Syntax. Kluwer AcademiePublishers: Dordrecht.

Noonan, M. 1989. Operator Licencing and the Case ofFrench Interrogatives.Proceedings ofWCCFL 8.

__,.1991. Case at S-strueture in Irish: An analysis of Ta. Ms., McGiIlUniversity.

_---'.1992. Case and Syntactic Geometry. Ph.D. Dissertation, McGiIlUniversity.

Cuhalla, J. 1990. Sentential Ne~f.;jn, Re1ativizOO Minimality and the AspeetualStatus ofAuxiIÏl:.;es. The Linguistic Review 7, 183-231.

Payne, J.R 1985. Negation. In T. Shopen (00.) Language Typology and SyntacticDescriptions: Clause Structure. Cambridge University Press: Cambridge.

Pearce, E. 1993. Tense and Negation: Competing Analyses in Middle French. In L.Dobrin (00.) Parasession on Negation: Proceedings ofthe 27th RegionalMeeting ofthe Chicago Linguistics Society.

Pesetsky, D. 1982. Paths and Categories. Ph.D. Dissertation, MIT.

256

•

•

•

Pilch, H. 1971. A Syntaetic Study ofColloquial Welsh. Studia Celtica VI, 138­157.

Pollock, J.-Y. 1989. Verb Movement, Universal Granunar, and the Structure ofIP.Linguistic Inquiry 20, 365-424.

Press, Ian. 1986. A Grammar ofModem Breton. Mouton de Gruyter: New York.

Progovaç, L. 1988. A Binding Approach to Po/arity Sensitivity. Ph.D.Dissertation, USC, Los Angeles.

___.. 1991. Subjunctive: Transparency ofFunctiona/ Categories. Ms., WayneState University.

___.. 1992. Subjunetive: Negative Polarity and Long-Distance Reflexives. Ms..Wayne State University.

Reuland, E.J. and AG.B. ter Meulen (eds.) 1989. The Representation of(In)definiteness. MIT Press: Cambridge, Mass.

Riemsdijk, H. van. 1978. A Case Study in Syntactic Markedness. Foris: Dordrecht.

Rizzi, L. 1986. Null Objects in Italian and the Theory ofpro. Linguistic Inquiry17, 507-557.

___.. 1990. Re/ativi::edMinima/ity. MIT Press, Cambridge, MA.

__~. 1991. Residual Verb Second and the Wh-Criterion. ms. Université deGenève.

Rizzi, L. and 1. Roberts, 1989. Complex Inversion in French. Probus l, 1-30.

Rochette, A 1988. Semantic and Syntactic Aspects ofRomance Sententia/Comp/ementation. Ph.D. Dissertation, MIT.

Ross, J.R. 1983. Inner Islands. Ms., MIT.

Rouveret, A 1990. X-bar theory, Minimality, and Barrierhood in Welsh. In R.Hendrick (ed.), Syntax andSemantics 23: The Syntax ofthe ModemCe/tic Languages. Academie Press: New York.

___" 1991. Functional Categories and Agreement. The Linguistic Review 8,353-387.

Sadler, L. 1988. We/sh Syntax: A Govemment-BindingApproach. Croom He1m:London.

Sadock, J. and A Zwicky 1985. Speech Act Distinctions in Syntax. In T. Shopen(ed.) Language Typo/ogy andSyntactic Description 1: Clause Structure.Cambridge University Press: Cambridge.

257

•

•

Safir, K. 1982. Syntaetie ChaÎlIS and the Dejil/itel/ess Effeet. Ph.D. Dissertation,MIT, Cambridge, Mass.

SelIs, P. 1983. Relative Clauses in Irish and Welsh, York Papas il/ Lil/guisties 10,159-172.

___.. 1984. Syntax and Semanties ofResumptive Prol/oulIS. Ph.D.Dissertation, University ofMassachusetts, Amherst.

Slùonsh.')', U. 1992. Resumptive Pronouns as a Last Resort. Lil/guistie II/quiry 23,443-468.

Skerrelt, R.AQ. (1971) Statement, Command, Question, and Wish. Studia CelrieaVI, 158-162.

Sportiche, D. 1988. ConditiollS 01/ Si/el/t Categories. Ms., UCLA.

___. 1990. Movement, Agreement and Case. Ms., UCLA.

___. 1992. Clitics, Voice, and SpeclHead Licensing. GLOW Newsletter.

Sproat, R. 1985. Welsh Syntax and VSO Structure. NalUral Language andLinguistie Theory 3, 173-216.

Stump, G. T. 1989. Further Remarks on Breton Agreement. Natural Languageand Linguistie Theory 7, 429-471.

TalIennan, M. O. 1983. Island Constraints in Welsh. York Papers in Linguisties10, 197-204.

___.. 1990a. VSO Word Order and Consonant Mutation in Welsh. Linguisties28,389-416.

___.. 1990b. Relativization Strategies: NP Accessibility in Welsh. Journal ofLinguistics 26, 291-314.

___.. 1991. The Directionality ofthe Head Subcategorization in Welsh. In 1. Fifeand E. Poppe (eds.) Current Issues in Linguistie Theory 83: Studies inBrythonie Word Order.

Taraldsen, T. 1992. SubjectJVerb Agreement and Word Order in Celtic andRomance. NELS 23.

Timberlake, A 1975. Hierarchies in the Genitive ofNegation. Slavie andEastEuropean Journal 19, 123-38.

Travis, L. 1984. ParametersandEffeetsofWordOrder Variation. Ph.D.Dissertation, MIT.

_----'. 1991. Inner Aspect. NELS 22.

258

•

•

•

Valois, D. and F. Dupuis (1989). On the Status ofVerbal TrJces in French: theCase ofStylistic Inversion. Ms. University ofCalifomia (Los Angeles) &Université de Québec à Montréal.

Williams, S. 1. 1980. A Welsh Grammar. University ofWales Press: Cardiff.

Zanuttini, R. 1989. The Structure ofNegative Clauses in Romance. Ms., MIT,Cambridge, Mass.

___" 1990. On the Relevance ofTense for Sentential Negation, Ms. Universitéde Genève.

___.. 1991. Syntactic Properties ofSententiaI Negation: A Comparative StudyofRomance Languages. Ph.D. Dissertation, University ofPennsylvania.

Zwicky, A. 1984. WeIsh Soft Mutation and the Case ofObject NPs. Proceedingsofthe Chicago Linguistics Society 20,387-402.

259