Sentential Complementation and the Lexicon: Studies in Honour of Wim de Geest 9783110878479, 9783110131215

Citation preview

Sentential Complementation and the Lexicon Studies in Honour of Wim de Geest

Linguistic Models The publications in this series tackle crucial problems, both empirical and conceptual, within the context of progressive research programs. In particular Linguistic Models will address the development of formal methods in the study of language with special reference to the interaction of grammatical components. Series Editors: Teun Hoekstra Harry van der Hulst

Other books in this series: 1 Michael Moortgat, Harry van der Hulst and Teun Hoekstra (eds.) The Scope of Lexical Rules 2 Harry van der Hulst and Norval Smith (eds.) The Structure of Phonological Representations. Part I 3 Harry van der Hulst and Norval Smith (eds.) The Structure of Phonological Representations. Part II 4 Gerald Gazdar, Ewan Klein and Geoffrey K. Pullum (eds.) Order, Concord and Constituency 5 W. de Geest and Y. Putseys (eds.) Sentential Complementation 6 Teun Hoekstra Transitivity, Grammatical Relations in Government-Binding Theory 7 Harry van der Hulst and Norval Smith (eds.) Advances in Nonlinear Phonology 8 Harry van der Hulst Syllable Structure and Stress in Dutch 9 Hans Bennis Gaps and Dummies 10 Ian G. Roberts The Representation of Implicit and Dethematized Subjects 11 Harry van der Hulst and Norval Smith (eds.) Autosegmental Studies on Pitch Accent

Sentential Complementation and the Lexicon Studies in Honour of Wim de Geest Dany Jaspers Wim Klooster Yvan Putseys Pieter Seuren (eds.)

¥

1989 FORIS PUBLICATIONS Dordrecht - Holland/Providence - U.S.A.

Published by: Foris Publications Holland P.O. Box 509 3300 AM Dordrecht, The Netherlands Distributor for the U.S.A. and Canada: Foris Publications USA, Inc. P.O. Box 5904 Providence RI02903 U.S.A. CIP-DATA

ISBN 90 6765 415 9 ® 1989 Foris Publications - Dordrecht No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission from the copyright owner. Printed in the Netherlands by ICG Printing, Dordrecht.

I

Preface Wim

de

Geest

has

complementary ways.

left his mark

on

linguistics,

in two different but

His now classical work on sentential complementation,

written in the tradition of early transformational grammar, has become a favourite among grammarians: Kaatje's singing has enlivened, and still does, many a linguist's hour.

But also, and in no lesser way, has Wim's personality

impressed itself on his colleagues and students.

Rarely have we known a

friendlier, a more understanding, or more hospitable person in our professional circles.

His total intellectual and personal integrity, together with his patience

and his kindness, have cemented lifelong loyalties.

It is for these reasons that

we felt we should manifest our loyalty in the shape of a Festschrift at Wim's sixtieth birthday. We wish to thank all the contributors to this volume for their generous and prompt response to our call for papers; the UFSAL Research and Development Council for their financial support; Sally Grobben and Bernadette Leon from the UFSAL Computer Centre and Veerle Uyttersprot for their technical assistance; Vivian Boulpaep, Marleen De Roover and Louisa Mathijs from the UFSAL typing pool for preparing a camera-ready manuscript with imperturbable patience and efficiency; Harry Van Paesschen for laying out the tables, charts, and diagrams with painstaking accuracy; Richard Baeyens of the UFSAL library staff for delving into Wim's bibliographical past with professional discretion;

Teun

Hoekstra, Harry van der Hulst and Michael Moortgat for their readiness to include this volume in their Linguistic Models Series; Henk J. La Porte for publishing the book in time for the celebration of Wim's sixtieth birthday; and last but not least our colleague and Dean Jozef D. Janssens for coordinating and supervising the whole enterprise with his customary zeal and enthusiasm.

Dany Jaspers Wim Klooster Yvan Putseys Pieter Seuren

III

List of Publications by Wim de Geest Author

(1960) "Ons moedertaalonderwijs op nieuwe, concentrische banen?," Doceo, 19591960, X, 71:8-10. (1960) "[Bespreking van] H.F.A. van der Lübbe, Woordvolgorde in het Nederlands: een synchrone, structurele beschouwing," Spieghel Historiael van de Bond van Gentse Germanisten, 1959-1960, II, 2:38-41. (1960) "Foniek of fonologie van het Nederlands," Wetenschappelijke XX, 9:423-428.

Tijdingen

(1963) "De tijd dringt, ook voor ons moedertaalonderwijs," Kultuurleven XXX, 3:192-202. (1966) "Een fonologie van het Nederlands," Le Langage et l'Homme, 1966-1967, I, 1:39-47. (1967) "Kennismaking met de transformationeel-generatieve grammatica," Le Langage et l'Homme, 1967, 3:53-62. (1967) "Een transformationeel-generatieve beschrijving van Nederlandse negatieve zinnen," Le Langage et l'Homme, 1967, 4:49-62. (1967) "Met de integrale methode van moedertaal aan het werk," Stem uit Nieuwland XXXVII, 2:160-163. (1967) "De huidige stand van de transformationeel-generatieve beschrijving van het Nederlands," in Droste, F.G., W.P.F. de Geest (1967) Linguistiek, Postuniversitair Centrum Limburg, Hasselt, Limburgse Akademische Bibliotheek - Deel Linguistiek, nr. 1:35-81. (1968) "Syntaxis voor vandaag en morgen," Wetenschappelijke 1:13-26.

Tijdingen XXVII

(1969) "Publikaties van het Instituut voor Toegepaste Linguistiek (K.U.L.)," Linguiste, 1968-1969, 6:22-24. (1970, 19742) "De Basiscomponent," in Dirven, R. (ed.) (1970) Kennismaking met de T.G.G., Acco, Leuven:60-72. (1970, 19742) "De samengestelde zin," in Dirven, R. (ed.) (1970) Kennismaking met de T.G.G., Acco, Leuven: 1-9. (1970) "[Bespreking van] W.G. Klooster, H J . Verkuyl en J.H.J. Luif, Inleiding tot de syntaxis. Praktische zinsleer van het Nederlands," Le Linguiste - De Taalkundige, 3-4:22.

IV

(1970) "Infinitiefconstructies bij verba sentiendi," Studia Neerlandica 3:33-59; ook in Hulshof, H. (1975) Transfomiationeel-generatieve grammatica in artikelen, Groningen:343-369. (1971) "[Besprekingen T.G.G., van] Jos Nivette, Grondbegrippen van Generatieve Grammatica, (188-189); Eddy Roulet, Syntaxe de la proposition nucléaire en français parlé, étude tagmêmique et transformationeile, (190); Raymond Lamérand, Syntaxe transformationeile des propositions hypothétiques du français parlé, (190-191)," Wetenschappelijke Tijdingen XXX, 3:kol. 188-189. (1972) "De historische dimensie in de huidige linguistiek," Handelingen van het 28e Vlaams Filologencongres 1971, Leuven: 189-196. (1972) "Dat Barbertje moet hangen (complete zinnen)," in Daems, F. (ed.) (1972) De Taal van de mens, W.G.G.T.-cahiers, nr. 2, Plantijn, Deurne:51-61. (1972) "Olie op de golven van de spellingstorm," Spectator XXVII, juli 1, 6-7. (1973) Complémentaire Constructies bij Verba Sentiendi in het Utrechtse Herdrukken 5, H.E.S. Publishers, Utrecht, 188 pp.

Nederlands,

(1973) "[Bespreking van] F.G. Droste, Algemene Taalwetenschap en Transformationele Grammatica," Le Langage et L'Homme, 1973, XXI:83-84. (1973) "Recent Advances in the Theory of Complementation and their Relevance for Contrastive Linguistics," Linguistische Perspektiven. Referaten des VII Linguistischen Kolloquiums, Nijmegen, 26-30 September 1972. Herausgegeben von A.P. ten Cate & P. Jordens Eds., Niemeyer, Tübingen:230-241. (1974) "De plaats van de TGG in de ontwikkeling van de Algemene Taalkunde," Recyclagecursus Docebo, UFSAL-publikatie, 7 pp. (1974) "De transformationeel-generatieve grammatica Recyclagecursus Docebo, UFSAL-publikatie, 34 pp.

I,

II

en

III,"

(1974) "De samengestelde zin," Recyclagecursus Docebo, UFSAL-publikatie, 15 pp. (1974) "Over de zogenaamde transformationele onrust in de grammatica en kritiek die geen hout snijdt," Nova et Vetera LI, 5:374-381. (1975) "Aspecten van de generatieve UFSAL-publikatie, 75 pp.

semantiek," Recyclagecursus

Docebo,

(1975) "On Lexical Islands, Predicate Raising and Dutch Infinitive Pro Participio Constructions," in S. de Vriendt, J. Dierickx and M. Wilmet (eds.) Grammaire Générative et Psychomécanique du Langage. Actes du Colloque organisé par le Centre d'Etudes Linguistiques et Littéraires de la Vrije Universiteit Brüssel, Bruxelles, 29-31 mai 1974, AIMAV, Bruxelles: 111-122. (1977) "Linguistics and the Language Teacher," Saint Louis University Research Journal, 1977, VIII, 3-4:511-515.

V

(1978) "The Relevance of Relational Grammar to Contrastive Linguistics," Seames Regional Language Centre Paper -Thirteenth Regional Seminar, Singapore, 17-21 April 1978, 10 pp.; ook in Saint Louis Research Journal, 1978, X, 34:386-411. (1979) Linguistische Theorievorming in de jaren Docebo, UFSAL-publikatie, 31 pp.

zeventig,

Recyclagecursus

(1979) "Naar een beter begrip van kale infinitiefconstructies," Handelingen XXXIII der Koninklijke Zuidnederlandse Maatschappij voor Taal en Letterkunde en Geschiedenis:53-12\ ook in GLOT, 1980, III, 1:29-46. (1980) "Rouwhulde Bernard Kemp - B.F. van Vlierden tijdens de uitvaart op 7 november 1980," Mededelingen van de Vereniging van Vlaamse Letterkundigen 105 bis; ook in Nieuws van de UFSAL VII-3, 6. (1981) "Ten Geleide," in Geest, W. de, R. Dirven, Y. Putseys (eds.) (1981) Twintig Facetten van de Taalwetenschap, Taal en Communicatie. Kennismaking met de Linguistiek nr. 5, Acco, Leuven:301-307. (1982) "De Saint Louis University in Baguio, Vlaams baken in Noord-Luzon (Filippijnen)," Academische Tijdingen XIII, nr. 1:26-27. (1984) "Grammatica als erfdeel," Ons Erfdeel, jaargang 27, 5/84:727-737. (1986) "Groene Ideeèn over Nederlandse Grammatica (1)," Nederlands Van Nu XXXIV, nr. 4:101-107. (1986) "Groene Ideeen over Nederlandse Grammatica (2)," Nederlands Van Nu XXXIV, nr.5:133-141. (1987) "Ten Geleide," in Geest, W. de, F. Debrabandere (eds.) (1987) Het Nederlands in Europa. (G)één van de Negen? Handelingen van het Nederlands Taalcongres - Vereniging Algemeen Nederlands, Brüssel 22 november 1986, Vereniging Algemeen Nederlands, Brussel:5-6. (1988) "Passive Bare Infinitival Complements in Dutch," in Liber Gratulatorius Prof. Dr. F. G. Droste

VI

Co-Author and/or Editor

Geest, W. de, H. Verstuyf (1961) De moedertaalvorming in het middelbaar en technisch onderwijs van de lagere graad, De Procure, Brüssel, 80 pp.Pedagogische Brochurereeks van de Sint-Thomasnormaalscholen. Droste, F.G., W.P.F. de Geest (1967) Linguistiek, Postuniversitair Centrum Limburg, Hasselt, Limburgse Akademische Bibliotheek - Deel Linguistiek, nr. 1, 87 pp. Geest, W. de, R. Dirven, G. Lurquin, Y. Putseys (1967) Initiation ä la grammaire generative et transformationeile, publikatie van het Instituut Marie Haps, Brüssel, 147 pp. Geest, W. de, R. Dirven, Y. Putseys (1969-1970) "Reflexive inchoatives revisited," Le Langage et l'Homme, I in nr. 11 (1969):31-33; II in nr. 12 (1970):53-64; III in nr. 13 (1970):64-75. Bousset, H., W. de Geest, J. Janssens (1972-1973) Taalgroei 4, De Procure, Brüssel, 2 vols. Bousset, H., P. Deblieck, W. de Geest (1973) Taalgroei 1, De Procure, Brüssel, 281 pp. Geest, W. de, R. Dirven (1973) "Nr 414 Grammatica," in Reinaert Systematische Encyclopedie 13, Taalkunde, D.A.P. Reinaert Uitgaven, Brussel:34-61. Bousset, H., P. Deblieck, W. de Geest (1974) Taalgroei 2, De Procure, Brüssel, 3 vols. Deblieck, P., W. de Geest, H. Bousset (1974) Taalgroei 1: handleiding bij werkboek en oefenboek, De Procure, Brüssel, 30 pp. Bousset, H., P. Deblieck, W. de Geest (1975) Taalgroei 3, De Procure, Brüssel, 2 vols. De Schutter, F., H. Bousset, W. de Geest (1975-1976) Taalgroei 1 en 3: bloemlezing, De Procure, Brüssel, 2 vols. Geest, W. de, F. De Schutter, H. Bousset (1976) Taalgroei : bloemlezing, De Procure, Brüssel. Geest, W. de, J. Janssens, H. Bousset (1977) Taalgroei. 6, De Procure, Brüssel, 222 pp. Geest, W. de, F. De Schutter, H. Bousset (1978) Taalgroei 5, De Procure, Brüssel, 288 pp. Geest, W. de, J. Janssens (1979) Handleiding voor de Leraar Nederlands van de Lagere Cyclus, De Procure, Brüssel.

VII

Geest, W. de, R. Dirven, Y. Putseys (eds.) (1981) Twintig Facetten van de Taalwetenschap, Taal en Communicatie. Kennismaking met de Linguistiek, nr. 5, Acco, Leuven. Geest, W. de, Y. Putseys (1984) Sentential Complementation. Proceedings of the International Conference held at UFSAL, Brussels, June 1983, Linguistic Models Series 5, Foris Publications, Dordrecht, 280 pp. Geest, W. de, H. de Jonghe (1985) Nederlands, je taal. Een overzichtelijke spraakkunst, Van In, Lier, 312 pp. Geest, W. de, D. Jaspers (1985) "De Regeer- en Bindtheorie," in Droste, F.G., (ed.) (1985), Stromingen in de Hedendaagse Linguistiek, Universitaire Pers, Leuven/Van Gorcum, Assen-Maastricht:26-57. Geest, W. de, F. Debrabandere (eds.) (1987) Het Nederlands in Europa. (G)één van de Negen? Handelingen van het Nederlands Taalcongres - Vereniging Algemeen Nederlands, Brüssel 22 november 1986, Vereniging Algemeen Nederlands, Brüssel, 77pp. Geest, W. de, D. Jaspers (1987) "Teoria del Governo e del Legamento," in Droste, F.G., Y. d'Hulst, (eds.) (1987), Nuove Prospettive Nella Linguistica Contemporanea, CLESP Padova:35-70. Geest, W. de, D. Jaspers (forthcoming) "Government and Binding Theory," in Droste, F.G., (forthcoming) (ed.), Mainstreams in Linguistics, John Benjamins, Amsterdam/Philadelphia.

IX

Table of Contents

Preface

I

List of Publications by Wim de Geest

III

Table of Contents

IX

Contributions: Bayer, Jozef : A Note on Alleged PRO-Government

1

Bennis, Hans and Teun Hoekstra : Why Kaatje was not heard sing a song

21

Besten, Hans den and Jean Rutten : On Verb Raising, Extraposition and Free Word Order in Dutch

41

Beukema, Frits and Marcel den Dikken : The Position of the Infinitival Marker in the Germanic Languages

57

Cinque, Guglielmo : On Embedded Verb Second Clauses and Ergativity in German

77

Declerck, Renaat : Tense in Complement Clauses

97

Dirven, René : A Cognitive Perspective on Complementation

113

Droste, Flip G. : On the So-called Innateness Hypothesis and the Lexicon

141

Grange, Corinne and Liliane Haegeman: Subordinate Clauses: Adjuncts or Arguments - The Status of het in Dutch

155

Haider, Hubert : Against Raising

173

Hale, Ken : The Causative Construction in Miskitu

189

X

Hantson, André : The Complementizer as

207

Hellan, Lars : On Lexical Compositionality

221

Jaspers, Dany : A Head Position for Dutch Clitics or: Wilma, Wim and Wackernagel

241

Klooster, Wim G. : Unpreposable Raising Complements

253

Koster, Jan : Left-right Asymmetries in the Dutch Complementizer System

271

Lewandowska-Tomaszczyk, Barbara : The Semantics and Discourse Function of Negative Verbs and their Complements

283

Longobardi, Giuseppe : Parametric Syntax and the Supposed Coordinating Function of the Gothic Complementizer ei

295

Platzack, Christer : The Swedish Supine : An Active Verb Form or the Non-agreeing Form of the Past Participle ?

305

Putseys, Yvan : Sentential Complementation and the Meanings of such

321

Riemsdijk, Henk van : Swiss Relatives

343

Seuren, Pieter A.M. : A Problem in English Subject Complementation

355

Taraldsen, Knut Tarald : Word Order in Tensed Clausal Complements

377

Tasmowski, Liliane : A Look at sembler + Infinitive from Different Angles

403

Wyngaerd, Guido Vanden : Verb Projection Raising and the Status of Infinitival Complements

423

List of References

439

A Note on Alleged PRO-Government Jozef Bayer

1. The Chomsky/Aoun/Sportiche Theory of Government Chomsky (1981) proposes a definition of government which modifies a proposal that was later published in Aoun and Sportiche (1983). The definition by Aoun and Sportiche, as quoted by Chomsky (1981), is as follows: (1) The Aoun/Sportiche definition of government LGB, (6), p.164 £Xei C NP clitic ]

c

NP clitic] c I P ] C ] C P

The definition of deletion under recoverability can be maintained under such an approach. 11. In a sentence like this, the sequence wo em contains avowel clash which is phonologically broken up by the insertion of an epenthetical n. Here and elsewhere such phonological effects have been disregarded. 12. a . Chomsky (1981:65). 13. The letter "e" is to be interpreted as "schwa" throughout this schema. 14. Fukui's (1986) proposal that languages like Japanese and Chinese lack functional projections and hence do not have a landing site for wh-movement has a rather brute force character to it as well. Furthermore, the proposal is controversial among those working on Japanese. Honda (1988), for example, argues that functional projections do play a role in Japanese syntax. 15. It turns out that there may be a potential answer to this question along the line of reasoning adopted above in dealing with questions ii. and iii. which may offer a more principled answer than the stipulative one offered here. The proposal above was that the Avoid Pronoun Principle is a global condition. What exactly does that mean? In other words, which aspects of a derivation must be taken into account? Suppose, now, that the choice of the correlative word (whword or resumptive pronoun) is also subject to the principle, even though each choice initiates a different derivation, one via wh-movement, the other via cliticization and deletion. One might say, perhaps, that this interpretation of the principle is a transderivational one. At any rate, on this interpretation the choice in favor of a resumptive pronoun is forced. Unfortunately, this approach is not entirely waterproof since it fails to apply to those cases in which cliticization followed by deletion is unavailable. One possibility would be to say that the Least Effort Principle of Chomsky (1988) discussed below is responsible for this: in languages that have a resumptive pronoun strategy, the resumptive pronoun that remains in situ must always be chosen over the wh-word, which requires the additional derivational step of movement. What remains unclear, however, is why the resumptive pronoun strategy is not available in all languages.

A Problem in English Subject Complementation Pieter A.M. Seuren Introduction In this paper I present a particular problem for the grammatical description and explanation of subject complementation in English. This problem has so far not found a solution in any of the current theories of grammar, in particular the theory of Government and Binding. I also present a principled solution of this problem, in terms of the theory of Semantic Syntax. The problem, and the essence of its solution, were presented by me in a talk at the MIT Linguistics Department almost twenty years ago. On that occasion, the audience, or at least those who set the tone for the audience, were unwilling to accept the reality of the problem, and thus even less willing to consider the solution proposed. One reason for this negative attitude, I felt at the time, was probably the fact that both the problem and the solution offered were awkward for the theoretical orientation which was then beginning to be developed, in particular X-bar theory. This awkwardness has remained: the problem has so far proved refractory in terms of MIT-based theory, and the solution is at loggerheads with X-bar theory, now as then.

1. The problem The problem at hand is illustrated by the different grammatical behaviour of the adjective likely and the verb seem, as can be seen from the following examples: (1)a.

Tom is likely to be ill.

b.

It is likely that Tom is ill.

c.

That Tom is ill is likely.

(2)a. b.

I expect Tom to be likely to be ill. I expect it to be likely that Tom is ill.

c. * I expect that Tom is ill to be likely. d.

That Tom is ill I expect to be likely.

356 (3)a. b. c. (4)a. b. c. (5)a. b. c. d. (6)a. b. c.

Tom is expected to be likely to be ill. It is expected to be likely that Tom is ill. That Tom is ill is expected to be likely. Tom seems to be ill. It seems that Tom is ill. * That Tom is ill seems. I expect Tom to seem to be ill. I expect it to seem that Tom is ill. * I expect that Tom is ill to seem. * That Tom is ill I expect to seem. Tom is expected to seem to be ill. It is expected to seem that Tom is ill. * That Tom is ill is expected to seem.

When we set off the sentences (l)-(3), which have the predicate likely, against (4)-(6), which have the predicate seem, we notice that the (a) and (b) cases behave similarly, as do (2c) and (5c), but while (lc), (2d), and (3c) are grammatical, the corresponding (4c), (5d), and (6c) are not. The question is: why not? Or more precisely: what can account for the fact that adjectives like likely allow for Subject Raising (all (a)-cases), ¡/-Placement (IT, as in all (b)cases), and Subject Clause (SC, as in (lc), (2d), (3c)), except when the subject clause is internal (as in (2c)), while verbs like seem allow for Subject Raising and IT, but not for SC? It will not do to dismiss the problem by saying that these are merely stylistic, and not grammatical, differences (as leading voices said when I presented this problem at MIT in 1970), because "English does not like sentences with subject clauses to end in a verb" in this case seems. That this cannot be the answer appears from cases like (7c): (7)a. * Tom follows to be ill. b. It follows that Tom is ill. c. That Tom is ill follows. Follows is a verb, like seems, yet the grammatical status of (7c) is unimpeachable. (7a) however, is clearly ungrammatical, while (7b) is fine. Apparently, there are verbs, like follow, which do not take Subject Raising

357 (SR), but do take IT and SC. Other verbs of this class are fit, hurt, and the complex verbs stand to reason, make sense. And we might as well use this opportunity to note that there are also verbs that take SR, but neither IT nor SC, such as tend, start, begin, continue, cease, and others: (8)a.

Tom tends to be lucky.

b. * It tends that Tom is lucky. c. * That Tom is lucky tends. The class of verbs that behave like seem includes: appear, turn out, happen, be believed, be expected, be thought,

be rumoured,

and others. The class of

adjectives or past participles that behave like likely includes: certain, sure, said, known, believed, expected, and a few more. If the reader wonders why and how predicates like be expected, be believed, can behave both like seem and like likely, he is asked to wait a little till the verbs expect and believe are discussed. It will then become clear that these verbs fall into the seem-class

or the

likely-class according to their subcategorization for one or another type of complement clause, with semantic differences. All the predicates (verbs or adjectives) discussed so far have in common that they subcategorize for an embedded subject clause (subject-S), or for an embedded object clause (object-S) turned into subject-S under Passive. We notice, in addition, that whenever a i/iai-clause finds itself in sentence-initial position, as in (lc), (2d), (3c), (7c), it has the discourse property of being either topic or comment. We assume that a sentence has a topic-comment modulation when it is uttered as an answer to a (mostly implicit) WH-question: the topic takes up the questioned element, and the comment is the answer. More will be said about this in the following section.

2. The essentials of Semantic Syntax The theory of Semantic Syntax holds that every natural language sentence has, besides its surface structure, a Semantic Analysis (SA), which contains all relevant semantic information of the sentence, without ambiguity, and in weakly1 compositional form. SAs are cast in a logical language akin to modern restricted quantification theory. The grammar of the language consists in the

358 set of rules that define the systematic mapping procedure between an SA and its surface structure (SS). Since both SAs and SSs are linguistic objects, a grammar is essentially a set of transformations, in the established sense of transformations as mappings between sets of linguistic (tree) structures and sets of linguistic (tree) structures. The transformations are, of necessity, semantically invariant, or meaning-preserving. They are required by the fact that SAs, with their deeply layered multiple S-embeddings and their compositional structure, are suitable objects for semantic interpretation, while their corresponding SSs, with their "flat" trees and largely linear order, are more suitable for acoustic transmission. It is not unlikely that the functional requirement of rapid acoustic transmission of semantically complex propositions places certain "autonomous" restrictions on transformations, especially on those nearer the surface. Yet the overall tendency is for the transformational rules to eliminate S-embeddings and "flatten" the trees, while introducing a greater variety of word classes and constitutent categories than is found in SAs. The rules seem to be subject to a general, perhaps universal, set of constraints, which will be presented below. The basic formation rule for SAs is as follows: (9)

S

V- { ^ - ( N P ) - « ^ } )

That is, an S consists of a (semantic) predicate V, followed by one, two, or three arguments. The first argument, which is NP or S, is always the subject. The last argument, in case there is more than one, is again either NP or S, and is the object. When there are three arguments the middle argument, always NP, is the indirect object. All sorts of abstract semantic elements in sentences are analysed on SA-level as (abstract) predicates: quantifiers, tenses, prepositions, sentence adverbials, modalities, coordinating and subordinating conjunctions, etc. The normal procedure for non-lexical abstract predicates is Lowering: they are incorporated in certain, language-dependent, ways into their argument-S, in a mapping from SA to SS. We shall adhere to the common practice of formulating the rules top-down, i.e., from SA-level to SS-level. Non-abstract, i.e. lexical, semantic predicates end up as verbs, adjectives, or as nouns, adverbs or prepositions. Verbs, adjectives and nouns can occur as surface predicates. These are characterized by the fact that they allow for tensing: a

359 verb or adjective in a full clause, i.e. with finite verb, takes two tenses, t^ and t 2 ; a verb or adjective in an infinitival clause takes either one tense, t^, or no tense at all, depending on the subcategorization properties of the higher predicate. The standard tree structure for a full clause in the language of Semantic Analysis is thus: (10) V

I tl

Embedded clauses are of the form S" (i.e. with two tenses), ending up as a full clause with a finite verb form, or S' (i.e. with just one tense, t^), ending up, normally, as an infinitive with or without the tense auxiliary have (which takes over the infinitive form and turns the main verb into a past participle), or S° (i.e. without any tense at all), ending up, normally, as an infinitive, but never with the tense auxiliary have + PP (i.e. past participle). It depends on the subcategorization properties of the higher, commanding, verb whether there is an embedded S", S' or S°. The difference can be read from the presence of a full embedded clause (usually with complementizer), or the possibility of having have + PP for an infinitive. Examples will be given below. Defining the semantics of t^ and t 2 in a language is not an easy matter. In most languages the use of one or both of the tenses gives rise to specific presuppositions and possibly other complicating factors. We shall limit ourselves to what is seen as the universal, language-independent, semantics of the two tenses. The highest tense, tj,, is indexical or anaphoric: it refers either to the indexically given moment of utterance ("PRES"), or to some contextually defined past time ("PAST'), in which sense it can be said to be anaphorical. It can be paraphrased as "at this/that time". The second tense, t 2 , quantifying and relative. It contains an existential quantification over either events (aorist aspect) or durations (durative aspect), placing them at the same time as t j ("SIM") or prior to t^ ("PREC"). It can be paraphrased as, roughly, "there is/was an event/duration simultaneous with/prior to that time". Typically, PRES

360 + SIM gives a simple present tense, PAST + SIM gives a simple past, PRES + PREC gives a present perfect, and PAST + PREC gives a pluperfect. Accordingly, the morphological realization of PRES and PAST consists in the tense marking for the finite verb form (present or past tense, respectively). The morphological realization of SIM is zero, whereas that of PREC is have + PP in practically all cases (very occasionally English allows for be + PP, as in: It is gone, or in archaic uses like: The gods are departed). In the transformational cycle, both tenses are lowered (L) onto the leftmost lower V. The highest tense, t j , moreover, induces SR (Subject Raising). As the details of these processes are given, it will become clear that this assignment of rules to the tenses automatically accounts for the change of the VSO-pattern into the standard NP-VP (or SVO) pattern of English surface structures. Languages with surface VSO-pattern differ from the NP-VP languages only in that t^ does not induce SR. Before we can go into the transformational rule system , however, a few things must be said about constraints on trees, whether on SA-level or on SS-level or on any level in between. First, and this is central to the solution proposed here for the problem described in section 1, an embedded S can also be an NP, depending on the subcategorization properties of the embedding predicate. In the SA-tree there will then be an NP over S. The number of tenses is irrelevant: S may be an S", an S\ or an S°. It should be noted that X-bar theory excludes, as a matter of principle, any configuration NP-over-S. In that theory, trees are structured as expansions of major categories: an NP is an Nexpansion, and an S is a V-expansion. Mixture of the two is excluded on principle. We do not accept, however, X-bar theory, and we are thus free to use the NP-over-S configuration as an explanatory principle. A further requirement is the following. When the cycle reaches S", then the subject of S', its argument-S, must be an NP (whether nominal or sentential), and not just an S. The reason for this is easy to see: t^, the (abstract) predicate of S", induces the rule of S(ubject) R(aising), which, as we shall see, operates only on NP-subjects. This constraint thus applies neither at SA-level nor at SS-level, but in between, at the level where the S"-cycle starts to work. Given that no further rule will destroy the NP-status of the subject, it follows that in SS, too, the subject will always be an NP (nominal or sentential).

361 Then there are constraints to do with the topic-comment modulation often found in sentences in discourse. As is well-known, there is a great deal of unclarity in the literature about questions of topic, comment, focus, theme, and related notions. We limit ourselves to the pair "topic-comment", and, as was said at the end of section 1, we assume that in living discourse a new utterance is often intended as an answer to a question, usually not explicitly phrased, that has arisen or that the speaker expects will arise in the listener's mind. The part of the utterance that reflects the question will then remain unaccented, whereas the remainder, i.e. the answer to the (implicit) question, will be assigned a salient accentual structure. This structural division we call "topic-comment modulation". It is sometimes connected with the grammar of the language in question, in that certain structural positions are mandatory, or forbidden, for topics, or comments. Take the sentence: (11) That we could not win the war was obvious from the start. The most normal intonation will have accent on the predicate part "obvious from the start". Under this intonation the tfiaf-clause is topic. But the sentence can also be read with accent on the ito-clause, in which case that clause is comment, as when the sentence answers the (implicit) question "What was obvious from the start?". In any case, the i/zai-clause has to be either clearly topic or clearly comment: a neutral reading is not possible, unlike: (12) It was obvious from the start that we could not win this war. which does allow for a reading that remains neutral as to topic or comment. It seems useful to stipulate that only NPs can be [ + top]. Furthermore, it seems that f/titf-clauses which are

[+top]

have to be in sentence-initial

position,

whereas comment i/iai-clauses may, but do not have to be, in sentence-initial position.

It

may well be

that

the

same

infinitivals, as in: (13) Tame that horse is what I saw him do. But this question is not relevant to this paper.

constraint

holds

for

embedded

362 Now to the transformational system, most of which may be considered common knowledge, especially the transformational cycle. What does need comment here are the constraints on the cyclic rules that operate on SAs. Semantic Syntax has always been hesitant to come forward with precise proposals in this regard, mainly to avoid the risk of premature statements, which will then be subject to constant public revision. Rather than indulge in this practice, as was done in at least one school of thought, it seems wiser to hold back a little and keep the inevitable process of testing and reformulating confined to one's study and one's classes until some reasonable degree of consolidation has been reached. It now looks as if that moment has indeed come: we can now formulate some significant and generally applicable constraints on the transformational mapping system from SAs onto SSs with a sufficient degree of confidence. It must be noted that the constraints proposed for rules of grammar in other schools do not apply to the SA-SS mapping system, but have been presented in the framework of "autonomous" syntax, where semantically relevant mappings play no part. Although it is usually accepted, in those circles, that an SA-SS mapping system must be taken to be part of the description of a language, no attempts seem ever to have been made to formulate constraints for this mapping system, all the effort in this regard being directed towards what is seen as the autonomous rules of syntax. It hardly needs to be argued that such a limitation cannot be justified: whatever arguments hold for there being constraints in autonomous syntax must also hold for there being constraints in SA-SS mapping systems. We thus formulate the following general (possibly universal) properties of the cyclic rules of transformational syntax. (The properties listed below do not form the complete set of constraints needed for an adequate theory of Semantic Syntax. Yet they cover more ground than is strictly necessary for the question at hand.)

I. Lexical and structural rule induction Most cyclic T(ransformational)-rules are induced by the V on the appropriate Scycle. Each V is specified in the lexicon for the cyclic T-rules it induces. Lexical rule specifications are placed between angled brackets. Round brackets indicate that the rule is optional.

363 Other T-rules are induced, obligatorily or optionally, by specific tree-structure configurations. A case in point is zf-insertion (IT), as given below.

II. Raising A lexical rule, induced by a higher V, may raise a lower V or NP. When V is raised it attaches itself to (the left or right of) the next higher V (Predicate Raising or PR). Only subject-NPs can be raised (Subject Raising or SR). When a subject-NP is raised, it occupies the place of its own S, the remainder of which shifts one position to the right. An NP can be raised only if its own S is not NP, and is S\

III.

Lowering

A lexical rule, induced by V on some cycle, may lower that V into an argumentS. NP-Lowering does not occur in the cycle. When a V is lowered it usually changes category: it becomes a preposition, adverb, particle, affix, etc. Abstract predicates, such as in particular logical operators and tenses, are all lowered. Quantifiers are lowered onto their variables. Tenses are lowered onto the lower V. Other landing sites for Lowering are language-specific.

IV.

Deletion

A lexical rule, induced by V on some cycle C , may delete an NP in an argument-S n + ^. Only a subject-NP can be deleted (Secondary Subject Deletion, or SSD). The deletion is controlled by an NP-argument in either S n (vertical SSD), or the subject-S n + ^ when V of S n takes both a subject-S and an object-S (horizontal SSD).

364 V. Effects of Raising, Lowering or Deletion If an S loses its subject-NP through Raising or Deletion it is demoted to the lower category /S ( = VP). If an S loses its V through Raising then the S-node in question is pruned and all its non-V material is reattached to the next higher S-node, at the right of the material already there. If an S loses its V through Lowering the S-node stays, unless it is idle (Sover-S without branching).

VI. Specific properties of SVO (= NP-VP) languages In SVO-languages an S' or an S° always loses either its subject-NP or its V. In SVO-languages, not only are, as elsewhere, the tenses lowered onto the lower V, but, in addition, t^ induces SR. t^ thus takes < S R , L > , and t2 takes < L > .

VII. Specific property of Copula-languages Predicates (Vs in SA) that are lexically categorized as non-verbs (i.e. adjectives, nouns, prepositions) get the copula be as verb under tense, i.e. just t^ or t^ and t^ What interests us here most directly are the rules of S(ubject) R(aising) and IT. The Deletion rules of (vertical or horizontal) SSD are not directly relevant to the question at hand, yet we shall illustrate how they work. We will, moreover, show how the combination of SR and L(owering) for t^ automatically brings about the desired NP-VP structure from the VSO-structure found in SA-trees. SR: S,

S, V (NP)

t-.

S" . . __—•—1—-——_ V NP (X) V

V

(NP)

NP

/c

"

^ y ^ t,

(X) v

365 IT: If the subject is S" or j^pfS"] and [-top], insert

in the position of

the subject-S, which is moved to the far right. Obligatory for S"; optional for N p [S"].

V

NP

(X)

(NP) S"„

(Vertical) SSD:

V NPX

Sn + , V

NP

(X)

(Horizontal) SSD:

V

NPX

(X)

V

NP

(Y)

In the specification of SR, the raised subject-NP has been italicized. The bracketed NP may or may not be there. If it is, we have what used to be called "Subject-to-Object Raising": the subject of S ' n + ^ becomes the grammatical object of S n . If the bracketed NP is absent, then S ' n + ^ is the subject of S n and we have Subject-to-Subject Raising. (Clearly, the bracketed X stands for any other, irrelevant, material that might be there). The rule IT requires little comment. It turns, e.g., surprise - that he failedJohn into surprise - it - John - that he failed.

366 Vertical SSD (traditionally Equi-NP-Deletion) involves one embedded objectclause, whose subject is either codenotational with some controlling NP in the higher clause or bound by the same quantifier (e.g. Everyone wants to win). The controlling NP is printed in italics and subscripted with x to indicate its coreferential or covariable link with the lower subject. The controlling NP is most often the higher subject, but, as is well-known, it may also be an (indirect) object, depending on the SSD-inducing verb and/or other factors that have so far not been identified with sufficient clarity. As with SR, the embedded clause is degraded to the status of /S (=VP), due to the loss of its subject-NP. Horizontal SSD is less well-known, owing, mainly, to the fact that it requires a type of syntactic analysis which is abstract to a degree found only in Semantic Syntax and not or hardly in other syntactic theories. It occurs in, e.g.: (14) John died while eating soup. analysed as follows (leaving out tenses as well as other derivational details): (15)

while

V | die

/VPX | John

V NP | | eat x

NP | soup

while

V

NP

V

NP

die John

eat

soup

It is now easily demonstrated that by assigning the rule SR and L(owering) to tj, and only L to t^, the VSO-structure of SAs is automatically converted into the NP-VP structure required for English surface structures:

367 (16) Tense Processing

S"„ V

t,

V NP

(X)

ti



First, on the S^-cycle (the symbol "C" is used for cycles), t2 must be lowered into S^. Tenses lower onto the lower V, as has been said. "Onto" means that the V-node is copied above itself, and the lowered element is (left)-attached to the copy. This process is called (Left-)Adoption. The S above the lowered t2 is pruned because it is idle. (The superscript "o" is changed into "'" to indicate that the S in question has one tense). Then we pass to the S^-cycle. The first rule to be applied is SR (if L applied first, there would be no way to apply SR later). The result is that the subject-NP of S^ occupies the place of the same S^, which is degraded to /S^ (due to the loss of its subject-NP), and shifts one position to the right. Finally, t^ must be lowered. This process is entirely analogous to the lowering of t2: t j is left-adopted by the (complex) V of /S' n Note that S^ is not pruned, because it contains the raised NP besides the embedded /S^. The result is an S n (with two tenses) which has the structure NP-/S^ or NP-VP. Note also that the difference between VSO-languages and NP-VP (i.e. SVO) languages can now be accounted for simply by stipulating that, in a VSO-language, t^ does not induce SR but only L: S° will then incorporate both tenses but will not change its VSO-structure. It is a striking general feature of this theory as a whole that it has a strong "flattening" effect on SA-structures: the resulting SS-stuctures show nothing like the degree of S-embedding (which makes for vertically oriented structures) found in SAs. SSs are relatively "flat", and also contain many more different

368 categories than SAs, which are limited to the categories S, V, NP. Surface structures have, in addition, adverbs, adjectives, prepositions and what not. This change from vertical to horizontal orientation in tree structures is highly functional, since the categorially simple and strongly S-cyclic SA-structures lend themselves

naturally to semantic processing, whereas

the "flat" SSs lend

themselves naturally to the highly linear process of rapid acoustic transmission. Finally, it must be observed that this restricted system of cyclic transformation is neutral between a top-down processing or a bottom-up processing. The formulation given here is in terms of top-down processing, i.e. from SA to SS, as has been customary in the theory of Transformational Grammar from the beginning. It is easily seen, however, that nothing prevents a bottom-up formulation, as long as the elements in question are marked for the kind of process or rule they induce. What we have called a raised V or NP must then be lowered: a V from a complex V-island, and an NP into the position of subject of the / S that stands to its right. An element that is lowered in our terminology will then be raised to become the V of a new S to be set up above the S in question. A deleted subject-NP is recovered on the basis of the predicate (V) in question and the designated controlling NP. Likewise, all effects as mentioned under V. above are unambiguously reversible. It would seem that the requirement of two-way processing is in itself a strong limiting condition on any theory of syntax. After these preliminaries we can now proceed to the presentation of the solution to the problem stated in section 1.

3. The solution The solution is now relatively simple: it consists in a judicious and factually correct assignment of subcategorization properties and lexically induced rules. Table (17) gives the surface category, the subcategorization, and the rules for the predicates likely, seem, tend and follow. All four are unary predicates: they take just one argument, in all cases a sentential subject: S' or S", the latter with or without a dominating NP. The only rule assigned is SR, which, as has been said, can apply only when the subject to be raised stands under an S'. The oblique stroke indicates a choice of category:

369 (17)

SS-category

subcategorization

rules

(for object-position) expect

verb

NP/S'/S"/Np[S"]

believe

verb

NP/S'/S"/Np[S"]

think

verb

S' / S"

As regards follow, it is clear that it can only take a that-clause,

either in

subject position (obligatory if the subject clause is [ + top]) or extraposed and with the dummy subject it, as was shown in (7) above). Likely has more possibilities: apart from those available also to follow,

likely can take, in

addition, an S'-subject, in which case SR applies obligatorily. Hence the grammatically of (la). The derivation of (la) is demonstrated in (18): (18)

PRES

NP I 0 be + ill Tom

Co.. S R

Or, ^

PRES T

o m

y to be ill 0

Cço

SR

0

be + likely

to be ill

be + likely

So V I likely

/Si I Tom

°

y

V I 0

V I be + ill

PRES 0

be + likely

370 Note that in (18) the embedded /S'j (i.e. VP) gets the infinitival complementizer to. Whether this is inserted during or after the cycle is not clear: it is added here only for the sake of completeness. To is a common feature of embedded infinitivals in English: the rule is that it occurs; the exceptions are to be stated. /Ss embedded after modal auxiliaries {may, must, can, need, will, etc.) notably lack this to. But also cases like the following: (19)a. I saw him fall. b. That will make him see the truth. c. He let the man go. These are cases where the embedded /S lacks tense altogether, as appears from the ungrammatically of these sentences with the infinitive replaced by have followed by the past participle of the verb in question: * J saw him have fallen, etc. It seems that, in general, tenseless /Ss are less prone to taking to, although here, too, the rule is not absolute: allow is followed by a toinfinitival without tense. Help takes optional to in the sense of "assist": She helped me (to) cut the grass, but embeds its /S without to in the sense "be instrumental in", as in: I helped the boat sink. Clearly, an /S that incorporates two tenses cannot take to because to is an infinitival complementizer, and an /S with two tenses is not infinitival but finite. Let us now pass on to the derivation of (lb) It is likely that Tom is ill: (20) V

I

PRES

SIM

likely

PRES

371

NP

s;

I

C

SR.L

S,

NP be+ill

/ST

T o m

v

V

c

S

IT

s°

I

. ^ F NP N P yV

I

PRES NP

I

LIKELY

V V

0

I

IT

Y be+ill

S

that

N P

I Tom

/S"

I

V

I

PRES+0+be+ill Co.,

00

T e n s e Processing

—>

NP

T

PRES+0+be+likely

that

Note that if the subject-NP of likely had been [+top], then S j would not have been extraposed by IT and would have occupied the normal subject position, resulting in (lc). Note also that the rule SR, associated with likely, remains idle: it can operate only when the complement-S has just one tense, as has been stipulated. It is now obvious why (4c) must be ungrammatical: the i/iaf-clause stands in subject position, which it is allowed to do only if it is NP: tensed clauses require NP-subjects. But seem is not subcategorised for an NP-subject clause, even though it may take a subject-S" clause. It follows that when seem takes an S" as subject the sentence must undergo IT, as in (4b). It is, moreover, in agreement with what has been said about topic-comment modulations that sentence (4b) does not seem to lend itself easily for a [+top] or even a [+comment] f/iai-clause.

372 The same requirement that fully tensed clauses, i.e. with a finite verb form, require an NP-subject also accounts for the ungrammaticality of (6c) That Tom is ill is expected to seem: here, too, the f/iaf-clause is grammatical subject and must, therefore, be an NP. Its SA-status, however, is that of subject-clause to seem, and it can, therefore, not be an NP. Let us now consider the derivation of (2a) J expect Tom to be likely to be ill. For

this

it

is

necessary

to

specify

first

the

subcategorization

and

rule

properties of the verb expect. We shall do this together with the specification for believe and think". (21)

SS-catcgory

subcategorization

rules

NP[S"]/S'

likely

adjective

seem

verb

S'/S"

tend

verb

S'

follow

verb

S

NP[ "1

A few observations are in order here. First, think has been assigned the rule of SR, together with the possibility of an S' as object-clause. This is correct only for certain, slightly archaic, varieties of English. Then, more importantly, both expect and believe, which have identical features, are subcategorized for four object options: NP (as in: I believe you),

S \ S", and jsfp[S"]. The interesting

point here is that one can detect a semantic difference between an embeded S"clause that is NP and one that is not. Consider the following two little dialogues: (22) A: D o you believe that the earth is flat? B: Yes, I believe that (*so). (23) A: D o you believe/expect/think that Tom will vote for Jack? B: Yes, I believe/expect/think so (*that). Not only does this show that so-pronominalization differs semantically from f/iaf/tf-pronominalization of i/zai-clauses, it also suggests that so pronominalizes non-NP clauses, whereas that or it pronominalize i/iai-clauses under NP. This suggestion is confirmed by the spread of ungrammaticality in (24) and (25): 2

373 (24)a. b.

So it seems. So it is thought/believed/expected.

c.

* So it is likely.

(25)a.

That is likely.

b. c.

That is believed/expected. * That seems.

d. (*) That is thought. 3 These observations lend support to the main hypothesis underlying this paper, that embedded Ss are either NP or not NP. X-bar theory is weakened to the extent that this hypothesis is confirmed. Now back to (2a), whose derivation is as follows: (26)a.

PRES

S1M

expect

The reader will have little difficulty now to work out the S^-cycle, which lowers SIM onto ill to give to be ill. Then, on the next cycle up, Sj, SR raises j^p[Tom] to the position of S^, which is demoted to /S^ and shifted one position to the right: (26)b.

CQ.

s;

L

Ji - >

v to be likely

NP

I

Tom

/s; I V

to be ill

374 W e are now on the S°-cycle, where, again, SR has to apply, lifting ^ [ T o m ] up, for the second time, to the position of its own S, this time S'^, and shifting /S'^ one position to the right: (26)c.

V

NP

NP

I

Tom

I

I

cxpect

I

/s; i v

V to be likely

to be ill

T e n s e Processing will then finally yield: (26)d.

V

NP

I

I

expect

Tom to be likely to be ill

T h e analysis as given so far accounts in principle for all the data in (1-8), but for the ungrammaticality of (2c) and the grammaticality of (2d). T h e case of (2c)

does

not

seem

too

difficult: it

is

explained

by

Ross's

"Internal

S

Constraint" (Ross 1967), according to which a nonrelative e m b e d d e d S cannot occur internally but must stand either at the beginning or at the end of a sentence. (2c) is in violation of that constraint. W e may, however, take (2c) to b e an intermediary stage in the generation of (2d), assuming that the

that-

clause has b e e n moved postcyclically to sentence-initial position, being marked [+top]. T h e bulk of the explanation, however, of the p h e n o m e n a observed in relation with the problem of English subject complementation lies in the rules of the transformational cycle, not in those of the postcycle.

375 Notes

*. My friendship with Wim de Geest, whom we honour in this volume, is even older than my solution to this problem of English grammar: our friendship goes back to the mid-60s. In the beginning our acquaintance was purely professional. But Wim's warm and caring personality made it impossible to keep it that way. We soon became solid personal friends. It is to this special friend and esteemed colleague that I dedicate this paper. 1. By "weak compositionality" we mean what is normally meant by "compositionality", i.e. the principle that the semantic effect of a construction is computable as the value of a function F such that one of the constituents is F and the other(s) form(s) the input to F, the value being assigned to the node dominating the construction. The principle is recursive in that the input to F can itself be the value of a compositional construction, - this principle being weakened, however, by possible appeals to background knowledge. That such appeals are sometimes required appears from cases like the following: (i) Each room has a shower. (ii) Each student has a supervisor. The former is true only if each of the rooms has its own individual shower, not shared with other rooms, but (ii) remains true even if students have to share supervisors. Clearly, the truth-conditions of such sentences depend on what is generally known about hotels and universities, or, in other words, the satisfaction conditions for have incorporate a parameter whose value is to be filled in by a knowledge base. Likewise for gradable adjectives like old: the truth-conditions of a sentence like John is old depend crucially on age expectations in the setting at hand. 2. Cp. Lindholm 1969, Seuren 1985:140. 3. For most speakers (25d) will not be as obviously ungrammatical as e.g. (25c). Yet an active sentence like I think it is clearly to be rejected. There is admittedly a problem here, which will need to be sorted out. One might surmise that some verbs have different subcategorizations for their active and their passive forms.

Word Order in Tensed Clausal Complements Knut Tarald Taraldsen Introduction In some languages, tensed clausal complements appear to have the same word order as declarative main clauses. In other languages, they do not. Compare the Icelandic and Norwegian embedded clauses in (l)-(4) with the corresponding main clauses in (5)-(12): (1)

ad han keypti ekki b6kina,... that he bought not the book

(2) * ad han ekki keypti b6kina, ... (3) * at han kjpfpte ikke boken,... that he bought not the book (4)

at han ikke kjrfpte boken,...

(5)

Han keypti ekki b6kina,... he bought not the book

(6) * Han ekki keypti bokina,... (7)

Han kjffpte ikke boken,... he bought not the book

(8) * Han ikke kjffpte boken, ... Assuming uniform adjunction of sentence adverbials (including the negation) to VP, 1 the word order contrast seen in (3)-(4) vs (7)-(8) reflects the obligatory application of V-movement to C in root clauses in Norwegian, according to the standard analysis of Germanic word order. As V-movement to C seems to be in general inapplicable in embedded tensed clauses, however, the superficially similar contrast between the Icelandic examples (1) and (2) is taken to result from obligatory V-movement to I in Icelandic. Now, the Icelandic/Norwegian contrast (l)-(2) vs (3)-(4) must show that whereas V-movement to I is possible and, in fact, obligatory in Icelandic tensed complements, it is neither licit nor necessary in Norwegian.

378

The first step in the process of explaining this cross-linguistic difference consists of finding a second difference that groups languages into the same two subclasses. Looking at Scandinavian, the following comes to mind: exactly those Scandinavian languages that have overt subject/verb agreement (in tensed clauses), also have V-to-I-movement, i.e. Faroese and Icelandic have both agreement and V-to-I, whereas mainland Scandinavian lacks both. Furthermore, the correlation is supported by historical evidence in that both agreement and Vto-I-movement existed in earlier stages of mainland Scandinavian, and disappeared at roughly the same time (cf. Platzack (1987)). The correlation appears solid also when languages outside the Scandinavian group are taken into consideration. Thus, Romance languages have both overt subject-agreement and V-to-I (most visible in French). (German and Dutch can be taken to have V-movement to I located on the right of VP. In English, Vto-I is overt at least with auxiliary verbs). Assuming the descriptive generalization to be valid, the remaining task is to explain why it should hold. In this article, we attempt to present one half of the answer, developing an explanation why V-movement to I is obligatory in tensed clauses in languages with overt subject/verb agreement.

1. Agreement across potential barriers We proceed as follows. First, we introduce certain general theoretical assumptions drawn from Chomsky (1986a) and applied to the analysis of Romance cliticization by Kayne (1987). Then, we show that these assumptions suffice to explain the correlation between subject/verb agreement and obligatory V-to-I movement, when combined with Koopman & Sportiche's (1985) analysis of subjects and a partial assimilation of subject/verb agreement to cliticization.

1.1. Cliticization to I Kayne (1987) shows that the ungrammaticality of (9)-(10) in (modern) French can be reduced to an ECP-violation, given the Barriers-framework (as described in Chomsky (1986a)) and certain auxiliary assumptions, in particular, that the only available analyses are those given in (11)-(12):

379 (9) * Jean soutient de les bien aimer (10) * Jean les veut aider (11) ... [jp PRO les+1 [yp bien aimer t... (12) Jean les veut [ I p PRO I [yp aider t... To satisfy the ECP, the trace of the clitic must be governed by its antecedent, but is not, since non-lexical I fails to L-mark its complement so that the VP containing the trace is a barrier to government. Interestingly, lexicalization of I through V-to-I movement permits L-marking of the VP, so that the structure in (13) is well-formed (see also Chomsky (1986a:70)):2 (13) Jean les + [j aime][yp bien V t... A similar analysis seems applicable to Scandinavian cliticization. In Scandinavian, an unstressed pronoun moves across preverbal adverbs just in case the V moves out of the head of VP, as illustrated by the following Icelandic:3 (14)

Konan elskare kki okkur the woman loves not us (15) Hann telur konuna ekki elska okkur he believes the woman not loves us (16) Konan elskar okkur ekki (17) * Hann telur konuna okkur ekki elska

We do not assume that Icelandic unstressed pronouns adjoin to I (when filled by V), since they are left behind under I-to-C movement ("verb-second"), unlike their French counterparts:4 (18) * Elskar okkur konan? loves you the woman (19) Elskar konan okkurl (20) Vous a/me-t-elle? (21) * Aime-t-elle vous? Rather, we suggest that Icelandic unstressed pronouns adjoin at the X'-level.5

380 More specifically, we tentatively take the structure of (16) to be like (22), adapting Pollock's (1987) proposal that instead of 1=AGR Tense, we have separate heads AGR and Tense: (22)

Konan elskar+T+AGR[ T p [ T ,oA*ur[ T ,T[ v p ekki V i]]]

If the trace of the fronted pronoun must be antecedent-governed, and T does not by itself L-mark VP, then V-movement to T is forced, as in Kayne's account of Romance clitics. In (16), the V has in fact moved to AGR, and we assume that it did so, passing through T. But in (17), the V does not leave its basic position, as shown by the fact that it follows ekki "not". Thus, (23), the representation associated with (17) is characterized as ill-formed by the ECP:

(23)

...

konuna AGR[rpp[rpofc/cwr[rpT[yp ekki elska i]]]

12. AGR and VP-internal subjects VP does not block NP-movement, even if not L-marked. Otherwise, (24) would violate the ECP just like (11)-(12):6 (24) Jean 6sp£re [Q>C[jpPi?0 ne I pas [yp arriver t en retard]]] Similarly, the Icelandic sentence (25) would be excluded:7 (25)

Mennirnir telja [jp skipin I[yp ekki koma t & morgun]] the men believe the ships not come tomorrow

For the moment, we shall disregard the question why the barrierhood of the VP does not block NP-movement, since the way we shall exploit this fact does not depend on its explanation. We now want to combine the core ideas of Kayne's analysis of cliticization with Koopman & Sportiche's (1985) proposal that the subject position, thought of as Spec-IP, is always filled by movement from a VP-internal position. In structures like (24), this corresponds to the standard analysis, since the only theta-role that can be associated with the subject is assigned to the object of the ergative V:

381 (24) Jean éspère [(--pC[jpPiîO ne I pas [yp arriver t en retard]]] Koopman and Sportiche extend this analysis to sentences like (26)-(27) by positing that "external" theta-roles are not assigned directly to Spec-IP, but rather to Spec-VP: (26)

Jean Jean (27) Jean Jean

éspère ne pas perdre son argent hopes not lose his money éspère ne pas dormir jusqu'à 9 heures hopes not sleep till 9 o'clock

Hence, such sentences must have S-structure representations like (28)-(29): (28) Jean éspère (29) Jean éspère

[^pC[jpf/?0 [^pC[jpPi?0

ne I pas ne I pas

[ y p [ y p

perdre son argent]]]] f [ y , dormir jusqu'à 9 heures]]]]

i[y,

Looking at the question from Kayne's (1987) point of view, these structures are licit, although I does not L-mark VP, since we have already seen in the case of (24) that the resulting barrierhood of the VP cannot suffice to block NPmovement to Spec-IP, regardless what the reason is. Likewise, there is no obstacle to assigning the Norwegian tensed clause in (4) a representation like (30) (instead of (16)), even if I is considered "weak": (4)

at han ikke kjtfpte boken,...

(30)

[ c p at [ I p han I [ypikke [ y p t[y,kj0pte boken]]]]]

We now assume that I does not contain AGR in (30), i.e. that the absence of morphological subject/verb agreement actually reflects the lack of syntactic agreement in Norwegian. Consider a structure with 1= ... AGR ... in the light of Kayne's analysis of cliticization: (31)

[

I p

M > . [

I

. . . A G R

i

. . . ] [ y p i

i

.

V']]

In particular, AGR is like a clitic insofar as standard analyses are correct in characterizing it as essentially an X°-element with certain nominal features. Furthermore, it is co-indexed with and c-commands an empty NP to its right, the trace of the NP moved to Spec-IP. Suppose we assimilate the analysis of

382 A G R even further to the analysis of clitics by positing that A G R actually must form a chain-link with (the head of) this empty NP in (31). Then, it follows from Kayne's assumptions that the VP must be L-marked, just as in structures like (11)-(12). Hence, I must be "strong" or else V must move to I, as in (13). Assuming that overt subject/verb agreement always corresponds to the presence of A G R in I, an Icelandic sentence like (2) must in fact have an S-structure representation similar to (31): (2)

* ad han ekki keypti bokina,... that he not bought the book

(32)

[ C p ad[ I p /ia/i i .[ I ...AGR i ...][ V pekki[ V p t^,keypti

b6kina]]]]

If I is "weak" in Icelandic, 8 it now follows that (32) is ill-formed, and (2) is ungrammatical, as desired. On the other hand, (1) is correctly predicted to be grammatical, since V-movement to I allows I to L-mark the VP in (33): (1)

ad han keypti ekki b6kina,... that he bought not the book

(33)

[cpa.6[lphani[l...keypti+AGRj...][ypekki[yp

^ [ y , V bokina]]]]

A similar analysis applies to French and other languages with overt subject/verb agreement. Thus, we have derived the generalization that a language has obligatory V-movement to I, if it has overt agreement.

1.3. AGR vs cliticization Our assimilation of the properties of A G R to those of the clitic pronouns (in Romance), is necessarily only a partial one. Although we claim that in each case, there must be a chain-like relation between a X°-element (AGR or a clitic) and the head of some c-commanded maximal projection, we shall have to maintain that the two types of X°-element differ in various crucial respects. On the one hand, we will assume that clitics and occurrences of A G R differ with respect to the structural relationship they bear to the heads they are attached to, an assumption that will be made precise in the next subsection. On the

383

other hand, we also posit that A G R and clitics impose different properties on the maximal projections whose heads they are chain-linked to. The motivation for the latter hypothesis is simply that AGR, but not a standard clitic can link to the head of a NP co-indexed with the subject. Thus, if i=j the configurations in (34)-(35) must be excluded, in contrast with (31): 9 (34)

[ I p N P i I [ v p t i [ v , C l j + V(ec j )]]]

(35)

[jpNP-Clj + I[ypt¿[y>V(eCj)]]]

We will depend on two assumptions to obtain this result. First, consider the effects of the Case-features associated with clitics. Suppose, for instance, that CL is marked [+acc] in (34)-(35). We assume that the head H forming a chainlink (C1,H) with the CI must share this feature. By standard feature projection, this feature is also associated with the maximal projection of H, which is therefore

licensed

only

in

a

position

assigned

accusative

Case.

Similar

considerations apply to other Case-features that can be marked on CI. It now follows that CI cannot link the head of the trace in the Spec-VP position in (34)-(35). Hence, it must link the head of the ec following the V. 1 0 Our second assumption is that CI is also marked [ + pro]. Again, we assume that the dependent head must share the features of CI, and projects them to its maximal projection. Correspondingly, a NP whose head H forms a chain-link (C1,H) is a pronoun. As formulated by Chomsky (1986b), the Binding Theory requires a pronoun P to be free in the least complete functional complex (CFC) in which it is governed, provided there is a Binding Theory (BT) compatible indexing of this CFC with respect to P. The least CFC in which the postverbal ec is governed in (34)-(35) is the VP, and the VP admits an indexing in which this pronominal element is free. Hence, the ec must be free in VP, and so we cannot have i=j. To allow for (31), we posit that AGR differs from CI being marked [ + nom]. But it is similar to CI in that the dependent head must share its features. Hence, the maximal projection of the dependent head is also marked [ + nom], and must occur in a position marked [ + nom]. Thus, AGR can licitly link the head of the Spec-VP trace in (31), provided we have nominative Case assigned by I under government (to its right).

384 Notice that this link remains licit even if we consider AGR to bear the feature [+ pro], just like CI, provided the V governs Spec-VP. Then, the VP is itself the least CFC in which Spec-VP is governed. The VP moreover always has a BT-compatible indexing for a pronominal element in its Spec-position, since no other element c-commands this position within VP. In particular, the indexing in (31) is BT-compatible. Hence, it is sufficient that the trace of Spec-IP, interpreted as a pronominal, is free in the VP. We will in fact assume that AGR always is marked [+pro] and, correspondingly, that V governs Spec-VP, i.e. that "government" is defined as follows (cf. Chomsky (1981:166)): (36) A governs B= ^ ^ (a) A c-commands B, and ... (37) A c-commands B= ^ ^ (a) no C distinct from A contains A, but not B, or (b) the least constituent C containing both A and B is a projection of A. In terms of this definition, a head may c-command, and, hence, govern every constituent within its maximal projection. Under a certain interpretation, the Binding Theory developed in Chomsky (1986b) would in fact be consistent with taking AGR to have both the feature [+pro] and [4-ana].11 By previous assumption, these features must be shared by the linked head of the Spec-VP in (31), implying that the NP in this position is both pronominal and anaphoric, i.e. an occurrence of PRO. The crucial assumption is that the BT-compatible indexing of PRO with respect to the feature [ + pro] can select a CFC different from the one selected for the feature [+ana]. In all cases where PRO is not in Spec-XP in the least CFC XP where it is governed, the classical "PRO theorem" is preserved, since the least domain having a BTcompatible indexing for [+pro] will also have one for [ + ana], so that the two independently selected binding domains coincide, and PRO must be both free and bound in XP: (38)

[ Xp NP i [ X) X...PRO j ...]... BT-compatible indexing for [ + pro]: i=j BT-compatible indexing for [ + ana]: i=j

385

But in a structure like (39), where PRO is in the Spec-position of the smaller CFC, this CFC has a BT-compatible indexing for [ + pro], but only the more inclusive CFC has a BT-compatible indexing for [+ana]: (39)

[xpNPj.-.fYpPROjY'...]

Consequently, YP is selected as the domain in which the Binding Theory applies with respect to [+pro] and XP is the domain relevant for [ + ana]. Hence, the structure is well-formed with respect to the Binding Theory, if i=j. Since (31) is an instance of (39), there is therefore no Binding Theoretic obstacle to reinterpreting the trace in Spec-VP as PRO, if we adopt the interpretation of the Binding Theory just described. Provided we do not consider PRO to be necessarily an inherent argument, but a potential "trace" subject to the binding conditions defining [+ana] and [+pro], there is no obstacle from the point of view of the Theta Theory either, since PRO can then always be part of a chain linking the Spec-IP to a theta-marked position. Accordingly, we will tentatively adopt the hypothesis that AGR is both pronominal and anaphoric. This decision may be used to exclude a structure like (40), where the Spec-VP, now interpreted as PRO, has no antecedent in Spec-IP: (40)

[ I p [ I ...AGR i ...][ v p PRO i V]]

Suppose we interpret the notion "A has a BT-compatible indexing for B", B an anaphor, in such a way that if A does not actually contain any element that could bind B, it still counts as having a BT-compatible indexing for B, provided it could have a possible binder without violating X'-theory. Then, since IP can have a Spec-position in (40), it is analyzed as the domain in which the Binding Theory applies with respect to the feature [ + ana] of PRO. Hence, (40) is excluded, because PRO is not actually bound in that domain. In general, the NP whose head is linked to AGR must occur in a Spec-position of some XP and be bound by the Spec of the least maximal projection containing this XP.

386 2. Relativized Minimality In this section, we first suggest an alternative to Kayne's (1987) analysis of cliticization. The alternative approach replaces the notion of "barrierhood" related to L-marking with a version of Rizzi's Relativized Minimality. We then want to show that our preceding analysis of AGR, formulated so far within Kaye's general approach, is also amenable to reinterpretation in terms of Relativized Minimality.

2.1. Clitic movement vs NP-movement Earlier, we left unexplained the fact that VP does not appear to be a barrier to NP-movement, even if not L-marked. We begin by suggesting a solution to this problem. Kayne (1987) assumes that clitics are not phrasal projections, but heads, e.g. N rather than NP. We see essentially two ways of distinguishing between (24) and (11)-(12) on the basis of this hypothesis. On the one hand, we could posit that antecedent-government and thetagovernment are both available modes of proper government. Then, the trace in (24) is properly governed by virtue of being theta-governed by the infinitival V, and need not be governed by its antecedent. Hence, the barrierhood of the VP is irrelevant. The traces in (11)-(12) must still be antecedent-governed, however, as they correspond to heads (of empty NPs), and heads are not theta-governed (since only maximal projections are assigned theta-roles). On the other hand, it may be possible to reinterpret the pertinent facts from the point of view of Rizzi's (1987) "relativized minimality". Then, we would say that a NP-trace must be bound from the closest Spec-position, a requirement that is in fact met in (24) (and (25)). A clitic, being a head, must be governed from the closest head-position, but is not, in (11)-(12), since the head of VP is closer than the head of IP. We must then find a way of allowing the head of VP not to count as the closest head when moved to I.

387 Various considerations favor the second option. Perhaps the most obvious one is that a NP also can be moved across I from a position which is not thetagoverned. Consider (41): (41)

II craint de [jpP.RO ne I [yppas sembler [jp t travailler assez]]]

In (41), the trace is the subject of a lower clause, and the subject position is not theta-govemed, even if theta-marked, by definition (cf. Chomsky (1986a: 15)). To adapt relativized minimality to our purposes, we will make use of the following formulations: (42)

A is locally xMinked to B=def(a)&(b)&(c)&(d): (a) A and B are X1 categories (b) A and B are co-indexed (c) A is c-commanded by B (d) No X1 distinct from B c-commands A, but not B

(43)

If A and B form a X'-chain -link (A,B), then B is locally xUinked to A

In terms of (42), (11)-(12) are ill-formed because in each V, an X°-category, intervenes between the clitic and (the head of) the trace, two elements that must form a chain-link: (11)

... [jp PRO les + l [yp bien aimer t...

(12) Jean les veut [ I p PRO I [ y p aider t... In (13), on the other hand, the intervening V is moved to I: (13) Jean te+[j aime][yp bien V i ... Accordingly, (13) can be reinterpreted in accordance with Relativized Minimality as having the clitic attached to V, prior to V-movement. Instead of les+[I aime], we then have [I[V les+aime]I]. Assuming that the adjoined V forms a chain-link with its trace, then interpretable as Cl + V, the chain associating the clitic under I with the head of the object NP can be broken up into two chain-links (les, CI) and (C1,N) both satisfying Relativized Minimality.12

388 Finally, a structure like (24) is well-formed with respect to relativized minimality ( = (42)-(43)), since there is no X max -element intervening between the moved NP and its trace: (24) Jean espere [^pQjpPiJO ne I pas [yp arriver t en retard]]] Before closing this subsection, we will briefly discuss two objections by Kayne (1987) against accounting for cliticization in terms of a minimality condition. Consider first the observation that in languages with "strong" I, clitics may move to I even when the V does not also move to I, so that sentences like (9)-(10) are characterized as well-formed, with underlying structures like (11)-(12): (9) * Jean soutient de les bien aimer (10) * Jean les veut aider (11) ... [jp PRO les+l [yp bien aimer t... (12) Jean les veut [jp PRO I [yp aider t... Here, the clitic has leap-frogged over an intervening head position, in violation of Chomsky's (1986a:42) minimality condition. Notice first, however, that by Kayne's assumptions, the only head position which the clitic skips across, is the V position, since he in fact seems to assume that in (12), the clitic attaches first to the lower I and is moved along with this I up into the higher I-position (and the lower I in (12) is therefore actually a trace). 13 On the other hand, it seems that those Romance languages that allow structures like (11)-(12), do not also allow clitics to adjoin to V. Thus, in Italian, for instance, clitics obligatorily precede infinitival Vs, contrasting in this respect with their French counterparts: (44)

Piero vuole veder la. Piero wants see+her (45) * Piero vuole la vedere. (46) * Pierre veut voir la. (47) Pierre veut la voir.

389

Kayne derives the contrast (44) vs (47) from the assumption that clitics can move to I without I being filled by V in Italian, but not in French. By this assumption, the representation in (48) is available in Italian, but not in French: 14 (48)

...[j,veder[j,/a + I [ w V t...

If, moreover, a clitic only adjoins on the left side of its host, and no proper subpart of a X°-element can be extracted, (48) is the only possible representation of (44) and (46), and the French/Italian contrast is accounted for. But the complementary contrast (45) vs (47) does not follow from any of Kayne's assumptions. To us, it seems minimally necessary to disallow clitics to adjoin to V in Italian, excluding representations like (49)-(50): (49) (50)

...I[yp la+vedere t... ...[j,/a+veder[j,I[yp V t...

This observation suggests bringing (42) closer to Rizzi's (1986) proposal by replacing condition (d) with (51): (51) No X*C distinct from B c-commands A, but not B, where C is a potential X1 -binder for A By virtue of the underlined part of (51), now to appear as the new clause (d) of (42), structures like (11)-(12) are compatible with Relativized Minimality in exactly those languages where a clitic cannot adjoin to V. Since in fact clitics apparently cannot adjoin to V in languages where the sentences corresponding to (11)-(12) are grammatical, there is no empirical argument against Relativized Minimality, although Kayne's argument against "rigid" minimality stands. As a matter of fact, we can make the stronger claim that Relativized Minimality is not only compatible with the Romance cliticization facts discussed by Kayne, but actually explains them, replacing Kayne's analysis. In French, it is necessary to allow representations like (31) to be associated with sentences like (47) (in particular, because the infinitival V does not move to I in such sentences). Hence, French must permit clitics to adjoin to V, and therefore, Relativized

390 Minimality does not allow the clitics to skip over the V in (11)-(12) in French. This approach presents one advantage over its competitor, since it correctly predicts that "long" cliticization (to I) is possible only in languages where "short" clicticization (to V) is impossible, a correlation not captured by Kayne's analysis. Relativized minimality, as modified above, is also compatible with the second case providing an argument against Chomsky's (loc. cit) minimality principle. In compound past tenses, clitics attach to the auxiliary in all Romance languages: (52)

Piero gli ha dato una mela Piero him+has given an apple (53) Pierre lui a donné une pomme

Presumably, the clitic has hopped over the lower V. It is noteworthy, however, that clitics cannot adjoin to participles:15 (54) * Piero ha gli dato una mela (53) * Pierre a lui donné une pomme This suffices to make (52)-(53) consistent with our present formulation of Relativized Minimality. One might still object that an analysis in terms of Relativized Minimality fails to establish a link between "long" clitic movement and the null subject property, a correlation that seems empirically well documented. Kayne suggests capturing the relevant generalization in the following way: since "long" clitic movement is clitic movement to I without concomitant V-movement to I, "long" cliticization is possible, only if I is sufficiently "strong" to L-mark VP. At the same time, a language is a null subject language only if it has "strong" I. Clearly, this line of reasoning is unavailable within an approach where L-marking plays no role. However, Kayne's suggestion does not carry much force, since it is not specified exactly in which ways the "strenght" of I is pertinent to the null subject property. Thus, although we are inclined to consider the correlation between "long" cliticization and the null subject property as a real one, it still seems an essentially open question how it is to be derived from a theory of syntax. On

391

the other hand, we will return to this issue in subsection 2.4. below to suggest a way of establishing the missing link in an approach based on Relativized Minimality.

2.2. Relativized Minimality and AGR From the point of view of Relativized Minimality, which we ended up stating as in (56), (32) would be well-formed, only if there were no X^-element ccommanding the head of Spec-VP, but not AGR, which could bind the head of Spec-VP: (32)

[ C p a d [ I p ^ a « . [ I . . . A G R i . . . ] [ v p e k k i [ V p P R O i [ v , ^ i / b6kina]]]]

(56)

A is locally xMinked by B = def(a)&(b)&(c)&(d): (a) A and B are X1 categories (b) A and B are co-indexed (c) A is c-commanded by B (d) No x ' c distinct from B c-commands A, but not B where C is a potential X1 -binder for A.

But the head of VP is a X°-element which c-commands the head of Spec-VP, but not AGR, given the definition in (27). Assume it could bind the head of SpecVP. Then, AGR must initially be inserted under V and can only reach the Iposition by movement of [yAGR+V], a requirement which cannot be met in (32), although it is met in (33): (33)

[ C p a.d[ l ^an i [ l ..AGR i +keypti...][ v v ettd[ V F PRO i [ v , V b6kina]]]]

We now take "C is a potential X'-binder for A" to have the following more specific meaning in the context of (56): (57)

No position C not containing B c-commands A, but not B, if B would be x'-linked from C, with A occurring licitly in C.

The intuitive content of this new formulation replacing clause (d) of (56) is that an element B locally xMinking another element A must be in the position closest to A where it could xMink it, unless the element in this position is itself XMinked by B.

392 This formulation has the desired effect with respect to (32) vs (33), if we assume that AGR is always analyzed as "amalgamating" with another head rather than being adjoined to it (see 3.2. for a more accurate formulation). Then, AGR occurring in C= the V-position in (32) would amalgamate with the V providing a new head of the VP-projection, and would therefore be able to c-command and bind the head of Spec-VP. Hence, Relativized Minimality has the effect described above. This property of AGR is not shared by clitics, we assume, making explicit an assumption hinted at above. Clitics can only adjoin to another head. Hence, the structure underlying the ungrammatical (58) is illicit, because, the IP not being a projection of the clitic, the clitic does not c-command the head it should link: 16 (58) * Je laisse les chanter I let them sing (59) je laisse [jpNPj/esj+Ifypt^y,chanter]]] Correspondingly, the grammatical (60) is not blocked by Relativized Minimality, since the clitic les appearing in C= the I-position, would be adjoined to I, and could therefore not link the head of Spec-IP from that position: (60) (61)

Je les laisse chanter Jej[jksJ-+laisse + I][ Vp PRO J .[ v ,V[ Ip M > i chanter]]]]

We assume that the outer "segment" of the head to which a clitic is adjoined does not count when c-command is determined. 17 Hence, the X'-projection of the host is the least constituent containing the clitic and the clitic c-commands all positions contained in this domain. In particular, it c-commands the head of the embedded Spec-IP in (61). More generally, we see that a clitic attached to V is able to bind any head position within V'. Therefore, our previous account of French cliticization in terms of Relativized Minimality is not affected by the modifications adopted here. On the other hand, we still have not found any reason why clitics cannot adjoin to V in Italian. This is the issue we now turn to.

393 2.3. Clitics as X max -binders We maintain that the NP whose head is linked to a clitic, is [+pro]. Hence, a structure like (62) is well-formed with respect to the Binding Theory just in case the object NP is not (A-) bound by any element in VP: (62)

[ I p N P i I [ v p e c i [ v > C l j + V NP-...

In (62), this requirement seems to be met if and only if i is distinct from j. Thus, a structure like (64), underlying the grammatical French sentence (63), is well-formed, when the subject is referentially disjoint from the object: (63)

Jean

semble Jean

(64)

Jeanj

semble

ne pas le connaître seems not [jp t ne

him

know

I pas [yp t.[y,

lej + connaître NPj]]] However, this presupposes that the clitic does not itself count as an A-binder for the NP whose head it links. Now, this assumption is of course warranted by the hypothesis that a clitic is an X°-element, and therefore can only bind the head of a maximal projection, not the maximal projection itself, giving exactly the right result for French. But suppose we adopt Muysken's (1983) feature system for projection levels. Then, an element could consistently be analyzed as [+ max, - proj], i.e. as simultaneously a head (not a projection) and a maximal constituent. In fact, Muysken suggests this analysis for clitics (cf. also Vikner & Sprouse (1987)). In the light of preceding remarks, we do not want to adopt this proposal for French clitics, which should continue to be only heads, i.e. [- max, - proj] in Muysken's system, but we shall now show that the classification [+ max, - proj] gives the correct prediction for Italian clitics. First, we note that because Italian clitics are [- proj], they are expected to show the effects of Relativized Minimality with respect to heads. They cannot move across X°-elements that they could possibly attach to (cf. footnote 13). In this respect, they are just like their French counterparts.

394 But since they are also [ + max] they have an important property that French clitics lack. They are potential A-binders for maximal projections. In particular, the clitic in an Italian instantiation of (62) always A-binds the object NP, in addition to binding its head. Hence, this structure now always leads to a Binding Theory violation in Italian, since the pronominal object NP is always bound in the VP, even if i is distinct from j. On the other hand, a structure like (65), where the clitic is attached outside the VP is well-formed from the point of view of the Binding Theory (provided that i and; are distinct): (65)

[ I p NP. CZ i+ Itypecjty, VNP.]]]

In order to make (65) also consistent with Relativized Minimality, it only remains to make the natural move, taking "licitly" in (57) to imply "without violating the Binding Theory". We now have an account of the fact that Italian clitics can adjoin to I, without V-movement to I, and cannot adjoin to V. 18 We will finally show that this account combines with previous assumptions to permit us to link the properties of Italian clitics to the fact that Italian is a null subject language, capturing the generalization defended by Kayne (1987). Again we draw on the partial similarity between AGR and the clitics. In particular, we assume that AGR has the feature specification [a max, b proj] as the clitics. Then, AGR is [+ max, - proj] in Italian, but [- max, - proj] in French. This assumption is consistent with previous result, provided we take AGR to have the c-command privileges of the head of IP or VP only with respect to the feature [- proj], i.e. only when analyzed as an X°-element, and not with respect to the feature [+ max], as seems natural. 19 Then, AGR, occurring in the head position of VP, could still bind the head of Spec-VP without incurring a Binding Theory violation, since the resulting pronominal NP in Spec-VP would not be ccommanded, and therefore not bound, by AGR analyzed as a maximal projection:

395 (66)[n,NP.I[vpPROi[v,

V...

I ...AGRj... Hence, Relativized Minimality still predicts that AGR cannot be located under I, unless V is moved to I, even for Italian. Incidentally, this result could not have been obtained within Kayne's (1987) framework, where the "strength" of I would eliminate the need for V-movement both under cliticization and agreement. But even though our analysis correctly predicts Italian to be just like French in that AGR under I forces V-movement to I, it also leads one to expect an important difference between the two languages. Consider again the structure in (31): (31)

[ I p[ I ...AGR i ...][ v p PRO i V']]

Above, we pointed out that this configuration is ill-formed, because PRO is free in a domain where it must be bound, namely the IP. However, this remains true only for French. If AGR is also [+ max] in Italian, PRO meets the binding requirement associated with its feature [+ ana] by being bound by AGR. 2 0 If we assume that there is no independent requirement that IP must have a Spec in (31), then it follows that (31) is well-formed in Italian, although not in French. Therefore, a grammatical Italian sentence like (67) can be associated with the well-formed representation (68), an instance of (31), but there is no similar structure available for the corresponding French sentence: (67)

Parlano molto they-talk much

(68)

[ I p parla+AGR { [

(69) * Parlent

w

PRO f V molto]]

beaucoup

If(70), where IP has a null Spec, is excluded in all languages, we can therefore conclude that Italian and French differ with respect to the null subject property for the same reason that Italian clitics behave differently from their French counterparts:21

396 (70) [ I p ecj parle+AGRj [

w

P R O j V beaucoup]]

More generally, "clitic climbing" and the other phenomena rendered possible by cliticization directly to I (without V-movement) are predicted to occur in exactly those languages that have the null subject property.

3. AGR and the Extended Projection Principle We now turn to two questions not adressed in the preceding discussion. First, we consider the question why AGR must eventually appear in the head position of IP, rather than in the head position of VP (or an intermediate projection, e.g. TP). Then, we suggest a solution to the problem of explaining why AGR must link the head of some NP, rather than form a single-membered head chain all by itself.

3.1. The Extended Projection Principle In the previous discussion, we have assumed, without explanation, that an occurrence of A G R must appear, in tensed clauses, under I in languages with overt subject/verb agreement. (71)

[jpiNPpp... AGR;...] I v p P R O i [ v > V . . .

Yet, nothing would prevent AGR from appearing in the V-position as well. In fact, this was the basic premise for our use of Relativized Minimality to force V-movement to I in structures like (71). Thus, our strategy must be not to prevent AGR from appearing under V, but rather to find an independent principle requiring an occurrence of AGR

in the I-position. We suggest that this

independent principle is to be identified with Chomsky's (1981) "Extended Projection Principle" (EPP), leaving open the possibility that this principle can itself be deduced from more general properties of the grammar. In essence, the EPP requires IP to always have a subject, i.e. IP must have its Spec filled. What we want to propose, is that having AGR in the head position of IP is another way of satisfying this condition.

397 The intuition behind this proposal is that the EPP reflects the need for certain features basically associated with N-projections to be associated with IP. Leaving aside the question what exactly these features are, we could then say that the appearance of an NP in Spec-IP helps satisfy this need, because features of Spec-XP for which the head X is not specified can project onto XP, adopting an idea of Holmberg (1986). But if this is on the right track, then the same requirement could also be met by adding the appropriate features in the head position. Thus we propose that some languages always satisfy the EPP by filling Spec-IP, whereas other languages, for some reason, must meet the same condition by inserting an N-type element, i.e. AGR, under I in tensed clauses. We note that apart from accounting for the obligatoriness of AGR in I, in certain languages, this approach also paves the way for our analysis of null subject languages, since the EPP no longer imposes the presence of a NP filling the Spec-IP in languages with AGR under I. Thus, (31) is licit with respect to EPP and therefore well-formed in Italian, but excluded in French by the Binding Theory, as discussed earlier:

(31)

[Ip[I...AGRi...][VpPROiV']]

3.2. AGR as Spec-I° Having suggested a way of accounting for the obligatory occurrence of AGR in I, it remains to see why this AGR must form a chain-link with some other nominal head. This leads us to consider in somewhat greater detail the relation between AGR and I. Previously, we posited that AGR is part of the host head in a way that clitics are not, so as to be able to account for the different c-command privileges of these categories. Yet, we cannot say that AGR is I, i.e. that IP simply is AGRP. For one thing, this would seem inconsistent with the claim that AGR is [+ max] in Italian (an assumption that still calls for some clarification). On the other hand, it would then be harder to explain why AGR must link another head, since it seems somewhat unlikely even that the head of a phrase can form a chain with the head of a constituent contained in that phrase.

398 Instead, we propose that AGR occurs in the Spec-I", transposing this notion of X'-theory to the word level. If we continue to assume that the element in Spec-X1 projects onto X1 features that are not specified on the head, then the appearance of AGR, an N-type item, in Spec-I°, satisfies the EPP, under our interpretation, by projecting the relevant features onto 1°, whence these project to IP by standard principles. Furthermore, we can now improve somewhat on our account of the c-command privileges of AGR. Our definition of c-command would not allow AGRj to ccommand the head of Spec-XP in the configuration (72): (72)

[xpNPi[x,[XoAGR.X]...

But if AGR projects the right features and its index onto X°, the head of NPj is nevertheless c-commanded and bound by the head of XP, i.e. X°. Thus, our claim that the appearance of AGR in the V-position would give an X°binder for the head of Spec-VP closer than AGR in I, is unaffected, provided, of course, no independent principle prevents AGR from filling the Spec-V°. Hence, we still expect V-movement to I containing AGR to be forced by Relativized Minimality, as explained previously. At the same time, it now follows directly that AGR, analyzed as a [+ max] constituent, cannot bind NPj in (72), simply because Spec-VP is not ccommanded by AGR in any event, but only by X°. Thus, another assumption needed for Relative Minimality to work correctly, falls out. We allow AGR to ccommand constituents of X', by stipulating that only phrasal projections block c-command under clause (a) of the definition. To distinguish between AGR and clitics, we continue to assume that clitics are adjoined to X°, rather than inserted in Spec-X°, taking this to mean that their features cannot project onto X°. Finally, we note that V-movement to I now must be viewed as movement to Spec-I°. Then, the relevant features of AGR, initially attached to V, as required by Relativized Minimality, project first from Spec-V° to V°, and then onto Spec-I°, as desired. (Similarly, we would consider I-movement to C or a higher I as movement to Spec-X°, in order to derive Kayne's (1987) idea about index sharing under I-to-C-to-I movement from Spec-X^to-X1 projection).

399 Hence, earlier accounts are preserved and even partially improved, if we consider AGR as a Spec of the head, rather than as the head of the projection in which it appears. On the other hand, it now becomes easier to claim that AGR must form a chain-link with the head of some NP. Basically, this ought to follow from the same principles that force clitics to form such links. Tentatively, we suggest the following: 22 (73) Every [- proj] element must be part of a chain (a^, ..., a Q ) such that a n heads a constituent marked [ + max, + proj]

Conclusion We are now in a position to claim that current versions of the Government-andBinding Theory provide a promising answer to the question raised at the outset. In particular, we have seen that the Koopman & Sportiche idea that Spec-of-IP always originates VP-internally combines with certain assumptions about the nature of AGR to induce V-movement to I containing AGR. 2 3 We have also considered two formulations of the locality principles used in this analysis, and shown that a version of Rizzi's Relativized Minimality appears to be a strong competitor to the framework of Kayne (1987). But our main result does not depend on a specific choice between the two alternatives.

Notes

* Gratulerer med dagen, Wim! 1. Irrelevantly, for present purposes, the negation may be more appropriately analyzed as the head of a projection containing VP, cf. Pollock (1987) and Kayne (1987). 2. Alternatively, the clitic could adjoin to V before V moves to I. On this analysis (13) provides no independent argument that V-movement to I allows I to L-mark VP. 3. (14) is grammatical, only if the pronoun is stressed.

400 4. We assume the analysis of French "subject clitic inversion" in Kayne (1983). 5. The possibility of adjoining heads to I' has been suggested by Kayne (oral tradition) for (infinitival) verbs in Italian. 6. Since there is nothing in the following discussion that depends on whether or not a separate projection appears between AGR en VP, we shall henceforth disregard the possibility that I should be split up into the autonomous heads AGR and T. 7. This example makes the point somewhat more clearly than (24), since the embedded subject, preceding ekki is visibly not inside the VP. 8. Under the analysis developed in Section 2, the "strength" of I is irrelevant, a desirable result, since on Kayne's (1987) approach, Italian must have "strong", given the cliticization facts, but still seems to have obligatory V-movement to I m tensed clauses. 9. This analysis may not be appropriate for reflexive clitics. 10. For (35), this follows also from lack of c-command. 11. This interpretation is suggested by Kayne (1986), but is not adopted by Chomsky, who stipulates that the CFCs selected as binding domains for PRO on the basis of [+pro] an [+ana] must coincide. 12. To preserve our proposal about Icelandic cliticization, we must consider adjunction to X' as equivalent to adjunction to X° from the point of view of Relativized Minimality. 13. Kayne claims that under "clitic climbing", the lower I must first move to the embedded C, to L-mark IP, and then to the higher I, along with the clitic. But the last step follows without further assumptions only if the clitic is already attached to the lower I when the lower I is in C. 14. Cf. footnote 5. 15. Following Kayne (1987), we take it that the structure of participle constructions like the one occurring in Italian Abbracciatala, Piero se ne ando "Having embraced her, Peter left" is hp...[yabbracciata[,,la + I[yp..., comparable to Italian infinitival constructions. 16. It could not link the head of Spec-VP, which is not in a position where [ + acc] is licensed. 17. We could make this more precise by reformulating clause (a) in the definition of "c-command" in terms of the predicate "dominates" interpreted in the sense of Chomsky (1986a:7). However, we suggest a somewhat different approach in 3.2. 18. A question remains as to why a clitic cannot adjoin to the higher V under "clitic climbing", since the pronominal "source" NP would then at any rate be free in the lower VP. Noting that the clitic must adjoin to the lower I before "climbing", so that the unwanted adjunction to the higher V would actually involve the entire lower I ( = C1 + I), we suggest that although V can adjoin to I, I

401 cannot licitly adjoin to V. As for the impossibility of adjoining the clitic to the higher V, rather than I, in Italian (i)

* Desidero [ ^ p C [jp PRO I [ y p la^+sentire [jpNPj cantare l'inno nazionale]]]] "I want to hear her sing the national anthem"

we must find a way of forcing the projection of sentire to be the binding domain selected for the subject of the lowest IP. This, however, is independently necessary, given (ii) *

Maria^ la- sente cantare l'inno nazionale "Maryj hears hen sing the national anthem".

19. This assumption is made more precise in 3.2. 20. We assume that A G R itself ( a l s o = P R O ) is ungoverned, so that the binding conditions do not apply to it. 21. (70} is of course well-formed, if the ec is the trace of an element in (the nearest) Spec-CP. 22. This presupposes that a NP associated with A G R or a clitic actually has internal structure, minimally distinct nodes corresponding to the head and its maximal projection. 23. We think our account could be extended to the obligatory V-to-I movement found in control infinitivals in Icelandic (but not in French), given a suitable theory relating syntactic and morphological AGR.

A Look at sembler + Infinitive from Different Angles Liliane Tasmowski

1. Semantics At least from a semantic point of view, the verb to seem, or for that matter, the French verb sembler, has been considered on a par with perception verbs such as to see, to look, to hear, to sound, and so on. Newman (1981) for instance suggests that to seem is a kind of perception verb neutral as to the particular sensation involved. As is well established (see i.a. Willems (1983) for French), perception verbs allow for two types of readings, physical or cognitive, partially depending on the construction in which they appear: (1)a. Je vois Jean tondre la pelouse/Je vois que Jean tond la pelouse I see Jean mow the lawn /I see that Jean mows the lawn b. Je vois a la machine que Jean a tondu la pelouse I see from the mower that Jean has mowed the lawn The l(a)-(b) opposition can easily be exemplified for sembler too: (2)a. Jean (me) semble tondre la pelouse Jean (to me) seems mow the lawn b. II (me) semble que Jean a tondu la pelouse It (to me) seems that Jean has mowed the lawn Willems (1983) posits a stronger parallellism between syntax and semantics: a tensed complement would always entail indirect perception (be it physical or cognitive), whereas a non finite clause implies direct perception. Whatever is understood by "direct perception", critical examples such as (3) adduced by De Geest (1972) and further discussed in Kirsner & Thompson (1976 section 3) and Koster (1987:130), prove beyond any doubt that it should not be equated with perception of the referent of the NP following the matrix verb (see (3a)) and that a complement denoting a cognitive activity is not incompatible with the use of an infinitivial construction (see (3b)):

404 (3)a. We hoorden de boer het varken slachten We heard the farmer the pig slaughter b. Ik zag geloof overal ontbreken I saw faith everywhere fail Material gathered in Cauwe (1982) evidences that both possibilities exist for sembler as well: (4)a. Le soleil semble avoir un effet tout à fait néfaste sur cette peinture The sun seems have an effect totally disastrous on this painting b. Le courage d'entreprendre semble manquer partout Courage to undertake seems fail everywhere Indeed, (4a) may well be uttered on a rainy day and (4b) can claim for the perception of no thing at all. In fact, although the Grand Robert dictionary and the Grand Larousse de la langue française give but one sense for all the sembler + infinitive cases,1 namely "donner l'impression/l'illusion", it is obvious that this gloss is awkward for (4a) and totally inadequate for (4b). In other words, the relation holding between the NP complement and the perception verb is paralleled by the relation holding between sembler and its NP subject. The notion of direct perception can however still be given content as far as the perception verbs are concerned because they do not allow an infinitive complement in the past. This argument does not hold for sembler as the following infinitive is not submitted to any such restriction: (5)a. # On voit Shakespeare avoir quitté Stratford on Avon vers l'âge de 25 ans One sees S. have left S.O.A. about the age of 25 years b. Shakespeare semble avoir quitté Stratford on Avon vers l'âge de 25 ans S. seems have left S.o.A. about the age of 25 years Clearly, sembler + infinitive lends itself to an indirect (IR) reading. We now proceed to illustrate the proposed proviso.

405 - Intended non ironically, an utterance allows for an IR reading in contexts where the infinitive amounts to a predication obviously at odds with an actual situation. In that case perception verbs are expected to be out: (6)a.

(To one's son, lazily extended on the sofa) Tu sembles travailler extrêmement dur You seem work very hard b. # Je te vois travailler extrêmement dur I you see work very hard

- A Direct Reading (DR) is incompatible with genericness and past time reference. Observe that in combination with sembler + infinitive both genericness and past time reference are available and acceptable, whereas perception verbs do not allow either: (7)a.

Le loup semble très mal supporter la chaleur The wolf seem very badly stand the heat b. # Je vois le loup très mal supporter la chaleur I see the wolf very badly stand the heat c. Paul semble avoir quitté la maison vers les 7 heures Paul seems have left the house by 7 d. # Je vois Paul avoir quitté la maison vers les 7 heures I see Paul have left the house by 7

- A direct reading is on the contrary enhanced when the subject is presented as involved in some perceptible activity, witness (6), if the son is sitting at his desk, with books all around him, in which case the sentence with the perception verb becomes perfectly natural too. On the whole, it turns out that perception verb constructions and sembler + infinitive constructions, though undeniably related, differ on several counts. - As is well known, scene-settings in thematic position set the limits within which a proposition holds true, in other words, they set the frame of reference for a judgment. They too impose a DR, thus possibly giving rise to unacceptable sequences, be it with sembler or with the perception verbs.

406 (8)a. # Quand le coucou doit pondre, il semble déposer ses oeufs dans le nid d'un autre oiseau When the cuckoo must lay, he seems lay his eggs in the nest of another bird b. # Quand le coucou doit pondre, on le vois déposer. When the cuckoo must lay, one him sees lay c. # Dès que Jean a un examen à passer, il semble vomir As soon as J. has an exam to take, he seems vomit d. # Dès que Jean a un examen à passer, on le voit vomir2 As soon as J. has an exam to take, one him sees vomit Let us now turn to semantic restrictions on the occurrence of sembler + infinitive proper and the conclusions they warrant with respect to the nature of sembler. - A first point to be recalled is that contrary to the perception verbs, sembler easily admits stative predicates: (9)a.

Jules semble être bien maussade Jules seems be very sad b. ?# Je vois Jules être bien maussade I see Jules be very sad

- Thoroughly significant is further the fact that a DR is unacceptable in contexts where the infinitive denotes an actual, unmistakably true qualification of the subject: (10)

(Description of the room one is sitting in, writing a letter) a. Des fleurs ornent la cheminée, quelques lithos égaient les murs Flowers decorate the chimney, some lithos brighten up the walls b. # Des fleurs semblent orner la cheminée, quelques lithos semblent égayer les murs Flowers seem decorate the chimney, some lithos seem brighten up the walls

407 (10b) is acceptable if vision is somehow impeded, for instance when uttered by somebody who is looking into the room from outside and who is not completely certain about what he is perceiving. In order for sembler to be acceptable in this construction, the possibility of misinterpretation must be sufficiently plausible. As a consequence, the difficulty exemplified in (10b) will not arise when the infinitive predicates actions or feelings, since both can easily be misunderstood: (11)a. Il semble boiter He seems lump b. Il semble être triste He seems be down We are thus led to believe that sembler always involves some deduction procedure - even when followed by an infinitive - and that in fact it shares more properties with a verb of cognition than with a verb of perception. This explains why contrary to the perception verbs, sembler is perfectly compatible with negation and modals: (12)a. Ces dames ne semblent pas apprécier l'exposé These ladies seem not appreciate the talk b. On voit ces femmes ne pas apprécier l'exposé One sees these ladies not appreciate the talk c. Ils semblent vouloir tout casser They seem want all break d. Je les vois vouloir tout casser I them see want all break As a verb of cognition, sembler asks for the coincidence of the judging mind with the signs for a state of affairs, not with the state of affairs itself: (13)a.

Les brontosaures semblent avoir été herbivores The brontosaurs seem have been herbivores b. # Les brontosaures semblaient être herbivores3 The brontosaurs seem be herbivores

408

From a semantic point of view we are thus entitled to consider sembler as an interpretation device applied to some circumstances expressed in a proposition: interpretatively: P Whether

there

are reasons

representation —

to paragon

this formula with

the

syntactic

sembler (§ -) is a question which will be tackled in the

following section.

2. Syntax Generative grammarians usually derive a sentence such as Jean semble tondre la pelouse from  semble (ip=s Jean tondre la pelouse). It is said that Jean is governed in S by sembler, but sembler cannot case-mark it. 4 To receive Case, Jean moves to the empty matrix subject position, leaving behind a trace. The resulting chain is grammatical since it is Case-marked in the matrix subject position and theta-marked in the position of the trace. So, Jean in matrix subject position is connected with a trace in the embedded subject position, yielding the S-Structure Jean\ semble (ip = s h tondre la pelouse). What compelling reasons are there to consider the complement of sembler an S? That there is no CP-projection is obvious since no indirect question can be formed, no overt complementizer can ever appear after sembler: (14)a. * Jean semble quoi tondre? Jean seems what mow? b. * Jean semble de tondre la pelouse Jean seems de mow the lawn Besides, when the behaviour of the semWer-sequence is compared to that of a vouloir-complement - which according to Kayne (1980) is an S' ( = CP) -, many differences manifest themselves: cleft vs no cleft, pronominalization vs no pronominalization, VP anaphora vs no VP anaphora: (15)a. C'est tondre la pelouse qu'il veut / * C'est tondre la pelouse qu'il semble b. Il le veut / *I1 le semble 5 c. Crois-tu qu'il voudrait? /* Crois-tu qu'il semblerait?

409 and further, selectional restrictions vs autonomous-role vs no autonomous-role:6

no

selectional

restrictions

and

(16)a. * Guy veut pleuvoir/Guy veut être aimé/*Il veut pleuvoir/*Il [impers.] veut aimer b. * Guy semble pleuvoir/Guy semble être aimé/Il semble pleuvoir/*Il [impers.] semble aimer All these data leave no doubt about the fact that sembler is not subcategorized for an S'. The problem we are left with is to adduce evidence that sembler is not subcategorized for a VP. We believe that apart from "the natural generalization that there is one-to-one correspondence between subjects and verbs" (Koster (1987, 126)), little definite proof is available here. Raising verbs are exceptional anyway, so it could equally well be stipulated that precisely that class of verbs just takes a VP-complement, thereby inheriting whatever selectional restrictions would be imposed by that VP. As for theta-role assignment, it is not impossible to argue that sembler enters into its calculus, as we will try to show. First, it is not at all clear that the agentive character of the participant in the action expressed by the infinitive clause is necessarily preserved in the sembler + infinitive construction: (17)a. Sur ce, Jean prend courageusement la parole Next, Jean courageously gets up to speak b. Sur ce, Jean semble prendre courageusement la parole Next, Jean seems to courageously get up to speak (18)a. Jean écoute sans broncher Jean listens without giving a sign b. Jean semble écouter sans broncher Jean seems listen without giving a sign We feel it rather hard to determine what sense exactly is to be assigned to the b. versions. Surely, if Jean merely seems to be acting as an Agent, he is not expected to be an Agent?

410

Second, and more convincingly, it appears that a non-agentive participant of the state of affairs described by the infinitive clause can get an agentive reading due to the action of sembler: (19)a. ?#Très intelligemment, Jean ne comprend pas le sens de notre décision ?#Very intelligently, John does not understand the meaning of our decision ? # Jean, très intelligemment, ne comprend pas le sens de notre décision ? # Jean ne comprend très intelligemment pas le sens de notre décision b. Très intelligemment, Jean semble ne pas comprendre le sens de notre décision Very intelligently, John seems not to understand the meaning of our decision Jean, très intelligemment, semble ne pas comprendre le sens de notre décision Jean semble très intelligemment ne pas comprendre le sens de notre décision (20)a. * Courageusement, Jean goûte la compagnie de Martine Courageously, John values Martine's companionship * Jean, courageusement, goûte la compagnie de Martine * Jean goûte courageusement la compagnie de Martine b. Courageusement, Jean semble goûter la compagnie de Martine Courageously, John seems to value Martine's companionship Jean, courageusement, semble goûter la compagnie de Martine Jean semble courageusement goûter la compagnie de Martine Third, imperative mood has often been claimed to require an agentive subject. Now observe the contrast in (21) : when tuer is the verb of a complement of sembler, an imperative is no longer available : (21)a.

b. * -

Un moustique! Tue-le! A gnat! Kill it! Un moustique! Semble-le tuer! A gnat! Seem to kill it!

411

The highest VP is thus apparently not totally alien to the theta-role calculus. As far as we can see, no solid evidence for the S-status of a sembler-complement can be based on the general behaviour of clitics either: it suffices to maintain that clitics attach to the verb they are arguments of 7 and that the anaphors are further bound in the minimal X m a x in which they are governed, i.e. S' following Koster (1987), 29, note 6) (22)a. Jean ne semble pas le trouver très sympathique John seems not to find him very sympathetic b. Jean semble se soumettre assez facilement John seems to submit rather easily It must be recognized that to be left with a "natural generalization" as the core argument in favour of a certain hypothesis is not very satisfying, the more so because other theoretical approaches such as relational grammar do well without it. Let us remind ourselves that in the latter framework, the axiom of the subject position is a necessary characteristic of the last stratum in a representation^ only. French, however, has for a while offered an empirical argument, the behaviour of the adnominal clitic en, which was capitalized on in Ruwet (1972). Ruwet observed that in some cases where en is attached to the embedded infinitive, it is an adnominal complement of the superficial matrix subject. He concluded that the superficial subject originated as an embedded one, witness the en left behind: (23)

La porte semble / - en être ouverte de la cathédrale semble / ((la porte-en) être ouverte) The door seems / -en be open of the cathedral

Analysing the restrictions on the occurrence of such an en, Couquaux (1980) observed that en cannot be related to the subject of an inergative verb, nor to the subject of an ordinary transitive verb. It can only be found with predicatives and with certain presentationals and movement verbs:

412 (24)a. * Le chef en téléphone The chief en telephones b. * Les hauts miroirs en réfléchissent sa silhouette {en = du living) The high mirrors en ( = of the living room) reflect its silhouette c. La porte en est ouverte d.

La confirmation en est arrivée {en = de la nouvelle) The confirmation en ( = of the piece of news) has arrived

The behaviour of en is then said to be highly systematic if trace theory is adopted: en must simply c-command its trace in S-Structure. The derivation proposed for (24) c. is: (24)c.

(est (la porte -en) (ouverte) — (est (la porte) (en) (ouverte) la porte en est (ej)(ej) ouverte 9

In the adopted perspective, the only claim substantiated by the behaviour of en is that the superficial subject has to be related to a post verbal position, in other words it does not produce the slighest evidence for the existence of an embedded subject position.

3. Semantically Oriented Syntax The peculiar behaviour of unaccusative verbs has been the object of much research since Perlmutter first raised the issue. We will concentrate on the diagnostic characteristic especially underscored in Burzio (1986), namely the selection of être as an auxiliary. We suggest that it is neither (i) a sufficient nor (ii) a necessary condition. (i) It is not a sufficient condition. It is generally thought that a verb conjugated with être cannot passivize. This can be explained if we admit with Burzio that the function of the passive morphology is to signal that the subject of the sentence is not the external argument of the verb, but an internal one. This would imply that an unaccusative verb has no external argument by definition, so that passive morphology would be entirely superfluous.

413 Nevertheless, Rivière (1981) raises inescapable counterexamples : there are cases of passivized verbs which are otherwise conjugated with être: (25)a.

Jamais encore il n'avait été descendu aussi bas Never before has there been gone down this deep b. - Etes-vous descendus aussi bas que nous? - Etes you gone down as deep as we have? c. * -Avez-vous descendu aussi bas que nous?

We add: C'est entendu, il sera donc intervenu en faveur de votre protégé. 10 OK, there will be intervened in favour of your charge b. Ils sont intervenus en faveur de mon protégé c. * Ils ont intervenu en faveur de mon protégé

(26)a.

(ii) It is not a necessary condition. Not all Romance languages have the être/avoir opposition. Romanian and Spanish are cases in point. Yet it is striking how much Romanian and Spanish presentationals and movement verbs have in common with their Italian counterparts. And even in those languages where the être/avoir opposition is available, avoir-ve rbs can exhibit unaccusative tendencies. Koster (1987, 255), citing discussion by Balk-Smit Duyzentkunst, recognizes the unaccusative status of the intransitive versions of certain transitive causatives. As illustrated in (27b) the same point can be made for French: (27)a. De aardappelen hebben gekookt (us Piet heeft de aardappelen gekookt) The potatoes have boiled (vs. Peter has brought the potatoes to the boil b. Ce tableau a pendu chez moi (vs Paul a pendu le tableau chez lui) This picture has hung at my place (vs. Paul has hung the picture at his place. Next, let us consider a second diagnostic characteristic, which is more akin to semantics: if a verb is unaccusative, it will not allow for an impersonal passive.11 This criterion applies to many avoir-verbs:

414

(28)a. Ca a duré très longtemps / *I1 a été duré très longtemps It has taken a very long time b. - Qu'est-ce qu'il a plu! /*- Qu'est-ce qu'il a été plu! It has rained like hell c. Le ton des pourparlers avait nettement refroidi / *I1 avait été nettement refroidi The tone of the negotiations had cooled off considerably And significantly, it distinguishes between the être-verbs too: (29)a.

L'équipe est descendue extraordinairement bas

Jamais encore il n'avait été descendu aussi bas (par aucune équipe au monde) b. Le thermomètre est descendu extraordinairement bas * Jamais encore il n'avait été descendu aussi bas (par aucun thermomètre) (30)a. Le directeur n'interviendra jamais en faveur de qui ce soit Il ne sera jamais intervenu en faveur de personne (par quelque directeur que ce soit) b. Ces considérations n'interviendront dans aucune décision * Il ne sera jamais intervenu dans la décision (par aucune considération de cette espèce). We are thus driving towards a notion of unaccusativity strictly tied up with non-agentivity of the internal argument. Some depart much further from the êire-unaccusativity connection than we do here, even for Italian. Belletti & Rizzi (1987) thoroughly question superficial transitivity as an absolute obstacle to unaccusativity. Returning to inspiring studies by Postal (1970) and Ruwet (1972), they single out a category of transitive psychological verbs whose lexical representation is seriously at variance with their superficial constructional properties: 12 (31)a. Ça a frappé Marie / Jean a frappé Marie par son intelligence It has struck Mary/ John's intelligence has struck Mary b. Jean a frappé Marie avec un bâton John has hit Mary with a stick

415 A great deal of psychological verbs enter into the a. use but ignore b.: (32)

Ça/Jean intéresse/intrigue/préoccupe...Marie It/Johninterests/intrigues/worries...Mary

Belletti & Rizzi mention a series of phenomena suggesting that the subject of (31a) and (32) is not a bona fide subject and that the object of the same sentence does not behave as a bona fide object either. 1 3 They propose the underlying structure (33) : (33)a.

e (yp (intéresse ça) Marie)) 1 4

analogous to (33) b.: (33)b.

e (yp (plâit la musique) à Jean))

Now, all the verbs considered so far as unaccusatives mostly happen not to reject adnominal en related to the superficial subject: (34)a. (Reporting a lenghty session at the dentist's) - Un

enfer,

cette

séance.

Chaque

instant

en

a

duré

une

éternité.

D'ailleurs, je n'en suis pas encore remis b. (Choosing a dress) - Evidemment, à toi de choisir. Mais à ta place, moi je ne prendrais pas cette robe. Le tissu en chiffonne très vite, tu vois bien, et comme tu voyages beaucoup... c. (Comment on a letter urging husband to be present next day in the director's office) Cette

lettre,

oui...

Le

ton

en

intrigue

comminatoire... We believe that is is possible to go still further: 15

beaucoup

mon

mari...

Si

416

(34)d. (A new road is planned) - Cette route est une très mauvaise affaire. Le tracé en suit malheureusement le cours de la Zwalm et ça menace tout le paysage. Encore une bataille de perdue pour les écologistes. e. (About a heavy file) - Vous m'étudierez ce dossier aujourd'hui. C'est évidemment beaucoup de papier, mais la plus grande partie n'en concerne qu'un seul et même problème, celui de la retraite des fonctionnaires d'Etat As far as (34d) and (34e) are concerned, it seems hopeless to try to reduce them to something like (33b). Either the problem of en is independent of unaccusativity or unaccusativity can have an expansion domain outside pure syntax.

4. Pragmatics Thus far, we have not considered two further constraints on the occurrence of adnominal en related to the superficial subject. Both are signalled in Ruwet (1983): the head of (NP - en) almost never refers to humans and (NP - en) must somehow describe a part- whole relationship. These conditions are seen to be at play in (35): (35)a. ?* C'est vrai que cet orchestre est plutôt médiocre mais le premier violon en est carrément génial, b. - Comment, vous n'avez pas ce que je demande? Mais voyez donc l'étalage! * - Le costume en est peut-être bleu, mais le modèle est différent. The pragmatic dimension of these two constraints is straightforward : the first amounts to a thematically low-graded head, whereas the second amounts to a thematically dominant en. In both cases, the same result obtains, namely NP (less thematic) - en (more thematic), in accordance with the dimensions listed in Givôn (1976) and represented here: 16

417

(i) (ii) (iii) (iv) (v)

human whole agent definite NP subject

more akin to thematicity than II

M

it

ti

non human part patient non-definite NP object

Let us examine (iii)-(v), in that order. To be an agent implies by and large to be human, so (iii) is not expected to give results at variance with (i). Since the converse does not hold true, (36a) should be better than (36b) : (36)a. Pas de village plus familial que celui-là? Le maire en est reçu par tout le monde et tous s'entre-aident s'il le faut b. Pas de village plus familial que celui-là. *Le maire en va chez tout le monde et tout s'entraident s'il le faut To illustrate item (iv) on the list, let us consider (34a), which seems fine. The same cannot be maintained for (34') a., where the head is introduced by a demonstrative pronoun: (34')a. - Vous avez apprécié le moment de l'extraction? - Un martyre, cette séance. *Et ce moment-là en a duré une éternité As for (v), it was seen at length in section 3 that en is only possible when the superficial subject has object-like properties. It can thus be concluded that when en is used, the head has a cluster of antithematic properties to allow thematicity for en. The hypothesis we infer from this situation is that the sentence en appears in is in fact about the referent of en. The part-whole relationship holding between NP and en acquires a particular significance in this perspective, since we suggest that the predication accomplished by the sentence must not be primarily - or at least not onlyabout the head of (NP - en). In other words we predict that where en is licensed, to predicate X of NP will amount to predicate X of (the referent of) en}1 Returning to the examples given, they can be seen to live up to expectation :

418 (24)c. = la cathédrale est accessible d. = la route suit le cours de la Zwalm c. = la convocation est comminnatoire Besides, contexts can be constructed to enhance, on discourse-level, topicality of en vs topicality of NP. In the latter case, en should be bad: (37)

(Judging competing pictures) a. - Votre avis? - Mon avis? L'auteur en restera inconnu. Voyons la suite. b. - Votre avis? - Merveilleux? *Dommage que vu le règlement, l'auteur en restera inconnu.

A last indication substantiating our hypothesis is the fact that although the predicative verb is a privileged context for en (see Couquaux 1980), en is nevertheless many a time at least clumsy when the AP is a predication restricted to NP itself: (38)a. - Ce livre me plâit, tu ne peux pas savoir. L'auteur en est génial. b. - Ce livre me plâit, tu ne peux pas savoir. ? "L'auteur en est un Juif de Brooklyn (39)a. - Allez jeter un coup d'oeil sur la cathédrale. La crypte en est fort intéressante. b. - Allez jeter un coup d'oeil sur la cathédrale. ?*La crypte en est un joyau de l'art roman ( = la cathédrale est un joyau de l'art roman) That examples are easier to find for (38) than for (39) is consistent with the necessary factual autonomy of a human referent. As a consequence, intersection of predications bearing at the same time on the referents of NP and of en are less likely. It should however be emphasized that we are hinting at a more subtle distinction: what is needed is not the prima facie intersection explicited here, but the fact is that the whole sentence in which en appears consists of a significant predication about the referent of en. It remains to determine the importance of the preceding observations for the sembler + infinitive construction. If our hypothesis is correct, following a last dimension for thematicity, namely that the theme comes early in the sentence, on discourse level the mere intervention of sembler must raise difficulties for

419

the apparition of the topic continuity mediating en. This certainly turns out to be the case. Consider: (34")a. * - Chaque instant a semblé en durer une éternité c. * - Le ton semble en intriguer beaucoup votre mari e. * - La

plus grande

partie

semble

en

concerner

la retraite

des

fonctionnaires d'Etat In fact, outside the scope of compelling contexts, en is readily acceptable only with a part-whole relationship concerning a thing and the predicative être followed by an adjective. The reason seems to be that in that case, the sembler + infinitive construction amounts to the assignment of a peculiar quality (the AP) to the referent of en (the whole), with no other semantic element interfering, except for the indecision carried along by sembler. Notwithstanding the speculative character of this tentative explanation, we believe there to be certain facts suggesting that the approach advocated here is very plausible. Example (2) introduced the possibility of an explicit indirect object, the interpreter of a certain state of affairs. But if the interpreter is indeed expressed, en appears to be still more constrained. In fact, although neither (24'c) nor (24'd) are really felicitous, informants tend to prefer (24'c) to (24'd) : (24')c. - Enfin la cathédrale!!! Et la porte m'en semble encore être ouverte! d. - Enfin la cathédrale! ?* Et la porte me semble encore en être ouverte! Now, such an interpreter, very much like scene-setting time and adverbials, establishes the reference frame for the predication to come.

place 18

As

such, it sets up a barrier with the preceding discourse, thus forming an obstacle for the topicality of en. To establish rules for en is to investigate what it means to be an anaphor. 5. Summary Starting from a semantic point of view, we saw that there are good reasons to represent the meaning of sembler + infinitive as a proposition in the scope of a cognitive operator. 19 From a syntactic point of view, the decison to label the proposition S or VP seems to be theory-bound. The behaviour of en has no bearing on the syntactic problem of the subject position but highlights the tremendous influence of context and discourse perspectives.

420 Notes

1. Here and further on, we shall consider only the "Raising" structure (i), not "Control" structures like (ii): (i) Jeanj (me) semble rêver; (ii) Il me; semble rêver; 2. These sentences should be compared to the perfectly normal (i)-(iii) with IR reading: (i) Le coucou, quand il doit pondre, semble déposer ses oeufs dans le nid d'un autre oiseau (ii) Quant au coucou, il semble déposer ses oeufs dans le nid d'un autre oiseau (iii) Jean semble vomir dès qu'il a un examen à passer Notice that the presence of an explicit indirect object has a largely similar impact, (iv) below is by and large contradictory, as opposed to (v) : (iv) # Madame Durand me semble avoir 60 ans, mais elle ne les parâit certainement pas (v) Madame Durand semble avoir 60 ans, mais elle ne les parâit certainement pas 3. The French imperfect provides for a reference time in the past, so that coincidence of judgment and state of affairs is expressed in (13b) (see Tasmowski 1980, 50). 4. It should be understood that sembler, as an unaccusative verb, has one internal argument, namely IP (= S), with the theta-role Theme and no Casemarking, in line with Burzio's generalization. The unaccusative character of sembler followed by an infinitive has been advocated for French by Fauconnier (1982-1983) in a relational framework. Koster (1987, 264) proposes an analogous analysis for the tensed complement of to seem, which is said to be a specification of the argument it: {it\ seems t[ (s; that John is ill)). 5. This is not entirely accurate. Special conditions can be found under which sembler acts more like a Control-verb. Consider for instance the following dialogue: (i) Je ne vous comprends pas. Qu'attendez-vous en somme? Le directeur ne semble-t-il pas toujours accueillir vos propositions avec beaucoup d'intérêt? - C'est bien là le problème. ?I1 le semble seulement, en réalité il n'écoute pas But this is a meta-linguistic use, bearing on the (stressed) word sembler. 6. For French, a recent exposition is provided by Rooryck (1988). 7. Admittedly, causatives and perception verbs behave differently. As for the tense auxiliaries avoir and être, it is perhaps sufficient to take advantage of the fact that in French the past participle is not a possible attachment site for clitics.

421

8. At least in Fauconnier's (1982-1983) interpretation. On the first stratum, before Union, a sentence such as Jean semble téléphoner would be represented by (semble (téléphoner Jean)). After Union, on the last stratum, it is (Jean semble (téléphoner)). See Fauconnier. 9. Pollock (1986) has shown that the whole of Couquaux's presentation cannot be accurate and that his "scission" hypothesis is ad hoc. It is not our contention to study the implications of Binding here. Suffices it to signal that Pollock's criticism and demonstration are thoroughly convincing. We have doubts though, about his own solution, which consists in replacing trace by PRO. 10. Ruwet (1986) signals the verb intervenir as a counterexample to the hypothesis that unaccusativity should be paired with a non-agentive internal argument. We, on the contrary, want to argue that a verb which takes être is not necessarily unaccusative. 11. The proposition was put forward by Perlmutter and Perlmutter-Postal. The references and discussion can be found i.a. in Tasmowski (1987) and Koster (1987). 12. Their overall theoretical point is that theta-roles do not influence syntax, but that they intervene at a pre-syntactic level where projection from theta-grids into syntactic positions applies. 13. For the non-subject character, see i.a. anaphora phenomena: (i) Un petit moment de concertation avec lui-mêmej ravigote toujours Bernard; Lui-même should be c-commanded by Bernardj, which is not the case. The nonobject character of the superficial object is posited on the basis of the difficulty to extract from it, a movement which is fully acceptable from inside a "normal" direct object. 14. Apparent passive clauses corresponding to (33a) are discarded because the participle then acquires typical adjectival properties, loosing its verbal ones. 15. As happens with many of the areas not fixed by normative grammar, there will be a certain range of variance among native speakers here. Pinchón (1986, 98, footnote) for instance rejects examples given by Ruwet. The material put forwards was submitted to the intuition of French speaking Belgians. 16. We stick to the following usage: thematic: "normally chosen as theme"; theme : "the segment whose referent the rest of the sentence is about"; topic: "the previously introduced referent the sentence is about". 17. The behaviour of adnominal en can profitably be compared to that of the dative clitics expressing possession in Romanian (see Popescu & Tasmowski 1988), as was suggested to us by C. Dobrovie (personal communication). 18. See footnote 2. 19. Newman (1981) considers NP sembler + infinitive a kind of synecdoche visa-vis the structure il semble que + tensed complement.

Verb Projection Raising and the Status of Infinitival Complements Guido Vanden Wyngaerd

Introduction In this paper I want to discuss a particular instance of verb-movement, notably the one called Verb Projection Raising (VPR) by Haegeman and Van Riemsdijk (1986) (henceforth H&VR). An outstanding feature of H&VR's analysis is that it relies on multiple tree representations, an option which they claim to be necessary to derive all possible orders of verbs and NPs in the verb cluster. Similar considerations lead them to the assumption that any projection of V, including V', can undergo Verb Raising. I shall argue that neither of these assumptions is necessary to give a full account of VPR. In doing so, I shall make and defend certain assumptions concerning the details of infinitival complementation.

1. Verb Projection Raising as VP Raising H&VR's assumption that V' is visible for the rule Move-« is at odds with the Barriers-framework, where it is assumed that Move-« can only affect heads or maximal projections. The restriction to the minimal and maximal elements of the X'-schema is obviously desirable in the context of limiting the power of the transformational component, and seems otherwise well-substantiated; consider (1-3) for A, P, and N, respectively: (1)a.

Very ill I don't believe John was.

b. * 111 I don't believe John was very. (2)a. High on the mountains hermits used to live, b. * On the mountains hermits used to live high. (3)a. Friendly French people I like, b. * French people I like friendly.

424 Direct evidence against V'-movement is hard to find in Dutch and German, in the sense that allowing it does not seem to lead to overgeneration, as far as I can tell.1 However, the null assumption obviously is that the same restriction holds. Nevertheless, H&VR argue that allowing V'-movement is necessary for the sentence in (4), from which three possible surface orders are derivable, as (5) shows: (4)

das [er [em Karajan en arie vorsinge] wil] (that he to-Karajan an aria sing wants)

(5)a. das er [em Karajan en arie t] wil vorsinge b. das er [em Karajan t] wil en arie vorsinge c. das er [t] wil em Karajan en arie vorsinge H&VR argue that these are derived by each time taking a different projection of V as input for the rule of V(P)R: (5a) instantiates V-movement, (5b) V'movement, and in (5c) the entire VP has moved to a position to the right of the matrix verb. In order for this analysis to work, however, it must be the case that in Dutch and German the structure of the double object construction is as in (6a), with a layered VP; in English, by contrast, word order considerations would force one to assume the structure (6b): (6)a. / IO

V" \ V' / \ DO V

b.

V" \ V' DO / \ V IO /

Whereas in Dutch and German the indirect object (IO) is structurally superior to the direct object (DO), the converse holds in English. This is a particularly unattractive situation, in that what seem to be identical thematic properties would be projected radically differently in English on the one hand and Dutch and German on the other. Moreover, the fact that the relative order of IO and DO is identical in these languages is entirely coincidental under this approach. Furthermore, the evidence discussed by Barss and Lasnik (1986) shows that the IO is structurally superior to the DO and not the other way round, as in (6b). The Small Clause (SC) analysis for double object constructions does not suffer from any of these defects. In particular, we shall assume that SCs are AGRPs with an abstract AGR node: 2

425 (7)a.

VP

/ \ AGRP V / \ IO AGR' / \

DO AGR

b.

/

V

VP

\ AGRP / \ IO AGR' / \ AGR

DO

Thematic properties are projected identically in the languages under consideration, the only difference being one which is known to exist independently, i.e. the relative order of the verb and its complement(s): complements occur preverbally in Dutch and German (7a) and postverbally in English (7b). The relative order of IO and DO mirrors that of clauses, which is known to be constant over English on the one hand and Dutch and German on the other. An additional advantage of this analysis is that it accounts for the left branch effects on IOs which are typical of subjects (Kayne 1984): (8) * Who did you give friends of books? Again, under the structure (6b) this would remain mysterious, the IO occupying a regular object position. If the SC analysis for double object construction seems well-supported, the question arises how (5b) should be derived: whereas in H&VR's analysis V and DO formed one constituent, such is not the case under the analysis in (7), and (5b) seems underivable. In order to solve this problem, we must take a look at (9): (9) dat ik Jan gisteren een boek stuurde (that I Jan yesterday a book sent) Suppose we took traditional terminology seriously and put the adverbial adjunct in a position adjoined to VP. This implies that (9), where IO and DO are nonadjacent at S-structure, does not reflect the D-Structure order but results from movement of the IO to a position higher than the adjunct. I shall refer to the rule at issue as Light NP Shift (LNS) (cf. Hoekstra 1984; Bennis and Hoekstra 1984; Den Dikken 1988). Apart from accounting for the nonadjacency between verb and object frequently observed in Dutch, the rule receives some striking evidence from parasitic gap constructions:

426 (10)a.

Jan heeft die boeken [zonder te bekijken] weggelegd. (Jan has those books without to look-at put-away) b. * Jan heeft [zonder te bekijken] die boeken weggelegd.

The possible occurrence of a parasitic gap in the adjunct clause in (10a) follows directly if in (10a), but not in (10b), there is a trace resulting from LNS, which moves the object across the adjunct, exactly as in (9). We now have the tools in hand to account for (5); (5a) and (5c) are not problematic, as they are instances of Head-to-Head Movement (HHM) and X m a x movement, respectively; following Chomsky (1986:4) we assume that heads can only adjoin to heads and maximal projections only to X m a x . Concretely, in (5a) the head vorsinge adjoins to the right of the head wil, whereas in (5c) VP adjoins to the right of another X m a x , i.e. the VP headed by the matrix verb. The VP thus created adjoins to TP, as is the case with Extraposition complements. In (5b), however, the IO is first moved out of its SC and adjoined to VP, and then the VP undergoes VP-Raising as in (5c). We thus have reduced alleged V'-movement to movement of the maximal projection VP.3 In other words, the dialects that allow VPR, now understood as VP-Raising, would differ from dialects that do not in having both V and VP-movement, whereas the more restrictive dialects only have HHM in this area. 4

2. Against multiple trees We shall now proceed to show that it is not necessary to have recourse to a multiple tree analysis to derive the relevant word orders in dialects with VPR. The main problems raised by this sort of analysis, i.e. its incompatibility with the Projection Principle and its extreme power, are thus eliminated.

Consider the DS tree (11) (Swiss German), from which the surface orders in (12) can be derived.

427 (11)

/

CP

C I das

\ / NP I er

TP \

T / \ VP1 T / \ AGRP2 VI I I AGR' hat / \ VP2 AGR / \ AGRP3 V2 I I AGR' wele / \ VP3 AGR / \ AGRP4 V3 AGR' chone / \ VP4 AGR / \ NP V4 / \ I en arie singe

(12)a. das er en arie hat wele chone singe (that he an aria has wanted can sing) b. das er hat en arie wele chone singe c. das er hat wele en arie chone singe d. das er hat wele chone en arie singe Standard Dutch and German only allow (12a), which is derived by applying HHM all the way: V4 moves to the adjacent AGR node, then adjoins to the right of V3; the resulting verb goes via AGR to adjoin to the right of V2, and so on until the final landing site, the topmost T, is reached. (12b) is derived in an analogous fashion, except that the result of adjoining to V2 does not undergo HHM; rather, the X m a x VP2 is now adjoined to the right of VP1 and the VP thus created adjoins to the matrix TP. For (3c) the same procedure is followed except that VP is selected already at the VP3 level; (12d), finally, is derived by applying VP-raising all the way through. The derivations of (12) are schematized in (13):

428 (13)a.

V4-V3-V2-V1/VP1

b. V4-V3-VP2-VP1 c. V4-VP3-VP2-VP1 d. VP4-VP3-VP2-VP1 Next consider the examples in (15), derived on the basis of the tree in (14); it differs from the previous tree in that it has a causative Acl verb at the V2 level with a lexical subject in its complement, AGRP3. 5 (14)

CP C I das

^ TP1 / \ T NP I / \ er VP1 T / \ AGRP2 V I I I AGR' wil / \ VP2 AGR / \ AGRP3 V2 / \ I NP AGR' laa I / \ sini VP3 A G R chind / \ NP V3 / \ mediziin studiere

(15)a. das er sini chind mediziin wil laa studiere (that he his children medicine wants let study) b. das er wil sini chind mediziin laa studiere c. das er wil sini chind laa mediziin studiere d. das er sini chind wil mediziin laa studiere e. das er sini chind wil laa mediziin studiere The derivations leading to the corresponding sentences of (15) are summarized in (16), as the reader may check:

429 (16)a. V3-V2-V1/VP1 b. V3-VP2-VP1 c. VP3-VP2-VP1 These exhaust the possible options, yet only three sentences of (15) are thus derived. How then do we derive (15d) and (15e)? Observe that nothing prevents the subject of AGRP3 to undergo LNS to VP1; once LNS has moved the subject of AGRP3 to VP1, we can apply the series of movements in (16b) and (16c) to derive (15d) and (15c), respectively. Let us finally consider (17), with surface orders (18). H&VR specifically argued against an adjunction analysis on the basis of this example, claiming that (18f) is underivable under an adjunction analysis; however, we shall demonstrate this claim to be incorrect.6 (17)

CP C I das

TP1 / \ NP T I / \ er VP1 T / \ AGRP2 VI I I AGR' wil / \ VP2 AGR / \ AGRP3 V2 I I AGR' chöne / \ VP3 AGR / \ AGRP4 V3 / \ \ NP AGR' vorsinge I /\ em Karajan NP en arie

AGR

430 (18)a. das er em Karajan en arie wil chöne vorsinge (that he to-Karajan an aria wants can sing) b. das er wil em Karajan en arie chöne vorsinge c. das er wil chöne em Karajan en arie vorsinge d. das er wil em Karajan chöne en arie vorsinge e. das er em Karajan wil chöne en arie vorsinge f. das er em Karajan wil en arie chöne vorsinge (19)a. V3-V2-V1/VP1 b. V3-VP2-VP1 c. VP3-VP2-VP1 The first three sentences of (18) are derived in the usual fashion by applying the series of movements in (19a-b-c), respectively; IO and DO are adjacent at the surface. In (18d-e-f), on the other hand, IO and DO are nonadjacent; as before, we claim that this is a result of applying LNS to the IO. Concretely, (18d) is derived by adjoining the IO to VP2 and applying VP-raising as in (19c); in (18e) and (180 the IO first adjoins to VP1; then, (19c) derives (18e), and (19b) derives (18f). Hence all the surface orders resulting from (17) are derivable without recourse to a multiple tree analysis.

3. VP Raising as AGRP Raising 3.1. A new analysis for VPR Thus far we have argued that sentences like (5) above are derived by moving VP. However, there is another possibility consistent with the data, viz. that the entire AGRP immediately dominating the embedded VP moves to the right of the matrix verb. As long as the subject of AGRP is empty, the resulting surface order reveals no difference. We shall examine some arguments for and against the thesis that AGRP rather than VP moves, in fact arguing that VP-Raising is impossible. To begin with the arguments against AGRP-Raising, consider a case where the difference between VP and AGRP is visible at the surface: this occurs whenever AGRP has a lexical subject, a situation which arises in Acl constructions, for instance with causative verbs. Consider the DS (20), and the relevant surface

431

(20)a. das er [[sini chind mediziin studiere] laa] wil b. * das er wil laa sini chind mediziin studiere This word order turns out to be ungrammatical; hence it would seem that AGRPraising is impossible. However, the question immediately arises why this should be the case. What I should like to suggest is that (20b) is barred for Case reasons, i.e. that the infinitival subject cannot receive Case in its derived position because the trace of a clausal constituent cannot transmit Case and structural Case is only assigned at S-structure. What this means is that the ungrammaticality of (20b) cannot be used as an argument against AGRP-raising: if indeed the subject of the infinitive is not lexical and therefore does not need Case, nothing should prevent AGRP from being able to move in principle. Let us now turn to an argument in favour of the AGRP-raising analysis. In order to do so, we must first briefly discuss the motivation for both V-raising and CPExtraposition in Dutch and (Swiss) German. In the theory of Bennis and Hoekstra (1988) both VR and Extraposition are triggered by the need for verbs to be T(ense)-linked to the matrix Tense. This can happen either through Vmovement, thus providing a direct T-link, or through CP-movement: in the latter case, a T-link is created between the matrix verb and the T-index in C; the latter percolates to T, where it is available for the V which has moved to T. In this theory, CP-Extraposition is triggered by V-to(-AGR-to)-I of the matrix V, as the trace of the moved V is unable to tense-link the embedded CP. Let us consider a concrete case: a verb triggering CP-Extraposition is the verb aagCC in (21) (the example is from Swiss German, but a similar argument can be made for West Flemish): (21) das er [[[[en arie singe] chone z] aagCC] wil] (that he an aria sing can to pretend wants) (cf. (9)) I assume that in (21) the constituent embedded under the verb aagCC is at least a TP (in fact, a CP); in this I again follow Bennis and Hoekstra (1988), who argue that the presence of the element z (Dutch te, English to) signals the presence of a tense-projection. The verb aagCC moves to the right of the verb dominating it, wil, and the resulting cluster into the matrix T-position; as a

432

result, the embedded clause must follow and it is adjoined to the right of the matrix TP. This gives rise to the two sentences in (22), depending on whether or not VPR has taken place on the lowest clause [en arie singe]: (22)a. das er wil aagCC, z chone en arie singe b. das er wil aagCG, en arie z chone singe Now consider (21) again: suppose that aagCC does not undergo V-raising but that its maximal projection undergoes VP-movement, so that its CP-complement is carried along; the resulting configurations will be as in (23): (23)a. * das er wil z chone en arie singe aagCC b. * das er en arie wil z chone singe aagCC These sentences are ungrammatical. Yet there is nothing that bars this configuration: the V aagCC is T-linked by VP-Raising as required; the link between aagCC and its CP-complement should be possible as CP is still in the strict government domain of aagCC, and the link between C and T of the CPcomplement should pose no problems either. What these facts suggest is that in (23) the motivating factor for CP-Extraposition is still present. The latter being V-to-AGR movement of the dominating verb, it must be the case that the V aagCC has moved out of its VP and is in the head position of the AGRP immediately dominating it. But given the surface order of these sentences, this implies that the relevant AGRP is to the right of the matrix verb wil, i.e. that AGRP-raising has taken place. One could now raise the question why VP-Raising is apparently excluded in principle, as suggested by (23). I should like to argue that this follows from a particular proposal concerning the internal structure of the AGRP-complement of modal verbs. In Vanden Wyngaerd (1988) I argued that elements of inflectional morphology, such as Passive Morphology (PM) and Infinitival Morphology (IM), head their own projections; concretely, they occupy the head position of AGRP, where they are assigned the external 9-role by VP. Now, if VP were to raise out of AGRP, the V would be separated from its inflectional ending IM and could not merge with it without violating the usual constraints on head movement; alternatively, if it does not move, Baker's (1988:140) Stray Affix Filter or some variant thereof will be violated.7

433

3.2. Consequences Given that to explain the ungrammaticality of (23), we crucially assumed that the complement of wil was an AGRP, these sentences also undermine the thesis that the complements of modal auxiliaries are VPs. They thus provide indirect evidence for the constituency of infinitival complements. Second, in view of our revision of the VPR analysis, our account of (5b), repeated here, must be amended. (24) das er [PRO [em Karajan t]] wil en arie vorsinge We assumed that the dative adjoined to the embedded VP and that subsequently the lower segment of this VP was raised, an option unavailable under the theory we are now adopting: if AGRP raises the dative is carried along, so that (24) seems underivable. Therefore, it must be the case that LNS moves the dative into the matrix clause. Concerning this movement, there is evidence for the fact that LNS makes no use of [SPEC,CP]. Relevant examples are found in (27): (25)a. b. * c. * (26)a. b. * c. *

dat ik geprobeerd heb Marietje een boek te laten lezen (that I tried have Mary a book to let read) dat ik Marietje geprobeerd heb om een boek te laten lezen dat ik Marietje beweerd heb dat ik een boek liet lezen dat ik geprobeerd heb Jan een boek te geven (that I tried have Jan a book to give) dat ik Jan geprobeerd heb om een boek te geven dat ik Jan beweerd heb dat ik een boek gaf

The sentence (25a) is derived by Extraposition of CP and yields a perfect sentence; (25b) on the other hand, must have been derived by prior application of LNS to the subject embedded under the causative, Marietje, moving it into the matrix. This operation crosses one subjacency barrier (CP by inheritance) and is indeed unacceptable. In (25c), then, LNS crosses two barriers: following Chomsky (1986:37) we assume that tensed TP adds a barrier; furthermore, [SPEC,CP] being unavailable, CP is a barrier by inheritance and the sentence is completely out. These judgments are reproduced in (26) and the same explanation holds.8

434 An objection that could be raised against the present analysis is that we have provided no explanation for the fact that some dialects apparently allow AGRPraising besides the usual V-Raising.9 Though admittedly the formulation of the parameter distinguishing dialects that allow VPR from those that do not has not become any easier, the VPR analysis to a certain extent faces the same problem, in that it must address the issue why AGRP complements cannot raise; hence I do not consider this to be a counterargument against the current proposal. Moreover, from a certain point of view the paradigm we claim to exist makes perfect sense: T-linking of an embedded verb takes place either through moving the verbal head of the complement or by extraposing the entire complement. In a number of cases this pattern follows either directly or indirectly from the usual constraints on movement (notably subjacency and the ECP), as the reader may verify. One case, however, warrants some further discussion. Consider a verb like beloven 'promise,' which can trigger either VRaising or Extraposition. (27)a. dat Jan de krant belooft te lezen (that Jan the paper promises to read) b. dat Jan belooft (om) de krant te lezen The former sentence is derived by Head-movement of the embedded verb, the latter by extraposing the entire complement. However, dialects with VPR should in principle allow a third option, i.e. movement of the embedded VP/AGRP. Assuming, as before, that te is located in the head position of TP, this should give rise to the following sentence: (28) * dat Jan te belooft de krant lezen This sentence is ungrammatical, also in VPR dialects. By consequence, the generalisation that either the verbal head or the entire complement moves, is well-supported.10 Another consequence is that one must allow for Adverb-movement rules. Consider (29):

435

(29)a. dase morgen wil kommen (that-she tomorrow wants come) b. dase wil morgen kommen An adverb may be stranded to the left of the verb under VP-Extraposition and still have narrow scope. This is most readily explained if the adverb has moved from its base position to the matrix VP.

3.3. VPR or Extraposition? H&VR present a number of arguments to the effect that VPR should not be analysed by means of some rule of Extraposition. Given that Extraposition does not exist as a rule but is rather a descriptive label for one particular instance of Move-«*, the debate whether or not VPR should be analysed by means of a rule of Extraposition is to a certain extent academic. However, under our analysis VPR shares one important feature with what is traditionally referred to as Extraposition, viz. it involves adjunction of a maximal projection to a maximal projection. On the other hand it differs from Extraposition in the categorial status of the element being moved, i.e. AGRP instead of CP. Let us then consider H&VR's arguments against the Extraposition analysis of VPR. The first concerns the example (20) discussed above; their claim is that under an Extraposition analysis it remains quite mysterious why (20b) is ungrammatical; however, we have just argued that such an explanation readily presents itself in the domain of Case theory; therefore, their objection is invalid. Their second argument concerns extraction facts: subextraction is impossible from unambiguous extraposed CP-complements (cf. (30)), whereas it is from complements to wich V(P)R has applied (cf. (31)): (30)a. ?* Was hät er aagCC das er e für büecher list? (What has he pretended that he for books reads) b. ?* Was hät er aagCC e für büecher z läse? (31)a. b.

(What has he pretended for books to read) Was hät er e für büecher wele läse? (What has he for books want read) Was hät er wele e für büecher läse?

436

It is not immediately clear what empirical content there is to the claim that VPR does not involve a rule of Extraposition; hence it is equally unclear what these examples actually show. However, if the argument is intended to demonstrate that the categorial status of the extraposed category in (31) is not CP, I can agree with the conclusion. Although it remains unexplained why long subextraction from CP-complements is impossible whereas it is from AGRPs, this problem is in no way specific to the AGRP-raising analysis. Finally, H&VR argue that the sentences in (23) above support their thesis that VPR should not be analysed by means of a rule of Extraposition. This is true, they claim, because the Extraposition trigger aagCC contrasts in this respect with VR-triggers, which do allow the surface orders of (23). However, we have seen that in the context of a theory which attempts to explain the ungrammaticality of (23) in terms of the motivation for V(P)R/Extrapositon, these sentences in fact constitute a strong argument in favour of the AGRP-raising analysis.

Conclusion In this paper, our initial concern was to demonstrate that in order to analyse VPR-patterns it was unnecessary to postulate V' movement rules: any seeming movement of V' could be analysed as a sequence of LNS and "remnant VP"movement. Closer examination showed that VP-raising could in turn be analysed as AGRP-raising, and that this analysis had certain distinct advantages. This led us to propose that VP-movement is even excluded in principle. The following picture thus emerges: verbs fall apart into three categories: first, those triggering Move-CP; second, those potentially triggering Move-V/AGRP, and third, those allowing both. Dialects, on the other hand, can be divided into those that have Move-CP and Move-V on the one hand, and those that allow Move-CP, Move-AGRP and Move-V on the other. I shall leave an explanation of these paradigms as a topic for future research.

437

Notes

* I acknowledge the support received from the Belgian National Fund for Scientific Research, and should like to thank for their comments the audiences at the BKL/CBL meeting in Mons (Belgium) and the Leiden University linguistic research club, the O.O.O. Marcel den Dikken, Teun Hoekstra and Dany Jaspers provided me with valuable suggestions, which I gratefully acknowledge. Last but not least, a word of gratitude goes to Wim de Geest, whose teachings are to be held responsible for my life as a generative linguist-to-be. 1. English might provide some evidence, in that VP-adjuncts must be carried along under VP-fronting: (i) (We forced John to eat the meat cold and) a. eat the meat cold he did b. * eat the meat he did cold These indicate that nonmaximal projections of V cannot undergo movement. 2. We follow Hornstein and Lightfoot (1987) in assuming that Small Clauses are IPs. Still, we rename their IP as AGRP in view of Pollock's (1988) analysis of IP as consisting of a TP and an AGRP and in view of the fact that SCs do not contain Tense (i.e. are not to be analysed as TP). 3. The idea that apparent movement of subparts of VP can be analysed as a succession of LNS and movement of the maximal projection VP is indebted to Den Besten and Webelhuth (1987), who employ it to analyse VP-topicalisation in Dutch and German. 4. The fact that VPR, like VR, gives rise to the IPP effect is one which stands in need of an explanation: (i) a. Ik heb willen/*gewild dat huis kopen (I have want/wanted that house buy b. Ik heb dat huis willen/*gewild kopen 5. For clarity, the matrix AGRP (AGRP1) has been omitted from the trees in (11), (14) and (17), it being irrelevant to our concerns. 6. An analysis for (18f) essentially identical to ours was proposed by Den Besten (1987). 7. It must be noted that the analysis which assigns V and IM to different projections is compatible with the VP-Raising analysis only at the expense of allowing IM to hop onto V in the syntax, an option for which no independent evidence seems to exist in Dutch and German. 8. Adding the complementizer om to the extraposed clauses of the b-sentences yields a considerably degraded result, for reasons unclear.

438 9. The variation among Germanic dialects is not restricted to the one reported here. The Groningen dialect, for instance, as discussed in Schuurman (1987), reveals a phenomenon that could be analysed as VPR under te, yielding sentences like the following: (i) dat e zat te drei eerabbels schillen that he sat to three potatoes peel "that he was peeling three potatoes" See Broekman and Den Dikken (1988) for some discussion. Sentences of this sort are excluded in West Flemish, as far as I can tell, and presumably also in Swiss German. 10. One might wonder what excludes (28). One possibility is that subjacency is violated, as Move-AGRP would cross the inheritance barrier CP; however, (28) is a much stronger violation than usually arises by crossing one subjacency barrier (cf. (25b), (26b)). Another possibility is that in addition to the embedded infinitive itself, the element te must be T-linked as well, which it cannot be in preverbal position.

439

List of References Abraham, W. (hrsg.) (1985) Erklärende Syntax des Deutschen, Tübingen. Allen, R.R. (1966) The Verb System of Present-day American English, Mouton, The Hague. Anderson, J.R. and Bowers, G.H. (1973) Human associative memory, Winston, Washington. Anderson, J.R. (1976) Language, memory and thought, Erlbaum, Hillsdale, N.J. Anderson, J.R. (1983) The architecture of cognition, Harvard U.P., Cambridge, Mass. Aoun, J. and D. Sportiche (1983) "On the formal theory of government," The Linguistic Review 2, 211-236. Attal, J.-P. (1987) Grammaire et usage de l'anglais, Duculot, Paris-Gembloux. Aviles, Alejandro, Kenneth Hale, and Danilo Salamanca (1987) Insubordinate Complements in Miskito. MIT Manuscript. Presented at SSILA Symposium at the AAA Meetings, Chicago, 1987. Awoyale, Yiwola (1987) Perspectives on Verb Serialization. Paper presented at the First Niger-Congo Syntax and Semantics Workshop, Boston University. Baker, M.C. (1985) Incorporation: A Theory of Grammatical Function Changing, Doctoral Dissertation, MIT (Published in 1987 by University of Chicago Press). Baker, Mark (1988) Object Sharing in Serial Verb Constructions. McGill University Manuscript. Baker, M. (1988) Incorporation, University of Chicago Press, Chicago. Bally, Ch. (1950) Linguistique générale et linguistique française, 3e éd., Francke, Berne. Barss, A. and H. Lasnik (1986) "A Note on Anaphora and Double Objects," Linguistic Inquiry 17, 347-354. Barss, A. (1986) Chains and Anaphoric Dependence: On Reconstruction and Its Implications, Ph. D. Dissertation, MIT. Barss, A. (1988) "Paths, Connectivity, and Featureless Empty Categories," in Cardinaletti-Cinque-Giusti (1988). Bartsch, R. (1987) Norms of language: theoretical and practical aspects, Longman, London. Barwise, Jon and John Perry (1983) Situations and Attitudes, MIT Press, Cambridge, Mass.

440 Bayer, J. (1982) Zur Syntax und Semantik von wie-Sätzen und Dialogen mit wie-Fragen, Papier #60 des Sonderforschungsbereichs 99, Universität Konstanz. Bayer, J. (1984) "COMP in Bavarian Syntax," The Linguistic Review 3,209-274. Bayer, J. (1986) "The role of event expression in grammar," Studies in Language 10,1-52. Bech, G. (1955) Studien über das deutsche verbum infinitum, Munksgaard Kopenhagen. (2. unveränderte Aufl., (1983) Niemeyer: Tübingen). Beheydt, L. (1986) "The 'semantic primacy principle' in motherese," in Beheydt, L. (ed.) Language enfantin, Cahiers de l'institut de Linguistique de Louvain 12, 3/4. Belletti, A. and L. Rizzi (1986) Psych-Verbs and Th-theory, preprint. Belletti, A. (1988) "Generalized Verb Movement: On some Differences and Similarities between Italian and French," GLOW lecture, Budapest, March 1988. Belletti, A. and L. Rizzi (1988) "Psych-Verbs and 0-Theory," Natural Language and Linguistic Theory, 6.291-352 . Bennis, H. and T. Hoekstra (1984) "Gaps and Parasitic Gaps," The Linguistic Review 4,29-87. Eennis, H. & T. Hoekstra (1985) "Gaps and parasitic gaps," Linguistic Review, 4. Bennis, H. (1986) Gaps and Dummies, Foris, Dordrecht. Bennis, H. and P. Wehrmann (1987) "Adverbial Arguments," Linguistics in the Netherlands, Beukema, F. and P. Coopmans (eds.), Foris, Dordrecht. Bennis, H.J. & T.A. Hoekstra (1988) "The Tense Connection," GLOW lecture, Budapest, March 1988. Bennis, H J . & T.A. Hoekstra (this vol.) "Why Kaatje Was Not Heard Sing a Song". Benthem, J. van (1986) Essays in logical semantics, Reidel, Dordrecht. Benveniste E. (1951) "La conjonction ci dans la syntaxe gotique," BSLP, 47, 52Besten, H. den (1977) "On the Interaction of Root Transformations and Lexical Deletive Rules," unpublished paper, reproduced in Groninger Arbeiten zur Germanistischen Linguistik 20,1-III and 1-78 (1981). Besten, H. den (1981) "On the interaction of root transformations and lexical deletive rules," in W. Abraham (ed.) On the formal syntax of the Westgermania, Benjamins, Amsterdam.

441

Besten, H. den and J. Edmondson (1983) "The Verbal Complex in Continental West Germanic," in W. Abraham (ed.) (1983) On the Formal Syntax of the Westgermania, Benjamins, Amsterdam. Besten, H. den (1987) "Where do all those verbs go: Continental Germanic vs Afrikaans," Paper read at GLOW Workshop on Dialectology, Venice. Besten, H. den and G. Webelhuth (1987) "Remnant Topicalization and the Constituent Structure of VP in the Germanic SOV Languages," paper read at GLOW conference, Venice. Besten, H. den et al. (1988) "Verb Raising, Extrapositie en de Derde Constructie," Unpublished Report of Research Group Institute of General Linguistics, University of Amsterdam. Bever, Th. A. (1980) "Broca and Lashly were right: cerebral dominance is an accidental growth," in Caplan, D. (ed.) Biological studies of mental processes, MIT, Cambridge, Mass. Bever, T.G. (1983) "Some implications of the nonspecific bases of language," in Wanner, E. & Gleitman, L.R., (eds.). Bloch, B. and Träger, G.L. (1942) Outline of linguistic analysis, L.S.A., Baltimore, Md. Blom, Alied, and Saskia Daalder (1977) Syntaktische theorie taalbeschrijving, Muiderberg. Bloomfield, L. (1950) Language, 2nd Impr., Allen & Unwin, London. Bolinger, D. (1968) "Entailment and the meaning of structures," Glossa 2:119-127. Bolinger, D. (1971) Meaning and Form, Longman, London and New York. Bolinger, D. (1912) Degree Words, Mouton, The Hague. Bolinger, D. (1978) "Yes/No Questions are not Alternative Questions," in H. Hirtz (ed.) Questions, Reidel, Dordrecht:87-105. Borer, H. (1984) "Restrictive Relatives in Modern Hebrew," Natural Language and Linguistic Theory 2, 219-260. Botin, A. af (1777) Svenska spräket i tal och shift, Stockholm. Bresnan J.W. (1972) Theory of Complementation in English syntax, Diss. MIT. Bresnan, J. W. (1976) "Evidence for a Theory of Unbounded Transformations," Linguistic Analysis 2, 353-399. Bresnan, J. (1976) "Nonarguments for Raising," Linguistic Inquiry 7, 485-501. Broekman, H. and M. den Dikken (1988) "Verb (Projection) Raising in Medieval English," ms. Leiden.

442 Brückner, A. (1970) ÇÎownik etymologiczny jezyka polskiego, Wiedza Powszechna, Warszawa. Burzio, L. (1986) Italian Syntax, Reidei, Dordrecht. Cardinaletti, A., G. Cinque, G. Giusti (eds.) (1988) Constituent Structure. Papers from the 1987 GLOW Conference, Venezia/Dordrecht. Cardinaletti, A. (1988) "On pro, es and sentential arguments in German," Università di Venezia. Carey, S. (1982) "Semantic development: The state of the art," in Wanner, R. & Gleitman, L.R. (eds.). Cauwe, A. (1982) Etude syntaxique et sémantique des verbes d'apparence, University of Antwerp-UIA, unpublished MA, department of Romance philology. Chomsky, N. (1964) Current issues in linguistic theory, Mouton, The Hague. Chomsky, N. {1965) Aspects of the theory of syntax, MIT, Cambridge, Mass. Chomsky, N. (1967) "The formal nature of language," in Lenneberg, E.H. (ed.) Biological foundations of language, Wiley & Sons, New York. Chomsky, N. (1968) Language and Mind, New York. Chomsky, N.A. and M. Halle (1968) The Sound Pattern of English, Harper and Row, New York. Chomsky, N. (1975) Reflections on language, Pantheon, New York. Chomsky, N. (1977) "On wh-movement," in P. Culicover, T. Wasow and A. Akmajian, (eds.) Formal Syntax, Academic Press: New York. Chomsky, N. (1981) Lectures on Government and Binding, Foris, Dordrecht. Chomsky, N. (1986) Barriers, MIT Press, Cambridge, Massachusetts. Chomsky, N. (1986) "Changing perspectives on knowledge and use of language," Leuvense Bijdragen 75, 1-71. Chomsky, N. (1986a) Barriers, MIT Press, Cambridge. Chomsky, N. (1986b) Knowledge of Language, Praeger, New York. Chomsky, N. (1988) "Some Notes on Economy of Derivation and Representation," MIT ms., Cambridge, Mass. Chomsky, N. (1988) "Prospects for the Study of Language and Mind," 2nd lecture, Israel, unpublished. Christensen, K., - Koch and T. Taraldsen (1987) "Expletive Chain Formation and Past Participle Agreement in Scandinavian Dialects," Unpublished ms., forthcoming in Proceedings of the Dialect Workshop at GLOW1987.

443

Cinque, G. (1981) "On Keenan and Comrie's Primary Relativization Constraint," Linguistic Inquiry 12, 293-308. Cinque, G. (forthcoming) Types of A'-Dependencies, MIT Press. Cinque, G. (1988) "Ergative Adjectives and the Lexicalist Hypothesis," to appear in Natural Language and Linguistic Theory. Comrie, B. (1981) Language universal and linguistic typology, Blackwell, Oxford. Comrie, B. (1986) "Tense in Indirect Speech," Folia Linguistica 20,265-296. Cooper, R. (1986) "Swedish and the Head-Feature Convention," in L. Hellan & K. Koch Christensen (eds.) Topics in Scandinavian Syntax, D. Reidel, Dordrecht, 31-52. Couquaux D. (1980) "Place de la transformation "Montée" dans la syntaxe du français moderne," Le français Moderne 48,193-218. Day, P.S. and Ulatowska, H.K. (1979) "Perceptual, cognitive and linguistic development after early hemisphere ectomy: two case studies," Brain & Language 7, 17-34. Dechaine, Rose-Marie (1988) "Thematic and Case Conditions on Haitian Double Complement Constructions," to appear in Isabelle Haik and Claire Lefebvre (eds.). Declerck, R. (forthcoming) Tense in English: its structure and use in discourse. Dikken, M. den (1987) "Secundaire predicatie en de analyse van small clauses," in GLOT10,1-28. Dikken, M. den (1988) The Syntax of Movement Rules Affecting Verbs and Verbal Constituents, MA thesis, University of Leiden. Dinsmore, J. (1982) "The Semantic Nature of Reichenbach's Tense System," Glossa 16,216-239. Dirven, R., Radden, G. (1977) Semantische Syntax des Englischen, Athenaion, Wiesbaden Dirven, R. (ed.) (1989)^4 User's Grammar of English, Lang, Frankfurt. Dowty, David, Robert Wall and Stanley Peters (1981) Introduction to Montague Semantics, Reidel, Dordrecht. Droste, F.G. (ed.) (1985) Stromingen in de Hedendaagse Linguïstiek, U.P. Leuven. Droste, F.G. (1987) Voorde Spiegel van het denken, U.P.L., Leuven. Emonds, J. (1976) A Transformatinal Approach to English Syntax, Academic Press, New York. Enç, M. (1987) "Anchoring conditions for Tense," Linguistic Inquiry 18:4, 633659.

444 Erdmann, P. (1974) "Factive, Implicative Verbs and The Order of Operators," Studia Linguistica (Revue de linguistique générale et comparée), XXVIII; 51-63 (Lund). Ervin-Tripp, S. (1971^ "An overview of theories of grammatical development," in Slobin, D.I., (ed.) The ontogenesis of language, Ac. Press, New York. Evers, A. (1975) The transformational Cycle in Dutch and German, Doctoral diss. Utrecht (distributed in Indiana University Linguistics Club). Evers, Arn (1981) "Twee functionele principes voor de regel 'verschuif V'," GLOT 5:1,11-30. Evers, A. (1984) Clause Union in French and German. Ms. Univ. Utrecht. Fabb, N. (1984) Syntactic Affixation, Unpubl. MIT-Dissertation. Falk, C. (1987) "Subjectless Clauses in Swedish," Working Papers in Scandinavian Syntax. Department of Scandinavian Languages, University of Lund, Sweden. Fanselow, G. (1987) Konfigurationalität, G. Narr. Fauconnier G. (1982-1983) "Generalized Union," in Tasmowski L. and D. Willems (eds.) Problems in Syntax, C & C and Plenum, Ghent, 195-229. Fauconnier, G. (1985) Mental Spaces, The MIT Press, Cambridge, Mass. Fillmore, Ch. (1968) "The case for case," in Bach, E. & Harms, R.T. (eds.) Universals in linguistic theory, Holt, Rinehart and Winston, New York. Finer, D. (1985a) The Formal Grammar of Switch-Reference, Garland. Finer, D. (1985b) "The Syntax of Switch-Reference," Linguistic Inquiry 16:35-55. Fodor, J.A. (1983) The modularity of mind, MIT, Cambridge, Mass. Fodor, J.A. (1985) "Precis of The modularity of mind," The behavioral and brain sciences 8, 1-42. Freed, A. (1979) The Semantics of English Aspectual Complementation, Reidel, Dordrecht. Fukui, N. (1986) A Theory of Category Projection and its Applications, Ph.D. dissertation, MIT. Gee,

J. (1975) Perception, Intentionality and Naked dissertation, Stanford University, Stanford.

Infinitives,

Doctoral

Geest, T. van der (1977) "Some interactional aspects of language acquisition," in Snow, C.E. & Ferguson, C.A., (eds.). Geest, W. de (1972) Complémentaire constructies bij Verba Sentiendi in het Nederlands. Ph. D. Dissertation, University of Nijmegen. Geest, W. de (1973) Complémentaire constructies bij verba sentiendi, Utrecht.

445 Gelder, B. De (1985) "The cognitivist conjuring trick or how development vanishes," in Bailey, C.J. & Harris, R. (eds.) Development mechanisms of language, O.U.P., Oxford. Giusti, G. (1985) Zur Syntax der Infinitivkonstruktion mit 'zu' im Deutschen, Tesi di laurea, Università di Venezia. Givon T. (1976) "Topic, Pronoun and grammatical Agreement," in LI Ch. (ed.) Subject and Topic, New York, Academic Press. Golinkoff, R.M., (ed.) (1983) The transition from prelinguistic to linguistic communication, Erlbaum, Hillsdale, NJ.-London. Goossens, L. (1977) "Complements of 'Significant' Predicates: The Semantics of 'For' Complements in the Light of their Dutch Counterparts," in Putseys (ed.) (1977). Graffi G. (1980) "Universali di Greenberg e grammatica generativa," Lingua e Stile 15,3, 371-387. Grange, C. (1987) Selected Aspects of the syntax of expletive constructions, University of Geneva Memoire de licence. Grebe, P. (hrsg.) (1973) Der Große Duden. Gegenwartssprache, Mannheim.

Grammatik

der deutschen

Grewendorf, G. (1985) Deutsche Sprache. Teil I, ms., Universität Frankfurt. Grewendorf, G. (1987) "Kohärenz und Restrukturierung. Zu verbalen Komplexen im Deutschen," in B. Asbach-Schnitker & J. Roggenhofer (eds.) Neuere Forschungen zur Wortbildung und Historiographie der Linguistik, Narr: Tübingen. Grewendorf, G. (1987) Ergativität im Deutschen. Teil I und II, ms., Universität Frankfurt. Grimshaw, Jane (1988) "Adjuncts and Argument Structure," Lexicon Project Working Paper #21, Center for Cognitive Science, MIT. Guéron, J. & T. Hoekstra (1988) "T-chains and the costituent structure of auxiliaries," in G. Cinque (ed.) Proceeding of the GLOW conference Venice 1987. Haan, G. de (1979) Conditions on rules, Foris, Dordrecht. Haegeman, L. (1986) "INFL, COMP and Nominative Case Assignment in Flemish Infinitivais," in Features and Projections, (eds.) H. van Riemsdijk and P. Muysken, Foris, Dordrecht. Haegeman, L. and H. van Riemsdijk (1986) "Verb Raising, Scope, and the Typology of Rules Affecting Verbs," Linguistic Inquiry 17, 41-466. Haider, H. (1986) "V-second in German," in H. Haider-M. Prinzhorn (eds.) Verb Second Phenomena in Germanic Languages, Foris, 49-75.

446 Haider, H. (1987) Deutsche Syntax - Generativ. Parameter der deutschen Syntax, Habilitationsschrift, Univ. Wien. Haider, H. (1987a) "Topicalization and other puzzles of German syntax," Ms. Univ. Wien (in press in: G. Grewendorf & W. Sternefeld, Scrambling and Barriers, Benjamins, Amsterdam.) Haider, H. (1987b) "Matching projections," (to appear in G. Cinque (ed.) Proceedings of the Xlth GLOW conference, Venice). Haik, I. and CI. Lefebvre (eds.) Niger-Congo Syntax and Semantics 2 (to appear). Hale, Ken (1988) Misumalpan Verb Sequencing Constructions, MIT Manuscript. Halle, M. and J.-R. Vergnaud (1987) An Essay on Stress, MIT Press, Cambridge, Massachusetts. Hantson, A. (1983) "For, with and without as non-finite clause introducers," English Studies 64, 54-66. Hayes, B. (1984) "The phonology of rhythm in English," Linguistic Inquiry 15, 33-74. Heath, George R. (1927) Grammar of the Miskito Language, Herrnhut, Lindesbein. Heath, G. and W.G. Marx (1961) Diccionario Imprenta Calderón, Tegucigalpa.

miskito-espanol/espanol-miskito.

Helbig, G.-F. Kempter (1981) Die uneingeleiteten Nebensätze, VEB Verlag Enzyklopaedie, Leipzig. Helbig, G.-J. Buscha (1984) Deutsche Grammatik, Leipzig. Hellan, Lars (1988a) Anaphora in Norwegian and the Theory of Grammar, Foris, Dordrecht. Hellan, Lars (1988b) "The Phrasal Nature of Double Object Constructions," ms., to appear in proceedings from Workshop on Germamc Syntax, Groningen, May 1988. Hesselman, B. (1919) "Ortografiska reformeri spräkhistorisk belysning,". Spr&k och still XIX, pp. 121-149. Higginbotham J. (1985) "On semantics,". Linguistic Inquiry 16:4, 547-595. Hoeksema, J. (1981) "Verbale verstrengeling ontstrengeld," Spektator 10-3. Hoekstra, T. (1984) Transitivity. Grammatical Relations in GB-Theory, Foris, Dordrecht. Hoekstra, T. (1986) "Passive and participles," Linguistics in the Netherlands 1986, Foris Publications, Dordrecht. Holmberg, A. (1986) Word Order and Syntactic Features in the Scandinavian Languages and English, Doctoral Dissertation, University of Stockholm.

447

Holmberg, A. & C. Platzack (1988) "The role of Inflection in Syntax," Paper presented at the Fifth Workshop on Comparative Germanic Syntax, Groningen, May 1988. To be published in W. Abraham & E. Reuland (eds.) Germanic Syntax Workshop. John Benjamins, Amsterdam. Holmberg, A. (1988) "The Head of S in Scandinavian and English," McGill Working Papers in Linguistics, Special issue on Comparative Germanic Syntax. 123-55. Holmberg, A. and C. Platzack (1988b) "On the Role of Inflection in Scandinavian Syntax," Working Papers in Scandinavian Syntax, 42, 25-42. Honda, M. (1988) "N'-gaps, the DP Hypothesis and the Theory of Empty Pronominals," unpublished manuscript, Tilburg University and Osaka. Hooper, J. B. (1973) "On Assertive Predicates" Unpublished Manuscript, State University of New York at Buffalo. Hoppenbrouwers, Cor, et al. (eds.), Proeven van taalwetenschap (Tabu 16, nr. 2., [Festschrift for Albert Sassen]), Groningen. Hopper, P.J. & Thompson, S.A. (1984) "The discourse basis for lexical categories in universal grammar," Language 60, 703-752. Horiguchi, I. (1989) "Complementation," in Dirven (ed.) (1989). Hornstein, N. and D. Lightfoot (1987) "Predication and PRO," Language 63, 23-52. Huang, C.-T. James (1982) Logical Relations in Chinese and the Theory of Grammar, MIT Doctoral Dissertation. Hüllen, W. and R. Schulze (eds.) (1988) Understanding the Lexicon - Meaning, Sense and World Knowledge in Lexical Semantics, Max Niemeyer Verlag, Tübingen. Humboldt, W. (1836) Ueber die Kawi-Sprache auf der Insel Jawa. Jackendoff, R.S. (1972) Semantic Interpretation in Generative Grammar, The MIT Press, Cambridge, Mass. Jakobson, R. (1970) "Linguistics," in Main trends of research in the social and human sciences I, Mouton, The Hague. Jeanne, LaVerne (1978) Aspects of Hopi Grammar, MIT Doctoral Dissertation. Jespersen, Otto (1968) Philosophy of Grammar, Allen and Unwin, London. Johannisson, T. (1945) HAVA och VARA som tempusbildande hjälpverb i de nordiska Spraken, Lunds unversitets ärsskrift N.F. Avd. 1. Bd. 41. Nr. 7. Gleerup: Lund. Johnson, K. (1988) "Verb Raising and 'Have'," McGill Working Papers in Linguistics, Special issue on Comparative Germanic Syntax. 156-67.

448

Karttunen, L. (1971) "Some Observations on Factivity," Papers in Linguistics, 4: 55-69. Katz, J.J. (1981) Language and other abstract objects, Rowman, Totowa, N.J. Kayne R. (1980) "De certaines différences entre le français et l'anglais," Langages 60, 47-64. Kayne, R. (1982) "Predicates and Arguments, Verbs and Nouns," Newsletters, 8, 24.

GLOW

Kayne, R.S. (1983) "Chains, Categories External to S, and French Complex Inversion," Natural Language and Linguistic Theory 1,107-139. Kayne, R. (1984) Connectedness and Binary Branching, Foris, Dordrecht. Kayne, R.S. (1986) "Participles, Agreement, Se/Si and PRO," abstract for the Princeton Symposium on Comparative Syntax. Kayne, R.S. (1987) "Null Subjects and Clitic Climbing," MIT ms., Cambridge, Mass. Kayne, R. (1987) "Facets of Romance Past Participle Agreement," Unpublished ms., forthcoming in Proceedings of the Dialect Workshop at GLOW 1987. Keenan, E. and B. Comrie (1977) "Noun Phrase Accessibility and Universal Grammar," Linguistic Inquiry 8, 63-100. Kempson, R. M. (1975) Presupposition and the Delimination of Semantics, Cambridge University Press, Cambridge. Kiparsky, P. (1966) "Über den deutschen Akzent," Studio Grammatica VII. Untersuchungen über Akzent und Intonation im Deutschen, Akademie-Verlag, Berlin. Kiparsky, P. and C. Kiparsky (1970) "Fact," in M. Bierwisch and K. Heidolph (eds.) Progress in Linguistics, Mouton, the Hague. Kiparsky, P. and C. Kiparsky (1970) "Fact," in D. Steinberg and J. Jakobovits (eds.), Semantics, Rinehart, Winston & Holt, New York. Kiparsky, P. (1979) "Metrical structure assignment is cyclic," Linguistic Inquiry 10, 421-441. Kirsner R. and S. Thompson (1976) "The role of pragmatic Inference in Semantics," Glossa 10, 200-240. Kiss, Katalin (1987) "More on Anaphora and Double Objects," ms., Budapest. Kitagawa, Y. (1986) Subjects in Japanese and English, Doctoral Dissertation, University of Massachusetts at Amherst. Klinghardt H. (1877) "Die Syntax der gotischen Partikel ei," ZfdPh, 8, 127-180; 289-329.

449 Klooster, W.G. (1984) "Ontkenning en noodzakelijkheid. Observaties met betrekking tot negatie en moeten," GLOT, 7,1,63-120. Klooster, W.G. (1986) "Problemen met complementen," in Cor Hoppenbrouwers et al. (eds.), 122-132. Koopman, H. (1984) The Syntax of Verbs, Dordrecht, Foris. Koopman, H. and D. Sportiche (1985) Theta Theory and Extraction, paper presented at the 1985 GLOW Colloquium. Koperberg, A. (1987) Small Clauses and Big Clauses Compared. MA Thesis, University of Amsterdam. Köster, J. (1974) "Het werkwoord als Spiegelcentrum," Spektator 3, 601-618. Köster, J. (1978) "Why Subject Sentences Don't Exist," in S.J. Keyser (ed.) Recent Transformational Studies in European Languages, MIT Press. Köster, J. (1984) "On Binding and Control," Linguistic Inquiry 15,417-459. Koster, J. (1987) Domains and Dynasties. The Radical Autonomy of Syntax, Foris, Dordrecht. Koster, J. (1988) "The Residual SOV Structure of English," Groningen Papers in Theoretical and Applied Linguistics, Tenk Nr. 5, University of Groningen. Kraak, A. and W.G. Klooster (1972) Syntaxis, Culemborg/Cologne. Langacker, R. (1985) "Cognitieve grammatica en grammaticale valentie," in Droste, F.G. (ed.) Stromingen in de hedendaagse linguistiek, U.P.L.-Van Gorcum, Leuven-Ássen, Maastricht. Langacker, R. W. (1987a) Foundations of Cognitive Grammar, Vol.1. Theoretical Prerequisites. Stanford University Press, Stanford. Langacker, R. W. (1987b) "Nouns and Verbs," Language, 63:53-94. Latour, B. (1979) "Zur Fakultativität des Pronomens es als Korrelat satzförmiger Ergänzungen," in A. Wierlacher, D. Egger, U. Engler (Hrsg.) Jahrbuch Deutsch als Fremdsprache 1, Heidelberg, 240-253. Leech, G. (1971) Meaning and the English Verb, Longman, London. Leisi, E. (1969) Das heutige Englisch, Carl Winter, Heidelberg. Lenneberg, E.H. (ed.) (1967) Biological foundations of language, Wiley & Sons, New York. Levin, Beth and Malka Rappaport (1986) "The formation of Adjectival Passives," Linguistic Inquiry 17, 623-662. Lewandowska-Tomaszczyk, B. (1987) Conceptual Analysis, Linguistic Meaning, and Verbal Interaction, Acta Universitatis Lodziensis, fadi.

450 Lewandowska-Tomaszczyk, B. (1988a) "The Increment Value of Predicates in the Semantic Lexicon," in Hüllen and Schulze (1988). Lewandowska-Tomaszczyk, B. (1988b) "Conceptualizations of Non-Assertive Meanings," paper submitted to Symposium on Cognitive Linguistics, Duisburg 1989. Liberman, M. (1975) The Intonational System of English, Doctoral dissertation, MIT, Cambridge, Massachusetts. Liberman, M. and A. Prince (1977) "On stress and linguistic rhythm," Linguistic Inquiry 8, 249-336. Lightfoot D. (1979) Diachronic Syntax, Cambridge UK. Lindholm, J.M. (1969) "Negative-Raising and sentence pronominalization," Chicago Linguistic Society 5:148-158. Longobardi G. (1978) La sintassi gotica. Aspetti e problemi, Diss. Univ. di Pisa. Manzini, R. (1983) Restructuring and Reanalaysis, Unpubl. MIT-Dissertation. Marantz, Alec (1984) On the Nature of Grammatical Relations, MIT Press, Cambridge, Mass. Marx-Moyse, J. (1983) Untersuchungen zur deutschen Satzsyntax. Es als vorausweisendes Element eines Subjecktsatzes, Heft 44, Zeitschrift für Dialektologie und Linguistik, F. Steiner Verlag, Wiesbaden. Massam, Diane (1985) Case Theory and the Projection Principle, MIT Doctoral Dissertation. Monk, R. (1980) "Productive reasoning and the structure of scientific research," in Nickles, T. (ed.) Scientific discovery, logic and rationality, Reidel, Dordrecht. Montague, Richard (1974) "The Proper Treatment of Quantification in English," in Montague, R. Formal Philosophy, Yale Univ. Press, New Haven, 247-70. Moortgat, M. (1988) "Generalized categorial grammar," in Droste, F.G. (ed.) Main streams in linguistics, Benjamins, Amsterdam-New York. Morris, W. (ed.) (1969) The American Heritage Dictionary of the English Language, American Heritage Publ. Company and Houghton Mifflin Company, Boston. Muysken, P. (1983) "Parametrizing the Notion "Head"," Journal of Linguistic Research 2,57-76. Nespor, M. and I. Vogel (1979) "Clash avoidance in Italian," Linguistic Inquiry 10, 467-482. Nespor, M. and I. Vogel (1986) Prosodic Phonology, Foris, Dordrecht. Newman J. (1981) The semantics of Raising Constructions, University of California, San Diego, PhD.

451

Nieuwenhuijsen, P. (1974) "Infinitief ipv Deelwoord," Spektator 3,6. Norwood, Susan (1987) Sumu. CIDCA Manuscript. Oh, Choon-Kyu (1974) "More on Degree of Factivity," University of Kansas, Mimeograph. Ota, A. (1963) Tense and Aspect of Present-day American English, Kenyusha, Tokyo. Palmer, F.R. (1974) The English Verb, Longman, London. Phillips, J. (1973) "Syntax and vocabulary of mother's speech to young children: age and sex comparisons," Child development 44,182-185. Pinchon J. (1986) Morphosyntaxe du français, Paris, Hachette. Pitz, Anneliese (1987) Middle Constructions in German, hovedoppgave, Univ. of Trondheim. Platzack, C. (1983) "Three Syntactic Changes in the Grammar of Written Swedish around 1700," Struktur och variation. Festskrift till Bengt Loman 7.8.1983. Publications of the Research Institute of the Abo Akademi Foundation, Abo 43-63. Platzack, C. (1986a) "The Position of the Finite Verb in Swedish," In Haider, H. & M. Prinzhorn (eds.) Verb Second Phenomena in the Germanic Languages, Foris, Dordrecht, 27-47. Platzack, C. (1986b) "The Structure of Infinitival Clauses in Danish and Swedish," in Dahl, Ö. & A. Holmberg (eds.) Scandinavian Syntax, Linguistics Department, University of Stockholm. Platzack, C. (1987) "The Case of Narrative Inversion in Swedish and Icelandic," Working Papers in Scandinavian Syntax 31. 9-14. Platzack, C. (1987) The Emergence of a Word Order Difference in Scandinavian Subordinate Clauses, ms., University of Lund. Pollock J.-Y. (1986) "Sur la syntaxe de en et le paramètre du sujet nul," in Ronat M. & D. Couquaux (eds.) La grammaire modulaire, Paris, Editions de Minuit, 211-246. Pollock, J.-Y. (1987) Verb Movement, UG and the Structure of IP, ms., Université de Haute-Bretagne, Rennes II. Pollock, J.-Y. (1988) "Verb Movement, UG and the Structure of IP," Ms., Université de Haute Bretagne, Rennes II. Popescu L. and L. Tasmowski (1988) "Thématicité et possession en roumain," Lingvisticae Investigationes 12, to appear. Postal, Paul (1972) "A Few Factive Facts," Linguistic Inquiry, 3:396-400. Postal, P. (1974) On Raising, MIT Press, Cambridge, Mass.

452 Prince, A. (1983) "Relating to the grid," Linguistic Inquiry 14, 19-100. Pullum, G.K. (1983) "How many possible human languages are there?," Linguistic Inquiry 14, 3, 447-467. Pustejovsky, J. (1988) Knowledge Representation and the Computational Lexicon, lectures delivered at Pisa Summer School of Computational Lexicology an Lexicography, Pisa, Aug. 8-13,1988. Putseys, Y. (ed.) (1977) Aspects of English and Netherlandic Grammar, Acco, Leuven. Putseys, Y. (1987) "A Modular Approach to the Grammar of English WhDeterminers," in A.M. Simon-Vandenbergen (ed.) Studies in honour of René Derolez, R.U. Gent, Ghent. Pütz, H. (19862) Ueber die Syntax der Pronominalform "es" im modernen Deutsch, Tübungen. Pylyshyn, Z.W. (1973) "The role of competence theories in cognitive psychology," Journal of Psycholinguistic Research 2, 21-51. Pylyshyn, Z.W. (1977) "What the mind's eye tells the mind's brain: a critique of mental imagery," in Nicholas, J.M. (ed.) Images, perception and knowledge, Reidei, Dordrecht-Boston. Quicoli, A.C. (1982) "Some Issues on the Theory of Clitics," Linguistic Analysis 10-3,203-273. Quine, W.V.O. (1960) Word and object, MIT, Cambridge, Mass. Quirk R., Greenbaum S., Leech G., Svartvik J. (1973) A Grammar of Contemporary English, Longman, London. Quirk R., Greenbaum S., Leech G., Svartvik J. (1985) A Comprehensive Grammar of the English Language, Longman, London. Ramat P. (1976) (ed.) La tipologia linguistica, Bologna. Ramat P. (1984) Linguistica tipologica, Bologna. Ransom, E. (1986) Complementation, Amsterdam.

its meanings and forms,

Benjamins,

Reis, M. (1974) "Assertion, Negation and "Fact". Anmerkungen zu P. und C. Kiparsky (1971): Kapitel 2," Papiere zur Linguistik 6: 73-89. Reis, M. (1985) "Satzeinleitende Strukturen im Deutschen: Ueber COMP, Hauptund Nebensätze, w-Bewegung und die Doppelkopfanalyse," in W. Abraham (1985), 269-298. Reuland, E. (1981) "On the Governing Properties of Infinitival Markers," in Fretheim, T. & L. Hellan (eds.) (1981) Papers from the sixth Scandinavian Conference of Linguistics.

453 Reuland, E. (1982a) "Why count your auxiliaries in Dutch," NELS12, MIT. Reuland, E. (1982b) "Dependencies and violations of configurationality," hand-out, GLOW-conference, Paris. Reuland, E. (1983) "Government and the Search for Auxes: A Case Study in Cross-Linguistic Category Identification," in F. Heny & B. Richards Linguistic Categories: Auxiliaries and Related Puzzles, Vol. 1, Reidel, Dordrecht. Riddle, E.M. (1978) Sequence of Tenses in English, Unpublished Doctoral dissertation, University of Illinois at Urbana-Champaign. Riemsdijk, H. C. van (1975) "A Case for a Trace: Preposition Stranding in Züritiiütsch," in A. Kraak (ed.) Linguistics in the Netherlands 1972-1973, Van Gorcum, Assen/Amsterdam. Riemsdijk, H.C. van (1978) A Case Study in Syntactic Markedness. The Binding Nature of Prepositional Phrases (doctoral dissertation), Lisse. Riemsdijk, H. van (1987) Against Adjunction to the Head. Ms. Tilburg Univ. Riemsdijk, H. C. van (1988) "Against Adjunction to the Head," McGill Working Papers in Linguistics, Special Issue on Comparative Germanic Syntax, 239259. Riviere N. (1981) La construction impersonnelle en français Editions J. Favard, Saint-Sulpice de Favières.

contemporain,

Rizzi, L. (1987) "Relativized Minimality," Ms., Université de Genève. Roberts, LG. (1987) The representation of the dethematized subjects, Foris Publications, Dordrecht. Rooryck J. (1988) Montée et contrôle revisités, University of Leuven - KUL, department of linguistics, preprint. Rosch, E. (1977) "Human categorization," in Warren N. (ed.) Studies in crosscultural psychology, Vol. 1, Academic Press, London. Rosenbaum, P. (1967) The grammar of English predicate constructions, MIT Press, Cambridge, (Mass.).

complement

Ross, J.R. (1967) Constraints on Variables in Syntax, Ph. D. thesis, MIT. Ruwet N. (1972) Théorie syntaxique et syntaxe du français, Le Seuil, Paris. Ruwet, N. (1978) "Une construction Investigationes II, 165-210.

absolue

en

français,"

Linquisticae

Ruwet N. (1983) "Montée et Contrôle: une question à revoir?", Revue romane n° spécial 24,17-37. Ruwet N. (1986) Les verbes météorologiques, preprint.

454

Salamanca, D. and Ernesto Scott (1985) Miskitu Bila Aisanka (Gramatica Mis/cita), Centro de Investigaciones y Documentación de la Costa Atlantica. Salamanca, Danilo (1988) Elementos de Gramatica del Miskitu. MIT Doctoral Dissertation. Scheurweghs, G. (1959) Present-Day English Syntax, Longmans, London. Schnelle, H. (1976) "Transcendentale und empirische Sprachgemeinschaften," in Apel, K.O. (ed.) Sprachpragmatik und Philosophie, Suhrkamp, Frankfurt. Schuurman, I. (1987) "Incorporation in the Groningen Dialect," in F. Beukema and A. Hulk (eds.) Linguistics in the Netherlands 1987, Foris, Dordrecht. Selkirk, E.O. (1984) Phonology and Syntax. The Relation between Sound and Structure, MIT Press, Cambridge, Massachusetts. Seuren, Pieter A.M. (1985) Discourse Semantics, Blackwell, Oxford. Sinclair, J. et al. (ed.) (1987) Collins COBUILD English language dictionary, Collins, Glasgow. Snow, C.E. and Ferguson C.A. (eds.) (1977) Talking to children. Language input and acquisition, C.U.P., New York. Snow, C.E. (1977) "Mother's speech research: from input to interaction," in Snow, C.E. & Ferguson, C.A. (eds.) Talking to children. Language input and acquisition, C.U.P., New York. Spears, A.K (1974) "Complements of Significant-Class Predicates: A study in the Semantics of Complementation," Papers from the ninth Regional Meeting of the Chicago Linguistic Society: 627-638. Stemmer, N. (1981) "A note on empiricism and structure-dependence," Journal of child language 8, 649-663. Stowell, T. (1981) Origins ofphrase structure, diss. MIT. Stowell, T. (1987) "As so, not so as," ms. UCLA. Svartnik, J. and R. Quirk (1980) A Corpus of English Conversation, CWK Gleerup, Lund. Tasmowski L. (1980) "Un "devoir" opérateur," Travaux de linguistique 1,43-58. Tasmowski L. (1984) "?*Lui faire téléphoner quelqu'un d'autre, une stratégie?", Lingvisticae Investigationes 8, 403-427. Ulvestad, B., H. Bergenholtz (1979) "Es als "Vorgreifer" eines Objekt-satzes," Deutsche Sprache, 7.97-116. Verhagen, A. (1986) Linguistic Theory and the Function of Word Order inDutch; a study on interpretive aspects of the order of adverbials and Noun Phrases, Foris, Dordrecht.

455

Vihman M.M., Macken M.A., Miller R., Simmons H. and Miller J. (1985) "From babbling to speech: a reassessment of the continuity issue," Language 61, 397-445. Vikner, S. and R. Sprouse (1987) "Have/Be Selection as an A-Chain Membership Requirement," Working Papers in Scandinavian Syntax 38, University of Lund. Vikner, S. and R. Sprouse (1988) "Have/Be-Selection as an A-Chain Membership Requirement," Working Papers in Scandinavian Syntax 38, Department of Scandinavian Languages, University of Lund, Sweden. Visser, F.Th. (1963-1973) An Historical Syntax of the English Language, Brill, Leiden. Wackernagel, J. (1892) "Über ein Gesetz der Indogermanischen Wortstellung," Indogermanische Forschungen 1, 334-436. Wanner, E. and Gleitman, L.R. (1982) Language acquisition: The state of the art, C.U.P., Cambridge. Webelhuth, G. & H. den Besten, (1988) "Adjunction and Remnant Topicalization in the Germanic SOV-Languages," ms. Wickboldt, J. (s.d.) "An Examination of a corpus of hate + verbal complement constructions and especially the "hate to" hedge," Manuscript, University Leuven. Wiese, R. (1986) "Syntax und Phonologie - Ein Überblicksartikel anhand von Elisabeth Selkirk "Phonology and Syntax," Linguistische Berichte 103, 252-276. Wilkinson, R. (1970) "Factive Complements and Action Complements," Papers from the Sixth Regional Meeting of the Chicago Linguistic Society. Willems D. (1983) "Regarde voir," in Mélanges MOURIN, Romanica Gandensia 20, 147-15. Williams, E. (1980) "Predication," Linguistic Inquiry 11, 203-238. Williams, Edwin (1985) "PRO and Subject of NP," Natural Language and Linguistic Theory 3, 297-216. Wilson, D. (1972) "Presuppositions on Factives," Linguistic Inquiry, 3. Woisetschlaeger, E. (1980) "A note on the autonomy of syntax," Journal of Linguistic Research 1, 55-70. Wood, F.T. (1956) "Gerund versus Infinitive," English Language Teaching XI, 11-16.

Wright, G.H. von (1963) Norm and Action, Routledge and Kegan, London. Wyngaerd, G. Vanden (1988) "Object-Shift as an A-movement rule," UFSAL/NFWO paper, Brussels.

456 Wyngaerd, G. Vanden (1988) "Passive and the Analysis of Auxiliary Verbs," in P. Coopmans and A. Hulk (eds.) Linguistics in the Netherlands 1988, Foris, Dordrecht. Zagona, K. (1988) "Proper Government of Antecedentless VP in English and Spanish," Natural Language and Linguistic Theory 6. 95-128. Zuber, R. (1973) "Decomposition of Factives," LAUT, Trier. Zubizarretta, M.L. (1985) "The Relation between Morphophonology and Morphosyntax: The Case of Romance Causatives," Linguistic Inquiry 16:247289. Zubizarretta, Maria Luisa (1987) Levels of Representation in the Lexicon and in the Syntax, Foris, Dordrecht.