Infinitives: Restructuring and Clause Structure [Reprint 2012 ed.] 9783110908329, 9783110178647

Citation preview

Infinitives

W G DE

Studies in Generative Grammar 55

Editors

Henk van Riemsdijk Harry van der Hulst Jan Köster

Mouton de Gruyter Berlin · New York

Infinitives Restructing and Clause Structure

by

Susanne Wurmbrand

Mouton de Gruyter Berlin • New York

2003

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

The series Studies in Generative Grammar was formerly published by Foris Publications Holland.

® Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability.

First published as a hardcover in 2001.

ISBN 3-11-017864-8 Bibliographic

information published by Die Deutsche

Bibliothek

Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at .

© Copyright 2001, 2003 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Prinding & Binding: Hubert & Co, Göttingen Printed in Germany. Cover design: Christopher Schneider, Berlin.

Acknowledgements

This book began as my doctoral dissertation and although the work has undergone substantial changes and revisions, many of the ideas presented here originate in the dissertation. I would therefore like to reiterate my thanks to all the people mentioned in the acknowledgements of my thesis. I am especially indebted to Noam Chomsky, Alec Marantz, and David Pesetsky not only for their support during my years at MIT but for continuing advice and encouragement over the past three years. I am also grateful to Henk van Riemsdijk who suggested reworking the dissertation into a book, and to an anonymous reviewer for many helpful and inspiring comments which in part led to the decision to substantially rewrite the dissertation. Over the years, I have had many interesting and fruitful (linguistic and non-linguistic) discussions with many fellow linguists which in one way or another highly influenced the work reported here. I would especially like to thank Elena Anagnostopoulou, Jonathan Bobaljik, Christine Czinglar, Danny Fox, Morris Halle, Kleanthes Grohmann, Martin Hackl, Hubert Haider, Meltem Kelepir, Katharina Köhler, Idan Landau, Winnie Lechner, Wilfried Öller, Claudia Plaimauer, Martin Prinzhorn, Mamoru Saito, Uli Sauerland, Philippe Schlenker, Lisa Travis, Kazuko Yatsushiro, and the members of the Dutch/Hungarian project on verb clusters. Portions of the material were presented at WCCFL 18 (Tucson), the workshop on Move and interpret alpha (Kanda University), CGSW 15 (Groningen), WCCFL 20 (Los Angeles), the 25th Penn Colloquim, GLOW 24 (Braga, Portugal), as well as at the University of Potsdam, University of Vienna, Tilburg University, Leiden University, University of Salzburg, Harvard University, Nanzan University, and McGill University. I am thankful to the audiences there for questions and comments. [Chapter 4 will also appear in the Proceedings of CGSW 15, and parts of Chapter 3 have been published in the Proceedings of WCCFL 18.]

vi

Acknowledgements

Some of the research reported here was supported by a fellowship from the Social Sciences and Humanities Research Council (Canada). I would also like to thank Erika Lawrance for her incredible and invaluable effort in proofreading the final manuscript. Ein grosses Dankeschön gebührt auch meinen Eltern und Tante Heli für die fortwährende Unterstützung aus der Heimat. Finally and most importantly, I would like to thank Jonathan Bobaljik for his invaluable advice, support and constant encouragement—Jonathan's contribution to this book is greater than I can express in words.

This book is dedicated to Auguste and Josef Kraker

Table of contents

Chapter 1: Introduction 1. 2.

Infinitives in brief Restructuring—its significance in history 2.1. The class of restructuring predicates 2.2. The structure of restructuring infinitives 2.2.1. Mono-clausal approaches 2.2.2. Bi-clausal approaches 2.3. The motivation for restructuring

Chapter 2: Lexical restructuring 1. 2.

Introduction Case assignment in restructuring constructions 2.1. A-movement—basic analysis 2.2. Other instances of vP-less infinitives 2.2.1. Unaccusative restructuring predicates 2.2.2. Some crosslinguistic evidence for vP-less infinitives 2.3. Restructuring vs. non-restructuring 2.3.1. German 2.3.2. Japanese 2.4. The lack of case and bi-clausal approaches to restructuring 3. Tense properties of restructuring constructions 3.1. Infinitivaltense 3.2. Tense vs. irrealis aspect 3.3. Restructuring infinitives are tenseless 3.4. Overt tense markers in Japanese infinitives 3.5. The restructuring/tense connection 4. Other functional projections 4.1. The restructuring configuration and functional projections 4.2. CP-properties of restructuring infinitives 4.2.1. Overt complementizers 4.2.2. ^-specifiers

1 1 5 6 9 10 11 14 16 16 18 19 25 25 31 35 36 46 57 62 62 72 79 85 91 99 99 101 102 105

viii

Table of contents

4.3. Infinitival marker 4.4. Negation 5. Verb raising in restructuring constructions 5.1. The verb raising/restructuring connection 5.2. Tense dependencies 5.3. Intervening heads 5.4. Dislocation of restructuring complements 6. Summary Chapter 3: Functional restructuring 1. 2.

Introduction The split IP parameter 2.1. English vs. German

2.2. German clause structure Motivating the lexical/functional distinction 3.1. Restructuring predicates as auxiliaries 3.2. Lexical/functional differences in German 3.2.1. Extraposition 3.2.2. The "infinitive for participle" effect 3.2.3. Thematic properties 3.3. Restructuring and functional structure (Cinque 2000) 4. German functional constructions 4.1. Modal constructions 4.1.1. Some ordering issues 4.1.2. Modal verbs must be raising verbs 4.1.2.1. The subject starts out below the modal 4.1.2.2. Passive in modal constructions 4.1.2.3. Do modals assign theta roles? 4.1.3. Summary 4.2. Raising constructions 4.3. Semi-functional constructions 5. Summary 3.

Chapter 4: The infinitival subject 1. 2.

Introduction Infinitives with and without a syntactic subject 2.1. Binding

109 115 118 120 126 127 132 135 137 137 139 139 143 145 146 155 156 162 167 176 182 182 184 187 188 196 203 204 205 215 224 226 226 227 229

Table of contents 2.1.1.

Transitive constructions

ix 230

2.1.2. Ditransitive constructions 2.2. Semantic control properties of restructuring infinitives 2.2.1. The obligatory/non-obligatory control distinction

233 236 236

2.2.2. Restructuring infinitives involve obligatory control 3. What is under control 3.1. Main proposal 3.2. Syntax * semantics 3.3. "It" anaphors 4. Conclusion

241 246 246 251 253 263

Chapter 5: Grades of (non-)restructuring 1. 2. 3.

Introduction Restructuring vs. non-restructuring Graded (non-)restructuring 3.1. Restructuring vs. reduced non-restructuring 3.2. Non-restructuring 3.2.1. Relative clause pied piping 3.2.2. Extraposition vs. intraposition 4. Semantic classification of infinitives 4.1. Two ways to forget 4.2. Factive (and propositional) infinitives

265 265 266 272 273 286 287 291 297 298 302

Chapter 6: Summary and conclusion

308

Appendix

317

1. 2. 3.

Overview Verb classification Distinguishing irrealis, propositional, factive 3.1. Irrealis properties 3.2. Factive, propositional 4. (Non-)Restructuring properties 4.1. Passive 4.2. Pronoun fronting 4.3. Scrambling 5. Verb classifications in different languages

317 318 319 319 323 327 328 332 337 341

χ

Table of contents 5.1. 5.2. 5.3. 5.4. 5.5. 5.6. 5.7. 5.8. 5.9. 5.10. 5.11.

Rizzi (1976, 1982) Napoli (1981) Zagona (1982), Aissen and Perlmutter (1976, 1983) Picallo (1985) Burzio (1986) Fanselow (1989) Rutten (1991) Haider (1993) Sabel (1994/1996) Cinque (1997) Grosse (2000)

343 343 343 343 343 344 344 344 344 345 345

References

346

Index

366

Chapter 1 Introduction

1. Infinitives in brief Infinitival complementation constructions provide a set of intriguing and deep challenges for contemporary linguistic theories. Not least among the questions raised are the various facts which contribute to the general observation that two apparently identical infinitival clauses may vary quite substantially in their syntactic behavior and their semantic interpretation. For example, an infinitival clause like [to read a book] may, depending on the environment in which it occurs, be opaque or transparent for extraction, may itself be movable or may be immobile, may or may not receive a variable interpretation for the understood infinitival subject, and it may or may not permit a tense interpretation distinct from that of the matrix clause (facts of this sort will be illustrated in detail as we proceed). Describing and explaining the correlations both between semantic properties and syntactic ones, and among various syntactic properties, has proven to be no easy task. In particular, a substantial literature has been devoted to cases in which infinitival complement clauses appear to be transparent domains for syntactic phenomena that are otherwise quite local, e.g., clause-bounded. (This is the phenomena of restructuring to which we return presently.) This book offers a detailed case study of infinitival constructions in German, though the properties we discuss can be generalized to other languages, provided the particular syntactic properties of any given language are taken into consideration. The primary thesis of the book is that infinitival complements do not all have the same functional (i.e., syntactic) architecture above the VP and that differences in the syntactic structure among different classes of infinitival

2

Introduction

complement clauses correspond closely (though not perfectly) to differences in interpretation. A particular emphasis is placed here on pursuing an explanatory approach, i.e., an approach which not only provides a framework for the description of the syntactic phenomena but one which attempts to predict the distribution of various properties across clause types and to explain the observed correlations between syntactic and interpretive differences among related sentences. The major claim of this study is that control and raising infinitives do not form uniform classes but are represented by a variety of configurations, which reasonably directly reflect the syntactic and semantic properties of the different infinitival constructions. Concentrating on German, we show that infinitival complements fall into four basic classes: lexical restructuring infinitives, functional restructuring infinitives, reduced non-restructuring infinitives, and non-restructuring infinitives. The distribution and properties of these types of infinitival constructions are summarized in Table 1. This study pursues four main goals: i) to provide evidence for (at least) a four-way classification of infinitival constructions; ii) to motivate the different syntactic structures for the different classes of infinitival constructions; iii) to derive the operations and properties listed in Table 1; and iv) to provide an explanatory account which not only derives but predicts the distribution in Table 1. There are two crucial points that we would like to direct the reader's attention to now concerning the distribution in Table 1. First, as the table shows, restructuring is not treated as a uniform phenomenon but is split into two subclasses—lexical vs. functional restructuring (an overview of the restructuring phenomenon will be provided in section 2). Thus, we argue against approaches such as the one suggested by Cinque (1997a, 1997b, 2000) which take all restructuring to be functional. Second, the table illustrates that the distinction between different types of infinitival constructions is not a simple distinction between VP and clause, but a more fine grained distinction. Consequently, there is no single phenomenon of restructuring per se, and there can be no tests that pick out a "restructuring property" or feature. Rather, different tests are shown to pick out distinct aspects of the syntactic structure in a principled manner.

Infinitives in brief Table 1. Type

3

German infinitival constructions Structure Properties, distribution

(Im)Possible operations

Restructuring: INF = • no embedded (PRO) subject possible: Lexical VP-layer • no embedded structural case • long object movement • scrambling • no embedded tense • pronoun fronting • no embedded negation impossible: • obligatory control • possible with: (strong) im- • ??extraposition of infinitive plicatives, aspectuals, irrealis * relative clause pied piping predicates • thematic properties are de- possible: Functional INF = main termined by the embedded • IPP effect • raising predicate predicate (except semifunctional predicates) impossible: • possible with: modal, rais- • extraposition of infinitive • matrix passive ing, aspectual, causative, perception, motion verbs • relative clause pied piping Non-restructuring: INF = possible: • embedded (PRO) subject Reduced • pronoun fronting vP or TP • embedded structural case • possible with: implicatives, • focus scrambling aspectuals, irrealis predicates • %extraposition of infinitive impossible: possible: • long object movement • embedded tense • (non-focus) scrambling * embedded negation • non-obligatory control • relative clause pied piping INF = Clausal • embedded (PRO) subject possible: (fall) CP • embedded structural case • relative clause pied piping • embedded tense • extraposition of infinitive * possible with: all lexical impossible: • long object movement predicates • scrambling • obligatory with: prepositional, factive predicates • pronoun fronting • %intraposition of infinitive

The clustering of these properties to yield the appearance of a restructuring phenomenon follows from general aspects of the clausal architecture. For example, tests for embedded infinitival (PRO) subjects and tests for embedded accusative case (e.g., long object

4

Introduction

movement) coincide because they both target the same domain of the clause, namely vP. Infinitival complements lacking vP will fail tests for PRO and will allow long object movement, thus instantiating the basic restructuring configuration. Variation in the literature as to the proper classification of restructuring results from discrepancies among the targets of different diagnostics. Long pronoun movement for example may cross vP though it is impossible out of CP. Therefore, the class of infinitives picked out by this test will be a superset of the class of infinitives picked out by long object movement. The book is organized as follows. Lexical restructuring is the topic of Chapter 2 and Chapter 4. In short, we argue that lexical restructuring is not defined by a single property or feature of certain predicates, but rather, lexical restructuring is a particular configuration. To qualify as a lexical restructuring configuration the following structural and semantic properties have to coincide: the infinitive denotes (roughly) an event or an action, lacks propositional and force properties such as an independent tense specification and complementizer material, lacks a structural case position/assigner, and does not include a syntactic subject. In Chapter 2, we concentrate on tense, negation, complementizers, and structural case and conclude that lexical restructuring infinitives are best analyzed as bare VPs—i.e., as infinitival complements to a lexical verb lacking the projections associated with tense, structural case, (sentential) negation, and complementizer material. In Chapter 4, we argue that lexical restructuring infinitives are obligatory control infinitives which lack a syntactic (PRO) subject. Chapter 3 investigates functional restructuring constructions. Functional restructuring refers to mono-clausal structures in which the restructuring verb represents a functional head and the infinitive is the main predicate of the clause. As such, functional restructuring is the core of restructuring since it is a direct (and unavoidable) result of the architecture of a clause. A detailed comparison of the properties of lexical restructuring constructions with the properties of functional restructuring constructions will motivate the distinction between lexical and functional restructuring and show that a uniform treatment is untenable. Non-restructuring infinitives are discussed in Chapter 5. Syntactic

Restructuring—its significance in history

5

properties such as pronoun fronting and relative clause pied piping split the class of non-restructuring infinitives into two subclasses: (full clausal) non-re structuring infinitives vs. reduced nonrestructuring infinitives, which correlate with different semantic classifications of the infinitival constructions. In particular, propositional and factive constructions differ from irrealis and aspectual infinitives in that the former can only be represented by full clausal structures whereas the latter allow a reduced clausal structure. Finally, Chapter 6 summarizes the major claims of the study including an illustration of the crucial examples. For reference purposes, the study also includes an appendix which lists the data used in this study, the properties tested and applied to a range of predicates, and a preliminary summary of verb classification in five languages (German, Dutch, Italian, Spanish, and Japanese). Before we turn to lexical restructuring, we will conclude the introduction with an overview of the history of the literature on the restructuring phenomenon. 2. Restructuring—its significance in history In an extensive study of infinitives, Gunnar Bech (1955) developed one of the first characterizations of infinitival complements in German. Bech showed that infinitives fall into two classes: infinitives that form an independent clausal domain and infinitives that do not exhibit clausal behavior. He labeled the former class kohärente Infinitive 'coherent infinitives', the latter inkohärente Infinitive 'incoherent infinitives'. The first study of the two classes of infinitives in a generative framework was provided by Evers (1975b). Evers observed that the split among infinitival constructions in Dutch and German correlates with a reordering process of the verbal elements in an infinitival construction. He proposed that this reordering is the result of a process of verb raising which applies in certain infinitival clauses but not in others. Evers' analysis which set the groundwork for most later analyses of clause structure in German and Dutch is the first work that builds on the correlation between mono-clausality and verb movement. In particular, he suggests a structure pruning princi-

6

Introduction

pie (also referred to as a Guillotine principle) which applies to the Snode of infinitives that have lost their head (i.e., by movement of the embedded verb to the higher clause). At the same time, Aissen and Perlmutter (1976) and Rizzi (1976) observed that in Italian and Spanish, certain infinitives lack clausal properties.1 That is, while in most cases infinitives constitute a boundary for processes that are restricted to apply within one clause, certain infinitives are transparent for the same processes. Aissen and Perlmutter suggest that certain infinitives undergo a process of clause union with the matrix clause and hence cease to function as independent clauses. Similarly, Rizzi proposes that what is special about infinitives lacking clausal properties is that they have undergone a process of restructuring. In Rizzi's analysis, restructuring is an optional rule according to which the embedded infinitive and the matrix verb are reanalyzed as one complex verb. Thus, restructuring transforms a bi-clausal structure into a mono-clausal one.2 Since then, the study of restructuring/coherence has received extensive attention in the Germanic and the Romance literature. The three main research questions addressed are: the determination and characterization of the class of restructuring infinitives, the structure of restructuring infinitives, and the motivation for restructuring or clause union. In the next sections, we will give an overview of the major insights and analyses. 2.1. The class of restructuring predicates In most studies on restructuring, it has been noted that the class of 1

Both articles—Rizzi (1976) and Aissen and Perlmutter (1976)—were subsequently republished. In most cases, we will refer to the newer versions. 2 The typical properties attesting to the transparency or lack of clause boundedness of certain infinitives in Romance are clitic climbing, object preposing, and auxiliary switch; the main transparency properties in Germanic are long distance scrambling, long passive, and verb raising. We will illustrate most of these properties in the course of this book. For the present purpose, it suffices to know that these operations are impossible out of finite clauses and non-restructuring infinitives, but are possible out of restructuring infinitives.

Restructuring—its significance in history

7

restructuring predicates varies across languages and also shows some variation among speakers of one language. However, it is also commonly accepted that there is a core of restructuring predicates that is not disputed and moreover found in all languages displaying restructuring effects. This core class of restructuring predicates is summarized in Table 2 for German, Dutch, Spanish, Italian, and Japanese (a list of actual predicates is provided in the appendix, section 5). As the table shows, infinitives selected by modal verbs (must, may, can, want etc.) motion verbs (come, go, return), aspectual verbs (begin, continue, finish), and causative verbs (let, make) are typically restructuring verbs. Table 2.

The core restructuring predicates

Verb

German

Dutch

Spanish

Italian

Japanese

modal verbs motion verbs aspectual verbs causatives

+ + + +

+ + + +

+ + + +

+ + + +

+ + + +

Turning to the less straightforward cases, speaker and language variation arises for instance for the predicates listed in Table 3. The table also shows, however, that there are predicates that are uniformly considered as non-restructuring predicates (in particular, these involve facti ve and propositional predicates). Table 3.

Variation

Verb

German

Dutch

Spanish

Italian

Japanese

try forget, manage dare, seem allow, permit decide, plan regret assume, claim

+ + + + % -

+ + +

% % -

% % %

+

% % -

+ -

The fact that the class of restructuring predicates shows both semantic cohesion as well as variation and apparent arbitrariness is re-

8

Introduction

fleeted in the general tension between two basic directions authors take regarding the question of how the class of restructuring predicates is determined. One type of approach considers the regularities of the class of restructuring predicates as the basic cases of restructuring, and language and speaker variation are assigned a special status or set aside. According to this view, restructuring is motivated through a semantic (and/or thematic) property found among the class of restructuring verbs. Another type of approach treats restructuring as a language-specific and irregular phenomenon, and the cohesion among the class of restructuring predicates is considered as an accident. According to this view, restructuring is generally considered as a lexical property that is assigned (arbitrarily) to a subclass of infinitive-taking verbs and parametrically restricted to certain languages. The major references addressing the question of whether restructuring is (mainly) a lexical/syntactic or (mainly) a semantic phenomenon are summarized in Table 4.3 Table 4.

Syntactic vs. semantic approaches to restructuring

Restructuring

References

Lexically/syntactically de- Aissen and Perlmutter 1976, 1983; DiSciullo and Wiltermined liams 1987; Fanselow 1989; Kayne 1989, 1990, 1991; Sternefeld 1990; Rutten 1991; Roberts 1993, 1997; Grewendorf and Sabel 1994; Sabel 1994/1996; Kiss 1995; Meurers 2000 Semantically/thematically Luján 1980; Napoli 1981; Strozer 1976; Zagona 1982; Rochette 1988, 1990, 1999; Rosen 1989, 1990; Picallo determined 1985, 1990; Rosengren 1992; Wurmbrand 1997, 1998a, 1998b; Cinque 1997a, 1997b, 2000; Roberts 19974 3

Note that the lists of references we will provide in this chapter are selective and we do not claim that they exhaustively represent the works on restructuring. In many cases, we will only mention works that explicitly discuss the questions or issues under consideration. 4 We list Roberts (1997) in both categories since Roberts claims that restructuring verbs are characterized as non-theta assigners, however, at the same time Roberts assumes that this property is essentially a language specific lexical property which does not necessarily correlate with the thematic and semantic properties of the verbs involved.

Restructuring—its significance in history

9

The approach we will take here is somewhat in-between. We will show that the class of restructuring predicates is characterized by uniform semantic properties, however, that the semantics of a configuration is not sufficient to determine whether a construction allows or disallows restructuring. Syntactic and language specific factors have to be taken into account to successfully characterize the restructuring phenomenon.

2.2. The structure of restructuring infinitives Central to most works on restructuring is the question of what the structure of a restructuring infinitive is. The various approaches fall into two general groups which we will label mono-clausal vs. biclausal approaches. According to mono-clausal approaches, a sentence with a restructuring infinitive is a single clause throughout the derivation and a restructuring infinitive never constitutes an independent clausal domain. According to bi-clausal approaches, the clause union effect is attained derivationally. Restructuring infinitives and non-restructuring infinitives start out with the same syntactic structure; however, a further application of restructuring (to be made precise below) alters the structure and/or properties of restructuring infinitives in a way that ultimately renders the clause boundaries ineffective. The advantage of mono-clausal approaches is that no mechanism of restructuring is necessary and that no additional language or construction-specific assumptions have to be made to account for the transparency of restructuring infinitives. Since there is no representation or stage of derivation in which a sentence with a restructuring verb consists of two clauses, the issue of unifying the clauses does not arise. The challenge for approaches of this sort is to motivate the existence of different initial structures for infinitival complements. The main motivation for bi-clausal approaches is the idea of uniformity of phrase structure; i.e., (control) infinitives project a TP, CP etc., irrespective of the syntactic or semantic content of these projections. The challenge for approaches of this sort is to provide evidence for the initial clausal structure of restructuring infinitives and to characterize and motivate the operation of restructuring.

10

Introduction

In this study, we will compare these two basic approaches and conclude that a mono-clausal approach is empirically and conceptually superior since it explains and predicts a range of correlations between the syntactic and semantic properties of restructuring contexts and hence allows us to gain some insight into the nature of restructuring, whereas bi-clausal approaches although descriptively adequate involve various unmotivated assumptions that question the explanatory value of these approaches. In the rest of this section, we will provide an overview of the major mono-clausal and bi-clausal proposals found in the literature. 2.2.1. Mono-clausal approaches The essential claim of mono-clausal approaches is that nonrestructuring infinitives are generated as clausal complements, whereas restructuring infinitives are smaller categories (typically it is assumed that they are VP complements as in (1)). Importantly, a restructuring infinitive is not derived from a non-restructuring infinitive in these approaches. (1)

Mono-clausal approaches FP

VP

to sing

a song

to sing

a song

The main questions addressed by mono-clausal approaches are whether restructuring verbs are lexical or functional categories (cf. (1)), whether there is an embedded infinitival (PRO) subject, and how the difference between VP-infinitives (restructuring) and CP/IPinfinitives (non-restructuring infinitives) is motivated. A selective summary of mono-clausal approaches is given in Table 5.

Restructuring—its significance in history Table 5.

11

Mono-clausal approaches

Property

References

Restructuring infini- Strozer 1976, 1981; Zagona 1982; Cremers 1983; Picallo tives are base1985, 1990; Haider 1986a, 1986b, 1986c, 1991a, 1991b, generated VPs 1993; DiSciullo and Williams 1987; Rochette 1988, 1990, 1999; Rosen 1989, 1990; Moore 1990, 1994; Rutten 1991 (some infinitives) Rosengren 1992; Broekhuis et al. 1995 (some infinitives); Cinque 1997a, 1997b, 2000; Wurmbrand 1998a, 1998b

An approach that is somewhat in-between mono-clausal and biclausal approaches is the reduced clausal approach (cf. Tappe 1984; Fanselow 1989; Li 1990; Rooryck 1994). The basic idea of this approach is that a sentence with a restructuring infinitive essentially involves a bi-clausal structure, however restructuring infinitives nevertheless differ from non-restructuring infinitives in the size of the infinitival complement. In particular, it is assumed that restructuring infinitives lack a CP-boundary (cf. (2)). (2)

Reduced clausal approach VP

V to sing

DP a song

2.2.2. Bi-clausal approaches Bi-clausal approaches maintain that restructuring infinitives start out as clausal (CP) complements and that the clausal status of restructuring infinitives is altered by one or more of the following mechanisms: structure changing operations, head movement, or topicaliza-

12

Introduction

tion. According to the first set of approaches, restructuring infinitives involve a special structure changing process that transforms or reanalyzes a CP-complement into a VP-complement. Evers suggests a pruning principle, Rizzi (1978, 1982) postulates a restructuring rule, Haegeman and van Riemsdijk (1986) assume a reanalysis process, and von Stechow (1990) speculates that restructuring infinitives might be created by deletion of the CP and IP-nodes.5 The common property of these approaches (see also Table 6) is that restructuring infinitives start out as sentential complements, then get reanalyzed or lose various projections in the course of the derivation, and finally end up as VP-complements. Arguments generally raised against structure changing processes of this sort are that they cause a violation of the Projection Principle (Chomsky 1981,1982,1986) and that they are to a large degree arbitrary rules. Table 6.

Structure changing approaches

Restructuring infinitives References Clauses that lose their clausal status

Evers 1975a, 1975b, 1986, 1988, 1990; Rizzi 1976, 1978, 1982; Aissen and Perlmutter 1976, 1983; Hoekstra 1984; Grewendorf 1987, 1988; von Stechow 1990 Reanalysis, multidimen- Manzini 1983b; Haegeman and van Riemsdijk 1986; sional structures DiSciullo and Williams 1987; Goodall 1987, 1991; von Stechow and Sternefeld 1988

The second (and predominant) mechanism to derive clause union is head movement. The variety of head movement analyses share one common property and assumption: it is assumed that some verbal head of the infinitive moves (overtly or covertly) to the matrix predicate (cf. (3)), and that this movement renders the infinitive transparent, either for government by the matrix verb (cf. Evers 1975a, 1975b; Rizzi 1978, 1982; Baker 1988) or for further movement of other elements from the infinitive (such as DPs, clitics). 5

Von Stechow does not commit himself to any assumption about the initial structure of restructuring infinitives (i.e., whether they are base-generated VPs or reduced CPs). However, since he talks about potential deletion of PRO and sentential nodes we include his reference here.

Restructuring—its significance in history

(3)

13

Head movement approaches

t

V ^ ^ ^ D P to sing a song

Since overt verb raising raises serious empirical questions, many authors assume that some form of covert or abstract head movement applies from restructuring infinitives. To give one example, Kayne (1989, 1990, 1991), suggests that restructuring involves overt movement of the infinitival tense head, leaving behind the actual infinitival verb (see Chapter 2, section 5 for a brief summary of various "covert" approaches). The major head movement approaches are listed in Table 7. Table 7.

Head movement approaches

Mechanisms

References

Overt verb (infinitive) raising; or Evers 1975a, 1975b; Rizzi 1978, 1982; Haider formation of a complex V 1986a, 1986b, 1991b, 1991a, 1993; Prinzhorn 1987, 1990; Sternefeld 1990; Rutten 1991; Guasti 1992, 1993,1996, 1997 Covert verb (infinitive) raising Grewendorf and Sabel 1994; Gonçalves 1998 Overt verb raising + pronuncia- Roberts 1997 tion of lower copy Overt raising of embedded Kayne 1989, 1990, 1991; Roberts 1993; BokT/INFL (without the infinitive) Bennema and Kampers-Manhe 1994; Terzi 1996 AgrS+T raising Rooryck 1994 Covert AgrO-raising Sabel 1994/1996

14

Introduction

The final mechanism to derive clause union goes back to Burzio (1986) who suggest that the CP-boundary of a restructuring infinitive is bypassed by moving the embedded verb phrase to the matrix clause (this approach can thus be seen as the ancestor of the more recent remnant movement approaches to restructuring by Hinterhölzl 1997, 1998, 1999; Koopman and Szabolcsi 2000). Furthermore, Baker (1988); Sternefeld (1990); Grewendorf and Sabel (1994); and Sabel (1994/1996) pursue a mixed head movement plus topicalization approach. In these approaches, it is assumed that headmovement is required in restructuring constructions, however, that head-movement cannot proceed through the embedded C°. Rather, head movement applies after the embedded VP or AgrOP has undergone topicalization to the embedded SpecCP (cf. (4) for a simplified structure). (4)

Topicalization plus head movement VP

2.3.

The motivation for restructuring

The last question we will address in this overview is the question of why certain infinitives but not others trigger restructuring. Most mono-clausal approaches involve the assumption that restructuring verbs are auxiliary-like in that they are less thematic than full verbs or are part of the functional structure of the clause. In bi-clausal approaches, the question has to be asked in a slightly different way,

Restructuring—its significance in history

15

namely why restructuring mechanisms can only apply in a subgroup of infinitival constructions. A common view that occurs throughout the works on restructuring is that restructuring verbs or restructuring infinitives are in some way or another deficient, and hence either lack certain projections (monoclausal approaches) or require support from the matrix clause which for instance can be achieved by head movement. Li (1990) for instance assumes that restructuring infinitives involve a deficient 'dummy' INFL, Sternefeld (1990) argues that restructuring infinitives lack an embedded complementizer position, whereas nonrestructuring infinitives involve a complementizer (which can be empty), and many researchers follow the idea that restructuring infinitives are deficient for tense. Table 8 summarizes some of the major proposals that suggest motivations for restructuring. Table 8. Motivation

Motivation for restructuring References

Restructuring verbs are Strozer 1976; Napoli 1981; Zagona 1982, Picallo 1985, not (fully thematic) lexical 1990; Prinzhorn 1987, 1990; Rochette 1988, 1990, verbs 1999; Rosen 1989, 1990; Rutten 1991; Cinque 1997a, 1997b, 2000; Roberts 1997 (but see fn. 4) Restructuring involves Napoli 1981; Rochette 1988, 1990, 1999; Rosen 1989, argument or event struc- 1990; Haider 1993; Rosengren 1992; Kiss 1995 ture unification Restructuring infinitives Guéron and Hoekstra 1988; Rochette 1988, 1990, 1999; are tense deficient Rutten 1991; Broekhuis 1992; Guasti 1992, 1993, 1996, 1997; Roberts 1993, 1997; Bok-Bennema and Kampers-Manhe 1994; Rooryck 1994; Wurmbrand 1997, 1998a, 1998b; Gonçalves 1998

Chapter 2 Lexical restructuring

1. Introduction The central thesis of this study is that infinitival constructions are not represented by a uniform structure but come in a variety of structures which are distinguished by different syntactic and semantic properties. The distinction we will concentrate on in this and the next chapter is the distinction between lexical and functional predicates. In this chapter, we will focus on the syntactic and semantic properties of lexical restructuring infinitives in German, which are found in constructions with the predicates listed in Table 9. Table 9.

Lexical restructuring predicates in German

Verb

Gloss

beabsichtigen beginnen empfehlen erlauben gelingen [unacc] gestatten mißlingen [unacc] untersagen verbieten vergessen vermeiden versäumen versuchen wagen

intend (want) begin recommend allow manage permit fail prohibit forbid forget avoid miss, neglect try dare

Lexical restructuring predicates ? + + + +

% +

% + +

% ?

+ +

For the present discussion, we will set aside the motivation for the

Introduction

17

claim that the predicates in Table 9 are lexical predicates; the discussion of the lexical/functional distinction as well as a detailed comparison of lexical and functional restructuring infinitives is postponed until Chapter 3. The major claim we will defend in this chapter is that the verbs in Table 9, when used as restructuring predicates, combine with syntactically and semantically very "small" predicates. We will show that restructuring infinitives denote bare events or actions—i.e., predicates that lack any kind of propositional or force properties (such as tense, complementizers, and negation). Further support for this claim will be provided in Chapter 4 where we will show that restructuring infinitives also lack an embedded syntactic (PRO) subject. This characterization of restructuring infinitives is straightforwardly implemented in a VP-approach to restructuring (see Chapter 1, section 2.2.1) which we will adopt here. In particular, we will assume that restructuring infinitives are represented syntactically by bare "VP" predicates as illustrated in (5), and, taking VP-approaches a step further, we will argue that restructuring infinitives not only lack tense and complementizer projections but also lack a structural object case position (e.g., AgrOP/vP). (5)

a.

weil Hans den Traktor zu reparieren versuchte since John the tractor-ACC to repair tried 'since John tried to repair the tractor'

b.

TP John

the tractor

T'

to repair

The chapter is organized as follows. In section 2, we will first dis-

18

Lexical restructuring

cuss the case properties in restructuring infinitives and conclude that no structural case position or assigner is available in restructuring infinitives in contrast to non-restructuring infinitives. In section 3, we will provide arguments for the lack of a tense projection in restructuring infinitives. Throughout this chapter, we will assume that restructuring infinitives also lack an embedded (PRO) subject; a thorough discussion of the properties of infinitival subjects, however, will be postponed until Chapter 4. Finally, in section 4, we will discuss the status of other clausal functional projections (CP, NegP, and TP as a host for the infinitival marker) in restructuring infinitives.

2. Case assignment in restructuring constructions Since Larson's (1988) work on double object constructions in English, various proposals have been made concerning the internal structure of the verbal projection of a clause. The general idea put forward by most contemporary works on the structure of "VP" is that the lexical VP is not a single projection but involves a more elaborate structure. Apart from Larson's various VP-layers that host the arguments of the verb, it has been proposed that the lexical "VP" also includes aspect, event, and/or voice phrases (see for instance Travis 1992, 1994, forthcoming; Kratzer 1994, 1996; Harley 1995 among others). Although the approaches to the internal organization of the "VP" differ in various respects, there seems to be consensus among many authors that there is an asymmetry between the external argument and the internal arguments of a verb. More specifically, many theories assume that the VP-layer hosting the external argument (usually labeled vP) has a different status (syntactically and semantically) from the VP-layers that host the internal arguments of the verb (cf. Marantz 1993; Kratzer 1994, 1996; Chomsky 1995 among many others). Generally, proponents of a v-type projection for the external argument assume that the v-projection has two functions (reminiscent of Burzio's Generalization): it hosts the external argument, and it is (directly or indirectly) responsible for structural accusative case. Assuming a vP-type analysis, the question that arises in the context of VP-approaches to restructuring (see Chapter 1, section 2.2.1

Case assignment

19

for a summary) is whether restructuring infinitives project a vP. In the course of this book we will show that restructuring infinitives indeed lack both a structural case position and an embedded subject. In this section, we will first concentrate on the question of whether restructuring infinitives include an embedded structural case position/assigner. The status of the embedded subject will be postponed until Chapter 4. 2.1. A-movement—basic analysis The strongest argument against a structural case position in restructuring infinitives comes from case driven Α-movement operations such as "long" passive (cf. Höhle 1978). As is illustrated in (6), passive of a restructuring verb like try affects the argument structure of the embedded predicate as follows: the embedded object is assigned nominative case and agrees with the matrix auxiliary. Note that the infinitive does not show passive morphology.6 (6)

"Long" passive a.

b.

c.

6

dass der Traktor zu reparieren versucht wurde that the tractor-NOM to repair tried was 'that they tried to repair the tractor' dass die Traktoren zu reparieren versucht wurden that the tractors to repair tried were 'that they tried to repair the tractors' dass der Traktor und der Lastwagen that [the tractor and the truck]-NOM zu reparieren versucht wurden to repair tried were 'that they tried to repair the tractor and the truck'

Following standard practice, we use embedded clauses to control for the verb second effect in German. Note that nominative arguments obligatorily agree with the finite verb in German. Since plural noun phrases do not distinguish between nominative and accusative and singular agreement is the default agreement, we use coordinated noun phrases as in (6)c to show that nominative case and agreement indeed go together.

20

Lexical restructuring

What is crucial about object movement of this sort is that in restructuring infinitives, the case assignment properties of the embedded verb are affected by passivization of the matrix verb; i.e., the suppression of structural case in the matrix clause (as a result of passive) causes the loss of accusative in the embedded complement. This behavior is quite unexpected if it is assumed that the restructuring verb and the embedded verb project their own argument structures—an unavoidable assumption under the CP-approach to restructuring where the infinitive is the head of a separate clause. As the following discussion will show, however, if restructuring infinitives are represented by a bare VP-structure, this form of Α-movement is not only straightforwardly accounted for but also in fact expected to occur. Note also that this form of long passive poses a serious problem for lexicalist approaches. The central claim of DiSciullo and Williams' (1987: 61) approach for instance is that morphological operations (like passive) on one verb cannot affect the argument structure or case properties of another verb. Moreover, since the infinitival verb and the matrix verb can be separated in German (e.g., by topicalization such as Zu reparieren wurde nur der Wagen versucht 'To repair was only the car-NOM tried') an approach that treats these constructions as morphological units seems untenable. Turning to our analysis, let us first consider a basic sentence with an active restructuring verb such as the one in (5)a. Under the assumption that restructuring infinitives lack an embedded structural case position/assigner, the obvious question that arises is how the embedded object is assigned accusative case. The proposal that we would like to make here is illustrated in (7) (the analysis is reminiscent of den Dikken's (1996) analysis of verb raising structures in Dutch which will be discussed below). The only structural object case position in a sentence with an active restructuring verb is provided by the matrix vP. We thus suggest that it is indeed the matrix vP that is responsible for the structural case of the embedded object—i.e., restructuring infinitives in a sense involve a form of Exceptional Case Marking (ECM). Assuming here for convenience a movement approach for case checking (though nothing hinges on it), the embedded object raises (overtly or covertly) to the "matrix

Case assignment

21

clause" in order to check its accusative case.7 (7)

vP-less structure for restructuring infinitives TP

the truck

to repair

Returning now to the passive examples in (6), the case and agreement facts are no longer surprising. Since restructuring infinitives do not involve a structural object case position or assigner, the embedded object is dependent on a case position/assigner in the matrix predicate. If the matrix predicate is passivized, structural accusative becomes unavailable, leaving the (matrix) Τ as the only case assigner for the embedded object (cf. (8))8. The analysis thus correctly predicts that in restructuring infinitives involving a passivized restructuring verb, the object enters into a case and agreement relation with the matrix T. 7

Given the structure we are proposing for restructuring infinitives, the term matrix clause is now obviously anachronistic. However, we will continue to use it as a purely descriptive label (i.e., without assuming that there are two distinct clauses) to distinguish the infinitival VP from the higher VP. Similarly, we will continue to refer to the passive construction in (6) as long passive, but the reader should be aware that under our analysis, this form of passive is a very local operation. 8 For the mere reason of simplicity we present passive constructions here as lacking a vP altogether. However, we do not commit ourselves to this structure. The only important fact for the present analysis is that there is no structural case position/assigner in passive constructions (i.e., no SpecvP). Whether a passive construction involves or lacks a passive voice head is orthogonal to the discussion of restructuring.

22

(8)

Lexical restructuring

Passive restructuring contexts TP

the trucks

to repair

As an aside, we would like to discuss a seemingly similar analysis proposed by den Dikken (1996) for verb raising structures in Dutch and West Flemish. Although our analysis seems to be anticipated by den Dikken's (1996) analysis of verb raising, we will see that our analysis also differs in important respects from his approach. Den Dikken assumes (like we do for German restructuring infinitives) that Dutch/West Flemish verb raising structures such as (9)a involve bare VP-complementation and case assignment for the object in the higher predicate. (9)

West Flemish: verb (projection) raising (den Dikken 1996: 71) a.

da

Jan dienen boek

wilt

that Jan that book wants 'that John wants to buy that book' b.

da

Jan wilt

dienen boek

that Jan wants that book 'that John wants to buy that book'

kuopen

buy kuopen

buy

The two analyses, however, differ significantly in their scope. While we are trying to explain the case and agreement properties in restructuring infinitives generally, the main aim of den Dikken's analysis seems to be to derive the differences in word order between verb raising constructions (cf. (9)a) and verb projection raising constructions found in West Flemish (cf. (9)b).9 Since den Dikken assumes a 9

Den Dikken also provides an account for the fact that verb projection raising complements are frozen for scope. While this fact can be derived under the struc-

Case assignment

23

head initial base for Dutch, the position of the object in (9)a leads him to the conclusion that the object is assigned case in the higher predicate, whereas it is assigned case in the lower predicate in the verb projection raising construction in (9)b. Den Dikken's structures are illustrated in (10). (10) a.

Verb raising

b.

AgrOP

Verb projection raising VP,

kuopen

d. boek

kuopen

d. boek

Although the analysis in (10) is attractive in its straightforwardness in deriving the word orders (and with certain additional assumptions the scope properties) in verb raising and verb projection raising constructions, it also raises some questions. Note that in our approach, the main motivation for the assumption that restructuring infinitives lack an embedded structural case position/assigner is the behavior of these constructions under passive. Since the embedded object cannot get case in the infinitive, it gets whatever case there is in the matrix predicate. Under den Dikken's analysis we would expect exactly the same—i.e., we would expect that in Dutch, an operation such as long passive should occur when the matrix verb is passivized. However, as the examples in (11) illustrate, while impersonal passive of a verb like try is in principle possible (cf. (1 l)a), long passive as in (1 l)b is ungrammatical (cf. Koster 1987; Broekhuis 1992). Note that in the active example, the embedded object can again occur to the left of the matrix verb (cf. (ll)c), which—under den Dikken's analytures suggested by den Dikken, the scope facts do not necessitate den Dikken's particular assumptions and system. We therefore leave the issue of scope aside (see Wurmbrand 2000a for a discussion of scope in long passive constructions).

24

Lexical restructuring

sis—would indicate that these constructions have to be treated on a par with (9)a. (11)

Dutch: long passive

(Broekhuis 1992: 39, 202)

a.

dat (er) geprobeerd werd dat boek te lezen that (there) tried was that book toread 'that they tried to read that book' b. *dat (er) dat boek geprobeerd werd te lezen that (there) that book tried was to read 'that they tried to read that book' c. dat Jan dat boek geprobeerd heeft te lezen that Jan that book tried has to read 'that John tried to read that book' While there are various ways to accommodate the facts in (11), they nevertheless seem to question the validity of the assumption that the case for the embedded object is dependent on a case assigner in the matrix predicate in Dutch. This is indeed the course we would like to take here. While languages like German (see below for an illustration of this phenomenon in other languages) allow VP-complements of restructuring predicates (i.e., infinitival complements lacking an embedded structural case position/assigner), languages like Dutch or English do not have this option.10 Thus, we reach exactly the opposite conclusion from den Dikken, namely that in Dutch, the embedded object is assigned case in the infinitive and not in the matrix An interesting question is why languages differ concerning the availability of vP-less infinitives. Although we do not have a fully worked out analysis of this fact, we would like to offer some speculations. As we will show in Chapter 4, German restructuring infinitives lack a syntactic PRO subject. However, these infinitives are nevertheless understood as propositions, and hence the syntactic structure does not fully match the meaning of these constructions. Assuming that this mismatch is dispreferred in languages such as Dutch and English, a structure including an embedded PRO subject would then be preferred in these languages. Under this assumption, the lack of long passive can then straightforwardly be reduced to a blocking effect of the intervening PRO subject or—assuming that the presence of a subject entails a structural object case position—of a closer case position in the infinitive.

Case assignment

25

predicate. Under this assumption (which seems unavoidable in light of the contrast regarding long passive between Dutch and German), den Dikken's analysis for verb raising structures cannot be maintained and an alternative derivation is called for (such as for instance a more traditional approach involving a head-final structure for Dutch and rightward movement of the embedded verb). To conclude, what is special about restructuring infinitives selected by a lexical restructuring verb is that thematically, the embedded object is the object of the embedded infinitive, however, for case and agreement purposes, the object is associated with a head of the higher predicate. In the following subsections, we will provide further evidence for the restructuring structure proposed in this section. In section 2.2, we will first illustrate that a structure lacking a (structural) object case position provides a straightforward account for a number of infinitival constructions in German as well as in other languages. In section 2.3, we will then show that in restructuring infinitives, the lack of an embedded structural case position/assigner is not simply possible but in fact necessary for a construction to display restructuring effects. Finally, in section 2.4, we will compare the analysis suggested here with approaches assuming a clausal structure for restructuring infinitives. 2.2. Other instances of ν Ρ-less infinitives 2.2.1. Unaccusative restructuring predicates The claim that the object in a restructuring infinitive enters into a case/agreement relation with the matrix vP makes a clear prediction concerning unaccusative restructuring predicates. Since unaccusative predicates do not have the ability to assign (structural) accusative case, the analysis outlined in the previous section predicts that an object embedded in a restructuring infinitive should not appear with accusative but rather with nominative case when the infinitive combines with an unaccusative restructuring predicate. In this section, we will discuss three constructions which confirm this prediction. The first type of construction involves infinitives selected by un-

26

Lexical restructuring

accusative matrix verbs such as gelingen 'manage'. As is illustrated in (12), the verb gelingen 'manage' is an unaccusative verb—it requires the copula be (cf. (12)a) and it cannot be passivized (cf. (12)b). Note that in contrast to the English construction involving the verb manage, the experiencer argument shows up as a dative noun phrase (i.e., as an internal argument rather than the subject) in the German construction. Furthermore, when the infinitive is extraposed, an expletive es 'it' is inserted which is optional for some speakers (though generally preferred). (12)

gelingen 'manage' a.

weil since 'since b. *weil since 'since

(es) ihm gelungen ist/*hat zu fliehen (it) him-DAT managed is/*has to escape he managed to escape' (es) (ihm) gelungen wurde zu fliehen (it) (him-DAT) managed was to escape he/they managed to escape'

Turning to the restructuring properties of constructions with manage, Haider (1993) has noticed the special case properties found in these constructions: as in passive constructions, the embedded object shows up with nominative case and agrees with the matrix auxiliary. Haider's observation is illustrated in (13). (13)

Case/agreement in gelingen 'manage' constructions a. Iweil mir since me-DAT zu entziffern to decipher 'since I managed b. Iweil mir since me-DAT zu entziffern to decipher 'since I managed

der Brief auf Anhieb the letter-NOM straightaway gelungen ist managed is straightaway to decipher the letter' die Briefe auf Anhieb the letters straightaway gelungen sind managed are straightaway to decipher the letters'

Case assignment

27

In light of the analysis proposed in the previous section, the case properties in (13) are again not surprising. Since unaccusatives (like passives) lack a structural object case position or assigner, the only head that the embedded object can enter into a case/agreement relation with is the matrix T. The structure for unaccusative restructuring contexts is depicted in (14) (for simplicity, we represent unaccusatives as VPs lacking a vP; however, see fn. 8). (14)

Unaccusative restructuring infinitive

=(13)b

TP

managed the letters

to decipher

As for the indirect dative argument, it suffices to assume that dative arguments do not compete for structural case in German. The dative argument in Spec VP therefore does not intervene when the embedded object moves to the nominative position (the same situation holds for dative arguments in a simple clause). The second type of construction involving unaccusative matrix predicates is the so-called easy-to-please construction (cf. (15)). (15)

Easy-to-please construction a.

b.

Dieser Text ist schwer This text-NOM is hard 'This text is hard to read' Diese Texte sind leicht These texts are easy 'These texts are easy to read'

zu lesen to read zu lesen to read

Assuming that easy-adjectives combine with a vP-less infinitival

28

Lexical restructuring

complement, the embedded object is dependent on a case assigner in the matrix predicate. Since easy-adjectives do not assign an external theta-role, they are essentially unaccusatives (see also Montalbetti, Saito and Travis 1982; Cinque 1990a), and hence do not involve a structural accusative position. The only head that the embedded object can establish a case/agreement relation with is thus again the matrix tense head. Treating easy-infinitives as restructuring infinitives thus accounts for the nominative case of the embedded object, the agreement between the underlying embedded object and the matrix auxiliary, as well as the thematic relation of the surface subject with the embedded predicate. The structure is illustrated in (16).11 (16)

Easy-to-please construction

=(15)b

TP

these texts

to read

Before turning to the third type of vP-less infinitives, let us point out one further property of the easy-to-please construction in German that can be seen as evidence for the analysis suggested above. In contrast to the English easy-to-please construction which has been argued to involve some form of A'-movement—so-called toughmovement (cf. Chomsky 1982, 1986a, 1986b; Browning 1987; Cinque 1990b; and Chomsky and Lasnik 1993), we have proposed above that the corresponding construction in German involves an in11

A peculiar property of the easy-to-please construction in German is that, although adjectives project head-finally in German, the easy adjective (in contrast to other restructuring verbs) has to precede the complement VP. We ignore this fact here since it does not seem to bear on the case and agreement properties discussed in the text. The structure in (16) can easily be accommodated to whatever will turn out to be the correct structure for easy-to-please constructions in German, as long as the infinitive is represented by a vP-less VP-complement.

Case assignment

29

stance of Α-movement (see also Haider 1993; Roberts 1993; Wurmbrand 1994). Assuming that an A'-movement analysis is essentially correct for the English easy-to-please construction, we expect that the two constructions should show different properties in English and German. A closer look at the locality conditions of fougft-movement shows that this prediction is borne out. 12 As is illustrated in (17)a, the easy-to-please construction is ungrammatical in German when an additional argument (the DP-object of überzeugen 'convince') appears between the thematic and the overt position of the surface subject. Under an Α-movement analysis for the German easy-to-please construction, this ungrammaticality is as expected since the embedded object crosses another A-position on its way to the matrix case position, resulting in a classic Relativized Minimality violation (cf. Rizzi 1990). Since no such violation is found in the English example in (17)b, we see that iowg/i-movement is not sensitive to intervening A-positions, and the assumption that the English construction involves A'-movement rather than Amovement seems justified. (17)

Easy-to-please construction a. *Dieses Buch ist schwer Hans zu überzeugen zu lesen to read This book is hard John to convince 'This book is hard to convince John to read' b. This book was easy to convince John to read

The third construction that can be argued to involve a vP-less infinitive is what we will refer to as the auxiliary-infinitive construction. 12

Although an A'-movement analysis for the English rowgft-construction seems to be justified by a number of A'-properties that this construction displays, it also has to be noted that the exact status of the /owgA-construction is still very much an open issue (see in particular Bresnan 1971; Lasnik and Fiengo 1974; and Montalbetti, Saito and Travis 1982 for discussions of properties of the foKgA-construction in English that are problematic for an A'-movement approach). In the discussion to follow, this intermediate status of the foKgA-construction and the persistent problems that this construction poses for nearly any analysis will be ignored. What is important here is simply that the English fowg/i-construction does display A'properties.

30

Lexical restructuring

As is illustrated in (18), the auxiliary verb be can combine with an infinitive, ße-infinitive constructions express modal meanings (possibility and necessity) and are as such highly context sensitive (in (18), only the most salient interpretations are given). The important property for our concerns here is that the ôe-infinitive construction again displays the case and agreement pattern typical for vP-less infinitives. Although the infinitive occurs in its active form, the embedded object is assigned nominative case and agrees with the matrix auxiliary. (18)

Auxiliary-infinitive construction a.

b.

weil diese Pilze nicht zu essen sind/*ist since these mushrooms not to eat are/*is 'since these mushrooms cannot be eaten' weil der Zaun bis morgen zu reparieren ist since the fence-NOM by tomorrow to repair is 'since the fence must be repaired by tomorrow'

Extending our analysis for restructuring infinitives to auxiliaryinfinitive constructions, we can derive the case/agreement properties without further assumptions. Since be is an unaccusative verb lacking a structural object case position, and (by assumption) restructuring infinitives lack an embedded structural case position, the only case assigner available for the embedded object is the matrix T. A case and agreement relation between the matrix Τ and the embedded object then yields the desired nominative case on the object and agreement with the matrix auxiliary. A simplified structure for auxiliaryinfinitive constructions is depicted below in (19).13 To summarize, we have seen that there are a number of constructions that motivate the assumption that restructuring infinitives lack an embedded structural case position. Passivized restructuring verbs, 13

We ignore the status of the projection headed by be for now. In Chapter 3, we will argue that modals and auxiliaries are functional categories in German. Under that analysis, the auxiliary-infinitive construction will then fall under functional restructuring. We nevertheless include the construction here, since at this point, the lexical vs. functional distinction is orthogonal to the analysis presented in the text.

Case assignment

31

easy-to-please constructions, and auxiliary-infinitive constructions all show that if the matrix predicate does not involve a structural accusative position, the embedded object does not show up with accusative case (as would be expected if the infinitive involves its own array of functional projections). Rather, the embedded object enters into a case and agreement relation with the matrix T. (19)

Auxiliary-infinitive construction

=(18)a

TP

X*

these mushrooms

to eat

2.2.2. Some crosslinguistic evidence for vP-less infinitives In this subsection, we will present some crosslinguistic evidence for the effects of vP-less infinitives. We will see that vP-less infinitives are not an idiosyncratic property of German, but a property of restructuring infinitives found in a number of languages. Although long passive with restructuring verbs like try, forget, or dare seems to be rather rare cross-linguistically (see Chapter 4 for some speculations and a possible explanation for the marked status of this construction), it can be found in some of the Romance languages in constructions involving aspectual verbs. As in German, passivization of the matrix verb (but not the infinitive) causes the embedded object to move to matrix subject position and agree with the matrix auxiliary. This is illustrated in (20)a for Italian and (20)b for Spanish (examples from Cinque 1997a).

32

(20)

Lexical restructuring

Long passive in Italian and Spanish a.

b.

Le nuove case furono iniziate a costruire negli anni '20 the new houses were started to build in the 20's 'They started to build these houses in the 20s' Estas paredes están siendo terminadas de pintar these walls were being finished to paint 'They were finishing painting these walls'

As for the easy-to-please construction, Romance patterns with German in that the kind of movement involved is subject to Relativized Minimality effects for A-positions (cf. Kayne 1989, Roberts 1997). As illustrated in (21)b for Italian and (21)d for French, the easy-toplease construction is ungrammatical when the embedded object crosses an intervening argument (the object of convince) on its way to the matrix subject position.14 (21)

Easy-to-please construction in Italian and French a.

Questo libro è difficile da finire prima di lunedì this book is difficult to finish before Monday 'This book is difficult to finish before Monday' b. *Questo libro è difficile da convincere Mario this book is difficult to convince Mario a finire prima di lunedì to finish before Monday 'This book is difficult to convince Mario to finish before Monday' 14

Note that our suggestion to treat the easy-to-please constructions in Italian and French in a parallel fashion to the easy-to-please construction in German (i.e., as vP-less restructuring infinitives) does not entail that these constructions are identical in all respects in these languages. As has been discussed for instance by Montalbetti, Saito and Travis (1982), easy-to-please constructions come in a wide variety in different languages. We cannot go into detail regarding the crosslinguistic variation of the easy-to-please construction, however, what is important for the present purpose is that the easy-to-please constructions in Italian, French, and German share the property of involving Α-movement (in contrast to the English type ioKg/i-movement), which can be attributed to restructuring (see also Kayne 1989; Roberts 1997) or a truncated structure (cf. Rizzi 1993, Haegeman 1995).

Case assignment

33

c.

Ce genre de livre serait difficile à lire this kind of book would-be difficult to read 'This kind of book would be difficult to read' d. * Ce genre de livre serait difficile this kind of book would-be difficult à convaincre Jean de lire to convince Jean to read 'This kind of book would be difficult to convince Jean to read' The final phenomenon we want to mention here as support for the lack of an embedded vP in restructuring infinitives is the nominative object construction in Japanese. Let us start with a few introductory comments on case assignment in Japanese. In Japanese, the case of the direct object depends on the stativity of the verb: non-stative verbs assign accusative case, stative verbs assign nominative case (cf. Kuno 1973). Thus, in a simple clause headed by a non-stative verb like eat, only accusative objects are licit, and nominative objects are prohibited (cf. (22)a vs. (22)b). (22)

Japanese: non-stative verbs assign accusative a.

Emi-ga ringo-o Emi-NOM apple-ACC 'Emi ate apples' b. *Emi-ga ringo-ga Emi-NOM apple-NOM 'Emi ate apples'

tabe-ta eat-PAST tabe-ta eat-PAST

If the verb is a stative predicate like be-capable, the opposite situation holds: accusative objects are prohibited, whereas nominative objects are licit (cf. (23)a vs. (23)b).15 ^ The claim that accusative is impossible with stative predicates might be too strong for some speakers. Moreover, the (un)availability of accusative seems to also depend on what kind of stative predicate is used (e.g., Koizumi 1995: 67 reports that accusative is more readily available in constructions with wakaru 'understand'; however, the same example is considered as ungrammatical by Saito and

34

(23)

Lexical restructuring

Japanese: stative verbs assign nominative (Koizumi 1995: 66) a. *Emi-ga Emi-NOM 'Emi speaks b. Emi-ga Emi-NOM 'Emi speaks

nihongo-o Japanese-ACC Japanese' nihongo-ga Japanese-NOM Japanese'

deki-ru be.capable-PRES deki-ru be.capable-PRES

The distribution of nominative objects provides us with a further test to establish whether in restructuring contexts, objects enter into a case relation with an element in the matrix clause or an element in the embedded clause. Under the analysis suggested here, a construction like the one in (22)b, which prohibits nominative objects in isolation, should become grammatical when the sentence is embedded under a stative restructuring verb (i.e., a verb that assigns nominative). Since—as we argue—restructuring infinitives lack a structural case assigner, the case of the embedded object is determined by the properties of the higher predicate. If the higher predicate in a Japanese restructuring construction involves a nominative assigner, our analysis thus predicts that the object is assigned nominative rather than accusative case, even in cases in which the lower predicate prohibits nominative objects. As the sharp contrast between (22)b and (24)a,b shows, the thematic object of the verb eat can indeed appear with nominative case when the infinitive is embedded under a stative restructuring verb as in (24)a,b.16 Similarly, if a stative complement such as the one in (23)a is embedded under a non-stative restructuring verb (i.e., a verb assigning only accusative case), the embedded object can show up with accusative (cf. (24)c, K. Yatsushiro, p.c.). A detailed analysis of the examples in (24) is postponed until section 2.3.2. What is important at this point is that the distribution of nominative vs. accusative objects in Japanese represents another instance of infinitival complements Hoshi 1998 and Nemoto 1993: 163). We will have to leave this issue aside. 16 As is well-known, the embedded object can also be marked accusative in these examples. In the next section, we will show, however, that the different cases on the embedded object correlate with restructuring vs. non-restructuring infinitives.

Case assignment

35

that are characterized by the lack of a structural object case assigner/position in the infinitive (24)

Japanese: stative restructuring contexts a.

b.

c.

Emi-ga ringo-ga tabe-rare-ru Emi-NOM apple-NOM eat-can-PRES 'Emi can eat apples' Emi-ga ringo-ga tabe-ta-i Emi-NOM apple-NOM eat-want-PRES 'Emi wants to eat apples' Emi-ga nihongo-o deki-hazime-ta Emi-NOM Japanese-ACC capable-begin-PAST 'Emi began to be able to speak Japanese'

To conclude, the Japanese facts as well as the other phenomena discussed in this section have shown that in restructuring infinitives, the embedded object is not assigned case by the embedded verb or a functional head in the infinitival complement, but rather the embedded object establishes a case and agreement relation with a head in the matrix predicate. The conclusion that we have to draw from this fact is that in restructuring contexts, the matrix case assigner (v° or T \ depending on the argument structure of the higher predicate) is the closest case and agreement head for the embedded object, or, in other words, that the embedded complement does not involve a (functioning) structural case assigner/position. 2.3. Restructuring vs. non-restructuring In the previous section, we have discussed cases in which the embedded object in a restructuring infinitive is not assigned case by an embedded structural case assigner but enters into a case and agreement relation with a head in the matrix predicate. In this section, we will argue that this case and agreement relation is not only possible but in fact obligatory. In essence, we will show that whenever there is reason to assume that the embedded object receives case in the embedded predicate, the construction does not allow restructuring

36

Lexical restructuring

properties at all and hence has to be considered as a non-restructuring infinitive. The facts that we will discuss will thus provide the final piece of evidence for the claim that restructuring infinitives lack a structural case position.

2.3.1. German Let us start with German. An important fact about long passive which we have not discussed so far is that long passive is restricted to restructuring infinitives. As the ungrammaticality of (25) shows, long passive is impossible in infinitival constructions involving a non-restructuring verb such as plan (for a complete list of examples illustrating the distribution of long passive in German we refer the reader to the appendix, section 4.1). The only way to express passive with a non-restructuring verb is with an impersonal passive construction as in (26)—i.e., a construction without long object movement.17 (25)

Long passive with non-restructuring infinitives a. *dass that 'that b. *dass that 'that

(26)

wurde was wurden were

Impersonal passive with non-restructuring infinitives a.

17

der Traktor zu reparieren geplant the tractor-NOM to repair planned they planned to repair the tractor' die Traktoren zu reparieren geplant the tractors to repair planned they planned to repair the tractors'

dass den Traktor zu reparieren geplant that the tractor-ACC to repair planned 'that they planned to repair the tractor'

wurde was

The examples in (26) require an intonational break before and after the infinitival complement. In general, German speakers prefer extraposition of nonrestructuring infinitives, however, the intraposed position of the infinitive as in (26) is also grammatical. See Chapter 5, section 3.2.2 for a discussion of the issue of extraposition vs. intraposition.

Case assignment

b.

dass die Traktoren zu reparieren geplant that the tractors to repair planned 'that they planned to repair the tractors'

37

wurde was

Thus, restructuring as well as non-restructuring verbs can be passivized in German (unless they are unaccusatives), however, matrix passive only has an effect on the case properties of the embedded object in restructuring contexts. In non-restructuring infinitives, the embedded object always enters into a case relation with an appropriate embedded case assigner. Assuming that non-restructuring infinitives have their own functional structure, the examples in (25)-(26) are as expected. The embedded v° is the closest functional head for the embedded object; case/agreement is hence established between the object and the embedded v°—resulting in accusative rather than nominative.18 Agreement between the embedded object and the matrix T", on the other hand, would violate the standard locality requirement on case/agreement relations. Turning now to restructuring infinitives, we have seen in the previous section that the embedded object can enter into a case and agreement relation with the matrix v° or T°. So far, we have only shown that this relation is possible; the question we will address now is whether it is indeed necessary. The answer to this question is clearly positive concerning the easy-to-please construction and the auxiliary-infinitive construction. As is illustrated in (27), the embedded object cannot appear with accusative case which indicates that there is indeed no structural object case position or assigner in the infinitival complement. (27)

Accusative in easy-to-please, auxiliary-infinitive construction a. *weil den Traktor leicht zu reparieren ist since the tractor-ACC easy to repair is 'since the tractor is easy to repair'

Although—as we will see in Chapter 5—there are different grades (i.e., sizes) of non-restructuring infinitives, all (transitive) non-restructuring infinitives involve a structural object case position/assigner.

38

Lexical restructuring

b. *weil den Traktor nicht zu reparieren ist since the tractor-ACC not to repair is 'since it is not allowed/possible to repair the tractor' The situation appears different, however, if we look at constructions involving passivized or unaccusative restructuring verbs. As the examples in (28) show, accusative objects are perfectly grammatical in infinitival complements combining with passivized or unaccusative restructuring verbs. Furthermore, (28) shows that agreement between the embedded object and the matrix auxiliary is also not required. (28)

No long object movement a.

b.

dass versucht wurde/*wurden It tried was/*were [den Traktor und den Lastwagen zu reparieren] [[the tractor and the truck]-ACC to repair] 'that they tried to repair the tractor and the truck' dass es ihm gelungen ist/*sind that it him-DAT managed is/*are [den Traktor und den Lastwagen zu reparieren] [[the tractor and the truck]-ACC to repair] 'that they managed to repair the tractor and the truck'

Given the claim that restructuring infinitives do not involve a structural object case/agreement position, we would expect that accusative should never be available for the embedded object and agreement with the matrix Τ should be obligatory when the matrix predicate does not license accusative. The examples in (28) thus appear to challenge the analysis presented here. Let us summarize where the analysis stands. We have argued that the case and agreement properties found in restructuring contexts point to the conclusion that restructuring infinitives do not involve an embedded structural case position. We have also shown that nonrestructuring infinitives do not allow long Α-movement, indicating that non-restructuring infinitives project their own case and agreement projections. To account for the lack of long passive it has to be assumed that case and agreement checking is a very local process

Case assignment

39

and that once checked no further checking is possible (otherwise we could not explain why long object movement is impossible in nonrestructuring infinitives). Assuming our analysis is correct, the obvious answer to the question of why accusative is possible and agreement with the matrix Τ is not necessary in the examples in (28) is that these examples are indeed instances of non-restructuring infinitives (i.e., infinitives that project a structural object case/agreement position). In other words, verbs like try, manage (in fact all lexical restructuring verbs) have to be considered as ambiguous (or perhaps underspecified, in some sense)—either they are restructuring verbs that combine with a bare VP-complement, or they are non-restructuring verbs that combine with a vP (or bigger) complement.19 In the rest of this section we will provide arguments for this claim. We will show that examples such as the ones in (28)—i.e., constructions in which there is evidence for an embedded object case position—have to be treated as nonrestructuring infinitives. The question of why easy-to-please constructions and auxiliary-infinitive constructions do not have this option but are unambiguous restructuring constructions will be attributed to the functional status of the matrix predicates (see Chapter 3). Before providing the crucial evidence for the non-restructuring status of the examples in (28), let us point out one further empirical fact which will be important for the discussion to follow. As is illustrated in (29), the non-restructuring versions of f/7-type infinitives (as we claim and will motivate momentarily) are not restricted to extraposed positions. The passive examples in (29)a,b und the unaccusative examples in (29)c,d which lack long object movement are perfectly acceptable for German speakers, provided that there is an intonational break before and after the infinitival complement (indicated by brackets in the examples). In (29)a,c we see again that the embedded objects are marked with accusative; (29)b,d illustrate the lack of agreement between the embedded objects and the matrix auxiliaries. 19

For the present discussion, we leave aside the question of whether examples such as the ones in (28) are CPs, TPs, or vPs; the only important concern is that these infinitives involve a structural case position (see Chapter 5 for criteria to distinguish between these options).

40

(29)

Lexical restructuring

Intraposed non-restructuring infinitives a.

b.

c.

d.

dass [ den Traktor zu reparieren ] versucht wurde that [ the tractor-ACC to repair ] tried was 'that they tried to repair the tractor' dass [ die Traktoren zu reparieren ] versucht wurde that [ the tractors to repair ] tried was 'that they tried to repair the tractors' dass ihm [den Traktor zu reparieren] gelang that him-DAT [the tractor-ACC to repair ] managed 'that he managed to repair the tractor' dass ihm [die Autos zu reparieren ] gelungenist that him-DAT [the cars to repair ] managed is 'that he (has) managed to repair the tractors'

To repeat, the claim that we are making here is that—despite the presence of a (potential) restructuring verb—the examples in (28) through (29) are non-restructuring infinitives. To determine whether this claim is correct we have to find some test that shows that the infinitives under consideration do not display transparency properties. The transparency property long passive obviously cannot be used in these examples. However, there is another property found in restructuring infinitives but not in non-restructuring infinitives in German, namely scrambling. Although the situation is slightly more complex for scrambling than for long passive, most authors agree that only restructuring infinitives allow (non-focus) scrambling of a phrase from the infinitive. As is illustrated in (30)a, the embedded object can be scrambled from an extraposed restructuring infinitive (see also Bayer and Kornfilt 1990 and Säbel 1994/1996).20 If, on the other hand, the matrix verb is a non-restructuring infinitive as in (30)b, scrambling is impossible (cf. (30)c). In constructions involving irrealis non-restructuring verbs (which we will argue in Chapter 5 are 'reduced' clausal infinitives—i.e., they are IPs/TPs lacking Cprojections), we find some speaker variation, however, for most speakers scrambling is clearly degraded (cf. (30)d). But see Chapter 5, section 3.2.2 for a qualification, some comments on speaker variation, and further discussion of the extraposition vs. intraposition issue.

Case assignment

(30)

41

Scrambling from extraposed infinitives21 a.

dass Hans den Traktor versucht hat zu reparieren that John the tractor-ACC tried has to repair 'that John (has) tried to repair the tractor' b. dass Hans bedauert hat that John regretted has den Traktor reparieren zu müssen the tractor-ACC repair to must 'that John (has) regretted that he had to repair the tractor' c. *dass Hans den Traktor bedauert hat that John the tractor-ACC regretted has reparieren zu müssen repair to must 'that John (has) regretted that he had to repair the tractor' d. Vodass Hans den Traktor geplant hat zu reparieren that John the tractor-ACC planned has to repair 'that John (has) planned to repair the tractor' Returning to the main question at stake here, we now have the tools to determine whether the examples in (28) through (29) are restructuring infinitives or non-restructuring infinitives. If they are restructuring infinitives, scrambling should be possible as it is in (30)a; if they are non-restructuring infinitives, scrambling should be impossible or at least degraded. As expected, we will see below that the examples in (28) through (29) become clearly ungrammatical when scrambling is applied—hence providing the crucial piece of evidence for the non-restructuring status of these constructions. The impossibility of scrambling can be straightforwardly illustrated for (28).22 In contrast to (30)a, the examples in (31)a,b which 21

As we will show in Chapter 5, the generalization is that the core cases of nonrestructuring (i.e., propositional and factive infinitives) clearly block scrambling; irrealis infinitives, on the other hand, allow focus scrambling and disallow nonfocus scrambling for most speakers. However, the two forms of scrambling are not always easy to distinguish and hence the speaker variation is not fully systematic. 22 As we will show in Chapter 5, section 3.2.2 extraposition adds a complication for restructuring (but not for reduced non-restructuring), and hence restructuring is

42

Lexical restructuring

correspond to the examples in (28)a,b (i.e., examples lacking a case and agreement relation between the embedded object and the matrix T) clearly block scrambling from the infinitive. Furthermore, if, as in (31)c, a dative argument is scrambled from an infinitive which arguably involves a structural case assigner (i.e., an infinitive that involves an accusative object in a matrix passive context), the result is degraded but perhaps slightly less so than in the clearly ungrammatical examples in (31)a,b. The reason for the difference between scrambling of a direct object vs. an indirect object is most likely due to the fact that scrambling of a dative argument is focus scrambling and hence subject to weaker restrictions than non-focus scrambling (the same kind of improvement can be achieved in (31)a,b when the accusative object is heavily focused; see Chapter 5, section 1 for further discussion). (31)

Non-restructuring infinitives a. *dass den Traktor versucht wurde that the tractor-ACC tried was 'that they tried to repair the tractor' b. *dass die Traktoren ihm gelungen ist that the tractors him managed is 'that he managed to repair the tractors' b'. *dass ihm die Traktoren gelungen ist that him the tractors managed is 'that he managed to repair the tractors' b " *dass die Traktoren ihm zu reparieren that the tractors him to repair 'that he managed to repair the tractors' c. Ί Idas s dem Peter versucht wurde that the Peter-DAT tried was einen Brief zu schicken a letter-ACC to send 'that they tried to send Peter a letter'

zu reparieren to repair zu reparieren to repair zu reparieren to repair gelungen managed

ist is

in fact excluded on two grounds in the examples in (31). This interfering factor does not arise, however, for the cases involving remnant topicalization to be presented below.

Case assignment

43

To use scrambling as a transparency test in (29)a,b requires a slightly more complex setup, since scrambling of the embedded object would be string vacuous in (29)a,b. However, this problem can be avoided by using an operation called remnant topicalization. Remnant topicalization refers to a two-step movement operation where first, an XP undergoes scrambling; second, a YP which contains the trace of the scrambled XP but not the moved XP is moved to topic position. An abstract illustration of remnant topicalization is provided in (32). (32)

Remnant topicalization CP YP

c*

XP

The importance of remnant topicalization for restructuring is that it enables us under certain conditions to determine whether a phrase has undergone scrambling. Abstractly speaking, the idea is that in order to move YP without XP (where XP originates inside YP) to topic position in (32), XP has to undergo movement first, in particular, it has to move to a position higher than the base position of YP. 23 Let us first look at remnant topicalization structures with restructuring infinitives. If—as in (33)a—the infinitive is topicalized together with the matrix verb, the topicalized phrase has to be (at least) the matrix VP (assuming that the topicalized material forms a constituent). As is shown in (33)b,c, in restructuring infinitives involving long passive, remnant topicalization is also possible. Since the embedded object in these examples is not carried along with the topicalized matrix VP but left behind, we can conclude that the object has undergone overt movement to the matrix predicate (which is as expected in restructuring constructions). There are of course ways of deriving remnant topicalization structures using discontinuous or non-configurational structures. We ignore these approaches here since we believe that they are too unrestrictive and lack explanatory value.

44

(33)

Lexical restructuring

Remnant topicalization of restructuring infinitives a.

b.

c.

[[Den Traktor zu reparieren ] versucht ]VP [[the tractor-ACC to repair ] tried ]VP wurde noch nie was yet never 'What has never happened before was that they tried to repair the tractor' [[toBj Zu reparieren ] versucht ]VP wurde [[t 0B j to repair ] tried ]VP was der Traktor noch nie the tractor-NOM yet never 'What has never happened before was that they tried to repair the tractor' [[tOBj Zw reparieren ] versucht ]VP wurden [[toBj to repair ] tried ]VP were die Traktoren noch nie the tractors yet never 'What has never happened before was that they tried to repair the tractors'

Going back to the examples in (29)a,b, remnant topicalization now offers a way of testing whether scrambling (and hence restructuring) is possible in these examples (see also Haider 1993 who makes extensive use of this tool). If the embedded object can be left behind when the matrix VP is topicalized (as in (33)b,c) there would be strong evidence for scrambling from the infinitive and hence the transparency of the infinitives. However, as the examples in (34) show, remnant topicalization is clearly ungrammatical. Note that (34)a only differs from (33)b in the case of the embedded object, and (34)b only differs from (33)c in whether the verb agrees with the embedded object or not. Thus, in contrast to the restructuring examples in (33)b,c, the same kind of movement is impossible in passive contexts when the embedded object appears with accusative ((34)a) and does not agree with the matrix auxiliary ((34)b).

Case assignment

(34)

45

No remnant topicalization a

· *[[t0Bj Zm reparieren ] versucht ]VP wurde [[toBj to repair ] tried ]VP was den Traktor noch nie the tractor-ACC yet never 'What has never happened before was that they tried to repair the tractor' b. *[[toBJ Zu reparieren ] versucht ]VP wurde [[toBj to repair ] tried ]VP was die Traktoren noch nie the tractors yet never 'What has never happened before was that they tried to repair the tractors' The same situation holds for the unaccusative constructions in (29)c,d- Remnant topicalization is possible when the infinitive lacks a structural case position and long object movement applies (i.e., when the embedded object appears with nominative case as in (35)a, and agrees with the matrix auxiliary as in (35)b). The same kind of movement is again ungrammatical when long object movement does not apply—i.e., when the infinitive involves its own structural case position (cf. (35)c,d). (35)

Unaccusative constructions a. [[tOBJ Zu reparieren ] gelungen]VP ist [[toBj to repair ] managed]VP is ihm der Traktor nicht him-DAT the tractor-NOM not 'What he didn't manage was to repair the tractor' b. [[tOBj Ζ" reparieren ] gelungen]WP sind [[toBj to repair ] managed]VP are ihm die Traktoren nicht him-DAT the tractors not 'What he didn't manage was to repair the tractors' c. *[[t0Bj Zm reparieren ] gelungen]VP ist [[toBj to repair ] managed]VP is

46

Lexical restructuring

ihm den Traktor nicht him-DAT the tractor-ACC not 'What he didn't manage was to repair the tractor' d. *[[toBj Zw reparieren ] gelungen]WP ist [[toBj to repair ] managed ]VP is ihm die Traktoren nicht him-DAT the tractors not 'What he didn't manage was to repair the tractors' To sum up, what the excursus on remnant topicalization has shown is that whenever the infinitive involves a structural object case position and the embedded object does not enter into a case/agreement relation with an element in the matrix predicate, the construction blocks scrambling. Taking scrambling as an indication for restructuring and the marginality or impossibility of scrambling as an indication for non-restructuring, we can conclude that infinitives involving an embedded vP are non-restructuring infinitives. To conclude, we have argued in this section that apparent counterexamples to the claim that restructuring infinitives do not involve a structural object case position, in fact provide support for the analysis proposed here: whenever an infinitive involves a structural accusative argument which cannot have been case-marked by the matrix verb, the infinitive is a non-restructuring infinitive (but see again Chapter 5 for a discussion of different grades of non-restructuring). In the next section, we will see that the same situation holds in Japanese. 2.3.2. Japanese As we have discussed in section 2.2.2, in Japanese restructuring constructions, the object of an embedded non-stative verb can show up with nominative when the matrix predicate is a stative verb (cf. the examples in (24); (24)a is repeated here as (36)a). However, what we have ignored so far is that in these contexts accusative is also possible for the embedded object (cf. (36)b).

Case assignment

(36)

47

Japanese: nominative/accusative objects a.

b.

Emi-ga ringo-ga Emi-NOM apple-NOM 'Emi can eat apples' Emi-ga ringo-o Emi-NOM apple-ACC 'Emi can eat apples'

tabe-rare-ru eat-can-PRES tabe-rare-ru eat-can-PRES

As with the German constructions involving accusative objects in infinitives selected by passive or unaccusative restructuring verbs (see the previous section), the example in (36)b appears to challenge the claim that restructuring infinitives lack a structural object case position. Assuming that stative verbs assign nominative rather than accusative and that restructuring infinitives lack a structural case assigner, accusative should not be available in these contexts. However, what we will show in this section is that examples such as the one in (36)b are non-restructuring infinitives, again supporting our analysis—in particular the claim that restructuring infinitives lack an embedded structural case position whereas non-restructuring infinitives involve such a position. Let us first outline the analysis we suggest for the examples in (36); arguments that support this analysis will be provided below. We assume that objective case (i.e., whether accusative or nominative) is assigned by v° (cf. Tada 1992, 1993 who argues that nominative is assigned by AgrO" and provides arguments against a TPposition for nominative objects). Furthermore, we assume that the choice between accusative and nominative is driven by a lexical feature in v°: [+stative] v° triggers nominative, otherwise accusative is assigned.24 24

We assume that the stative feature is a lexical rather than a true semantic property—i.e., t+stative] is not necessarily present with all stative predicates. There are two reasons for this assumption. First, as was pointed out to us by Mikinari Matsuoka, not all semantically stative predicates allow nominative objects (e.g., believe, know only occur with accusative objects). Second, as mentioned in fn. 15, accusative objects are also possible with stative verbs for at least some speakers and some predicates, indicating that the stative feature can be optional.

48

Lexical restructuring

Turning now to the structures for (36), we propose that (36)a involves a restructuring verb which combines with a VP-complement (cf. (37)a). The example in (36)b, on the other hand, is a nonrestructuring verb which combines with a vP-complement (cf. (37)b). (37)

a.

Restructuring infinitive

=(36)a

TP

ringo

b.

tabe

Non-restructuring infinitive TP

=(36)b

Case assignment

49

Since restructuring infinitives do not contain a structural object case position, the embedded object has to raise to the matrix predicate to check its case. If the matrix verb is a stative verb as in (37)a, the embedded object shows up with nominative. In non-restructuring infinitives such as (37)b, on the other hand, the infinitival complement projects its own vP. The closest case position for the embedded object is hence the embedded vP (i.e., an accusative position) rather than the matrix vP (we assume here that no double case checking applies). The major difference between (36)a/(37)a and (36)b/(37)b is thus the position where the embedded object checks or is assigned case—in (37)a, the case position is higher than the matrix verb, in (37)b, the case position is lower than the matrix verb.25 Assuming an analysis along these lines, we expect to find differences between the two examples in (36) that can be attributed to the different structural positions of the embedded object in these examples. We will see that this is indeed the case: the examples in (36) display different scope and binding properties that reflect the different positions of the embedded object in the structures in (37). As was noted by Tada (1992), in sentences like (38) which differ minimally in the case of the embedded object, nominative objects take scope higher than accusative objects (see also Tada 1993; Koizumi 1995). The example in (38)a, involving an accusative object, conveys that John has the ability to close only his right eye (and to leave his left eye open). That is, can takes scope over the object.26 The sentence in (38)b, involving a nominative object, on the other hand, conveys that only John's right eye is such that he can close it (i.e., he cannot close 25

See Saito and Hoshi (1998, 2000) for a different analysis. Saito and Hoshi assume that Japanese restructuring constructions involve a form of complex predicate formation. Since complex verb formation cannot be maintained for German restructuring constructions (cf., the possibility of separating the two verbs by XPmovement such as in Zu reparieren wurde nur der Wagen versucht 'To repair was only the car-NOM tried'), we will not pursue this approach here. Furthermore, since we believe that the similarities between the Japanese and the German restructuring constructions are not accidental, a unified account seems desirable. 26 The example in (38)a is ambiguous for some speakers when the object is focused (cf. Koizumi 1995: 68, fn. 6). We assume that an inverse scope reading results from focus scrambling which is irrelevant for the present discussion.

50

Lexical restructuring

his left eye). In this case, the object takes scope over can. (38) Japanese: scope of embedded object a.

John-ga migime-dake-o John-NOM right-eye-only-ACC 'John can close only his right eye' can>only; V.only>can b. John-ga migime-dake-ga John-NOM right-eye-only-NOM 'John can close only his right eye' *can>only; only>can

tumureru close-can

tumureru close-can

Thus, the importance of the scope properties in (38) is that in these examples, accusative objects obligatorily take scope under the restructuring verb (modulo fn. 26), whereas nominative objects obligatorily take scope over the restructuring verb. Assuming that scope reflects a hierarchical (c-command) relation between two elements, the two sentences in (38) have the LF-representations in (39): (39) Japanese: LF positions of objects in (38) a.

Accusative objects XP

b.

Nominative objects XP can

Assuming that the position of the modal verb is the same in both structures in (39), we can conclude that nominative objects appear in a higher position at LF than accusative objects. What are the positions of the objects in (39)? Under the account we have suggested above, the answer is straightforward: the scope positions are the case positions.27 In restructuring infinitives, case is assigned in the higher 27

This is of course a simplified picture of scope in Japanese (see for instance Yatsushiro (1999) for a more detailed analysis in particular of the interaction of case and scope). However, it seems to be in line with the core of most analyses of scope in Japanese—namely the claim that quantifiers are assigned surface scope

Case assignment

51

predicate—hence the object also takes scope in the higher predicate. In non-restructuring infinitives, case is assigned in the lower predicate and the object takes scope in the lower predicate. The scope properties thus provide strong support for the claim that the infinitives in (36)b and (38)a involving accusative objects are nonrestructuring infinitives. A crucial point of our analysis is that the different scope positions are not associated with inherently different positions for accusative vs. nominative arguments (but see fn. 28 for some qualification of this claim). Rather, we claim that the scope positions correspond to the case positions. In restructuring infinitives, the object takes scope over the matrix verb since it is assigned case in a position higher than the matrix verb; in non-restructuring infinitives, the object takes scope under the matrix predicate since it is assigned case in a position lower than the matrix verb. Further support for this approach comes from non-stative restructuring verbs such as begin. Since nonstative verbs do not license nominative objects, the object in an infinitive embedded under the non-stative verb begin can only show up with accusative (cf. (40)). (40)

Japanese: accusative objects a.

Emi-ga Emi-NOM 'Emi began b. *Emi-ga Emi-NOM 'Emi began

ringo-o apple-ACC to eat apples' ringo-ga apple-NOM to eat apples'

tabe-hazime-ta eat-begin-PAST tabe-hazime-ta eat-begin-PAST

According to our analysis, examples such as the one in (40)a nevertheless have two structural representations—a restructuring and a non-restructuring version. In the former, the object is assigned case by the matrix vP, in the latter, it is assigned case by the embedded vP. However, both the restructuring version and the non-restructuring version end up with an accusative marked object. The structures are illustrated below in (41). and do not undergo QR (see for instance Hoji 1985 among many others).

52

(41)

Lexical restructuring

a.

=(40)a

Restructuring infinitive TP

T'

NOM Emi-ga

r

vP

-ta •ACC -o

ν

VP

[-stative]

V' Vo hazime

VP OBJ ringo

b.

Vo tabe

Non-restructuring infinitive TP

Emi-ga

=(40)a

Case assignment

53

Given the structures in (41), our analysis—correctly—predicts that examples such as the one in (40)a are ambiguous: the embedded accusative object can take scope higher or lower than the matrix predicate. Examples which illustrate the ambiguity of accusative objects are provided in (42). (42)

Japanese: scope of accusative objects a.

b.

Emi-ga ringo-dake-o Emi-NOM apple-only-ACC 'Emi began to eat only apples' begin>only; only>begin Emi-wa niku-dake-o Emi-TOP meat-only-ACC 'Emi overdid eating only meat' overdo>only; only>overdo

(Koizumi 1995: 61-62) tabe-hazime-ta eat-begin-PAST

tabe-sugi-ta eat-overdo-PAST

We can thus conclude that it is not an inherent difference between accusative and nominative that determines the LF-position of the embedded object but rather a distinction along the lines of restructuring vs. non-restructuring.28 28

The situation is in fact slightly more complex. As was pointed out by Yatsushiro (1999), constructions involving a non-stative matrix predicate and a stative embedded predicate allow nominative or accusative objects. However, in contrast to what would be predicted by the analysis presented here so far, nominative objects take scope over the matrix predicate in these contexts. A possible way to accommodate these facts is by the following two assumptions: i) stativity can be contributed by either predicate in a restructuring construction, and ii) nominative is not assigned by v° but by a higher functional head (e.g., by T°, following Koizumi 1995). As we will see in section 3.4, Japanese affixal constructions cannot involve tense material in the infinitival complement, indicating that these infìnitives lack a T-projection. Thus, complements to affixal verbs are at most vP complements. Given this fact together with the assumption that nominative is assigned in a projection higher than vP, it then follows that nominative can never be assigned inside an infinitival complement to an affixal verb (even when the complement is a non-restructuring infinitive—i.e., a vP). Thus for the problematic restructuring construction under consideration, we would assume that the [+stative] feature from the lower verb percolates up to T°, licensing nominative case on the embedded object. However, since nominative is assigned in the matrix TP, the object will always take scope

54

Lexical restructuring

The second argument for the claim that examples such as the one in (36)b (i.e., sentences involving an accusative object with a stative matrix predicate) are non-restructuring infinitives comes from the binding properties of these constructions. As has been noted by Miyagawa (1987), restructuring infinitives in contrast to nonrestructuring infinitives do not constitute an independent domain for binding. In particular, Miyagawa shows that pronouns embedded in a restructuring infinitive cannot be construed as coreferential with a ccommanding R-expression in the matrix predicate. Miyagawa's observation is illustrated in (43). The affixal motion verb go which combines with an infinitival complement is a stative restructuring verb, and hence the embedded object can show up with either accusative or nominative case (cf. (43)a,b). However, as is indicated by the indices in (43), the binding properties of the two examples are crucially different: when the object occurs with accusative, coreference between the embedded pronoun and a matrix antecedent (Hanako) is possible (cf. (43)a); when the object occurs with nominative, coreference between the embedded pronoun and a matrix antecedent is blocked (cf. (43)b). (43)

Japanese: binding properties in infinitives a.

b.

Taroo-ga Hanakorto daigaku-ni Taroo-NOM Hanako r with university-to kanozyOty -o syookaisi-ni ikeru [she^j -ACC introduce-to go-can 'Taro can go with H. to the university to introduce her' Taroo-ga Hanakorto daigaku-ni Taroo-NOM Hanako r with university-to kanozyo,·ϋ] -ga syookaisi-ni ikeru [she*i/j -NOM introduce-to go-can 'Taro can go with H. to the university to introduce her'

over the matrix predicate. At this point, we will not commit ourselves to either of the two analyses. What is important for our purposes here is that both analyses support our main approach: restructuring entails that the object is assigned case by a case assigner in the matrix predicate. If the object is assigned case by a head in the infinitival complement the infinitive is a non-restructuring infinitive.

Case assignment

55

Under the analysis we have proposed here for restructuring infinitives, these facts receive a straightforward account. The impossible binding relation in the restructuring example in (43 )b constitutes a standard principle Β violation. Since the embedded object is assigned case by the matrix v° or T, it ends up in the matrix predicate. Assuming that binding takes place at S-structure or LF, the matrix argument with Hanako and the embedded object are then part of the same binding domain, resulting in a principle Β violation. We will not commit ourselves to any particular binding theory here, but simply note that the violation we find in (43)b should be accounted for in the same way as the prohibition of coreference between the matrix subject and an ECM subject pronoun in English (cf. (44)a). (44)

Binding in ECM constructions a. b.

Johnj expects him^ to win the race Johns expects Mary to like him/^himself^

The lack of a Principle Β violation in (43)a, on the other hand, indicates that an infinitive involving an accusative object in a stative context constitutes an independent binding domain in Japanese, just like English ECM constructions from the perspective of the embedded object (cf. (44)b). To account for the contrast in (43), let us assume a weak version of Burzio's generalization, namely that structural case is tied to the presence of an external argument. In order to license an embedded structural accusative case position, an infinitival subject has thus to be projected in non-restructuring infinitives. Furthermore, following standard approaches to binding, we assume that an external argument demarcates an independent binding domain (cf. for instance Chomsky 1980, 1981, 1986b; Lasnik 1989; Reinhart and Reuland 1993 among many others). Under these assumptions, the object pronoun and the matrix antecedent Hanako are in different binding domains in examples such as (43 )a, and hence no principle Β violation arises when these two arguments are coreferential. The structure we suggest is depicted below in (45). To sum up, the scope and binding properties discussed in this section have shown that in stative contexts, nominative objects are part

56

Lexical restructuring

of the matrix predicate, whereas accusative objects are part of the embedded predicate at the level where scope and binding apply. We have argued that this contrast reflects two different structures—restructuring infinitives which lack an embedded case position/assigner vs. non-restructuring infinitives which involve an embedded case position/assigner. (45)

Non-restructuring infinitive

=(43)a

TP

VP

kanozyo

ν

syookaisi-ni

Moreover, we have seen that in non-stative contexts which only allow accusative objects, the embedded object can either be part of the matrix predicate or the embedded predicate, which argues strongly for two different structures rather than solely an inherent difference between nominative and accusative case positions. The correlations between scope, binding, and case discussed in this section thus again confirm the claim that whenever an infinitive involves its own structural object case position, the infinitive is a non-restructuring infinitive.

Case assignment

57

2.4. The lack of case and bi-clausal approaches to restructuring In this section, we will compare the analysis we have proposed with clausal approaches to restructuring—i.e., approaches that assume that restructuring as well as non-restructuring infinitives involve an underlying clausal structure. Let us first summarize the major facts that have to be accounted for. First, in restructuring constructions, the embedded object (obligatorily) enters into a case and agreement relation with a functional head of the matrix predicate. Second, in non-restructuring constructions, the embedded object (obligatorily) enters into a case and agreement relation with a functional head of the embedded predicate. The key examples are repeated in (46); (46)a vs. (46)b shows that only certain predicates can combine with restructuring infinitives and that non-restructuring infinitives block long passive; (46)c vs. (46)d shows that restructuring is not obligatory but that predicates that allow a restructuring complement can also combine with a nonrestructuring complement which then, however, crucially blocks all restructuring properties. (46)

Restructuring vs. non-restructuring a.

dass der Traktor zu reparieren versucht wurde that the tractor-NOM to repair tried was 'that they tried to repair the tractor' b. *dass der Traktor zu reparieren geplant wurde that the tractor-NOM to repair planned was 'that they planned to repair the tractor' c. Es wurde versucht den Traktor zu reparieren It was tried the tractor-ACC to repair They tried to repair the tractor' d. *Es wurde den Traktor versucht zu reparieren It was the tractor-ACC tried to repair "They tried to repair the tractor' According to the analysis we have suggested, restructuring infinitives lack an embedded case position/assigner, whereas non-restructuring infinitives involve an embedded case position/assigner. Long object

58

Lexical

restructuring

movement is thus always forced in the former and blocked in the latter by standard locality (i.e., closeness) considerations. Under a clausal structure for restructuring infinitives, it seems less obvious how the facts in (46) can be accounted for. The major challenge that most clausal analyses face is to account for the nonoptionality of long object movement in restructuring constructions (as discussed in the previous section) and to provide a motivation for these types of movement. Assuming a structure as in (47) (details about the number and label of the various projections in the infinitive are irrelevant), the main questions are: i) why are the case properties of the embedded object affected by the presence vs. absence of structural case in the matrix clause (or put differently, why is this movement operation obligatory); ii) how can object movement apply across an embedded case position and the embedded subject; and iii) why is long object movement blocked in non-restructuring infinitives. (47)

Clausal structure for restructuring infinitives TP

V· tried

c

ν

VP the truck

to repair

Case assignment

59

Let us discuss these questions in more detail. In general, argument structure alternations like passive do not apply across clauses. Since the infinitive and matrix clause project independent argument structures in a structure like (47), it seems unclear what the motivation for long object movement would be. Recall from the discussion in the previous sections that accusative is only possible in a restructuring infinitive when the matrix predicate involves an accusative assigner. If accusative cannot be assigned in the matrix predicate and the object nevertheless shows up with structural accusative case, the infinitive is a non-restructuring infinitive. Since in restructuring infinitives nothing alters the argument structure of the embedded predicate in cases where the matrix predicate is unaccusative or passivized (i.e., in particular, the embedded predicate shows up in the active form rather than the passive form), it is not clear why the ability to assign accusative should be lost in the infinitive under a structure like (47). As for the locality conditions of long object movement, it has to be explained why the embedded subject does not intervene when the object raises to the matrix subject position (i.e., why Relativized Minimality is not violated in restructuring infinitives), and why the closest case position for the embedded object—the embedded object case position—can and in fact has to be skipped in restructuring infinitives. Let us first consider the head raising approaches to restructuring, i.e., approaches that assume that restructuring infinitives are fullfledged CPs or TPs, and that restructuring essentially means that a verbal head (T°, v", or Vo) raises from the infinitive to the matrix clause to make the infinitival projections transparent for further movements (see Chapter 1 for references). While verb raising of this sort eliminates boundaries for movement operations such as scrambling or clitic climbing, it generally does not have any effect on the argument structure of a predicate. In other words, verb raising could account for the possibility of long head and A'-movement, but it does not provide an answer to the questions long Α-movement raises (in particular, the question of how the embedded subject can be skipped and why the embedded object has to skip the embedded case position). One way of saving head movement accounts would be to modify

60

Lexical restructuring

the theory of verb raising. Possibilities that come to mind are to assume that verb raising eliminates the embedded PRO subject (e.g., as a result of government by the trace of the verb in C°), or that it 'carries away' the structural case (feature) of the infinitive. However, crucially, an embedded PRO subject or accusative case would then be impossible in passive and active restructuring contexts, since verb raising applies from restructuring infinitives in general. In other words, it should not be possible to ever assign accusative to an object inside a restructuring infinitive. Thus, what these modified verb raising approaches would end up with is exactly what we are proposing here: the object in a restructuring infinitive does not check case in the infinitive. The question then obviously is whether there is evidence or reason to assume an initial representation with an embedded structural case position, and—provided there is such evidence—to balance whether the evidence is good enough to invoke a powerful process like deletion (of PRO or the case features) or feature hijacking. The VP-movement approach and the topicalization/headmovement approach, on the other hand, offer a more principled way to account for the lack of Relativized Minimality effects of the embedded PRO subject for long object movement. Recall that in these approaches, the infinitival verb phrase or some projection of it is moved to the matrix clause or to the specifier of the embedded CP (depending on the actual version of this type of approach). As a result, the object is removed from the c-command domain of the infinitival PRO subject. In terms of Relativized Minimality, this then means that the embedded subject does not intervene anymore when the embedded object moves to the matrix clause. The structure in (48) is a simplified version of the structure proposed in Sabel (1994/1996). Note that it remains undetermined which argument (the embedded object or the embedded subject) is closer to the matrix T° in the structure in (48) below since the two arguments do not enter into any c-command relation with each other.29 29

The question of being closer does not arise in Grewendorf and Sabel's (1994) and Sabel's (1996) approach since their accounts are rooted in the Barriers framework; according to these analyses, movement of PRO from the specifier of the em-

Case assignment

61

While the VP-movement and topicalization/head-movement approaches provide an account for the lack of blocking effects of the embedded PRO subject, the question of why the embedded object has to move to the matrix subject position remains mysterious in this set of approaches as well. Although there are attempts to capture the fact that the object cannot check case in the infinitive in passive contexts, they seem to simply restate the facts. Sabel (1996: 205-206) for instance asserts that the restructuring feature when transferred from a passivized matrix restructuring verb to the embedded AgrO° absorbs the case assigning property of the embedded AgrO°. Besides the question of what a restructuring feature is, it is not clear why and how it eliminates structural case in the infinitive. (48)

Topicalization and overt/covert verb raising TP NOM^^

T'

To sum up, the major problems that clausal approaches to restructuring are faced with is to circumvent the infinitival subject and to develop a restructuring-specific mechanism that renders the embedded structural case position inactive. While we are not claiming that these tasks are unachievable, we agree with Strozer (1981), who points out that theories that postulate a clausal structure for restructuring infinitives (and a structurally present infinitival subject) have to be adjusted in non-obvious ways to explain why the PRO subject does not intervene when the embedded object raises to matrix subject bedded IP to the matrix clause would cross two barriers, which would be illicit.

62

Lexical restructuring

position. Moreover, the question of what drives object movement in restructuring infinitives, and in particular, what forces movement of an argument that should be perfectly happy within the infinitival complement is a serious challenge for this set of approaches that appears to be less trivial to overcome. The approach proposed here, on the other hand, has the advantage of not requiring any additional assumptions or mechanisms to account for the variety of long object movement operations discussed in this section. Most importantly, however, the approach presented here not only accounts without further stipulations for the contrasts summarized in (46) but it also succeeds in motivating the existence of long movement operations in restructuring constructions.

3. Tense properties of restructuring constructions In the previous section, we have seen that lexical restructuring infinitives do not involve a structural case position or assigner and that the embedded object enters into a case and agreement relation with the closest head in the matrix predicate. In this section, we will investigate the tense properties of infinitives and show that restructuring infinitives also lack an independent internal tense specification which will be taken to reflect the lack of a T-projection. 3.1.

Infinitival tense

Since Stowell's (1981, 1982) work on the tense of infinitives, many studies distinguish between two classes of infinitival constructions: irrealis vs. propositional infinitives. Irrealis infinitives refer to situations in which the event denoted by the infinitive is necessarily unrealized or uncompleted at the time of the matrix event. That is, examples like (49)a,b cannot refer to situations in which John has already gone to Kamchatka at the time of trying or deciding. Propositional infinitives, on the other hand, do not presuppose or assert anything about the embedded event. Thus, in a propositional infinitive like (49)c, the embedded event could have occurred at the time of John's

Tense properties

63

believing (i.e., Mary could have gone to Kamchatka). (49)

Irrealis vs. propositional infinitives a. b. c.

John John John

tried decided believed

to go to Kamchatka to go to Kamchatka Mary to have gone to Kamchatka

Stowell attributes the distinction between irrealis and propositional infinitives to the presence vs. absence of infinitival tense (see also Martin 1996; Boskovic 1995, 1996, 1997). Propositional infinitives are tenseless, whereas irrealis infinitives involve infinitival tense which is characterized as an unrealized or possible (quasi-)future tense (Stowell 1982: 562). While the actual implementation of the irrealis/propositional distinction is subject to some debate (see below), most authors agree that the distinction between different semantic types of infinitives is essential for an appropriate characterization of the syntactic properties of infinitival constructions (see in particular Pesetsky 1992). We will come back to this syntax/semantics connection in Chapter 5 where we will show that the distinction between propositional and factive infinitives, on the one hand, and irrealis infinitives, on the other hand, correlates with various syntactic properties of these constructions. In particular, the semantic types of infinitival constructions will prove to be of high importance for the determination of the different grades of (non)restructuring. Concerning infinitival tense, however, we will argue here that the irrealis/propositional distinction cannot be reduced to the presence vs. absence of infinitival tense. We will see that despite the fact that the examples in (49)a and (49)b both involve irrealis complements, the tense properties of these infinitival complements are quite different. With Stowell, we will assume that the tense properties divide infinitival complements into (at least) two classes, however, contrary to his conclusion, we will claim that these classes do not correlate with the irrealis/propositional distinction but rather with the restructuring/non-restructuring distinction.30 As we will argue in Chapter 4, restructuring infinitives do not involve an em-

64

Lexical restructuring

Let us start with some basic differences between irrealis and propositional infinitives. Following Pesetsky (1992: 143), a property of propositional infinitives is the possibility of truth/falsity predication, i.e., of interpreting the tag which was true as asserting the truth of the infinitival complement alone. As is illustrated in (50)a (see also the appendix, section 2 for a more detailed list of examples), complements to verbs like claim (i.e., propositional complements) allow predication of the truth of the complement. Non-propositional complements such as infinitives combining with irrealis verbs like decide, on the other hand, do not allow predication of the truth of the complement (cf. (50)b). Note that this claim is independent of the tense of the truth predicational statement (cf. the impossibility of a future statement; the way to convey a type of confirmation of the embedded event is not by truth predication but rather by statements like which he will do). (50)

Truth/falsity predication a.

b.

Hans behauptete im Lotto gewonnen zu haben John claimed in-the lottery won to have was auch stimmte which also was-true 'John claimed to have won the lottery, which was true' => it is true that John won the lottery Hans beschloß ein Fahrrad zu kaufen John decided a bicycle to buy was auch stimmte which also was-true 'John decided to buy a bicycle, which was true' => it is true that John decided to buy a bicycle φ it is true that John bought/will buy a bicycle

bedded PRO subject and hence are technically not control infinitives. In this sense, Stowell's generalization—namely that only control infinitives involve tense—will again be valid. However, the empirical domains captured by the two generalizations (i.e., Stowell's proposal that only irrealis control infinitives involve tense; vs. the proposal here, namely that only non-restructuring infinitives involve tense) are quite different.

Tense properties

65

*was auch stimmen wird which also be-true will *'John decided to buy a bicycle which will be true' A second property that distinguishes irrealis and propositional infinitives in German is the (im)possibility of embedded past modifiers such as gestern 'yesterday' or vor zwei Tagen 'two days ago'. As is illustrated in (51)a, propositional and factive complements allow modification of the embedded event by past indexicals—i.e., modifiers that are deictic in that they take the utterance time of the sentence as the reference time (note that the infinitive is extraposed which shows that the modifier is unambiguously in the embedded clause). Irrealis complements, on the other hand, are incompatible with this kind of modification as is illustrated in (51)b. 31 (51)

Embedded past modifiers a.

Hans hat behauptet behauptet / bedauert John has claimed / regretted gestern/vor zwei Tagen Kirschen gegessen zu haben yesterday/two days ago ehernes eaten to have 'John claimed/regretted that he ate cherries yesterday/two days ago' *Hans hat beschlossen / geplant John has decided / planned

•Ι Unfortunately, the borders between different classes of infinitival constructions are not always as clear-cut as one would like them to be. In many cases, infinitives (or verbs embedding infinitives) can be assigned to more than one semantic class. The verb decide, for instance, which in an unmarked context is interpreted as an irrealis verb, can also (at least marginally) be used as a propositional verb: in a context where John has been looking for his wallet for two days, one might say Eventually, John decided to have lost his wallet, with an interpretation along the lines 'John decided eventually to believe/accept that he had lost his wallet'. In the discussions to follow, we will largely restrict ourselves to the unmarked usages of the constructions under consideration and ignore these kinds of re-interpretations or coerced readings. However, the possibility of class-switching should be kept in mind since it has obvious consequences for the characterization of the properties of different types of infinitival constructions.

66

Lexical restructuring

gestern/vor zwei Tagen zu verreisen yesterday/two days ago to go-on-a-trip 'John decided/planned to go on a trip yesterday/two days ago' It is important to point out that the impossibility of (51)b cannot simply be reduced to the claim that these infinitives have to refer to a time that is after the utterance time (which would then be incompatible with past adverbials). As is illustrated in (52)a, infinitives combining with a verb like decide can refer to a time in the past (i.e., a time prior to the utterance time), as long as the infinitive is irrealis/future with respect to the (event) time of the matrix predicate, and the temporal modifier is not a past indexical (cf. (52)a vs. (52)b). Thus, the contrasts in (51) and (52) show that it is indeed the past modifier that is causing the problem in the ungrammatical examples, and not the fact that the tense of the irrealis complement is understood to be before the utterance time in these examples. (52)

Past future interpretation a.

b.

Vor einer Woche hatte Hans beschlossen A week ago had John decided am nächsten Tag zu verreisen on-the next day to go-on-a-trip Ά week ago, J. had decided to go on a trip the next day' Vor einer Woche hatte Hans beschlossen A week ago had John decided (*gestern) zu verreisen (*yesterday) to go-on-a-trip Ά week ago, John had decided to go on a trip yesterday'

Although we will not be able to provide an account for the distribution of tense modifiers in German, we want to point out the descriptive generalization that is behind this restriction of modification in irrealis contexts (see also Wurmbrand 2001). To do so, let us first look at examples with finite embeddings which show a similar restriction. If an embedded clause receives a past future interpretation (i.e., an interpretation where the time of the embedded event is un-

Tense properties

67

derstood to be after the time of the matrix event but before the utterance time), past modifiers are impossible (cf. (53); the same appears to be the case in French in examples similar to (53)c which was pointed out to me by Philippe Schlenker). The sentence in (53)a can be saved (in at least certain dialects) by changing gestern 'yesterday' to so wie gestern 'such as yesterday'. (53)

Past modifiers in finite complements a. *Hans hatte schon vor einem Monat angekündigt John had already a month ago announced dass er gestern heiraten werde/würde/wird that he yesterday get-married will-COND/would/will 'John had already announced a month ago that he would get married yesterday' b. *Hans hatte schon vor langem angekündigt dass John had already a while ago announced that er vor zwei Tagen heiraten werde/würde/wird he two days ago get-married will-COND/would/will 'John had announced already a while ago that he would get married two days ago' c. *1986 hat Hans gedacht dass er 1986 has John thought that he vor drei Jahren in Pension gehen werde/würde/wird three years ago retire will-COND/would/will 'In 1986, John thought that he'd retire three years ago'

Note that this problem cannot be reduced to a tense clash in these examples or a special property of indexicals. As is evident from the wellformed paraphrases in (53), past modifiers are possible in the same past future contexts in English. Furthermore, the examples in (54)a,b demonstrate that past future interpretations are in principle possible in German, however, only if the sentences do not involve past modifiers (future oriented modifiers are licit). Finally, in (54)c, we find that future indexicals can be used in clauses that are dependent on a past event. The deictic adverbial tomorrow in (54)c can only be interpreted as 'the day after today' and not as 'the day after the

68

Lexical

restructuring

announcement was made'. This example thus shows that indexicals can access the utterance time in embedded contexts in German. (54)

Temporal modifiers in finite complements a.

b.

c.

Hans hatte vor einem Monat angekündigt dass er John had a month ago announced that he eine Woche später heiraten werde/würde/?wird a week later get-married will-COND/would/?will 'John had announced a month ago that he would get married a week later' 1986 hat Hans gedacht/gesagt/beschlossen 1986 has John thought/said/decided dass er 1997 in Pension gehen werde/würde/?wird that he 1997 retire will-COND/would/?will 'In 1986, John thought/said/decided that he would retire in 1997' Hans hatte vor einem Monat angekündigt John had a month ago announced dass er morgen heiraten werde/würde/?wird that he tomorrow get-married will-COND/would/?will 'John had announced a month ago that he would get married tomorrow'

The generalization about temporal modification in German thus seems to be that past modifiers can only occur in clauses with a basic past orientation (i.e., clauses that express a before relation). In the past future examples in (53), the temporal orientation of the embedded clauses is future/irrealis—i.e., the embedded event is unrealized or set in the future with respect to the matrix event (cf. the obligatory conditional in English). Crucially, in these contexts, past modifiers—i.e., modifiers expressing a before relation—are prohibited. With this generalization in mind, we can now draw certain conclusions about the nature of irrealis infinitives. Since irrealis infinitives—in contrast to propositional and factive infinitives—never allow past modifiers, we can conclude that these constructions are inherently incompatible with a past specification. Going back to Stow-

Tense properties

69

ell's analysis, this observation would follow naturally since irrealis infinitives by definition involve future tense. While we agree with the claim that irrealis infinitives are incompatible with a past specification, we will see below that this incompatibility is of a more subtle nature and cannot be reduced to the presence of a quasi-future tense. Before doing so, we will first look in more detail at Stowell-type accounts of the distribution of infinitival constructions. Stowell claims that the distinction between irrealis and propositional infinitives is directly reflected in the syntax of infinitival constructions. He notes that irrealis infinitives correlate with control infinitives, whereas propositional infinitives correlate with ECMinfinitives. To account for this correlation, Stowell assumes that infinitival tense is represented in COMP and that the lack of tense corresponds to the lack of a CP. Since ECM is only possible with CPless infinitives, the correlation between the lack of tense and ECM vs. the presence of tense and control follows. Similarly, Martin (1992, 1996) and Boskovic (1995, 1996, 1997) assume that irrealis control infinitives are [+tense], whereas propositional complements (i.e., raising and ECM-infinitives) are [-tense], and that this correlation between the type of the infinitive and the value of tense is expressed in terms of case assignment in infinitives. Following Chomsky and Lasnik (1993/1995), these authors assume that infinitival PRO requires Null Case; furthermore, only [+tense] infinitival Τ can assign this special case. Thus, under these type of approaches, it again follows that ECM and raising constructions cannot involve a PRO subject. There are, however, a number of reasons that speak against this classification of tense and the connection between infinitival tense and the type of the infinitival complement (i.e., ECM, raising, control). First, as has been pointed out by Pesetsky (1992), both correlations are not perfect: on the one hand, there are irrealis ECMinfinitives (e.g., infinitives combining with verbs like expect or want; see Pesetsky for arguments that wanf-constructions involve ECM); and on the other hand, not all control infinitives receive an irrealis interpretation—infinitives selected by implicative and factive verbs are control infinitives rather than ECM or raising infinitives, but these constructions are not assigned an irrealis interpretation. Fur-

70

Lexical restructuring

thermore, the claim that prepositional complements are [-tense] (and hence prohibit PRO) has to be restricted to English. French and German, for instance, allow propositional control complements, and hence, as Martin and Boskovic admit, propositional complements have to be classified as [+tense] in these languages. Since it is not obvious that the tense interpretation of &e/ieve-constructions is different in English and German (but see Martin 1996 for some potential evidence from French), language variation along these lines seems to weaken the analysis. A similar problem arises for constructions that are assumed to be ambiguous between a raising and control structure (i.e., infinitives combining with aspectual verbs such as begin, or verbs like promise or threaten). It would have to be assumed that the tense properties are different in the raising vs. the control version, which does not seem to be motivated. Another observation we would like to point out is that some authors end up with exactly the opposite tense values for the same type of infinitival construction. Raposo (1987) considers infinitives selected by volitional predicates (i.e., certain irrealis infinitives) as [-tense], whereas epistemic and declarative predicates (i.e., propositional infinitives) are assumed to select [+tense] complements. Without taking sides on which is the correct tense value, the existence of this inconsistency seems to cast some doubt on the legitimacy of the [±tense] specification as a semantic property and point to the conclusion that the [±tense] feature as used in the above analyses has to be seen as some diacritic regulating the case properties of the embedded subject. Finally, a closer look at the temporal properties of the examples in (49)a,b reveals another problem for the assumption that an irrealis interpretation corresponds to future tense. As is illustrated in (55)a, the embedded predicate can be modified by a future oriented adverbial phrase like in two months. Infinitives selected by the verb try, on the other hand, do not allow future oriented adverbials (cf. (55)b). (55)

Adverbial modification of embedded tense a.

Hans beschloß Maria in zwei Monaten zu besuchen John decided Mary in two months to visit 'John decided to visit Mary in two months'

Tense properties

b. *Hans versuchte María in zwei Monaten John tried Mary in two months 'John tried to visit Mary in two months'

71

zu besuchen to visit

Importantly, however, both constructions involve irrealis infinitives. Complements combining with the verb try behave just like the irrealis examples discussed above in that truth predication of the complement (cf. (56)a) and embedding of past modifiers (cf. (56)b) are impossible (note that the latter is also ungrammatical when the time of the matrix predicate and the embedded time are understood as simultaneous; we will come back to the temporal orientation in these infinitives in the next section). (56)

Irrealis properties of try a.

Hans versuchte ein Fahrrad zu kaufen John tried a bicycle to buy was auch stimmte which also was-true 'John tried to buy a bicycle which was true' => it is true that John tried to buy a bicycle it is true that John bought/will buy a bicycle b. *Hans hat versucht gestern zu verreisen John has tried yesterday to go-on-a-trip 'John tried to go on a trip yesterday' The contrast in (55) thus raises a question for the claim that irrealis infinitives involve infinitival tense. If irrealis infinitives are specified for a (quasi) future tense, it would not be clear why in examples such as (55)b, the infinitival event cannot be modified by a future adverbial, given that this type of modification is available in other irrealis complements (e.g., (55)a). If we assume that irrealis is defined in terms of future tense, we would be faced with a paradox: the impossibility of past modifiers in ^-infinitives indicates that these infinitives are irrealis, hence involving future tense; the impossibility of future adverbials indicates that these infinitives are incompatible with a (true) future specification. In the next section, we will propose an analysis that avoids this problem.

72

Lexical restructuring

3.2. Tense vs. irrealis aspect The analysis we would like to suggest is built on the assumption that the property of being irrealis is independent of the property of infinitival tense. In particular, we claim that infinitives fall into two classes—tensed and tenseless infinitives—and that both types of constructions can be interpreted as irrealis infinitives. The choice between a tensed or a tenseless infinitive is determined as part of the meaning of the higher predicate. Following standard assumptions, tense is represented structurally as part of the tense-projection (whether as tense features, a before/after relation, or a tense operator will not be crucial for the discussion here). Furthermore, embedded tense is dependent on the tense of the higher predicate in that it takes the event time of the higher predicate as the reference time for the embedded event (cf. Klein 1994). The function of infinitival tense is then to order the time of the embedded event with respect to the (event) time of the higher situation. For the property irrealis, on the other hand, we assume that it is an aspectual (or mood) property, which essentially expresses that an event is not realized at the reference time, which in dependent clauses, is the event time of the higher predicate. The two structures we suggest are illustrated in (57) (the position of vP is ignored here for simplicity). (57)

a.

Tensed irrealis infinitive b. Tenseless irrealis infinitive AspP

TP

V'

ν

Tense properties

73

Note that we represent tenseless infinitives as lacking a tense projection altogether; alternatively, one could assume that tenseless infinitives involve a tense projection, however, that this projection is deficient and lacks content (see section 5.2 for a discussion of this issue). We also assume that irrealis infinitives do not necessarily project a C-domain (motivation for this claim will be provided in Chapter 5). Under these assumptions, the only possible setting for infinitival tense in irrealis contexts is [-PAST]. A past tense in the infinitive would lead to a contradiction, since it would specify that the time of the infinitival event is ordered before the time of the matrix event. This situation, however, would then be incompatible with irrealis aspect in the infinitive which specifies that the embedded event is unrealized at the time of the higher event. It thus follows that irrealis infinitives are incompatible with a past orientation of the embedded event. The final assumption we make to accommodate the distribution of temporal modifiers in German is that past adverbials cannot occur in irrealis contexts in German. As mentioned above, we do not have an explanation for this property at this point and further investigation of the interaction of tense and aspect in German is necessary to draw any firm conclusions. We will show, however, that this assumption is on the right track. Let us now see how this system accounts for the facts discussed so far. The two types of irrealis infinitives are represented here by a dec¿de-infinitive (tensed) and a ^-infinitive (tenseless). As illustrated again in (58), future modification is only possible in the former but not in the latter (but see section 3.3 for some qualification of this claim). (58)

Two classes of infinitives a.

b.

Hans John 'John Hans John 'John

hat beschlossen (morgen) zu verreisen has decided (tomorrow) to go-on-a-trip decided to go on a trip (tomorrow)' hat versucht (*morgen) zu verreisen has tried (*tomorrow) to go-on-a-trip tried to go on a trip (*tomorrow)'

74

Lexical

restructuring

Assuming that ¿fedde-infinitives involve a future-type tense, the possibility of a future adverbial as in (58)a is not surprising. Infinitival future tense specifies that the time of the embedded event is after the time of the matrix event, which itself is in the past (i.e., before the utterance time). The indexical adverbial tomorrow specifies further that the embedded time is (contained in) the day after the utterance time. Since these two temporal specifications are compatible, nothing blocks future adverbials in this context. In case of a //7-infinitive, however, the situation is different. The crucial claim we make is that fry-infinitives are tenseless (though irrealis) and that the temporal location of the embedded event is fully determined by the tense of the higher predicate. Under this (simplified) assumption (see below for a refined analysis), it is easy to see already that examples such as (58)b represent a contradiction. In the absence of an infinitival tense specification, the matrix past tense determines that both the time of the attempt and some aspect (to be made precise below) of the time of the embedded situation are before the utterance time. The indexical adverbial, on the other hand, specifies that the time of the embedded event is after the utterance time. Since both temporal specifications cannot be met at the same time, the sentence is impossible (in English and German). Note that we make an important distinction about tense dependencies here. In both decide and try-type infinitives, the infinitival tense is in a sense dependent on the tense of the matrix event. In decidetype infinitives, the time of the infinitive is specified as being after the time of the matrix event. As discussed above, this is the result of the infinitival future tense which establishes a temporal ordering relation between the time of the infinitival event and the time of the matrix event. In rry-type infinitives, on the other hand, the matrix predicate selects a tenseless complement (i.e., a complement that does not involve any temporal ordering relation). In other words, the infinitive involves neither a before nor an after specification that would order the time of the infinitival event with respect to another time. Assuming that temporal relations are represented structurally, the lack of a temporal ordering relation in the infinitive means that the time of the infinitival event is neither before nor after the time of

Tense properties

75

the higher event. The only possible interpretation is thus a simultaneous interpretation. The notion of a simultaneous interpretation has to be seen in light of the intensional interpretation of the infinitive. An important factor of the meaning of irrealis infinitives such as (59) is that try or decide describe intensional contexts—i.e., what is attempted or decided is evaluated in a possible world rather than in the real world. To determine the meaning of these examples, it is not sufficient to ask whether it is true that John switched on or will switch on the light in the real world (i.e., the world in which the utterance takes place). Rather, we have to look at the relevant possible worlds or situations—namely those where what John tries/decides actually succeeds or happens. Similarly, the irrealis nature of the infinitive has to be seen relative to the world of trying/deciding and not the actual world. The sentences in (59) do not (necessarily) convey that the embedded events are uncompleted at the time of the matrix event. Rather, they express that according to John's beliefs, the situation that he tried/decided to achieve is uncompleted at the time of his attempt/decision. In other words, at the time of trying/deciding, John has to believe that the light is off—whether the light is indeed off or not in the real world does not affect the truth conditions of the sentence. The claim that in iry-infinitives, the time of the matrix event and time of the embedded event are simultaneous is thus misleading in that the embedded event is not (necessarily) an event of the real world. However, taking into consideration the intensional character of the construction, the claim of simultaneity still holds. What fryexamples such as (59)a express is that some aspect of what John thinks will bring about a situation in which the light is on has to coincide temporally with John's attempt. That is, at the time of John's attempt he has to perform some action that according to his beliefs will make the light go on. Thus, the simultaneous character of trycontexts is not a simultaneity between two actual events, but rather between the (actual) event of trying and whatever the subject thinks will bring about the situation described in the infinitive. In a decidecontext, on the other hand, the embedded situation is entirely situated in a future world.

76

(59)

Lexical restructuring

English irrealis infinitives a. b.

John tried to switch on the light John decided to switch on the light

Although we will refer to the tense interpretation in fry-type infinitives as a simultaneous interpretation, this qualification should be kept in mind. To come back to the main question we are addressing here—the relation between tense and aspect—the interpretation we get for examples such as (59)a thus shows that it is indeed necessary to distinguish between tense and aspect. While ¿/^-infinitives involve simultaneity as described above (which we argue is due to the lack of tense), they are also interpreted as irrealis, indicating that irrealis cannot simply be reduced to future tense. The second tense-related property discussed above is the impossibility of past modifiers in German (the relevant examples are repeated here as (60)). (60)

Past modifiers a. *Hans John 'John b. *Hans John 'John

hatte beschlossen gestern zu verreisen had decided yesterday to-go-on-a-trip had decided to go on a trip yesterday' hat(te) versucht gestern zu verreisen has/had tried yesterday to-go-on-a-trip (has/had) tried to go on a trip yesterday'

From a meaning point of view, nothing seems to block an adverbial such as yesterday in the embedded complement in (60)a, since—as we have seen in the previous section—a past future interpretation is in principle possible in decide-'mïimûvts in German. Furthermore, the fact that this example is possible in English also confirms that the (temporal) structure is perfectly interpretable and hence the ungrammaticality of (60)a cannot be the result of a tense clash or some inherent semantic incompatibility. As mentioned before, we do not have an elaborate account for this problem in German, but simply stipulate that past adverbials are incompatible with irrealis aspect in German. Since try and decide-infinitives involve irrealis aspect, it

Tense properties

77

follows from this restriction on temporal modifiers that past adverbials are prohibited in both examples in (60).32 Although this analysis is based on the unmotivated stipulation as stated above, there is reason to assume that it is on the right track. Note first that an attempt to attribute the ungrammaticality encountered in (60) to the requirement that temporal modifiers be licensed by a matching tense (i.e., past modifiers require a past tense, future modifiers require a future tense) would fail for (60)b. Assuming that ^-infinitives lack future tense (which we have seen is necessary to account for the impossibility of future adverbials), the tense responsible for the temporal location of the embedded event in (60)b is the matrix tense. However, since the matrix tense is past tense, the past adverbial should then be possible (note again that this example is ungrammatical even when the adverb is understood to modify both the matrix and the embedded event). Furthermore, the assumption we make to account for the ungrammaticality of (60)—namely that past modifiers are incompatible with irrealis aspect—says nothing about other modifiers. In particular, it allows in principle for non-past modifiers in irrealis contexts. As we have seen in (58), repeated as (61)a,b, future oriented modifiers are possible in decide-contexts, but impossible in iry-contexts. However, as we have suggested, the reason for the ungrammaticality of (61)b is due to a tense clash between the past tense of the matrix clause and the indexical future adverbial (essentially like in a simple sentence such as John ate the cake tomorrow). The (correct) prediction our analysis makes is that if this tense clash is avoided, the structure becomes grammatical, which is illustrated in (61)c. The contrast between (61)b and (61)c is important since it shows that future modifiers are not inherently banned from fry-infinitives, and moreover, it appears to be strong evidence for the claim that the tense of these types of infinitives is determined by the matrix tense. 32

Our analysis would predict that examples such as (60)b are grammatical in English. Sentences like John tried to marry the woman yesterday that he had dreamt about all his life could be taken to show that this predication is correct. However, it is not clear to us if it can be proven that examples of this sort necessarily involve a structure in which the adverbial is in the embedded clause.

78

(61)

Lexical

restructuring

Future modifiers a.

Hans John 'John b. Hans John 'John c. IHans John 'John

hat beschlossen (morgen) zu verreisen has decided (tomorrow) to go-on-a-trip decided to go on a trip (tomorrow)' hat versucht (*morgen) zu verreisen has tried (tomorrow) to go-on-a-trip tried to go on a trip (*tomorrow)' wird versuchen morgen zu verreisen will try tomorrow to-go-on-a-trip will try to go on a trip tomorrow'

Note that the example in (61)c also shows that the presence of a temporal adverbial such as tomorrow cannot be taken as evidence for a(n effective) tense projection. The only interpretation that (61)c can receive is again a simultaneous reading—i.e., both the time of the attempt and the time of the embedded situation have to be in tomorrow. Thus, although the adverb is attached to a very low position inside the infinitive in this example, it nevertheless has to take scope over the matrix tense. Contrary to approaches as suggested for instance by Cinque (1999), examples of this sort seem to suggest that adverbs do not necessarily have to occur in fixed positions in German. The slight markedness of examples like (61)c (most speakers prefer the adverb in the higher predicate) can be seen as support for the claim that the adverb does not appear in its 'designated' position in this example, however, the (universal) ordering suggested by these approaches has to be seen as a relative rather than absolute notion. To conclude, what we have seen in this section is that the notion of irrealis has to be kept separate from the notion of infinitival tense. Assuming that irrealis is a property of aspect rather than tense allows us to account for the distribution of adverbial temporal modification in infinitival constructions, as well as the tense properties of different types of infinitival constructions. In the next section, we will return to the issue of restructuring and we will see that restructuring infinitives are tenseless infinitives.

Tense properties

79

3.3. Restructuring infinitives are tenseless Regarding the tense properties of infinitival constructions, it has been observed by a number of researchers that infinitival constructions do not display a uniform tense behavior. While certain infinitives involve an independent tense orientation, other infinitives appear to lack any kind of tense contribution. Karttunen (1971) and Pesetsky (1992) for instance observe that implicative constructions (i.e., constructions with verbs like manage or dare) display a tense conflict similar to the one we mentioned above for fry-constructions (e.g., an utterance such as John managed to win the game tomorrow is infelicitous). Furthermore, many authors claim that tense properties such as the ones discussed are also reflected in the syntactic structure. In this sense, Guasti (1993) argues against a tense projection in causative constructions; Rizzi (1993) and Haegeman (1995) make a similar claim for the easy-to-please construction; and Boivin (1998) and Felser (1998) provide evidence for the lack of a tense projection in (certain) perception verb constructions. What we would like to add here to the discussion of the interaction of the syntax and semantics of tense is that the constructions mentioned by these authors have one common property—namely that they are restructuring infinitives (but see section 3.5). Looking at the list of lexical restructuring verbs listed in Table 9, we find that a number of constructions show the same tense properties as the /^-constructions discussed in the previous section. That is, the examples in (62) which all involve restructuring verbs in German (i.e., they allow long object movement) are inherently incompatible with a tensed infinitive (see also Cremers 1983 for Dutch). (62)

Tenseless infinitives a.

b.

Hans John 'John Hans John 'John

begann (*morgen) einen Brief began (*tomorrow) a letter began to write a letter (*tomorrow)' wagte (* morgen) einen Brief dared (*tomorrow) a letter dared to write a letter (*tomorrow)'

zu schreiben to write zu schreiben to write

80

Lexical restructuring

c.

d.

e.

f.

g.

Hans gelang es (*morgen) einen Brief zu schreiben John managed it (*tomorrow) a letter to write 'John managed to write a letter (tomorrow)' Hans mißlang es (*morgen) einen Brief zu schreiben John failed it (*tomorrow) a letter to write 'John failed to write a letter (*tomorrow)' Hans vergaß (*morgen) einen Brief zu schreiben33 John forgot (*tomorrow) a letter to write 'John forgot to write a letter (*tomorrow)' Hans versäumte es (*morgen) einen Brief zu schreiben John neglected it (*tom.) a letter to write 'John neglected/failed to write a letter (*tomorrow)' Der Kuchen war leicht (*morgen) zu backen The cake was easy (*tomorrow) to bake 'The cake was easy to bake (*tomorrow)'

The claim we would like to make is that this distribution is not accidental but attributable to a property of restructuring—the requirement that restructuring infinitives be tenseless. There are, however, some constructions that seem to contradict this claim. The examples in (63) show that some verbs from the class of restructuring verbs can combine with a tensed infinitive as evidenced by the possibility of future adverbials. (63)

Tensed infinitives a.

b.

Hans erlaubte dem Kind John allowed the child-DAT ('Jmorgen) einen Kuchen zu essen (?tomorrow) a cake to eat 'John allowed the child to eat a cake (tomorrow)' Hans verbot dem Kind John forbade the child-DAT

For some speakers, this example is grammatical under a factive interpretation of forget (i.e., under a planned event interpretation such as He forgot that he will write a letter tomorrow, see Chapter 5, section 4.1). Since factive infinitives are nonrestructuring infinitives, this interpretation is irrelevant for the present discussion.

Tense properties

c.

d.

81

(?morgen) einen Kuchen zu essen (?tomorrow) a cake to eat 'John forbade the child to eat a cake (tomorrow)' Hans empfahl dem Kind John recommended the child-DAT (Imorgen) einen Kuchen zu essen (?tomorrow) a cake to eat 'John recommended to the child to eat a cake (tomorrow)' Hans beabsichtigte (morgen) einen Brief zu schreiben John intended (torn.) a letter to write 'John intended to write a letter (tomorrow)'

Despite the occurrence of tensed infinitives with (apparent) restructuring verbs as in (63), we will show that there is a (one-way) correlation between restructuring and tense. In particular, we will see that the possibility of future tense in an infinitive is restricted to the nonrestructuring versions of these constructions and that restructuring properties require tenseless infinitives. Let us start with the examples in (63). Recall from the discussion of embedded structural case that lexical restructuring verbs are ambiguous in that they can combine with either a restructuring (i.e., case-less) or a non-restructuring complement (i.e., an infinitive with a structural case position/assigner). What we have shown is that this claim is not simply made out of convenience but motivated by a clear correlation between the case properties of the two constructions and the possibility of restructuring. What we will argue for now is that the same situation is found with respect to tense. That is, upon further inspection, we will find that the examples in (63) which involve an independent future tense in the infinitive (as evidenced by the possibility of a future adverbial) clearly qualify as non-restructuring infinitives. Hence, we will conclude that there is also a clear correlation between restructuring and tense. The argument is based on the examples in (64) (to keep the amount of data as low as possible, the examples below only involve the verb allow; a list of examples making the same point for the other constructions in (63) is provided in the appendix, section 4.1). To summarize (for details about each of the examples see below), what

82

Lexical restructuring

the data in (64) show is that long passive (hence restructuring) is only possible in the absence of future adverbials, and future adverbials are only possible in the absence of long passive. (64)

Long passive and tense a.

b.

c.

d.

Dem Kind wurden nur Kekse the child-DAT were only cookies (*morgeh) zu essen erlaubt (•tomorrow) to eat allowed 'The child was only allowed to eat cookies tomorrow' [[toBj {*Morgen) zu essen ] erlaubt ]VP wurden [[toBj {Tomorrow} to eat ] allowed ]VP were dem Kind {*morgen} nur die Kekse the child-DAT {*tomorrow} only the cookies Only cookies were such that the child was allowed to eat them tomorrow' Dem Kind wurden (,gestern) the child-DAT were (yesterday) nur Kekse zu essen erlaubt only cookies to eat allowed 'The child was only allowed to eat cookies (yesterday)' Dem Kind wurde erlaubt the child-DAT was allowed (Imorgen) einen Kuchen/Kekse zu essen (?tomorrow) a cake-ACC/cookies to eat 'The child was allowed to eat a cake/cookies (tomorrow)'

The examples in (64)a and (64)b show that in contrast to (63 )a, future adverbials are impossible in the context of long passive—i.e., a context that requires a restructuring infinitive. The examples also show that long passive is possible with the verb allow (though some speakers are a bit uneasy with this construction) when the future adverbial is omitted. Furthermore, (64)c illustrates that the matrix predicate can involve a temporal adverbial such as yesterday. In light of this example it becomes evident again that it is crucial to distinguish between tense and aspect in infinitives. While (64)c involves

Tense properties

83

an apparent after relation of the infinitive, the impossibility of future adverbials in (64)a or (64)b shows that this effect cannot be attributed to future tense in the infinitive. Rather, we would argue that it is due to the irrealis nature of these infinitives. Thus, the interpretation we get in (64)c is in fact a simultaneous interpretation (as described in the previous section) and the apparent after orientation comes from the irrealis aspect of this construction. Finally, the impersonal passive example in (64)d shows that an embedded future adverbial is licensed under matrix passive when no long object movement takes place. Since in this example, the embedded object shows up with accusative and does not agree with the matrix auxiliary, the infinitive involves its own structural case position and is thus a nonrestructuring infinitive. As expected under our analysis, future adverbials (hence infinitival tense) are possible in this context. The examples in (64) thus strongly support the claim that restructuring infinitives are tenseless. If a restructuring infinitive were to involve infinitival (future) tense, the contrast between (64)a,b and (64)d would not be trivial to explain. The observation in (64) illustrates an important theoretical and methodological point which we would like to stress again since it is often misunderstood. We do not claim that infinitives combining with the verbs in Table 9 are necessarily tenseless. Rather, we claim that whenever an infinitive qualifies as a restructuring infinitive, it has to be tenseless. Since the infinitival constructions we are considering in this chapter allow two structures (restructuring vs. non-restructuring complements) it is of particular importance to distinguish between the two constructions. However, as soon as the examples are construed unambiguously, the correlations are strikingly clear. The second piece of evidence for our claim that restructuring infinitives are tenseless comes from a special interpretation available (at least for some speakers) in iry-contexts. An interesting property that complicates the facts and is also responsible to an extent for some of the speaker variation found with respect to the restructuring/non-restructuring distinction is that iry-infinitives also allow a marked reading. In examples like (65), an interpretation like The thieves tried to make arrangements so that they would be able to smuggle the car across the border the next day is marginally avail-

84

Lexical

restructuring

able as well. Assuming the marked reading is indeed available (at least marginally) in fry-constructions, the presence of future adverbials in (65) indicates that the infinitives involve future tense. (65)

Marked tense reading a.

b.

dass die Diebe versuchten den Wagen that the thieves tried the car-ACC (%morgen) über die Grenze zu schmuggeln (%tomorrow) across the border to smuggle 'that the thieves tried to smuggle the car across the border (%tomorrow)' dass versucht wurde den Wagen that tried was the car-ACC (%morgen) über die Grenze zu schmuggeln (%tomorrow) across the border to smuggle 'that they tried to smuggle the car across the border (%tomorrow)'

The crucial question then is whether these infinitives also still qualify as restructuring infinitives. The answer is clearly negative. If we try to construct the examples as restructuring infinitives involving long passive the result is unacceptable. As is illustrated in (66), long passive is impossible in examples such as (65)b, confirming again that infinitives involving tense are non-restructuring infinitives. (66)

Marked tense reading and restructuring a.

b.

dass der Wagen (*morgen) that the car-NOM (*tomorrow) über die Grenze zu schmuggeln versucht wurde across the border to smuggle tried was 'that they tried to smuggle the car across the border (%tomorrow)' dass die Autos (*morgen) that the cars (*tomorrow) über die Grenze zu schmuggeln versucht wurden across the border to smuggle tried were 'that they tried to smuggle the cars .. .(%tomorrow)'

Tense properties

85

To conclude, what we have shown in this section is that there is a strong correlation between restructuring and tense in infinitives: restructuring is only possible with tenseless infinitives. As soon as the infinitive involves its own tense specification (i.e., tense features or operator, or before/after relation), the structure blocks restructuring properties and can only be a non-restructuring infinitive.

3.4. Overt tense markers in Japanese infinitives Comparing German or English infinitival constructions with the corresponding constructions in Japanese, we find that Japanese 'infinitives' come in a variety of forms. In this section, we will show that among the various constructions two classes can be distinguished that are of particular interest for the issue of tense in infinitives: infinitives without an overt tense marker on the embedded verb and 'infinitives' with an overt tense marker on the embedded verb. The crucial observation will be that the former—i.e., the class of tenseless infinitives—correspond by and large to the class of restructuring verbs. A terminological question that arises immediately but will be set aside here is whether constructions that involve embedded tense marking are to be considered as infinitives. If infinitives are defined as lacking tense morphology (note that—as the discussion in the previous sections has shown—infinitives cannot simply be described as lacking tense), constructions involving embedded tense markers would obviously not be considered as infinitives. However, it is equally plausible to define an infinitive as a construction lacking an embedded overt subject. Under this conception, embedded constructions which involve tense marking but lack an overt subject would then be considered as infinitives. We will not take any position in this labeling quarrel here but will rather concentrate on the properties of the constructions under consideration. To avoid confusion, however, we will use the term quasi-infinitive to refer to constructions that correspond to infinitives in English or German, but that show overt tense marking in Japanese. Let us begin with quasi-infinitives that appear with an overt tense-

86

Lexical

restructuring

marker. A common way of expressing a quasi-infinitive in Japanese is by embedding it in a complex noun phrase. What is important for the discussion at hand is that the embedded verb in constructions under koto 'fact' obligatorily shows up with a tense marker (cf. (67)). (67)

Japanese: quasi-infinitives with koto a.

b.

c.

d.

Emi-ga ringo-o tabe-ru koto-ni kimeta Emi-NOM apple-ACC eat-PRES fact-DAT decided 'Emi decided to eat an apple' Emi-ga ringo-o tabe-ru koto-o Emi-NOM apple-ACC eat-PRES fact-ACC yakusoku sita promise did 'Emi promised to eat an apple' Emi-ga Susi-ni ringo-o tabe-ru Emi-NOM Susi-DAT apple-ACC eat-PRES koto-o settoku sita fact-ACC persuade did 'Emi persuaded Susi to eat an apple' Emi-ga ringo-o tabe-ta koto-o kuyanda Emi-NOM apple-ACC eat-PAST fact-ACC regretted 'Emi regretted that she had eaten an apple'

Furthermore, we find quasi-infinitives with overt complementizers (cf. (68)). Here again, the embedded verb shows up with a present or past tense marker (depending on whether the quasi-infinitive allows a past orientation or not). As illustrated in (68)d, tense marking is obligatory. (68)

Japanese: quasi-infinitives with complementizers a.

b.

Emi-ga ringo-o tabe-ru to yakusoku sita Emi-NOM apple-ACC eat-PRES COMP promise did 'Emi promised to eat an apple' Emi-ga ringo-o tabe-ru to happyou sita Emi-NOM apple-ACC eat-PRES COMPannouncem't did 'Emi announced that she would eat an apple'

Tense properties

87

c.

Emi-ga ringo-o tabe-ta to happyou sita Emi-NOM apple-ACC eat-PAST COMP announcem't did 'Emi announced that she had eaten an apple' d. *Emi-ga ringo-o tabe to happyou sita Emi-NOM apple-ACC eat COMP announcement did 'Emi announced that she had eaten an apple' Finally, there are quasi-infinitive constructions involving the infinitival element yoo(ni).34 Ignoring the exact nature of these constructions, the property we want to point out here again is that the embedded verb is marked for tense. (69)

Japanese: quasi-infinitives with yoo a.

b.

c.

Emi-ga Susi-ni ringo-o tabe-ru yoo(ni) itta Emi-NOM Susi-DAT apple-ACC eat-PRES YOO told 'Emi told Susi to eat an apple' Emi-ga Susi-ni ringo-o tabe-ru yoo(ni) Emi-NOM Susi-DAT apple-ACC eat-PRES YOO kyoosei sita force did 'Emi forced Susi to eat an apple' Emi-ga Susi-ni ringo-o tabe-ru yoo(ni) Emi-NOM Susi-DAT apple-ACC eat-PRES YOO settoku sita persuade did 'Emi persuaded Susi to eat an apple'

Furthermore, in many cases, quasi-infinitive taking verbs in Japanese can take more than one kind of complement (e.g., the verb promise can combine with a koto quasi-infinitive as well as a quasi-infinitive According to the literature and the native speakers consulted, the status of this element is somewhat mysterious (cf. Nemoto 1993: 44, fn. 43). There seems to be a consensus, however, that it is some kind of infinitival element rather than a complementizer, since in the examples in (69), an additional complementizer can occur in certain dialects (cf. Nemoto 1993: 194). One speaker described yoo as a volitional marker. In light of our analysis in section 3.2 (see also section 3.5), an option that comes to mind is that yoo is in fact an irrealis marker.

88

Lexical

restructuring

with an overt complementizer; the verb persuade and most verbs with an additional DAT argument can combine with a koto quasiinfinitive as well as a yoo(ni) quasi-infinitive). We will set aside the exact distributional aspects of quasi-infinitives with overt tense markers as well as the question of whether the constructions in (67) through (69) are finite complements or infinitives. What is important for the discussion here is that constructions corresponding to nonrestructuring infinitives in Germanic or Romance display one common property in Japanese: the embedded verb is marked for tense, and in none of the examples can the tense marker be left out. Under the approach taken here, this is not surprising since our central claim is that infinitives involve tense, however, only if the infinitive is a non-restructuring infinitive. Let us now turn to the class of restructuring verbs in Japanese. As in German, there are various ways restructuring infinitives can be represented (see for instance the distinction between functional and lexical restructuring). Although the constructions we will discuss below involve different structural properties, the important property that all these restructuring constructions share is that the embedded verb cannot involve tense marking. Thus, in contrast to the quasiinfinitives discussed above, constructions that qualify as restructuring infinitives prohibit tense marking. To illustrate the lack of tense marking in restructuring infinitives, we will summarize various constructions that display restructuring properties. The first construction we want to mention here is the motion verb construction (cf. Miyagawa 1987; see section 2.3.2 for an illustration of the restructuring properties). As shown in (70), verbs combining with motion verbs (ignoring the prepositional ni) do not involve tense marking. (70)

Japanese: motion verbs a.

b.

Emi-ga ringo-o Emi-NOM apple-ACC 'Emi went to eat apples' Emi-ga ringo-o Emi-NOM apple-ACC 'Emi came to eat apples'

tabe-ni eat-to

itta went

tabe-ni eat-to

kita came

Tense properties

89

The addition of a tense marker as in (71) results in ungrammaticality. (71)

Japanese: motion verbs and tense markers a. *Emi-ga ringo-o Emi-NOM apple-ACC 'Emi went to eat apples' b. *Emi-ga ringo-o Emi-NOM apple-ACC 'Emi went to eat apples' c. *Emi-ga ringo-o Emi-NOM apple-ACC 'Emi came to eat apples' d. *Emi-ga ringo-o Emi-NOM apple-ACC 'Emi came to eat apples'

tabe-ta-ni itta eat-PAST-to went tabe-ru-ni itta eat-PRES-to went kita tabe-ta-ni eat- PAST-to came tabe-ru-ni kita eat- PRES-tocame

Secondly, it has been pointed out by many authors that certain verbs in Japanese are affixes in the sense that they appear attached to the lower verb and that no phrasal material may intervene between the two verbs (cf. Shibatani 1978; Nishigauchi 1993; Kageyama 1993; Koizumi 1995). A (partial) list of these verbs is given in (72) (see also the appendix, section 4). Affixal predicates also include easyadjectives as in (73). Importantly again, the embedded verb cannot involve any tense marker in all of these constructions. (72)

Japanese: affixal verbs Emi-NOM apple-ACC eat-VERB-PAST a. b. c. d. e. f. g· h. i.

Emi-ga Emi-ga Emi-ga Emi-ga Emi-ga Emi-ga Emi-ga Emi-ga Emi-ga

ringo-o ringo-o ringo-o ringo-o ringo-o ringo-o ringo-o ringo-o ringo-o

tabe-rare-ta 'can' tabe-taka-tta 'want' tabe-hazime-ta 'begin' tabe-dasi-ta 'start' tabe-tuzuke-ta 'continue' tabe-oe-ta 'stop' tabe-age-ta 'finish, complete' tabe-sugi-ta 'overdo' 'be about' tabe-kake-ta

90

Lexical restructuring

j· k. 1. m. n.

Emi-ga Emi-ga Emi-ga Emi-ga Emi-ga

ringo-o ringo-o ringo-o ringo-o ringo-o

tabe-owe-ta tabe-naos-ta tabe-wasure-ta tabe-sokone-ta tabe-toge-ta

'come to an end' 're-do' 'forget' 'fail' 'manage' 35

Emi-NOM apple-ACC eat-VERB-PAST (73)

Japanese: easy-adjectives a.

b.

Ringo-ga tabe-yasuka-tta apple-NOM eat-easy-PAST 'The apple is easy to eat' Kono hon-ga yomi-niku-i this book-NOM read-hard-PRES 'This book is hard to read'

Finally, we want to mention constructions involving the modal verb must. As is shown in (74), a must-statement requires a more complex structure in Japanese. However, this construction can nevertheless be taken to confirm the generalization proposed here. Like affixal verbs or motion verbs, the embedded verbs in (74) lack tense markers. This is expected under our analysis since modal verbs are the prototypical case of restructuring infinitives. (74)

Japanese: must a.

b.

Emi-ga ringo-o Emi-NOM apple-ACC 'Emi must eat an apple' Emi-ga ringo-o Emi-NOM apple-ACC 'Emi must eat an apple'

tabe-nakere-ba naranai eat-if not-COND must-do tabe-nakute-wa eat-if not-TOP

naranai must-do

To sum up, the crucial property of the constructions in (70) through (74)—i.e., restructuring infinitives—is that in contrast to the quasi35

To express the meaning of manage, an additional adverb nantoka 'barely, with an effort' is usually added.

Tense properties

91

infinitives in (67)-(69), tense markers on the embedded verb are prohibited and that none of the elements found in quasi-infinitives (i.e., koto, yoo(ni), or complementizers) can appear in these constructions. Under the assumption that in Japanese, tense is marked morphologically in finite as well as non-finite constructions, the lack of tense markers in these constructions can be taken as additional empirical support for our modification of Stowell's claim—infinitives can involve tense, however, restructuring infinitives are tenseless.

3.5. The restructuring-tense connection To conclude the discussion of tense, we will summarize the observations made in this section and discuss the role tense plays in the determination of restructuring. Let us start with a comparison of the verbs that combine with tenseless complements in Japanese with restructuring predicates in other languages. As illustrated in Table 10, all tenseless constructions in Japanese correspond to restructuring constructions in Germanic and/or Romance (see also the appendix for a more detailed list of predicates). Note that this correlation is not to be understood as a strong correlation. That is, it does not say that all Japanese restructuring verbs are restructuring verbs in German or that all German restructuring verbs are restructuring verbs in Japanese. It merely illustrates that all restructuring verbs in Japanese are restructuring verbs in some other language. As mentioned before, a comparison of restructuring predicates across languages shows that the classes of restructuring verbs while quite similar also display significant differences (see also the appendix, section 4). Table 10.

Verbs taking tenseless complements in Japanese

Verbs

Japanese

Romance and/or Germanic

want, can, must come, go begin, continue, finish be about to, re-do easy-adjectives manage, fail, forget

+ + + + + +

+ + + + + +

92

Lexical restructuring

The generalization we find in Table 10, however, does have some significance that we believe should not be overlooked. Recall that Japanese restructuring predicates are instantiations of predicates that are incompatible with a tensed complement. Since all of these predicates are found among the class of restructuring predicates in some language (although not necessarily all in one language), it seems safe to say that tenselessness plays a crucial role in the context of restructuring. Furthermore, in light of the German facts discussed in section 3.3, this role can be made precise as the following one-way implication between tense and restructuring: restructuring entails a tenseless complement. We have seen that it is not the case that all (potential) restructuring verbs are inherently (e.g., semantically) incompatible with a tensed infinitival complement in German. However, we have also shown that it is the case that restructuring is incompatible with a tensed infinitive. In the remainder of this section, we will discuss the distribution of restructuring in Japanese and German and provide further support for the correlation pointed out above. We will see that the tense properties are crucial for the actual make up of restructuring infinitives, however, they are also not sufficient to predict whether a certain predicate in principle allows a restructuring complement or not. To start with, consider Table 11 which lists various predicates, their restructuring status in German and Japanese, and the (inherent) compatibility of these predicates with a tensed infinitival complement (i.e., the possibility of a tense interpretation of the infinitive that is different from the tense interpretation of the matrix predicate).36,37 The criteria used to distinguish between restructuring and nonrestructuring are long object movement for German and the lack of 36

As for the notation, in the columns labeled "German/Japanese", '+' indicates that the predicate allows a restructuring complement (provided the construction meets all restructuring requirements); '-' indicates that restructuring is always impossible; and ' 0 ' indicates that infinitival constructions are not available with the predicate under consideration (i.e., the question of whether restructuring is possible or impossible cannot be asked). 37 Following Wurmbrand (1998b); Erb (2001), we assume that complements to modal and raising verbs do not involve true tense but that the apparent tense effect is caused by the inherent semantics and pragmatics of modal statements.

Tense properties

93

tense morphology in the infinitive for Japanese. Table 11. Tense and restructuring Verb

German

Japanese

Tense

modals raising causatives come, go begin continue, start, finish, stop easy know how forget try manage fail dare intend {-want, mean) recommend allow, permit forbid decide, choose, plan order, offer, propose prefer, wish propositional, factive verbs

+ + + + +

+

-

-

-

-

+ + + + + + +

?

+ + +

+ + + + ηΝ/Α +

±

+ + N/A N/A -

-

-

-

-

-

-

-

-

-

+ + + + + + + +

As the reader can verify, Japanese shows the following correlation between restructuring and tense: restructuring predicates only combine with infinitives that are incompatible with an independent tense interpretation. However, the correlation goes again only one way since not all verbs that block an independent tense interpretation are (necessarily) restructuring verbs. This point can be illustrated with constructions involving the verb(s) try. As is shown in (75), there are two ways to express a /^-statement. In (75)a, we find an infinitive that lacks morphological tense marking and includes the infinitival element yoo. Assuming that this element marks irrealis mood or aspect, the presence of y o o in a íry-context is not surprising, and moreover, can be taken as additional evidence for our claim that tense and (irrealis) aspect have to be distinguished. The example in

94

Lexical restructuring

(75 )a is thus exactly what we would expect given our analysis of tense and aspect in infinitives. However, as is shown in (75)b, there is another way of expressing a fry-relation—the same meaning can be conveyed by a morphologically tense marked infinitive embedded under koto. The example in (75)b hence patterns with the nonrestructuring constructions discussed before (i.e., the examples in (67) involving verbs like decide, promise etc.). (75)

a.

b.

John-ga ringo-o tabe-yoo-to John-NOM apple-ACC eat-yoo-COMP 'John tried to eat an apple' John-ga ringo-o taberu koto-o John-NOM apple-ACC eat-PRES fact-ACC 'John tried to eat an apple'

sita did kokoromi-ta try-PAST

Thus, Japanese try is compatible with both a restructuring infinitive and a non-restructuring infinitive, which is reminiscent of the general ambiguity of (lexical) restructuring predicates in German. Importantly, however, we have to conclude that while the tense properties play an important role for the determination of restructuring, they also do not seem to be sufficient to predict whether a predicate requires a restructuring or non-restructuring complement. Turning to the distribution of restructuring in German, a similar situation arises. As Table 11 shows, there is certainly a close connection between restructuring predicates and the incompatibility of tense. However, we also find that a priori, the distribution of restructuring and tense does not allow for any absolute correlation (in either direction). Nevertheless, the course we would like to take here is that the similarities are not accidental and that tense does play an important role in the determination of restructuring. Our crucial claim is that restructuring is not defined as a single property or feature of certain predicates, but is rather a particular configuration—a monoclausal structure. As such it will then not be surprising that restructuring has to meet various structural and—given the assumption that structure has a direct effect on interpretation—semantic criteria, among them the lack of tense in the infinitive. In other words, from the assumption that restructuring infinitives are simple VP-layers

Tense properties

95

lacking clausal functional projections (such as vP, TP, and CP), it follows that infinitives lacking tense are only potential restructuring contexts; i.e., they are restructuring infinitives only if all other requirements are met as well and nothing else blocks restructuring independently. Let us illustrate this claim with two cases where the generalization that tenseless infinitives are restructuring infinitives does not hold. One such case has already been discussed at length in section 2.3.1. Recall that iry-infinitives that involve an embedded structural case position/assigner are non-restructuring infinitives (a relevant example is repeated in (76)a; for arguments that this example can only be a non-restructuring infinitive see section 2.3.1). However, as is illustrated in (76)b, the tense properties of this example are no different from the tense properties of the restructuring version of a tryinfinitive. That is, the only way the non-restructuring infinitive in (76)b can be interpreted (in its unmarked form) is with a quasisimultaneous interpretation. (76)

No long object movement a.

Es wurde versucht den Traktor zu reparieren It was tried the tractor-ACC to repair 'They tried to repair the tractor' b. #Es wurde versucht It was tried morgen den Traktor zu reparieren torn. the tractor-ACC to repair #'They tried to repair the tractor tomorrow' The fact that the infinitives in (76) are tenseless hence does not entail that these infinitives are restructuring infinitives. Since the case properties of the infinitives in (76) (i.e., the presence of vP) do not meet the restructuring requirement (i.e., the monoclausal restructuring configuration), restructuring is nevertheless blocked. Similarly, these examples show that the fact that a construction is a nonrestructuring configuration does not entail that the infinitive involves tense. In our approach, non-restructuring only means that a property

96

Lexical

restructuring

necessary for restructuring (in case of the examples in (76), the lack of embedded structural case) is not met; it does not mean that none of the properties are met. In other words, given our analysis, there is no reason to assume that there is a pre-determined structure for nonrestructuring infinitives. Rather, we claim that non-restructuring infinitives can be either vPs, TPs, or CPs, however, the choice of structure has clearly defined consequences for both the syntactic and semantic properties of a construction. Thus, given a minimalist approach to phrase structure, examples such as the ones in (76) would be represented as vPs, rather than as TPs/CPs. The second case we would like to discuss involves the aspectual verbs continue, start, stop (see Table 11). As is illustrated in (77), these aspectual verbs can only combine with a tenseless complement. (77)

Tenseless infinitives a.

b.

c.

Hans John 'John Hans John 'John Hans John 'John

fing an (*morgen) den Brief started PART (*tomorrow) the letter started to read the letter (*tomorrow)' fuhr fort (*morgen) den Brief continued PART (*tomorrow) the letter continued to read the letter (*tomorrow)' hörte auf (*morgen) den Brief stopped PART (*tomorrow) the letter stopped reading the letter (*tomorrow)'

zu lesen toread zu lesen toread zu lesen toread

However, as shown in (78), these aspectual constructions do not allow long passive and are hence to be considered as non-restructuring verbs in German. (78)

Long passive a. *als der Brief zu verlesen angefangen when the letter-NOM to read PART-started 'when they started to read the letter' b. *als der Brief zu verlesen fortgesetzt when the letter-NOM to read PART-contin. 'when they continued to read the letter'

wurde was wurde was

Tense properties

97

c. *als der Brief zu verlesen aufgehört wurde when the letter-NOM to read PART-stopped was 'when they stopped reading the letter' The verbs in the examples in (78) contrast with the aspectual verb beginnen 'begin' which readily allows long passive (cf. (79)). (79)

Long passive als der Brief zu verlesen begonnen when the letter-NOM to read begun 'when they began to read the letter'

wurde was

Interestingly, the same contrast has been noted for Dutch. A phenomenon found in certain Dutch infinitival constructions (which roughly correspond to restructuring constructions in German) is the so-called verb raising effect (cf. Evers 1975b). Assuming a headfinal organization of the verb phrase in Dutch, verb raising structures are configurations in which the infinitival verb appears to the right of the matrix verb (this is an oversimplification but sufficient for the discussion here). Importantly, verb raising is only possible when the matrix verb is not a particle-verb. As is shown in (80), a construction with the verb beginnen 'begin' allows verb raising, whereas a construction with the verb aanfangen 'begin, start' blocks verb raising. (80)

Dutch: begin vs. start a.

dat that 'that b. *dat that 'that

(van Riemsdijk 1998b: 642)

hij Nederlands begint te leren he Dutch begins to learn he begins to learn Dutch' hij Nederlands aanvangt te leren he Dutch PART-starts to learn he starts learning Dutch'

As has been pointed out for Dutch by a number of authors (cf. Evers 1975b; Hoeksema 1988; Rutten 1991; Neeleman 1994; van Riemsdijk 1998b to name just a few), the difference between examples such as (80)a and (80)b lies in the presence of a particle in the

98

Lexical restructuring

latter. That is, particle verbs (i.e., verbs that require stranding of the preverbal element in verb second contexts) generally block verb raising in Dutch. There are various ways to account for this blocking effect, depending crucially on what theory of verb raising and what assumptions about locality, adjacency etc. one assumes. We will not go into detail about this restriction here, but simply note that there is a (fatal) competition between restructuring complements and particles for the complement position of the verb—i.e., only one of them can survive. Thus, if a verb requires a particle, there is no space for a restructuring complement. The only way a verbal complement can be realized is as a non-restructuring infinitive. Returning to the question at stake (i.e., the relation between restructuring and tense), we find that all the aspectual verbs in the examples in (78) involve a particle which cannot be omitted. Under the assumption that particles also block restructuring in German, we have an explanation for the fact that these predicates do not allow passive. Thus, whatever the motivation for the competition between particles and restructuring infinitives is, it shows that some independent factor is at work in (78) which blocks restructuring configurations. In other words, although the tense properties of these aspectual verbs would predict that these verbs are restructuring verbs (as is the case in Romance and Japanese; see the appendix, section 4), the incompatibility of particle verbs with restructuring complements nevertheless excludes restructuring. To sum up, what we can conclude from the discussion in this section is that restructuring is not a simple feature assigned to certain predicates, but rather the result of a number of syntactic and semantic properties that have to play together in a certain way to license restructuring. So far, we have discussed two crucial properties—the lack of tense and the lack of an embedded structural case assigner/position. We have shown that both properties have to be met for an infinitive to qualify as a lexical restructuring infinitive. In Chapter 4, we will see that there is another property, namely the lack of an embedded syntactic subject which is tied to the lack of a nonobligatory control interpretation. We will see that verbs like intend, allow, permit (i.e., verbs that are in principle compatible with an independent infinitival tense but that can also function as restructuring

Other functional projections

99

verbs, however, only when they lack tense) fulfill this property. The (inherent) lack of a non-obligatory control interpretation will then allow us to understand the remaining 'mismatches' in Table 11—i.e., the fact that some (but not all) predicates that are compatible with a tensed infinitival complement are among the class of restructuring predicates in German.

4. Other functional projections We have argued so far that lexical restructuring infinitives are simple VP-layers lacking a tense projection and a structural case position. In this section, we will discuss restructuring in light of other clausal functional projections. In section 4.1, we will first provide some general comments on the question of restructuring and functional projections. In section 4.2, we will show that restructuring infinitives cannot involve material associated with a C-projection. In section 4.3 we will address the question of the status and position of the infinitival marker zu 'to'. We will conclude that while restructuring infinitives have to be extended to include a projection hosting this element, this projection does not carry any tense information and hence the oft-mentioned connection between the infinitival marker and tense is not justified. And finally, in section 4.4, we will show that restructuring infinitives lack embedded negation. 4.1. The restructuring configuration and functional projections We would like to start this section with a general point concerning the labeling of functional projections that might help to clarify some confusion regarding the validity and importance of our analysis. As has been pointed out by a reviewer, our analysis (in particular the claim that restructuring infinitives lack a TP) might have less significance under a syntactic analysis of tense along the lines of Stowell (1996). The reviewer's concern is that if the phrase structure involves more than one tense projection (i.e., a "TP" for each aspect of the temporal structure such as an event time TP, a reference time TP, and

100

Lexical restructuring

a speech time TP) a restructuring infinitive could lack a certain TP but still include another TP. This concern seems to be part of the bigger question of how the analysis we presented here fares with respect to phrase structure accounts that postulate not only a CP, TP, and vP but various other functional categories. To answer this question, we simply have to separate two aspects of this question: what is the nature of a restructuring configuration and what is the syntactic label of a restructuring infinitive? Let us emphasize the crucial points of our analysis again. We have claimed that clause union or transparency properties are licensed by the restructuring configuration. We have chosen to refer to the restructuring configuration as "VP" since it appears to be the most neutral label (syntactically and semantically) for a category with the following properties (which we take to be the defining properties of a restructuring configuration):38 a lexical restructuring infinitive (i.e., a complement to a lexical restructuring verb) denotes (roughly) an event or an action (cf. Cremers 1983; Rochette 1988, 1990; see also Chapter 3, section 3.1); it lacks propositional and force properties such as an independent tense specification and complementizer material (see below); it lacks a structural case position/assigner; and, as we will show in Chapter 4, it does not include a syntactic subject. Thus, in our account, the restructuring configuration, which we take as the pre-requisite to license the full range of transparency properties, is a construct of clearly defined syntactic and semantic properties. The syntactic classification of a restructuring configuration as a "VP", however, is a label and can in principle be replaced by any other label. In other words, when we claim that restructuring infinitives lack "TP", we claim that restructuring infinitives lack the functional structure associated with tense (such as tense features, a before/after relation, or a tense operator). In light of recent explosions of phrase structure, one might refer to the restructuring configuration as "FPn". However, crucially, whatever functional projections one postulates in a restructuring infinitive, it seems that under any approach, the properties of the restructuring configuration as described here have to be accounted for. In particular, the properties 38

But see Chapter 5 for a discussion of different grades of (non-)restructuring.

Other functional projections

101

of a restructuring configuration (such as the lack of an independent tense contribution) are not claims that only hold under certain assumptions about clause structure, but they are simply empirical facts. As such, it seems desirable for an explanatory account to encode these facts in some form or another in the syntactic/semantic structure. We have chosen a model in which there is a close connection between phrase structure and meaning. Given this assumption and the idea of a "motivated" syntax, the lack of an independent tense contribution for instance entails the lack of a TP. If one adopts a model in which the projection of syntactic structure can be (fully or partially) deprived of a semantic correspondence (most notable in this respect are approaches that make use of Kayne's recent "WP"—i.e., word order phrases that have no syntactic and/or semantic contribution except to provide a landing site for movement; cf. Kayne 1998), or in which the projection of syntactic structure is templatic and not motivated by the actual properties of a configuration, our conclusion that tenseless infinitives are TP-less obviously does not follow. However, what is nevertheless valid and what we see as the core of our proposal is the fact that restructuring is subject to the restructuring configuration. Thus, however the phrase structure of a restructuring configuration is described, one has to make sure that a restructuring infinitive only involves projections that do not bear any of the functions or involve any of the type of syntactic and/or semantic features that are incompatible with the restructuring configuration.

4.2. CP-properties of restructuring infinitives In this section, we concentrate on the CP-properties of restructuring infinitives, in particular, the distribution of complementizers and whspecifiers. Since German allows neither w/i-infinitives nor infinitival complementizers in complement clauses,39 we will restrict the dis39

Infinitival complementizers (i.e., um 'in order to') can only appear in purpose clauses in German. However, since adjunct clauses are non-restructuring infinitives, these contexts are irrelevant for the discussion here. On some apparent exceptions to the claim that German does not allow wA-infinitives see section 4.2.2.

102

Lexical restructuring

cussion to languages that show overt material associated with the Cprojection of infinitival complements. 4.2.1. Overt complementizers The first argument against complementizers in restructuring infinitives comes from Dutch. Although Dutch does not allow long passive (cf. section 2.1) and hence lacks true restructuring according to our classification (but see Chapter 5 for extensive discussion of different grades of (non-)restructuring and terminological clarifications), Dutch does display two properties that attest to (some degree of) clause union. The first property is verb raising as in (80)a which is repeated here as (81)a. As mentioned before, in verb raising structures, the infinitive undergoes movement to the right of the matrix verb. An important property of verb raising is that this re-ordering process goes hand in hand with a change in the morphological properties of the matrix verb. As is illustrated in (81)b, in the perfective construction the matrix participle is changed to an infinitive (i.e., proberen 'try' instead of geprobeerd 'tried'). This phenomenon is called Infinitivus Pro Participio 'Infinitive for participle' (IPP) and found in a number of West Germanic languages. Returning to the main issue of this section, (81)b also shows that verb raising and the IPP effect are incompatible with the complementizer orti. While in principle overt complementizers are possible in Dutch infinitival constructions (cf. (81)c), complementizers are prohibited in verb raising and IPP constructions (cf. Hoekstra 1984; Bennis and Hoekstra 1989; Rutten 1991, Broekhuis et al. 1995 among others). (81)

Dutch: verb raising and IPP a.

b.

dat that 'that dat that

hij Nederlands begint he Dutch begins he begins to learn Dutch' Jan (*om) zijn broer Jan (*COMP) his brother

te leren to learn die brief the letter

tv tv

Other fiinctional projections

c.

heeft

proberen

te schrijven

has

try-IPP

to write v

103

'that John has tried to write the letter to his brother' dat Jan heeft geprobeerd / *proberen that Jan has tried / *try-INF (om) zijn broer die brief te schrijven (COMP) his brother the letter to write 'that John has tried to write the letter to his brother'

The second property we want to mention briefly is remnant extraposition (cf. den Besten and Rutten 1989; Kroch and Santorini 1991; Rutten 1991). Remnant extraposition (which is sometimes also referred to as the Third construction) refers to constructions such as (82), which involve leftward movement of some constituent (the embedded direct object) and rightward movement of a larger constituent (the infinitive) containing the trace of the scrambled phrase.40 (82)

Dutch: overt complementizers a.

b.

dat Jan die brief heeft geprobeerd that Jan [the letter]SCR has tried (*om) zijn broer tSCR te schrijven (*COMP) his brother tSCR to write 'that John has tried to write the letter to his brother' dat Jan (*om) zijn broer die brief that Jan (*COMP) his brother the letter heeft besloten / *besluiten te schrijven has decided / *decide-ipp to write 'that John has decided to write the letter to his brother'

Importantly again, remnant extraposition is incompatible with an infinitival complementizer. Thus, remnant extraposition as in (82)a differs crucially from examples such as (81)c involving extraposition of the whole infinitival complement without prior scrambling. Note also 40

This analysis was first proposed by Reuland (1981) and later adopted for Dutch by Bennis and Hoekstra (1989); den Besten and Rutten (1989); Rutten (1991); Broekhuis et al. (1995) among many others.

104

Lexical restructuring

that remnant extraposition and verb raising are not co-extensive (see Chapter 5, section 3.1 for a preliminary analysis). The IPP effect is restricted to a small class of predicates which by and large correspond to restructuring verbs in German. While verbs like try allow the IPP effect (cf. (81)b), constructions with verbs like decide prohibit this phenomenon (cf. (82)b). Nevertheless, remnant extraposition is possible in these contexts as the grammaticality of (82)b without the IPP effect illustrates. However, again, remnant extraposition is only licensed when the infinitive does not include a complementizer. Assuming that verb raising, the IPP effect, and remnant extraposition require some degree of clause union, the incompatibility of these operations with overt complementizers is expected. On the one hand, the presence of a complementizer requires a full clausal CPcomplement. On the other hand, as we will suggest in Chapter 5, section 3.1, verb raising and the IPP effect are restricted to vP (or smaller) infinitival complements, and remnant extraposition is restricted to TP (or smaller) complements. Thus, under this analysis it follows that verb raising, the IPP effect, and remnant extraposition are incompatible with complementizers. A similar blocking effect of complementizers can be found in Italian. As is illustrated in (83), the transparency properties object preposing and clitic climbing which are found only in restructuring constructions are blocked by an intervening overt complementizer. As argued by Kayne (1989, 1991) se like English if is a complementizer rather than a w/z-specifier. (83)

Italian: overt complementizers

(Rizzi 1982: 47)

a. * Certe risposte non si sanno mai se dare certain answers not SI knows ever if give They don't know whether to give certain answers' b. *Non lo so se fare not it I-know if do Ί don't know whether to do it' To conclude, these facts can be taken to support our analysis of re-

Other functional projections

105

structuring (as well as the account of reduced non-restructuring which we will provide in Chapter 5). Assuming that restructuring infinitives are VP-predicates (and reduced non-restructuring infinitives are vP or TP complements), it follows that these constructions cannot involve a complementizer which would require the projection of a CP. In the next section, we will see that restructuring infinitives also cannot involve CP-specifiers. 4.2.2. Wfc-specifiers It has been claimed by many authors (cf. Rizzi 1982, Kayne 1989, Roberts 1997) that in contrast to w/i-complementizers as discussed in the previous section, w/z-specifiers are marginally acceptable in restructuring infinitives. (84)

Italian: w/i-specifiers

(Kayne 1989: 16)

ΊΝοη ti saprei che dire not to-you I-would-know what say Ί wouldn't know what to say to you'

tCL tCL

Taking clitic climbing as an indication of restructuring, the whinfinitive in (84) would count as a restructuring infinitive, and hence provide evidence for the assumption that restructuring infinitives can involve a C-projection. The assumption that w/z-specifiers are possible in restructuring infinitives, however, has also been questioned by many authors. Rooryck (1994: 437, fn.6) for instance reports that there is a "tendency to exclude sentences with a matrix tense other than the conditional" in Italian (cf. (85)) and a "strong tendency to exclude these structures with embedded verbs other than decir 'say'" in Spanish (cf. (86)). All the examples are fully grammatical when clitic climbing does not occur (i.e., in non-restructuring infinitives). (85)

Italian: restrictions on w/i-specifiers

(Rooryck 1994: 437)

a. ΊΝοη ti saprei che not to-you I-would-know what Ί wouldn't know what to give to you'

regalare give

106

Lexical restructuring

b. *Non ti sapevo che not to-you I-knew what Ί didn't know what to give to you' (86) Spanish: restrictions on w/z-specifiers

regalare give

(Rooryck 1994: 437)

a.

No te sabía/sé qué decir not to-you I-knew/I-know what say Ί don't/didn't know what to tell you' b. *ΊΝο te sabía/sé qué regalar not to-you I-knew/I-know what give Ί don't/didn't know what to give to you' Furthermore, Rooryck shows that w/z-specifiers in Italian restructuring constructions are if at all only marginally possible in left dislocation contexts or when the wft-element is che 'what' ; in all other contexts, w/z-specifiers are prohibited when clitic climbing occurs (cf. (87)a,b which are again fine without clitic climbing). (87)

Italian: restrictions on ννΛ-specifiers

(Rooryck 1994: 421)

a. *Non ti saprei come dire (che...) not to-you I-would-know how say (that...) Ί wouldn't know how to tell you (that...)' b. *Non lo saprei a chi dire not it I-would-know whom say Ί wouldn't know who to tell it to' Rooryck's analysis is built on the assumption that clitic climbing is only possible if the verb+clitic complex raises first to the highest head position in the infinitival complement. In Rooryck's system, clitics have to be governed by the matrix tense in order to excorporate to the matrix clause. This condition is only met if the clitic is in the head position of the sister projection of the verb. To explain the contrast between cases like in (85)a vs. (87), Rooryck assumes that che 'what' cliticizes onto the embedded verb. The need of the whelement to check its features in the C-projection forces the complex head formed by the verb and all clitics to raise to C°. In contrast to

Other functional projections

107

the clitic che, full w/i-phrases move to SpecCP independently to check their w/z-features which takes away the motivation for the verb (+clitics) to raise to C°. Restructuring infinitives without a whelement, on the other hand are AgrSPs which involve overt raising of the infinitive to AgrS0 in Italian. In this position again the clitics attached to the verb are governed by the matrix T° and can excorporate. Although the contrast between clitic wh-words and full whphrases can be derived in Rooryck's system it does not seem to provide room for the contrasts in (85)a,b and (86)a,b. Interestingly, a similar contrast is found in German. In general, German does not allow wft-infinitives. However, there are certain instances of infinitival constructions that look like w/i-infinitives. As in Italian restructuring infinitives, the combination w/i-expression plus infinitive is only possible with the pronoun was 'what'; as in Spanish restructuring infinitives, there are strong limitations on the verb that appears in the infinitive (although there is some variation; for most speakers, examples like (88) are only natural with tun 'do' and sagen 'say'). (88)

German: w/z-infinitives a. lieh weiß I know Ί don't know b. V.Ich weiß I know Ί don't know c. *Ich weiß I know Ί don't know d. *Ich weiß I know Ί don't know

nicht was tun not what do what to do' nicht was essen not what eat what to eat' nicht wann/wo/warum not when/where/why when/where/why to do it' nicht wann essen not what eat when to eat'

(es) tun (it) do

We conclude from the highly restricted distribution of w/i-infinitives in Italian and Spanish restructuring infinitives and of German whinfinitives in general that these constructions are not productive wh-

108

Lexical

restructuring

movement constructions but rather fixed idiomatic expressions (see also Napoli 1981: 854f, fn. 18). This claim is supported by the observation that the w/i-infinitives under consideration are less natural the more specific the meaning of the verb involved is. Setting aside details of the structure of these idiomatic w/z-word plus infinitive constructions, the relevance for restructuring is that examples such as (84) (i.e., cases of wft-restructuring infinitives) are only apparent cases of w/i-movement constructions. To be more specific, we assume that the wh-word (which is part of the idiomatic expression) is in fact inside the verb phrase in Italian and Spanish restructuring infinitives and in German w/z-infinitives—i.e., it is inside the idiomatic domain of the clause (cf. Marantz 1995, 1997) rather than in sentence-initial operator position. In other words, no syntactic whmovement to SpecCP is involved in the grammatical examples in (84) through (88).41 Assuming that the idiomatic examples with apparent w/z-specifiers are not instances of true w/z-movement, we can conclude from the impossibility of restructuring in the clear cases of w/z-movement constructions that restructuring infinitives are indeed only well-formed when the infinitive does not contain a true whelement. To conclude this section, we have shown that restructuring is incompatible with infinitives that (clearly) involve CP-material. In our approach, this is not surprising since the basic claim is that restructuring requires a certain restructuring configuration—i.e., a VPcomplement. Since complementizers or w/z-specifiers require the presence of a CP, the restructuring configuration could never be met in the presence of CP-material. In bi-clausal approaches to restructuring, on the other hand, these restrictions do not necessarily fall out naturally. While the blocking effect of complementizers receives a straightforward account in head movement approaches (but see section 5.3 for some discussion), the blocking effect of w/z-specifiers is less straightforward in these types of approaches. In the topicalization plus head raising approach, on the other hand (see Chapter 1, section 2.2.2), the blocking effect of 41

This assumption seems to be supported by the fact that nothing is allowed to intervene between 'what' and the verb in Italian as well as in German.

Other functional projections

109

w/z-phrases receives a natural explanation, whereas the blocking effect of complementizers seems surprising. According to this approach, a w/z-phrase in SpecCP would prohibit topicalization of VP, TP or AgrOP (provided that only one phrase can appear in SpecCP), and hence a necessary prerequisite for restructuring could not be met. The prohibition against wA-specifiers thus follows directly from this approach. However, since in these types of approaches, head movement applies after the relevant phrase has undergone topicalization—i.e., head movement applies from the XP in SpecCP—head movement should not interfere with a complementizer in the infinitive. Thus, the prohibition against complementizers requires additional assumptions such as e.g., a version of the Doubly Filled Comp filter.

4.3. Infinitival marker A common assumption about the structure of infinitives in English is that the infinitival marker to occupies the tense head. The major motivations for this claim are co-occurrence restrictions between to and elements that are generally assumed to be in INFL/T (i.e., modals and auxiliaries). Since certain restructuring infinitives in German—namely constructions with lexical restructuring verbs such as try, begin as well as raising verbs—obligatorily project the infinitival marker zu 'to', the question arises of how the presence of the infinitival marker can be reconciled with our claim that restructuring infinitives do not involve a T-projection. While we will not contribute any new insight concerning the function and position of the infinitival marker, we will show in this section that the presence or absence of the infinitival marker does not correlate with any syntactic and/or semantic property of infinitival constructions. In particular, we will argue that the claim that the infinitival marker represents tense is untenable. We will conclude that i) the presence of an infinitival marker does not challenge the claim that restructuring infinitives lack tense; and ii) restructuring infinitives can include a projection that hosts the infinitival marker, however since this projection bears no semantic content and does not seem to fulfil any syntactic function, it

110

Lexical restructuring

is essentially 'invisible' for the computation of the restructuring configuration. Let us start with a short summary of the distribution of the infinitival marker which is by and large the same in German and English. Modal verbs, active perception verbs and causative verbs combine with bare infinitives (i.e., infinitival complements without the infinitival marker); control verbs and raising verbs require an infinitival marker. In both languages, there is a small class of verbs that have an intermediate (or dual) status: verbs such as brauchen 'need' and helfen 'help' can show up with or without an infinitival marker. The main difference between English and German is found in passive perception verb contexts which require the infinitival marker in English but lack it in German. Finally, ECM-constructions which do not exist in German show up with an infinitival marker in English. This distribution is summarized in the following table. Table 12. Distribution of the infinitival marker in English and German Construction Modals Perception verbs, causatives (active) Perception verbs, causatives (passive) Semi-modals (help, need) Control verbs Raising verbs ECM-verbs

to

Til

-

-

-

-

+

-

±

±

+ + +

+ + N/A

A common assumption is that constructions with infinitival markers involve a TP, whereas constructions without infinitival markers lack a TP. Looking in more detail at the properties of various infinitival constructions however reveals that the presence or absence of the infinitival marker does not correlate in any sense with the tense properties of infinitives (nor any other property). The claim that certain infinitives project a TP whereas others don't thus merely serves the purpose of encoding the distribution in Table 12 and lacks an empirical and theoretical foundation. Crucially, however, as such it can then be replaced by any other stipulation, and hence the distribution of the infinitival marker does not provide any insight into the ques-

Other functional projections

111

tion of whether an infinitival construction involves a (meaningful) TP or not. Besides attributing the distribution of the infinitival marker to a selectional property of the higher predicate we are aware of no account that successfully derives the distribution in Table 12 from independent properties attested in infinitives. The challenge seems to be to find a property that is common to either the class of constructions that take a zw/to-infinitive, or the class of constructions that take a bare infinitive. Intuitively, the infinitival marker appears in contexts that can be described as non-finite. However, apart from the fact that not all non-finite constructions involve an infinitival marker (modals, perception verbs), one seems to run into even bigger problems, when one tries to define what "non-finite" means. Concerning the distribution of the infinitival marker, let us start with a few options that come to mind, but that turn out to be unsatisfactory in light of the distribution in Table 12. We can observe for instance that overt subjects can originate in to-infinitives (ECM), as well as bare infinitives (causatives, perception verbs). Subject traces are possible in to-infinitives (raising, passivized perception verbs in English) as well as bare infinitives (modal constructions). Intensional infinitives can be to-infinitives (ECM and control infinitives like believe, try, decide etc.) or bare infinitives (modals). Non-intensional infinitives can be to-infinitives (factive or aspectual infinitives; e.g., hate, begin) or bare infinitives (perception verbs, causatives). Two further options have been mentioned as potential explanations for the presence or absence of the infinitival marker: first, the infinitival marker is responsible for infinitival case and/or agreement; and second, the infinitival marker represents infinitival tense. We will discuss these two options in more detail. The assumption that the function of to is to assign case to the infinitival subject (e.g. Null Case as suggested in Chomsky and Lasnik 1993/1995 and subsequent works) raises the question of why Null Case (or in other words PRO) is generally not available in raising and ECM constructions (cf. *John believes PROjohn to be smart vs. John tried to be smart; or *It seems PROarb to be smart vs. It is impossible to be smart). Since these infinitives obligatorily include the infinitival marker, Null Case should be available for the infinitival subject

112

Lexical

restructuring

contrary to fact. If raising and ECM constructions require—for whatever reason—an overt subject, it would not be clear why to has to be present nevertheless. Finally, even if an answer to these questions is found, the assumption that to is responsible for case seems to unavoidably run into problems when we consider modal constructions. Ignoring the debate of whether modal constructions are ambiguous between a raising and a control structure (see Chapter 3, section 4.1.2), either structure would predict that modal constructions should display the infinitival marker. If modal constructions involve a control configuration, they should behave like other control constructions—i.e., to should be required to assign Null Case. If modal constructions involve a raising configuration, they should behave like other raising constructions—i.e., to should be required for whatever reason it is required in seem constructions. Thus, the fact that only some control constructions and only some raising constructions project the infinitival marker is a problem for any account that tries to tie the absence or presence of to to the case properties of the infinitival subject (position). Since the case properties of the infinitival subject are identical in raising and modal constructions and in control and modal constructions, they should behave uniformly with respect to the presence or absence of the infinitival marker. Turning to the second claim—the assumption that the infinitival marker represents infinitival tense—we will now show that the presence or absence of the infinitival marker does not correlate with tense. That non-finiteness is not equivalent to the lack of (semantic) tense, has more or less been accepted since Stowell (1981, 1982). As discussed above, Stowell (1981, 1982), Martin (1992, 1996) and Boskovic (1995, 1996, 1997) assume that irrealis control infinitives are tensed since they involve a quasi-future interpretation, whereas propositional complements (i.e., raising and ECM-complements) are tenseless since they lack this kind of quasi-future interpretation. We have pointed out in section 3.1 that these approaches suffer from a number of inconsistencies, and we have argued that irrealis control infinitives are not necessarily tensed but can also be tenseless. For the purpose of the discussion here, we will abstract away from this dispute since under either approach the claim that to is responsible for infinitival tense is untenable. The crucial fact is that ECM and

Other functional projections

113

raising infinitives (except modal constructions) as well as control infinitives require the presence of the infinitival marker. Assuming a Stowell/Martin/Boskovic type approach, both tensed (control) and tenseless (ECM, raising) infinitives occur with the infinitival marker; assuming our approach, again both tensed (non-restructuring) and tenseless (potential restructuring) infinitives occur with the infinitival marker. Thus, in both approaches, the distribution of the infinitival marker does not correlate in any obvious way with tense; both tensed and tenseless infinitives show up with the infinitival marker and the assumption that infinitival to represents (infinitival) tense is hence unfounded. In sum, to account for the presence vs. absence of the infinitival marker it seems necessary (under any account) to make reference to category specific or selectional specifications of the selecting verb (as for instance in the more traditional view developed by Bech 1955 who suggested that different verbs ask for a different "status"—such as a bare infinitive status, to-infinitive status etc.). Assuming that the distribution of the infinitival marker is determined by a lexical or selectional property of the matrix verb (cf. Burzio 1986; Rochette 1988) certain tendencies typical for thematic selection are then not unexpected. Rochette for instance notes that the choice of infinitival marker in French shows regularities (though not absolute ones) for verbs that convey the meaning of the beginning of an action—these verbs require the preposition à (cf. Rochette 1988: 247, fn. 33). Although these regularities as well as the close to identical distribution of the infinitival marker in English and German are of potential importance, the generalizations seem to be far from clear at this point. We thus follow the approaches mentioned above and assume that until we can gain a better understanding of the issue at hand, the best way to account for the distribution of the infinitival marker is to invoke lexical or selectional restrictions imposed by the higher verb. The final issue we would like to touch on is the question of the position of the infinitival marker in the clause. Concerning the infinitival marker in Romance, there is some consensus among authors that the infinitival markers in Romance are not INFL or tense elements but rather heads of prepositional phrases outside the infinitive (cf. Manzini 1982; Rizzi 1982; Burzio 1986; Rosen 1989; Rochette

114

Lexical

restructuring

1988 among others). For English, Travis (1994, 2000) argues that the infinitival marker is not situated in T/INFL but in a lower functional projection inside the verb phrase. The main argument against the assumption that the infinitival marker to occupies the tense head comes from the position of to with respect to negation. As illustrated in (89)a, in the most neutral context, to follows negation in English. In structures where the infinitival marker shows up before negation (cf. (89)b), Travis assumes that the negation is an instance of constituent negation (an intuition that seems to be shared by most English speakers we have consulted). Some further evidence for this assumption can be drawn from examples involving VP-ellipsis. While the infinitival marker can precede or follow negation in (89)a,b, it has to follow negation when the infinitival VP is elided (cf. (89)c,d). Stranding of the negation after an auxiliary, on the other hand, is possible in a finite VP-ellipsis context (cf. (89)e). (89)

VP-ellipsis a. b. c.

John tried not to win John tried to not win John tried to win the game but Mary tried not to (in order to make him happy) d. *John tried to win the game but Mary tried to not e. John tried the cake but Mary did not A Travis-style analysis for the infinitival marker together with the assumption that constituent negation cannot modify an empty category (constituent negation requires some focus which presumably cannot be realized by a phonologically empty phrase), can account for the ungrammaticality of (89)d: (89)c involves sentential negation, however, (89)d can only involve constituent negation. Given the assumption that constituent negation (but not sentential negation) requires an overt constituent, (89)d is correctly ruled out. In sum, we have shown in this section that the presence of infinitival markers in restructuring infinitives does constitute a problem for our claim that restructuring infinitives do not involve tense. As mentioned in the introduction to this section, syntactic theories have be-

Other functional projections

115

come quite diverse, and one thus has to separate the issue of what constitutes a restructuring configuration from the question of what syntactic label a restructuring infinitive has. While we concede that there is a projection hosting the infinitival marker in restructuring infinitives, we have also argued on semantic and syntactic grounds that this projection cannot be assumed to be TP in the traditional meaningful sense. Although we did not adopt any specific analysis for the infinitival marker, we have suggested that the distribution of the infinitival marker is driven by selectional properties of the higher verb and that the projection hosting the infinitival marker is part of the lexical VP. The presence of the projection hosting the infinitival marker, however, is invisible for the restructuring configuration.

4.4.

Negation

In many recent works it is assumed that sentential negation is represented by a clausal functional projection (NegP) above VP (cf. Pollock 1989; Belletti 1990; Haegeman andZanuttini 1991; Zanuttini 1991). The question that arises for the issue of restructuring and the restructuring configuration is whether restructuring infinitives can involve sentential negation. The literature appears to be split regarding this question. While Sabel (1994/1996) argues for a full clausal structure of restructuring infinitives which can function as independent domains for negation, Haider (1993) claims that a restructuring infinitive does not constitute an independent domain for sentential negation, but rather a sentence with a restructuring infinitive consists of only one (sentential) negation domain. We would like to suggest that the reason for this dispute is that most analyses do not distinguish between the different grades of (non-)restructuring. To anticipate the analysis we will motivate in detail in Chapter 5, we assume that infinitival constructions that allow pronoun fronting and focus scrambling but that prohibit long passive and non-focus scrambling, are not instances of restructuring infinitives, but rather what we will call reduced non-restructuring infinitives. Syntactically, reduced non-restructuring infinitives are represented as vPs or TPs that lack any kind of C-projections. If—as we suggest in this book—(true) re-

116

Lexical restructuring

structuring is distinguished from reduced non-restructuring the situation concerning negation becomes suddenly quite clear: restructuring infinitives prohibit embedded negation, whereas reduced nonrestructuring infinitives allow embedded negation. Some authors have recently proposed that sentential negation has to be licensed by tense (cf. Zanuttini 1991, 1996; Kayne 1992; Haegeman 1995, 1996). Although it is not clear that this claim extends straightforwardly to German, let us assume for now that it does indeed hold for German as well. Since the restructuring configuration requires a tenseless infinitive and we assume that this property corresponds to the lack of a TP in restructuring infinitives, it would be predicted that (true) restructuring infinitives cannot involve an embedded negation. Assuming that reduced non-restructuring infinitives, on the other hand, can involve embedded tense and hence an infinitival TP, it would be predicted that reduced non-restructuring infinitives allow embedded negation. The examples in (90) through (91) confirm this prediction. The sentences in (90) which involve a passivized or unaccusative matrix predicate but lack long object movement (i.e., case is assigned inside the infinitive and hence the infinitives are non-restructuring infinitives) undoubtedly license embedded negation.42 The most salient (and as far as it is possible to show for some speakers the only) interpretation (90)a can receive is that John got permission to not eat the spinach (i.e., embedded negation). Similarly, (90)b refers to a situation in which the implicit agent, trying to lose weight, made an explicit attempt not to eat the cake. Finally, (90)c receives the interpretation that John succeeded in not eating the cake. The wide scope readings of negation, on the other hand (i.e., interpretations in which John did not get permission to eat the spinach, did not make any attempt to eat the cake, or did not succeed in eating the cake), are only 42

As mentioned before, pre-verbal non-restructuring infinitives such as the ones in (90) require an intonational break before and after the infinitive. Assuming that "larger" complements have to be set off from the matrix predicate for prosodie or parsing reasons, this intonational property can be seen as further support for the claim that examples of this sort involve more structure than restructuring infinitives. See also Chapter 5, section 3.2.2 for the issue of extraposition vs. intraposition.

Other functional projections

117

marginally available in these examples. As we will argue in detail in Chapter 5, section 3.1, the lack or marginality of this interpretation follows from the assumption that non-restructuring infinitives block (regular) scrambling and allow (if at all) only focus scrambling. In order to get a wide scope interpretation of the negation in (90), the embedded objects, which appear to the left of negation, have to have undergone scrambling from the non-restructuring infinitives. However, since this form of scrambling is not readily available for all speakers (and is only licensed when the scrambled phrase is heavily focused, which can be facilitated by adding nur 'only' to the scrambled phrase), a structure with matrix negation is not or only marginally licensed. (90)

Reduced non-restructuring and negation a.

weil dem Hans [den Spinat nicht zu essen] since the John-DAT [the spinach-ACC not to eat] erlaubt wurde allowed was %'since John was not allowed to eat the spinach' 'since John was allowed not to eat the spinach' b. weil [den Kuchen nicht zu essen] versucht wurde since [the cake-ACC not to eat] tried was %'since they didn't try to eat the cake' 'since they tried not to eat the cake' c. weil dem Hans [den Kuchen nicht zu essen] since the John-DAT [the cake-ACC not to eat] gelungen ist managed is %'since John didn't manage to eat the cake' 'since John managed not to eat the cake' Turning to restructuring infinitives, we find exactly the opposite situation. Examples involving long passive (i.e., true restructuring infinitives) clearly block an interpretation with embedded negation. The examples in (91) which differ minimally from (90) in the case of the embedded object only allow interpretations in which negation

118

Lexical restructuring

takes matrix scope (i.e., the examples refer to contexts in which John did not get permission, did not make an attempt, did not succeed). (91)

Restructuring and negation a.

weil dem Hans [der Spinat nicht zu essen] since the John-DAT [the spinach-NOM not to eat] erlaubt wurde allowed was 'since John was not allowed to eat the spinach' *'since John was allowed not to eat the spinach' b. weil [der Kuchen nicht zu essen] versucht wurde since [thecake-NOM not to eat] tried was 'since they didn't try to eat the cake' *'since they tried not to eat the cake' c. weil dem Hans [der Kuchen nicht zu essen] since the John-DAT [thecake-NOM not to eat] gelungen ist managed is 'since John didn't manage to eat the cake' *'since John managed not to eat the cake' To sum up, the contrast mentioned above (i.e., the impossibility of embedded negation in a restructuring infinitive vs. the possibility of embedded negation in a non-restructuring infinitive) is exactly as we would expect if different types of infinitival constructions have different structures: restructuring infinitives lack an embedded TP and hence do not license embedded negation, whereas non-restructuring infinitives have the possibility of projecting an embedded TP and hence also license negation (see Chapter 5, section 3.1, for further evidence for the structures suggested and additional examples illustrating the behavior of negation with different kinds of restructuring properties such as scrambling and pronoun fronting). 5. Verb raising in restructuring constructions The main claims we have argued for so far are that restructuring in-

Verb raising

119

finitives (i.e., infinitival constructions displaying transparency properties) lack a structural case position/assigner, an independent (semantic and morphosyntactic) tense contribution, and other clausal functional projections such as CP. In our approach to restructuring, the lack of these properties follows from the restructuring configuration—i.e., the claim that restructuring infinitives are VP-predicates rather than independent clauses that share all the clausal functional categories such as tense and structural case with the matrix predicate. This approach is corroborated by the (morpho)syntactic and semantic properties of restructuring constructions. Under a bi-clausal approach, on the other hand, it is assumed that the underlying structure of a restructuring infinitive includes T/vprojections which either lack any kind of features or lose their features in the course of the derivation. The obvious question that these approaches are confronted with is how the building up of vacuous structure can be motivated in cases where the projections do not show any semantic, syntactic or phonological correspondence. Assuming a restrictive theory of syntax that goes beyond description of structures (i.e., a theory that seeks to provide explanations and motivations for structures and operations) this question presents a challenge that does not seem to be trivial to overcome. Besides the question of motivating vacuous projections, which as far as we are aware of has never been addressed, bi-clausal approaches have to develop a mechanism to derive clause-union. As discussed in chapter one, the core idea in many bi-clausal approaches to restructuring is that restructuring involves some sort of headmovement (i.e., V/T-raising) which unifies the infinitive and matrix predicate and renders the infinitive transparent. The main aim of this section is to investigate the evidence that has been suggested in favor of head movement in restructuring constructions. Since under the VP-approach as proposed here, restructuring and the transparency of the infinitive follow from the restructuring configuration, a head movement operation becomes superfluous. Thus, the question we will address is whether there is independent evidence for head movement in restructuring infinitives or whether head movement is simply suggested as a means to derive clause union. If it turns out that the sole motivation for head movement in restructuring infini-

120

Lexical

restructuring

tives is to derive clause union, the approach suggested here which accounts for the properties of restructuring constructions without recourse to head movement is evidently superior to an approach that requires this additional (unmotivated) mechanism.

5.1. The verb raising/restructuring connection The idea of head movement in restructuring infinitives goes back to Evers (1975b). Assuming a basic head-final verb phrase structure for Dutch, the lowest infinitives in (92) show up on the 'wrong' side of the matrix verbs in Dutch. Evers proposes that this reordering effect is the result of movement of the infinitival verb. That is, the infinitive moves up to the matrix clause and attaches to the right of the matrix verb in Dutch in (92). (92)

Dutch: verb raising a.

b.

Jan heeft John has 'John (has) Jan heeft John has 'John (has)

Marie gisteren op proberen te bellen Mary yesterday up try-INF to call tried to phone Mary yesterday' Marie gisteren op willen bellen Mary yesterday up want-INF call wanted to phone Mary yesterday'

As a short aside, an oft-discussed problem concerning sentences like (92) is the question of whether the material to the right of the matrix verb is indeed a head or whether it is a bigger constituent. That is, one could imagine that no head movement applies in these examples but that the infinitive is part of a remnant phrase that appears in extraposed position. One of the arguments for a head movement analysis of (92) comes from the distribution of particles (cf. den Besten and Rutten 1989; Rutten 1991; Broekhuis 1992; Broekhuis et al. 1995 among many others). As can be seen in (92), the particle op which is associated with the embedded verb, appears to the left of the matrix verb. If the examples in (92) were to involve phrasal movement, one would have to assume that the embedded verb phrase (i.e., the projection that ends up to the right of the higher verb) has been

Verb raising

121

emptied (partially) prior to extraposition. In other words, the particle would have to move out of the verb phrase before the latter is extraposed. One fact about particle placement in Dutch, however, is that particles cannot undergo leftward movement such as scrambling. On the other hand, what is possible is stranding of a particle (as for instance in a verb-second context). Thus, given these facts, it is generally assumed that the dissociation of a particle from its corresponding verb provides evidence for verb raising in examples such as (92) rather than for extraposition. Returning to the issue of verb raising in restructuring contexts, Evers (1975a, 1975b) was also the first who suggested that there is a close relation between verb raising and restructuring or clause union. He suggests the Guillotine Principle in (93).43 (93)

Guillotine/S-pruning Principle

(Evers 1975a: 147)

An S that has lost its head (i.e., its V-constituent) does not survive Thus, in Evers' account, verb raising—i.e., the formation of a complex verb consisting of the matrix verb and the infinitive—creates a sentence that has lost its head and therefore the S-node cannot survive and is deleted. The close relation between clause union or restructuring and some form of verb complex formation is also an essential ingredient of Rizzi's (1978,1982) analysis of restructuring in Italian. In contrast to Dutch, however, the formation of a complex head is less straightforward in Italian. While the verb cluster in Dutch cannot be separated by any elements other than particles, suggesting that there is indeed a very tight (e.g., head) relation between the verbal elements, phrasal elements may intervene between the matrix verb and the infinitive in Italian. As Rizzi notes, clitic climbing (cf. (94)a), object preposing (cf. (94)b) and auxiliary switch (cf. (94)c) are possible in cases where adverbs show up between the two verbs. Rizzi therefore suggests that "Restructuring creates a syntactic constituent 'verbal complex', and 43

Evers attributes this principle to a proposal made by Kuroda in an unpublished paper (cited in Ross 1967: 56) which we could not track down.

122

Lexical restructuring

that this constituent cannot be simply a V. [...] Italian syntax makes use of a syntactic category, distinct from V, dominating non-lexical verbal compounds." (Rizzi 1982: 38). Since head raising creates Xo categories rather than X' categories, it is not clear how Rizzi's syntactic complex verb could be implemented in a more recent Government and Binding or Minimalist style analysis of head movement. (94)

Italian: verb raising and adjacency a.

b.

c.

(Rizzi 1982: 38)

Lo verrò subito a scrivere it-CL I-will-come at-once to write Ί will come to write it at once' Gli stessi errori si continuano stupidamente the same errors SI continue-3PL stupidly a commettere to make 'People continue to make the same errors in a stupid way' Maria è dovuta immediatamente tornare a casa Mary is must immediately return at home 'Mary has had to come home immediately'

To account for the discontinuity of the two verbs found in Romance restructuring infinitives (while keeping Rizzi's assumption that the infinitive and the matrix verb form a complex head) various analyses have been proposed that despite their technical differences all seem to express one basic idea: restructuring constructions involve covert head movement. One set of approaches assumes that head movement in restructuring infinitives does not apply in overt syntax but takes place in the covert component (i.e., as LF-incorporation of the infinitive; cf. Sternefeld 1990; Grewendorf and Sabel 1994 and Gonçalves 1998). The second set of approaches assumes that head movement in restructuring infinitives does take place in overt syntax, but that for various reasons, this movement is not reflected in the (phonological) output of the structure. First, adopting a version of the copy theory of movement in which either copy may be accessible to PF or LF (cf. Bobaljik 1995, Brody 1995, Groat and O'Neil 1996, Pesetsky 1997), Roberts (1997) suggests that head movement applies overtly, but that

Verb raising

123

in the phonological component, the tail rather than the head of the chain is pronounced. Second, many authors assume that the head that undergoes movement is not the actual infinitival verb but rather the (abstract) infinitival tense or Agr-node. This form of overt (though abstract) head movement thus creates a complex head in restructuring constructions but at the same time leaves behind the infinitival verb (cf. Kayne 1989; Roberts 1993; Bok-Bennema and KampersManhe 1994; Rooryck 1994; Sabel 1994/1996; Terzi 1996). Finally, there are approaches that assume that overt head movement of the infinitive takes place, followed by further excorporation of the higher verb (cf. Roberts 1991; Guasti 1992,1993, 1997). Let us briefly mention one other fact concerning the adjacency between a restructuring verb and the infinitive. Guasti (1997: 136) suggests that examples such as (95)a—i.e., examples in which the embedded verb precedes a floating quantifier associated with the matrix subject—provide evidence for overt verb raising in causative constructions in Italian. (95)

Italian: causative constructions a.

b.

I professori facevano commentare The professors make comment tutti quel libro a Ugo all that book to Ugo 'The professors all made Ugo comment on that book' (my paraphrase) I professori non fanno più commentare The professors NEG make anymore comment tutti quel libro a Ugo all that book to Ugo 'All the professors do not make Ugo comment on that book anymore'

While this example can be taken to show that (short) verb movement is possible in causative constructions, this fact does not show that verb raising is required in restructuring constructions.44 Assuming as 44

The argument for verb movement is based on the claim that floating quantifiers

124

Lexical restructuring

we will argue for in Chapter 3, that causati ves are functional elements and that a causative construction is an obligatory mono-clausal construction, movement of the infinitive to the left of the subject in (95)a (i.e., to a functional projection higher than vP such as for instance Kayne's 1989 InfP) reduces to the well-known fact that (Pollock-style) verb movement is possible in Italian infinitival constructions (cf. Kayne 1989, Belletti 1990).45 Crucially, however, this short verb movement is possible in all infinitival constructions and hence does not bear on the question of restructuring. Moreover, as Guasti points out, it cannot be assumed that the matrix causative verb and the infinitive form a complex head in examples such as (95)a, since various elements can occur between the two verbs (cf. (95)b). To account for the lack of adjacency Guasti assumes that the finite verb excorporates from the complex head in these contexts. This is certainly a possible account, however, it also makes the argument for obligatory verb raising less compelling. Turning to the major issue we want to address in this section, there are three types of arguments that can be found in the literature for a head movement operation in restructuring contexts (i.e., movement that unifies the infinitive with the matrix predicate). First, it is argued by a number of authors that the fact that restructuring infinitives are in a tense dependency with the matrix tense is an indication of T-to-T movement. Second, it is claimed that contexts in which a ccommanding head appears between the matrix and embedded verb/tense are incompatible with restructuring. Assuming some version of the Head Movement Constraint (cf. Travis 1984, Baker 1988), the prohibition of restructuring in these contexts would follow straightforwardly if restructuring requires head movement. And third, it is argued in a similar way that restructuring complements cannot be displaced and that this restriction can be attributed to the require-

occur in trace positions. Although this view is highly controversial and has been shown to be untenable (see Bobaljik 1998 and references therein), we will grant for the sake of the argument here that tutti in (95)a marks the underlying matrix subject position. Note also that Guasti herself assumes that causatives combine with VPcomplements and not with clausal complements.

Verb raising

125

ment of head movement in restructuring contexts which would be blocked from a dislocated infinitival complement. In the next sections, we will discuss and evaluate these arguments in turn and show that none of them can be maintained as an argument for the necessity of head movement in restructuring constructions.46 The conclusion we will reach is that the sole purpose of head movement in restructuring constructions is to unify the two clauses—i.e., to derive restructuring. In a monoclausal approach as suggested here, on the other hand, no such special restructuring mechanism is needed and head movement can thus be dispensed with. (This claim of course only refers to restructuring head movement—i.e., movement from the infinitive to the matrix predicate to create clause union. That is, we do not claim that there is no Pollock-style short verb movement in infinitives.)

46

The fact that overt head movement (i.e., verb raising from the infinitive to the matrix predicate) is found in restructuring constructions in some languages is sometimes postulated as additional indirect evidence for the claim that restructuring constructions generally involve this form of head movement. Note, however, that this claim can (if at all) only be maintained if it can be shown that verb raising in languages like Dutch or West Flemish is triggered by a syntactic or semantic property which can be argued to be present universally. Although a discussion of verb (projection) raising in West Germanic is beyond the scope of this study let us point out that so far no such property seems to have been detected. Surveying the empirical situation of verb (projection) raising structures and contexts in West Germanic, all considerations seem to point to the conclusion that verb (projection) raising is in fact not a syntactic operation but rather a post-syntactic PFphenomenon that does not affect the syntax/semantic computation in any obvious way. As is demonstrated in Wurmbrand (1999b, 2000b), the different word orders in verb (projection) raising structures found in different languages do not correlate with any difference in meaning; the "inversion rules" are language specific, construction specific, and category specific; and finally, the movement operations one has to invoke to derive the different word orders (and to exclude the impossible orders) are not connected to other syntactic or semantic properties (neither in general nor within a particular language) but seem to be best stated in terms of (arbitrary) rules or parameters (as for instance suggested by the system developed in Haegeman and van Riemsdijk 1986). We thus assume that the existence of overt verb raising does in no way force the conclusion that verb raising or head movement is a necessary or universal mechanism of restructuring constructions.

126

Lexical restructuring

5.2. Tense dependencies One of the core ideas of head movement approaches to restructuring is that head movement is triggered by some deficiency of the infinitival tense. The basic idea as developed in detail in Guéron and Hoekstra (1988) and Bennis and Hoekstra (1989) is that verbs have to be "tense licensed"—i.e., a verb has to be in a local relation with a tense head. Generally, the local relation is achieved via movement of the verb to the closest tense head. What is special about restructuring constructions is that the infinitival tense head is deficient and thus not capable of licensing the infinitival verb; hence the verb (or the infinitival tense) has to move to the matrix clause. Apart from the problem of how the presence of a featureless projection can be motivated in a restrictive theory of syntax, head movement approaches that consider the lack of tense as the trigger for restructuring are faced with a number of questions. For instance, if head movement takes place covertly (i.e., if the infinitive only becomes transparent at LF), one might wonder how overt restructuring properties such as clitic climbing, long object movement, or auxiliary switch are licensed (this problem has also been noted by Roberts 1997). Furthermore, many languages display an asymmetry between V-to-T movement in finite clauses and V-to-T movement in restructuring infinitives. Most authors assume that in Italian, verb movement takes place overtly (cf. Kayne 1989, Belletti 1990); in particular, it is claimed that short overt verb movement also applies in infinitives (restructuring as well as non-restructuring infinitives). Since according to head movement approaches, infinitival Τ in restructuring infinitives is deficient and hence cannot check features, movement of the infinitival verb in a restructuring infinitive (which takes place overtly) has to target a functional head other than Τ (in order to avoid the rather unattractive claim that the tense head in a restructuring infinitive despite being unable to license the infinitive nevertheless attracts it). An interesting question for these approaches is then why finite verbs undergo overt movement to T, whereas the non-finite verb in a restructuring infinitive—which by assumption also has to move to the finite Τ of the higher predicate—can move to some infinitival head overtly, but can only undergo covert movement

Verb raising

127

to the matrix T. While these questions are not insurmountable, they seem to point to the special nature of head movement in restructuring infinitives. Finally, the most serious problem for the claim that the lack of tense is the trigger for restructuring is that this claim appears to be too strong. As we have shown in section 3.5, not all tenseless infinitives are restructuring infinitives. The lack of tense is a necessary condition for restructuring, however, it is not a sufficient one. To qualify as a restructuring infinitive, other properties have to be met as well. Thus, the question arising for head movement approaches is how restructuring can be excluded in certain tenseless infinitives (and—depending on the answer to this question—how the embedded verb can be tense licensed in a tenseless non-restructuring infinitive).47 To accommodate these problems, one might suggest that restructuring infinitives are "tense deficient", whereas nonrestructuring infinitives (whether tenseless or tensed) are not "tense deficient". Note however, that this modification essentially dissociates head movement from the actual tense properties of a construction. In other words, while this is a possible move which derives restructuring, it seems to defeat the original idea that head movement is motivated by the tense properties. We thus conclude that head movement is not motivated by the tense properties of infinitival constructions but would have to be considered as an arbitrary restructuring property or feature (as for instance suggested by Sabel 1994/1996), and that the whole purpose of head movement in restructuring infinitives is to derive restructuring in a bi-clausal structure.

5.3.

Intervening

heads

The second set of arguments for head movement in restructuring 47

Note that under our approach, this problem does not arise since a TP is only projected when there is a reason for it—e.g., when there are tense features. Thus, even in the non-restructuring versions of fry-infinitives, a TP need not be present. As mentioned before, non-restructuring infinitives can be CPs, TPs, or vPs depending on their syntactic and semantic properties.

128

Lexical

restructuring

constructions comes from contexts in which a head appears between the matrix and embedded verb/tense. It has been argued that these contexts are incompatible with restructuring and hence provide an argument for head movement in restructuring contexts (see for instance Bok-Bennema and Kampers-Manhe 1994; Sabel 1994/1996; Terzi 1996; and Roberts 1997). In this section, we will show first that the argument for head movement is weakened by the necessity to make certain non well-established assumptions, and second, that the claim that intervening heads block restructuring is empirically incorrect. There are two contexts that have been taken to motivate the existence of verb raising in restructuring contexts: overt complementizers and negation. Let us start with the first case. A general crosslinguistic property of restructuring is that clause union effects are not found in the presence of complementizer material and across finite clauses (see also section 4.2). In our approach, this restriction follows from the restructuring configuration: restructuring infinitives are VPpredicates and hence cannot include a C-domain. As discussed in section 4.2, approaches that derive restructuring by topicalization of an infinitival projection to the embedded SpecCP (plus subsequent head movement) can only account for this generalization by an additional assumption such as the Doubly Filled Comp filter. "Pure" head movement approaches to restructuring, on the other hand, account for the prohibition against complementizers in restructuring infinitives by invoking some version of the Head Movement Constraint (cf. Travis 1984, Baker 1988). That is, it is assumed that overt material in C° blocks head movement and that intervening heads cannot be skipped. However, as straightforward as this approach might appear, it also raises some technical questions. Let us consider how head movement proceeds in a restructuring construction. According to most head movement approaches, it is assumed that restructuring infinitives are CPs. Assuming that head movement cannot skip intervening heads (otherwise the explanation for the intervention effect of complementizers would be lost), the embedded verb/tense moves from its base position to INFL (and other functional inflectional heads) and then to COMP. At this point, we seem to have two options of how the derivation can proceed. First, the whole V/T+C

Verb raising

129

could move to the matrix clause; or second, the V/T could excorporate (cf. Roberts 1991,1994) and leave behind the embedded C on its way to the matrix clause. Under either assumption, however, it is not clear what would distinguish overt from covert complementizers. Why can an overt complementizer not move along with the verb? Or why can the verb only excorporate when the C it attaches to is empty? A similar question arises for Roberts' (1997) approach. Roberts states that V/T-movement in restructuring infinitives is blocked in the same way verb second movement is prohibited in embedded clauses with a complementizer in German or Dutch. Note, however, that in Roberts' system the crucial constraint regulating verb movement is a morphological constraint that prevents the pronunciation of a complex head consisting of two morphological words. Given this system, it is not clear why movement of the verb to C would be blocked by an intervening complementizer. That is, one could imagine that in both Germanic embedded clauses as well as restructuring infinitives, verb movement to C does take place (even in the presence of a complementizer), however, the morphologically complex head created by adjunction of the verb to an overt complementizer would force pronunciation of the verb in a lower position. Thus, while Roberts can account for the fact that a verb cannot form a complex head (morphologically) with a complementizer, it seems that his system does not actually prevent movement of the verb. Although these questions are not insurmountable, they nevertheless show that additional assumptions or mechanisms are necessary, which however, make the claim that complementizers provide an argument for head movement in restructuring infinitives somewhat less compelling.48 A more important problem for the head movement approach to restructuring arises concerning the second point—the claim that negation blocks restructuring. Examples illustrating this claim are provided in (96) for Italian (cf. Radford 1977, Kayne 1989 among many 48

An elaborate system of different types of head movement as developed in Roberts 1991, 1994 (in particular, the idea that head movement can be substitution) might avoid some of the problems mentioned, however, as has been pointed out by Iatridou (1994), Roberts' system itself is faced with various inconsistencies that would carry over to the issues at hand.

130

Lexical restructuring

others) and (97) for Spanish (cf. Luján 1980). While embedded negation is possible in an infinitive when the clitic stays inside the infinitive (cf. (96)a, (97)a), negation is prohibited when a clitic climbs to the matrix predicate (cf. (96)b, (97)b).49 (96)

Italian: negation and clitic climbing a.

b.

(97)

Voglio non farlo I-want not to-do+it-CL Ί want not to do it' Lo voglio (*non) fare it-CL I-want (*not) to-do Ί want (*not) to do it'

Spanish: negation and clitic climbing a.

b.

Quisiera no verte I-would-like not see+you-CL Ί would like not to see you again' Te quisiera (*no) you-CL I-would-like (*not) Ί would like to (*not) see you again'

más again ver see

más again

While the contrast in (96) and (97) shows an interesting effect regarding the interaction of clitic climbing and negation, we will see that it cannot be used as an argument for head movement in restructuring constructions. If it were the case that the negation in (96)b and (97)b blocks V/T raising (i.e., if negation blocks restructuring and not simply clitic climbing; rather clitic climbing is blocked as a result of the lack of restructuring), it should be the case that restructuring is generally impossible with embedded negation. This predication, however, is not borne out (see also fn. 49). While clitic climbing is prohibited when the infinitive includes negation, other restructuring properties such as object preposing are perfectly well-formed across Note that the validity of the claim that negation blocks clitic climbing has also been questioned by various authors. Napoli (1981) and Cinque (2000) for instance provide examples involving clitic climbing across negation that are considered grammatical (at least by these authors).

Verb raising

131

an embedded negation (cf. (98) which is from Watanabe 1993: 366 and has been confirmed by all of my informants). Note that the only way (98) can be interpreted is with narrow scope for the negation; i.e., the negation cannot have matrix scope and is thus clearly in the embedded predicate. (98)

Italian: negation and object preposing Quei libri si potrebbero non leggere subito these books SI would-be-able not read immediately 'What people could do is to not read these books immediately' * 'What people could not do is to read these books immediately'

Examples such as the one in (98) thus show that negation in the embedded complement does not block restructuring in Italian.50 Since so far, we have not found any reason to assume that clitic climbing and object preposing define different classes of (non-)restructuring (but see fn. 50), we assume that these properties characterize the same type of construction, and hence the contrast between (96) and (98) can be taken to indicate that negation does not interfere with restructuring but only affects clitic climbing. Assuming that clitic climbing in contrast to object preposing is (or at least involves a step of) head movement, the blocking effect of an intervening negative head is straightforwardly accounted for (see for instance Kayne 1989, 1991; Watanabe 1993; Roberts 1994 for accounts along these lines). 50

Object preposing is thus similar to focus scrambling and pronoun fronting in German. Assuming that negation works the same way in Italian and German and that the example in (98) does not involve constituent negation, this example would count as a reduced non-restructuring infinitive rather than a true restructuring infinitive in our terminology (see the discussions in section 4.4 and Chapter 5, section 3.1). This claim might be further supported by the speaker variation we find in Italian (see fn. 49) and the fact that long passive (i.e., full restructuring) is significantly more restricted in Italian than in German (see Cinque 1997a). However, any firm conclusion would be premature at this point since an exhaustive survey involving a detailed comparison of different restructuring properties in different constructions is still outstanding for Italian.

132

Lexical

restructuring

Crucially, however, negation has no effect on restructuring, which defuses the argument for head movement in a restructuring infinitive based on the intervention effects of negation. 5.4. Dislocation of restructuring complements The final set of phenomena used to argue for head movement in restructuring infinitives involves constructions in which the infinitive is displaced (by operations such as pied-piping, cleft formation, or right node raising). As is illustrated in (99) (cf. Rizzi 1982: 7), come constructions allow clitic climbing in Italian (cf. (99)a). Furthermore whpied-piping of an infinitive (i.e., movement of a constituent containing the wfr-phrase) is possible (cf. (99)b). However, what appears to be blocked is a combination of the two phenomena: when the infinitive involves a restructuring property such as clitic climbing, pied piping is blocked (cf. (99)c). An illustration of the structure of these examples is provided in (100). (99)

Italian: pied-piping and clitic climbing questi argomenti... these topics... a.

dei quali ti verrò a parlare al più presto about which you-CL I-will-come to talk asap 'these topics, about which I will come talk to you as soon as possible' b. a parlarti dei quali verrò al più presto to talk+you-CL about which I-will-come asap 'these topics, about which I will come talk to you as soon as possible' c. *a parlar dei quali ti verrò al più presto to talk about which you-CL I-will-come asap 'these topics, about which I will come talk to you as soon as possible' Proponents of a head movement approach to restructuring suggest

Verb raising

133

that the ungrammatically of (99)c is caused by the impossibility of head movement from an infinitive that is not in its base position. That is, in a (simplified) structure such as (100), the head movement constraint would block movement of the embedded verb/tense from the infinitive in SpecCP to the (now lower) matrix verb/tense, and hence restructuring (and as a result clitic climbing) is prohibited (see for instance Roberts 1993/1997 for an analysis along these lines). (100) Italian: pied piping and restructuring

=(99)b,c

CP INF

In what follows, we will show that both the account of the (apparent) prohibition against displacing a restructuring infinitive and the empirical characterization of the phenomenon are problematic. Let us start with a closer look at the analysis in (100). Recall that head movement in restructuring infinitives in Italian has to be covert (or some version of covert movement as mentioned above). Thus, to base the argument for head movement in restructuring infinitives on the ungrammaticality of (100), it has to be the case that head movement is blocked at LF in (100). A question that arises in this respect is why head movement would not be possible after the infinitive is reconstructed to its base position. This question appears even more urgent in light of the German restructuring examples in (101) which involve dislocated (i.e., topicalized) restructuring infinitives (evidenced by long passive in (101)a and scrambling in (101)b). Importantly, both examples are grammatical.

134

Lexical restructuring

(101) Topicalization and restructuring a.

b.

Zu reparieren wurde nur der Wagen versucht To repair was only the car-NOM tried 'They only tried to repair the car' Zu reparieren hat Hans nur den Wagen versucht To repair has John only the car-ACC tried 'John only tried to repair the car'

To account for the grammaticality of the German examples in (101), there are two options: either restructuring does not require head movement, or if it does, covert head movement is possible after reconstruction. Under either assumption, however, the explanation for the ungrammaticality of the Italian examples (and hence the argument for head movement) would be lost. If head movement is possible after reconstruction in (101), it should be possible in (100) as well (see also below). If head movement is not required for restructuring, the ungrammaticality of (99)c/(100) must have a different source. The latter claim (i.e., that pied piping is subject to special licensing conditions) is further confirmed by the fact that pied piping is not generally ruled out in restructuring contexts. As mentioned by Cinque (1998, 2000), pied piping seems to be subject to various contextual and pragmatic restrictions that are not well understood so far. This can be seen in (102)a, which according to Cinque (2000: 33, fn. 11) contrasts sharply with (99)c. Furthermore, Cinque notes that topic or focus movement structures also readily allow clitic climbing in Italian (cf. (102)b). (102) Italian: dislocation structures and clitic climbing a.

b.

Maria presentare alla quale non lo vorrei Mary introduce to whom not him-CL I-would-like '.. .Mary to whom I wouldn't like to introduce him' A PARLARE DEI SUOI PROBLEMI To speak about his problems ti verrà. Vedrai. to-you-CL he-will-come. You-will-see 'He'll come speak to you about his problems. You'll see.'

Summary

135

In sum, in all the cases in (101) through (102) restructuring infinitives (i.e., infinitives involving a transparency property) appear in dislocated positions. To keep a head movement analysis it would have to be assumed that head movement applies after reconstruction in these examples. However, if this is the case, nothing would block head movement in (99)c, and hence the head movement approach would make the wrong predication concerning the grammaticality of this example. While a determination of the factors that regulate pied piping is still outstanding (but see Cinque 2000 for some speculations), it seems clear that the (im)possibility of head movement does not provide any obvious way to account for the difference between (99)c and (102)a. We thus conclude that dislocation contexts cannot be taken to support the claim that restructuring constructions have to involve head movement. To conclude, the discussion in this section has shown that the assumption of head movement in restructuring constructions does not seem to be beneficial, neither on empirical nor on theoretical grounds. Despite the claim that there are arguments for head movement in restructuring constructions, we have seen that head movement approaches either make wrong predications (e.g., the lack of blocking effects of embedded negation for some restructuring properties in Italian), or have to be supplemented by additional stipulations to properly cover the facts. Thus, since the assumption of head movement does not provide any empirical advantage over approaches that do not rely on head movement, but rather introduces further complications and theoretical inelegances, an approach that does not require this mechanism seems superior.

6. Summary In this chapter, we have argued that (lexical) restructuring is only licensed when the construction forms a restructuring configuration. A restructuring configuration is characterized by the following syntactic and semantic properties. A restructuring infinitive does not involve 'propositional' or 'force' properties such as tense, negation, or complementizers and lacks an embedded structural case posi-

136

Lexical

restructuring

tion/assigner. That is, restructuring is only possible when the infinitive can be interpreted as a predicative event or action rather than a proposition (this claim is also supported by the lack of an infinitival (PRO) subject, which will be discussed in Chapter 4). Furthermore, we have shown that an object in a (lexical) restructuring construction is assigned case in the functional domain of the matrix predicate rather than inside the infinitive or directly by the embedded verb. We have argued that this property follows directly from the lack of an embedded structural case position/assigner in the infinitive. Finally, we have seen that restructuring requires that all properties of the restructuring configuration are met, and that for instance tenselessness alone is not sufficient to license restructuring. Assuming that phrase structure is meaningful and motivated by the syntactic and/or semantic properties of a construction (rather than being templatic and vacuous), the obvious syntactic implementation of a restructuring configuration is a structure in which a lexical restructuring infinitive corresponds to a "VP"-predicate. We have shown that the assumption that a restructuring infinitive represents a simple VP-layer of the matrix predicate not only provides a straightforward account for the syntactic and semantic properties of lexical restructuring constructions but also allows us to predict and explain why restructuring is the way it is. As a final note of caution, we want to stress again that the crucial contribution of our analysis is the characterization of the restructuring configuration. The syntactic label of a restructuring infinitive (e.g., "VP" as we have used it here) is merely a notation and can thus be replaced by any other label. In particular, our system allows for as many functional projections as one might want to posit for a restructuring infinitive. However, crucially only those functional projections are possible that do not represent any of the properties excluded by the restructuring configuration. Before we turn to further evidence for our analysis of lexical restructuring infinitives in Chapter 4 where we will discuss the properties of infinitival subjects, we will first introduce and discuss a second class of restructuring constructions, namely fimctional restructuring constructions.

Chapter 3 Functional restructuring

1. Introduction In this chapter, we will discuss a second class of infinitives which we will refer to as functional restructuring infinitives. A list of predicates that can function as functional restructuring predicates in German is given in Table 13. Table 13.

Functional restructuring predicates in German

Verb

Gloss

Verb

Gloss

beginnen drohen dürfen dürfte gehen haben hören kommen können lassen

begin threaten may might go have hear come can let

möchte müssen pflegen scheinen sehen sein sollen versprechen werden wollen

would like to must used to seem see be shall/should promise will want

The major claim we will defend in this chapter is that constructions involving the predicates in Table 13 have to be distinguished from the lexical restructuring constructions discussed in the previous chapter. Like lexical restructuring constructions, functional restructuring constructions are mono-clausal configurations. However, in contrast to lexical restructuring infinitives (which result from the option of combining a lexical restructuring verb with a very "small" complement), functional restructuring is a direct (and unavoidable) result of the architecture of a clause. The (simplified) structure of

138

Functional restructuring

functional restructuring constructions is illustrated in (103).51 As can be seen in the diagram, a functional restructuring construction is a simple clause with the infinitive as its main VP predicate and the restructuring verb as a head of its functional domain. (103) Functional restructuring constructions FP„ F vP SUBJ

F° fiinctional restructuring verb v'

VP

v° V'

DP

V" main verb (infinitive)

The main aim of this chapter is to motivate the lexical vs. functional distinction and to provide an analysis for the different constructions in Table 13 that will allow us to account for the inherent restructuring nature of these predicates. Although we will sketch an analysis for all constructions, some of the discussions will have to be kept short and certain constructions will not receive the attention they might deserve. The chapter is organized as follows. In section 2, we will give an overview of our analysis and lay out the clause structure we suggest for German. In section 3, we will discuss and motivate the distinction between lexical and functional restructuring predicates. We will first show that alternative approaches (in particular approaches that consider all restructuring predicates as auxiliaries, thematically deficient verbs, or functional predicates) are not viable. A detailed comparison

Our analysis is similar to a recent approach suggested by Cinque (1997a, 1997b, 2000). However, in contrast to Cinque, we will argue in this chapter that not all restructuring can be reduced to functional restructuring. We return to a comparison of our approach and Cinque's in section 3.3.

The split IP parameter

139

of the properties of different restructuring predicates in German will yield the conclusion that the distinction between lexical and functional restructuring is indeed necessary. Finally, in section 4, we will suggest an analysis for the different types of functional restructuring constructions in German with particular focus on the syntactic properties of modal constructions. 2. The split IP parameter 2.1. English vs. German A well-known restriction on modal constructions in English is that modal verbs cannot co-occur with inflectional material (cf. (104)a) and cannot occur in embedded constructions (cf.(104)b,c). (104) English modal constructions a. *John musts/musted (to) go home early b. *John must can swim c. *John seems to can swim A common account of the distribution in (104) which goes back to the 1950s is that co-occurrence restrictions of this kind indicate competition of the elements involved (or a subset thereof) for the same syntactic position (see for instance Chomsky 1957; Jackendoff 1972, 1977; Fiengo 1974; Akmajian, Steele and Wasow 1979). Thus, in English, modals and inflectional material are in competition and only one element can win out. A similar effect is encountered for the English inflectional affixes representing agreement and tense. While English has an obligatory agreement marker (cf. -s in (105)a) and an obligatory tense marker (cf. -ed in (105)b), these two affixes cannot co-occur (cf. (105)c). (105) English tense and agreement a. John walks/*walk to the store b. John walked/*walk to the store c. *John walkeds to the store

140

Functional restructuring

The distribution in (104) through (105) can be accounted for if it is assumed that modals and inflectional affixes compete for one position—the head of "IP" (cf. Bobaljik 1995; Bobaljik and Thráinsson 1998). That is, assuming that in English the domain responsible for inflection and modality includes only one functional projection, only one of these elements can be realized (cf. (106)). 52 Which element is inserted is determined as part of the morphological insertion rules (i.e., modals are ranked higher than inflectional affixes; and tense is ranked higher than agreement).53 (106) English IP IP Γ

SUBJ

Γ modalst -ed -s

vP

VP

Note that as mentioned in the text "IP" refers to the tense/agreement and modal domain and does not include categories like aspect. That is, we assume that English clauses can project independent aspect phrases which are lower than the IP (cf. examples such as John must have been elected). 53 Under this approach, the modals could, should, would, and might cannot be analyzed as complex modal+PAST elements. This problem has been addressed by a number of authors and there are two possible ways to account for these constructions: these modals can be analyzed as lexicalized non-complex forms or they can be analyzed as C+I combinations. The first option is supported by the (idiosyncratic) semantics of some of these modals: might cannot refer to a past may situation; could can but does not have to refer to a past can situation; and finally, most North American speakers do not use the modal shall but nevertheless use the modal should. The second option is supported by the subjunctive properties of these modals; since subjunctive is generally seen as a feature of the CP rather than the IP domain, (certain) modals (e.g., would) could be treated as modal+PAST combinations with the modal originating in C \ We will not engage further in this question here (for discussion see McGinnis 1993; Rowe 1994 among others).

The split IP parameter

141

Turning to German, we find that German differs from English in all respects mentioned above. As is illustrated in (107)a, agreement and tense affixes are simultaneously realized on the verb. (107)b,c show that modals are inflected in German; and (107)d shows that modals can be embedded (e.g., under other modals). (107) Modals and inflectional elements in German a.

Hans sag-t a', die Kinder sag-en John say-AGR the children say-AGR a " . Hans sag-te a ' " , die Kinder sag-te-n John say-PAST the children say-PAST-AGR b. Hans muß / muß-te nach Haus gehen John must-PRES / must+PAST to home go 'John must/had to go home' c. Die Kinder muß-te-n nach Haus gehen The children must+PAST+AGR to home go 'The children had to go home' d. Hans muß schwimmen können John must-PRES swim can 'John must be able to swim' The account we will adopt here to characterize the differences between English and German is based on the idea that languages may vary in their inventory of functional projections and in particular the organization of the "IP" (cf. Iatridou 1990; Johnson 1990; Ouhalla 1991; Bobaljik 1995; Thráinsson 1996; Bobaljik and Thráinsson 1998). The basic analysis we suggest for German is illustrated in (108) (see the next section for refinements). Following this idea, we assume that tense and agreement are split in German, and that modality is expressed by independent modal projections which can be iterated. The structures in (106) vs. (108) straightforwardly account for the differences between English and German (i.e., the different morphological properties of modal verbs in these two languages as well as the possibility vs. impossibility of stacking modals). Note that we consider the bundling of features in different syntactic projections to

142

Functional restructuring

be a language specific issue. To be more specific, we assume that there is a universal pool of features (and perhaps, these features are even organized in a universal hierarchical order as suggested by Cinque 1999). However, the way the syntax combines these features in terms of syntactic projections is a language specific and largely arbitrary issue (however, it has clearly defined syntactic consequences, such as for instance the number of specifier positions; cf. Bobaljik and Thráinsson 1998). (108) German modal and inflectional domain AgrSP

od° können

Thus, while in German in contrast to English, tense, agreement, and modality are represented by different functional projections, this does not entail that all languages are either of the English type or the German type. Rather we expect that there are languages that for instance group tense and agreement together but exclude modality. Possible instantiations of this scenario are the Mainland Scandinavian languages which involve only one projection for tense and agreement (cf. Bobaljik 1995; Thráinsson 1996; Bobaljik and Thráinsson 1998), but nevertheless allow inflected modals. This approach thus provides room to account for the range of inflectional and verbal combinations we find across languages. Assuming that phrase structure is universal would leave open the question of why

The split IP parameter

143

tense and agreement can be expressed simultaneously in German and Icelandic but not in English and Mainland Scandinavian, or why modals can occur with inflectional affixes in German, Icelandic, and Mainland Scandinavian, but not in English. In the next section, we will give an overview of our analysis of German clause structure; arguments and a detailed discussion of the properties of different functional constructions will be provided in the following sections. 2.2. German clause structure Although German allows verbal elements such as modals to co-occur we will see that there are also a number of restrictions (both syntactic and semantic) that influence the number and order of verbal elements in a clause. The analysis which we introduce here is intended to capture these restrictions as well as to account for the properties of different functional constructions. In essence, we claim that a German "IP" consists of three major parts (the full structure is given in (109)). First, as argued in the previous section, tense and agreement are split in German. Following standard assumptions, tense and agreement do not assign theta roles (the specifier positions of these projections are thus non-thematic positions). Second, we assume a "modal domain" which is split into two projections—labeled here as "AuxP' and "ModP". We will argue that the modal domain is also non-thematic (i.e., we will provide evidence against that claim that modal verbs assign theta roles). Finally, we assume a "thematic domain" which consists of the functional vP or AspP and the lexical verb phrase. Note that there are two lexical/functional distinctions that should not be confused. On the one hand, we assume that all projections except the VP heading a clause are functional projections (this should be seen as a purely technical syntactic distinction). On the other hand, we follow approaches that distinguish a lexical and a functional domain of a clause on lexical or semantic grounds. Following Marantz (1997) the lexical domain of a clause is the domain of special meaning, which structurally corresponds to v' (see also Harley 1995 and Travis 2000 for a similar idea

144

Functional restructuring

but slightly different implementation of the lexical/functional distinction). Thus the lexical domain of a clause (in the latter sense) can involve lexical as well as functional projections (in the former sense). For the present purpose, we consider the voice or aspect head as a functional head, however, since it is part of the lexical domain of a clause (i.e., since it establishes a thematic relation with the argument in its specifier position), this head is also a thematic head. 54 (109) German clause structure AgrSP ->©

Àgrs* TP

- ι Θ ^ ^ ^

functional, non-thematic

=>

functional, non-thematic

=ΐ>

functional, non-thematic

=>

functional, non-thematic

=Φ

functional, thematic

=φ

lexical, thematic

Τ'

A u x Î 5 ^ ^ ^ ^ T° —ιΘ

=>

Aux' Mod^^^^Aux"

-πΘ^^^Ιοά' vP/AspP^"^^fvlod° Θ '^^v'/Asp' VP Θ " " " " " " " V' V°

The syntactic distinction between lexical and functional heads and the semantic distinction between thematic and non-thematic heads thus yields three types of categories (see also van Riemsdijk 1998a 54

We will not distinguish between voice and aspect here since it does not seem to be motivated for the German constructions we will consider in this chapter. However, extending the empirical domain of our analysis might show that a separation of these categories is necessary.

Motivating the lexicaUfiinctional distinction

145

who reaches a similar conclusion). In section 3.2, we will show that this is a welcome result. We will argue that verbal categories fall into three subgroups: lexical predicates that establish a thematic relation with their arguments; purely functional predicates that do not establish any thematic relations with the arguments of the clause they appear in; and semi-functional predicates which behave like functional predicates concerning their syntactic properties, but like lexical predicates concerning their thematic properties. The structure and properties suggested in (109) will thus provide a straightforward way to accommodate these three classes. Lexical predicates are V-heads; (fully) functional categories are Mod or Aux-heads; and semifunctional categories are v/Asp-heads. Returning to the classes of restructuring predicates, we will argue for the distribution summarized in Table 14. In short, our claim is that in German, lexical restructuring verbs are main verbs, raising verbs are Aux elements, (non-dynamic) modal and auxiliary verbs are Mod or Aux elements, and finally motion, perception, and causative verbs will be treated as voice/aspect elements. A motivation for these claims will be provided in section 4. Table 14. Predicates

Category and position functional and lexical predicates in German Examples

Lexical restructuring verbs try, dare, forget Modal verbs must, can, want Auxiliary verbs be, have, will Raising verbs seem, promise Motion verbs go, come Causatives let Perception verbs see, hear

Position

Category

V Aux/Mod/v Aux/Mod Aux/some V ν ν ν

lexical functional/semi-functional functional functional/some lexical semi-functional semi-functional semi-functional

Before we turn to a discussion of the different restructuring constructions in German, we will first discuss the motivation for a lexical/functional distinction in particular in light of restructuring. 3. Motivating the lexical/functional distinction In this section, we will motivate the necessity of the lexical vs. fune-

146

Functional restructuring

tional distinction for restructuring. We will show that alternative approaches—in particular approaches that consider all restructuring verbs to be auxiliaries or functional categories—although attractive in their simplicity are nevertheless untenable. Among the many approaches to restructuring, there are two sets of accounts that attempt to provide a uniform structure for all restructuring constructions without postulating two classes of restructuring (such as lexical and functional restructuring). First, we find approaches that assume that restructuring is triggered by a thematic deficiency of restructuring predicates (cf. Napoli 1981; Rochette 1988, 1990; Rosen 1989,1990; Roberts 1997). The major claim of these approaches is that restructuring predicates are types of auxiliaries rather than full verbs semantically. Second, in light of Cinque's (1999) recent study on the ordering of clausal projections, Cinque extends his analysis of adverb phrases to restructuring predicates. The main claim of Cinque (2000) is that all restructuring predicates are functional categories syntactically that are ordered according to a universal hierarchy. We will show in this section that both types of approaches are inadequate and that neither the assumption that all restructuring predicates are auxiliaries semantically nor the assumption that all restructuring predicates are auxiliaries syntactically is capable of characterizing the class of restructuring predicates. We will first discuss the thematic deficiency approach to restructuring (section 3.1). In section 3.2„ we will provide an overview of the differences between functional and lexical restructuring constructions in German which will motivate the three-way distinction of the clausal architecture as suggested in the previous section. Finally, in section 3.3, we will briefly summarize the argument for the lexical status of certain restructuring predicates and against Cinque's unified functional treatment of restructuring predicates.

3.1. Restructuring predicates as auxiliaries A number of semantic approaches to restructuring are based on the assumption that restructuring predicates form a coherent semantic class that can be characterized semantically as "auxiliaries". The first

Motivating the lexical/fiinctional distinction

147

study in this line of research was provided by Napoli (1981). Napoli proposes that what distinguishes restructuring from non-restructuring verbs is that the former do not function as full lexical verbs but rather as auxiliaries. Napoli for instance proposes that the relation between a restructuring verb and the infinitive is in many respects (syntactic and semantic) identical to the relation between an auxiliary and a participle. Auxiliaries are characterized by the following properties (Napoli 1981: 859): i) auxiliaries offer supplemental information about the action or state of the main verb, introducing no additional independent action or state; ii) auxiliaries add conceptually basic or simple information; and iii) a modifier of the auxiliary must be interpreted as a modifier of the main verb, as well. Thus, in essence, Napoli states that auxiliaries do not express independent events but necessarily denote one single event together with the lower predicate. Restructuring predicates, in this approach, are considered as auxiliaries as defined above—i.e., Napoli claims that restructuring constructions denote only a single event. Similarly, Rochette (1988, 1990) proposes that restructuring predicates are like auxiliaries in that they are underspecified with respect to their event structure. Rochette claims that restructuring verbs lack an event position in their theta-specification. Following Higginbotham (1985), Rochette assumes that event variables have to be bound by INFL. If the matrix verb does not have an event position that has to be bound, i.e., if it is a restructuring verb, the matrix INFL binds the event position of the embedded verb. Underspecification also plays a crucial role in Rosen's (1989, 1990) account. Rosen suggests that the common property of restructuring verbs is that they are light verbs. Similar to Napoli's or Rochette's approaches, she assumes that restructuring verbs are characterized by an empty argument structure—i.e., by the lack of argument and event specifications. In order to obtain an argument and event structure, restructuring verbs have to undergo argument structure merger, the empty argument/event structure of a restructuring verb is superimposed on the fully specified argument/event structure of the embedded verb. Despite various theoretical differences, these approaches share the basic assumption that restructuring predicates have a deficient argu-

148

Functional

restructuring

ment structure or event structure and that restructuring constructions involve some kind of 'event unification'. That is, all three accounts deny the existence of an independent event structure for the restructuring predicate—rather, a sentence with a restructuring infinitive is characterized as involving a single event. In this section, we will show that the assumption that restructuring predicates do not denote independent events and that restructuring involves 'event unification' cannot be maintained. Although one of the essential claims in this study is that restructuring constructions (both lexical and functional) are mono-clausal constructions, we will nevertheless conclude (in contrast to the Napoli/Rochette/Rosen-type approach) that restructuring constructions can involve a complex event structure. Before we turn to a discussion of the Napoli/Rochette/Rosen-type approach, let us mention a slightly different account of restructuring which also makes use of the idea of thematic deficiency. Roberts (1997) tries to reduce restructuring to the claim that restructuring predicates do not theta-mark their complements (however, Roberts acknowledges that these predicates can establish a thematic relation with their subjects). Roberts' approach is different from the semantic approaches mentioned above in that the thematic deficiency does not constitute a true semantic property of the predicates involved but is essentially determined lexically. That is, if a predicate in a particular language blocks restructuring that predicate obligatorily theta-marks its complement, whereas the possibility of restructuring (of for instance the same predicate in another language) entails that that predicate does not theta-mark its complement (at least not obligatorily). As Roberts points out this account "is unsatisfactory in that it postulates an otherwise unmotivated difference between semantically and thematically very similar elements" [p. 454] in different languages. Thus, since in this account "thematic deficiency" does not represent a true semantic or thematic property but basically reduces to a lexical diacritic, we will not discuss this approach further here. Let us now return to Napoli's approach. The main support for the claim that restructuring constructions denote only one event is provided by adverbial modification: in restructuring constructions (like in auxiliary-verb constructions), modification of the matrix predicate (the auxiliary) entails modification of the embedded predicate (the

Motivating the lexical/flinctional distinction

149

infinitive or main verb). This point is illustrated in (110). The sentence in (110)a which does not involve any restructuring properties (in particular, it does not involve clitic climbing) can describe two scenarios. One scenario presupposes that there has been a previous act of imprisoning (however not necessarily a previous desire to do so). The other scenario presupposes that there has been a previous desire (but not necessarily a previous act of imprisoning). Under the first interpretation, the adverbial modifies the embedded predicate, under the second interpretation, the adverbial modifies the matrix predicate. Napoli claims that the situation is different in a restructuring context—i.e., a sentence that involves a restructuring property such as clitic climbing (cf. (110)b). According to Napoli, the sentence in (110)b has to be interpreted with the adverbial modifying the embedded verb. In other words, it is claimed that there is no reading in which di nuovo 'again' modifies the matrix verb but not the embedded verb. Thus, in a situation where the speaker in (110) has never committed an act of imprisoning but has had previous desires to do so, sentence (110)a could be uttered but (110)b would be false. What is possible, however, is a situation in which the adverb modifies only the embedded predicate (i.e., a previous imprisoning but no previous desire is presupposed) (110) How many events? a.

b.

(Napoli 1981: 874)

Voglio di nuovo imprigionarli I-want once more to-imprison-them 'We want to imprison them again' Li voglio di nuovo imprigionare them I-want once more to-imprison 'We want to imprison them again'

Since adverbs apparently cannot modify just the matrix predicate in a restructuring context, Napoli (1981) (and following her Rochette 1988, 1990 and Rosen 1989, 1990) concludes that restructuring constructions denote a single event. In the rest of this section, however, we will show that this conclusion is too strong. First, as was pointed out above, the sentence in (110)b does not require that the adverbial

150

Functional restructuring

modify both predicates simultaneously; an interpretation in which the adverbial modifies only the embedded predicate is available. Thus, the embedded predicate clearly constitutes an independent event to the exclusion of the higher predicate. This is not a serious problem for the claim that the restructuring predicate is not an independent event, but it raises some questions about what exactly the restructuring predicate is. 55 The second and more serious problem with the conclusion drawn from the examples in (110) is that the lack of the relevant reading in (110)b seems to be excluded for reasons other than the event structure of this example. That is, it appears to be the case that in restructuring contexts, adverbial modifiers following the matrix verb strongly favor a narrow scope interpretation of the adverbial. If the adverb is placed in a different position (e.g., at the end of the sentence), both sentences in (110) become ambiguous—i.e., both restructuring and non-restructuring constructions denote two events.56 Let us look at examples involving event modification in more detail. The argument for two events is based on the assumption that modifiers like again presuppose the existence of a previous event which is identical to the event they modify. In a sentence like (11 l)a, the presence of again which modifies the marry-Liz event presupposes the existence of another marry-Liz event in the past. Since the second part of the sentence in (11 l)a denies the existence of such an event, the sentence expresses a contradiction. In ( l l l ) b , on the other hand, despite the fact that Richard and Liz had never been married, an event modifier like again is possible since it modifies the matrix event (thus, ( l l l ) b presupposes a previous desire to get married, but no previous marriage). 55

This point is addressed for instance by Picallo (1985, 1990) who assumes that restructuring predicates are essentially modifiers of the embedded event. 56 Although interesting in itself, we have to set aside the question of how exactly the position of the adverb interacts with the scopai properties in restructuring constructions (i.e., how the contrast between (110)b and (112)b below is accounted for). What is sufficient for the discussion here is that in the right context, restructuring does not enforce embedded modification (i.e., a restructuring predicate constitutes a separate event). Since there is also some debate among Italian speakers about whether (110)b is unambiguous, we leave this issue for future research.

Motivating the lexical/functional distinction

151

(111) Presuppositional event modifiers a. #Richard married Liz again, but she had never been married before b. Richard wanted to marry Liz again, but they had never been married before Importantly for the discussion here, we will now show that the same situation holds in German and Italian restructuring contexts. The context given in (112) enforces modification of the matrix event since modification of the embedded predicate would constitute a contradiction. To ensure that the infinitives are restructuring infinitives, we used scrambling of the embedded object to the left of the matrix subject in German (cf. (112)a)57 and clitic climbing to the matrix predicate in Italian (cf. (112)b). Since both examples are appropriate utterances in the context provided, it can be concluded that both German and Italian restructuring constructions allow matrix event modification (without enforcing modification of the embedded predicate) and hence that restructuring predicates constitute independent events. (112) Matrix event modification a.

Context: A year ago, John and Mary were engaged, but John got cold feet and broke off the engagement; now, he has suddenly realized that he is terribly in love with Mary. dass die Maria der Hans wieder heiraten since the Mary the John again marry 'that John wants to marry Mary again'

b.

57

will wants

Un anno fa, Gianni ha rotto il fidanzamento con Maria. Ma recentemente, si dice che

In the German example, the repetition of the names is somewhat unnatural in the given context (pronouns would be preferred). However, to make sure that scrambling takes place, full DPs have to be used since pronouns always have to appear in the underlying order in German. See below, for an example with long passive which makes the same point but avoids this and another complication arising with scrambling (namely the possibility of focus scrambling).

152

Functional restructuring

Ά year ago, John broke off his engagement with Mary. But recently, people say that...' lui la voglia sposare di nuovo he her-CL would-want to-marry once more 'He wants to marry her again' To further illustrate the lack of a correlation between restructuring and the event interpretation, let us consider the core cases of restructuring in German—examples with long passive. The examples in (113) are clear cases of restructuring as evidenced by the nominative case on the embedded object. In (113)a, the context (i.e., the fact that John's car is new and has never been repaired before) is incompatible with embedded event modification by the adverbials. Since the sentence is a wellformed utterance in this context, (113)a shows again that the matrix predicate can be selectively modified and thus is accessible as an independent event. Similarly, (113)b includes two occurrences of an η-times modifier—i.e., a modifier that expresses how often an event takes/took place (which as such presupposes the existence of an event). A sentence such as (113)b is thus only possible if there are two events that can be counted independently. The wellformedness of examples of this sort then provides a further argument for the claim that restructuring predicates denote independent events. (113) Restructuring and events a.

Context: Last week, John's new car (which has never been repaired before) broke down. He brought it to the garage, the mechanics looked at it and tried to repair it. However, they couldn't fix the problem. Yesterday, they had an expert come from a different garage. Der Wagen wurde aufs Neue/zum zweiten Mal The car-NOM was again/the second time zu reparieren versucht to repair tried They tried again/for the second time to repair the car'

b.

Schon zwei Mal wurde der Schachweltmeister Already 2 times was the chess worldchampion-NOM

Motivating the lexical/fiinctional

drei Mal hintereinander three times in-a- row 'Somebody tried already champion three times in a

distinction

153

zu besiegen versucht to beat tried twice to beat the chess world row'

There is a subset of restructuring predicates, however, that do not allow modification of only the matrix event by a modifier like again. As is illustrated in (114), implicative constructions indeed force a single event interpretation. That is, it is impossible to utter an example such as (114)a in a situation in which John and Mary had never been married to each other before (we will discuss (114)b,c below). The reason for the lack of ambiguity in (114), however, is not related to restructuring but rather to an inherent property of implicatives (cf. Karttunen 1971, Pesetsky 1992). The crucial property of implicative predicates is that the truth of the sentence implies the truth of the embedded infinitive. Thus, whenever John managed to ρ is true, ρ has to be true. If manage is modified by the event modifier again, the presupposition requirement of again and the entailment relations imposed by manage interact as follows. A sentence like John managed to marry Sue entails that John married Sue (due to implicative entailment). In John managed again to marry Sue, the presupposition requirement of again presupposes a previous managing event (i.e., it has to be true that John managed to marry Sue at some time before the current managing event). This previous managing event, however, involves again an implication that John married Sue (before). The implicative nature of verbs like manage thus makes it impossible to test whether the matrix predicate denotes an independent event, since it is impossible to construct a context in which the embedded event has never occurred. Whatever the correct analysis of implicative constructions is, important for the discussion here is that the unambiguity of examples like (114)a is not caused by restructuring but rather by the implicative properties of the verb involved. (114) Implicative contexts and events a.

weil since

die Maria dem Hans the Mary-ACC the John-DAT

auf s Neue once more

154

Functional

b.

c.

restructuring

zu heiraten gelang to marry managed 'since John again managed to marry Mary [#but they have never been married before]' weil die Maria der Hans aufs Neue since the Mary-ACC theJohn-NOM once more zu heiraten wagte to marry dared 'since John again dared to marry Mary' Der Wagen wurde aufs Neue/zum zweiten Mal The car-NOM was again/the second time zu reparieren gewagt to repair dared 'They dared again/for the second time to repair the car'

Finally, sentences involving the restructuring verb dare provide further support for the assumption that the matrix predicate in a restructuring context denotes an independent event. The verb dare is (marginally) ambiguous between an implicative and a nonimplicative (irrealis) predicate. The latter does not impose any implications on the embedded predicate and can be paraphrased as 'get up the nerve to, risk, hazard, convince oneself to do' etc. We then expect that a sentence involving the verb dare and the event modifier again should be ambiguous under this non-implicative agentive reading. This prediction is borne out. The examples in (114)b,c are felicitous (at least marginally) in contexts in which the embedded event has never occurred or been completed, however, only when dare is interpreted as 'John convinced himself...'—i.e., when there is no implication relation and when dare is interpreted agentively (and as an irrealis predicate) rather than as an implicative predicate. To sum up, it seems that analyses that assume a single event structure for restructuring constructions are too restrictive in that these approaches do not allow certain modifications of the matrix predicate that are indicative of the existence of a separate event denoted by the matrix predicate. On a final note, we would like to point out that the fact that restructuring constructions consist of two separate events does in no way entail that a sentence with a restructuring

Motivating the lexical/functional distinction

155

infinitive involves two clauses. Following recent approaches on event structure that posit events within the VP (e.g., Travis 1994, forthcoming; Harley 1995), we assume that a simple clause can involve more than one event. Restructuring contexts then provide an interesting area for the investigation of the relation between tense and event structure: since a restructuring configuration involves only one tense but two events, these constructions argue strongly for the assumption that events cannot be reduced to tense.

3.2. Lexical/functional differences in German We will now turn to a first illustration of the claim that the predicates we referred to as lexical restructuring predicates (repeated in the first part of Table 15) systematically differ in certain syntactic properties from predicates that we claim are functional restructuring predicates (see the second and third part of Table 15; as we will see below, the verb begin is ambiguous and hence listed twice). With van Riemsdijk (1998), we will conclude that a binary lexical/functional distinction is not sufficient to characterize verbal categories, but that certain predicates display an intermediate status (i.e., they simultaneously display both lexical and functional properties). Table 15.

Lexical and functional restructuring predicates in German

Lexical predicates

Gloss

Lexical predicates

Gloss

beabsichtigen beginnen empfehlen erlauben gelingen [unacc] gestatten mißlingen [unacc]

intend (want) begin recommend allow manage permit fail

untersagen verbieten vergessen vermeiden versäumen versuchen wagen

prohibit forbid forget avoid miss, neglect try dare

Semi-functional predicates Gloss

Semi-functional predicates Gloss

gehen hören kommen können

lassen möchte sehen wollen

go hear come can

let would like to see want

156

Functional restructuring

Functional predicates

Gloss

Functional predicates

Gloss

beginnen drohen dürfen dürfte haben können möchte müssen

begin threaten may might have can would like to must

pflegen scheinen sein sollen versprechen werden wollen

used to seem be shall/should promise will want

In this section, we will motivate this three-way distinction between lexical, functional, and semi-functional predicates as in Table 15, and show that this three-way split can be derived from the assumptions about clause structure together with the definition for the lexical vs. functional distinction provided in section 2.2. 3.2.1. Extraposition The first property we would like to discuss is the possibility vs. impossibility of extraposition of the infinitival complement. As is illustrated in (115), the predicates we classify as lexical predicates in Table 15 allow constructions in which the infinitival complement appears to the right of the matrix verb (i.e., in extraposed position, assuming a head-final structure of the German verb phrase). (115) Extraposition with lexical restructuring predicates that John VERB a. b. c. d. e. f. g.

dass dass dass dass dass dass dass 'that

the cake

to eat

Hans beabsichtigte den Kuchen zu essen Hans begann den Kuchen zu essen Hans vergaß den Kuchen zu essen Hans vermied den Kuchen zu essen Hans versäumte den Kuchen zu essen Hans versuchte den Kuchen zu essen Hans wagte den Kuchen zu essen John intended/began. .. to eat the cake'

'intend' 'begin' 'forget' 'avoid' 'miss' 'try' 'dare'

Motivating the lexical/functional

h. i.

that dass dass 'that

that j. dass k. dass 1. dass m. dass η. dass 'that

distinction

157

it John VERB the cake to eat es Hans gelang den Kuchen zu essen 'manage' es Hans mißlang den Kuchen zu essen 'fail' John managed/failed to eat the cake' her John VERB the cake ihr Hans empfahl den Kuchen ihr Hans erlaubte den Kuchen ihr Hans gestattete den Kuchen ihr Hans untersagte den Kuchen ihr Hans verbot den Kuchen John recommended/allowed... her

to eat zu essen 'recomm.' zu essen 'allow' zu essen 'permit' zu essen 'prohibit' zu essen 'forbid' to eat the cake'

The grammaticality in (115) contrasts sharply with the prohibition against extraposition in constructions involving what we will refer to as functional (cf. (116)) and semi-functional (cf. (117)) predicates. The examples in (116) include modal and raising verbs. (116) Extraposition with functional predicates that a. dass b. dass c. *dass d. *dass e. *dass f. *dass g· *dass h. *dass i. %dass j· *dass k. dass 1. *dass m. *dass 'that

(the) cake John VERB den Kuchen Hans begann Hans drohte den Kuchen Hans darf den Kuchen Hans dürfte den Kuchen Hans hat den Kuchen Hans kann den Kuchen Hans möchte den Kuchen Hans muß den Kuchen Hans pflegte Kuchen Hans soll den Kuchen Hans versprach den Kuchen Hans wird den Kuchen Hans will den Kuchen John begins/threatened... to eat

(zu) eat 'begin' zu essen zu essen 'threaten' essen 'may' essen 'might' zu essen 'have to' essen 'can' essen 'would like' 'must' essen zu essen 'used to' essen 'shall' zu essen 'promise' essen 'will' essen 'want' the cake'

n. *dass Hans schien den Kuchen gegessen zuhaben that John seemed the cake eaten to have 'that John seemed to have eaten the cake'

158

Functional restructuring

o. *dass der Kuchen nicht ist zu essen that the cake-NOMnot is to eat 'that the cake cannot/may not be eaten' Two comments are necessary concerning these examples. First, note that the ungrammaticality of extraposed complements of modal verbs is independent of the interpretation of the modal (i.e., stranding of root/deontic or dynamic modals [see below] is as bad as stranding of epistemic modals). Second, (116) also includes the verbs begin, threaten, and promise which are in fact ambiguous between a lexical and a functional version of these verbs. Following the widely accepted view (originally suggested for begin by Perlmutter 1970), these predicates are ambiguous between a raising and a control version. In section 4.2, we will argue that in German, the raising vs. control distinction corresponds to the lexical (control) vs. functional (raising) distinction. Thus, the possibility of extraposition of infinitives combining with begin, threaten, and promise (cf. (116)) is expected since these predicates can be lexical predicates. Finally, semi-functional predicates which include motion verbs, causatives, and perception verbs, pattern with modal and raising verbs in that extraposition of the infinitival complement is prohibited. (117) Extraposition with semi-functional predicates a. b. c. d. e.

that *dass *dass *dass *dass *dass 'that

John VERB the Peter Hans geht Hans hörte den Peter Hans kommt Hans ließ den Peter Hans sah den Peter John goes/hears Peter...

the cake den Kuchen den Kuchen den Kuchen den Kuchen den Kuchen eat the cake'

eat essen essen essen essen essen

'go' 'hear' 'come' 'let' 'see'

In sum, extraposition splits verbal predicates into two groups: while lexical predicates can be stranded under extraposition, modal verbs, (unambiguous) raising verbs, motion verbs, causatives, and perception verbs cannot be separated from their complements by extraposition. To account for the prohibition against extraposition it is necessary

Motivating the lexical/fiinctional distinction

159

to point out two further facts. First, the (im)possibility of extraposition cannot be related to the presence or absence of the infinitival marker in the infinitival complement. While it is the case that most functional predicates combine with zu 'to'-less infinitives, there are also infinitives that require the infinitival marker (e.g., infinitives combining with the verbs scheinen 'seem' and haben 'have'). As can be seen in (116)e,n, extraposition is nevertheless prohibited in these constructions. Second, movement of infinitives combining with functional predicates is not generally excluded in German. Extraposition sharply contrasts with topicalization in that movement of the infinitive is possible in both lexical and functional constructions (cf. (118)). Thus, the difference between topicalization and extraposition and the general availability of remnant movement in German make an account that attempts to derive the ungrammaticality of (116) and (117) from the claim that the infinitives include an unbound trace (see for instance Rizzi 1990) less viable. (118) Topicalizationwith lexical and functional predicates a.

b.

c.

Den Kuchen essen hat nur der Hans müssen The cake eat has only the John must Only John had to eat the cake' Den Kuchen zu essen schien nur der Hans The cake to eat seemed only the John Only John seemed to eat the cake' Den Kuchen zu essen hat nur der Hans versucht The cake to eat has only the John tried Only John tried to eat the cake'

We thus conclude that the impossibility of extraposition of an infinitive combining with a functional predicate cannot be related to the internal structure of the infinitive (e.g., the presence of traces or the lack of the infinitival marker), but is rather caused by a particular property or requirement of extraposition.58 58

As we will show in Chapter 5, section 3.2.2, extraposition is sensitive to the restructuring/non-restructuring distinction. In particular, for many speakers extraposition is only possible in non-restructuring infinitives, and the examples in (115)

160

Functional restructuring

The account we would like to suggest is based on the analysis of extraposition provided in Truckenbrodt (1995). Truckenbrodt argues that extraposition in German is subject to the following prosodie condition: movement can only target the edge of a prosodie phrase—i.e., a prosodie phrase cannot be interrupted by moving material into it. Truckenbrodt shows that modals and auxiliaries when they appear in sentence-final position form one prosodie phrase with the main verb, and hence movement between two clause final verbal elements is prohibited (extraposed material has to attach to the right of the highest verbal element). We will adopt this system and assume that all (final) functional heads of a clause form a single prosodie phrase (PPh) with the main verb (cf. (119); we leave aside the question of the prosodie phrasing of the nominal arguments in the clause). Furthermore, we assume that extraposition is defined over prosodie categories, and in particular, that only prosodie phrases can undergo extraposition. Topicalization, on the other hand, has an effect on the interpretation and hence applies to syntactic projections (roughly any XP can undergo movement to SpecCP provided that the pragmatic and discourse requirements necessary for topicalization are fulfilled). 59 Thus, in a structure like (119), the VP complement and the different FPs can move to topic position, but they cannot extrapose since these categories do not form independent prosodie phrases; only the CP complement of the main verb can undergo extraposition (see also Chapter 5, section 3.2.2). would thus have to be analyzed as non-restructuring infinitives. In this sense, the question of why functional predicates are incompatible with extraposition would have to be rephrased as the question of why functional predicates cannot combine with non-restructuring infinitives. Various answers to this question come to mind: one could either assume that there are selectional restrictions among different functional categories, or one could relate the impossibility of functional categories combining with clauses to a violation of the idea of Grimshaw's (1991) extended projections (i.e., the claim that each CP/TP/IP has to dominate in its own VP). Importantly, however, the possibility vs. impossibility of non-restructuring complements is again a matter of the lexical vs. functional status of the higher predicate (see also the discussion of relative clause pied piping in Chapter 5, section 3.2.1). But see Wurmbrand (to appear) for a restriction against TP fronting in German.

Motivating the lexical/functional distinction

161

(119) Prosodie structure of a German clause FP

pph

««-cp

Under this system of extraposition, the possibility vs. impossibility of extraposition of infinitival complements follows from the lexical vs. functional status of the matrix verb. If the matrix predicate is a functional verb, it is in the same prosodie phrase as the infinitive, and hence extraposition of the infinitive—i.e., a partial prosodie phrase—is impossible. If the matrix predicate is a lexical verb, the infinitival complement is an independent prosodie phrase and hence extraposition of the infinitive is possible. To sum up, extraposition of infinitival complements splits infinitival constructions into two classes. We have suggested that the two classes are defined by the lexical vs. functional status of the matrix predicate. Furthermore, as is summarized in Table 16, the presence or absence of an infinitival marker does not correlate with the (im)possibility of extraposition. Table 16. Predicates

Extraposition and infinitival marker in German Examples

Lexical restructuring verb try, dare, manage Modal verbs must, can, want Auxiliary verbs be, have, will Raising verbs seem, promise Motion verbs go, come Causatives let Perception verbs see, hear

Extraposition

Infinitival marker

YES NO NO NO NO NO NO

YES NO YES YES NO NO NO

162

Functional restructuring

3.2.2. The "infinitive for participle" effect A peculiar property found in a number of West Germanic languages is the so-called infinitivus pro participio "Infinitive for participle" (henceforth IPP) effect. In certain constructions which we will characterize below for German, the matrix verb which semantically corresponds to a participle shows up as an infinitive (cf. (120)a). For many speakers, this effect is obligatory and examples such as (120)b are ungrammatical. 60 Note that the participle form is in principle available as can be seen in constructions such as (120)c in which the modals do not combine with an (overt) infinitival complement but with a noun phrase. In this case, most speakers prefer the participle, however, the infinitive is also possible in certain dialects (cf. (120)d). (120) The IPP effect Hans John 'John b. %Hans John 'John c. Hans John 'John d. %Hans John 'John a.

hat nach Hause gehen wollen has to home go want-INF wanted to go home' hat nach Hause gehen gewollt has to home go want-PART wanted to go home' hat ein neues Auto gewollt want-PART has a new car wanted a new car' hat ein neues Auto wollen want-INF has a new car wanted a new car'

The distribution of the IPP effect in German can be characterized as follows: functional verbs in the scope of perfective have are realized as infinitives rather than as participles. In contrast to Dutch, the IPP effect is not found in lexical restructuring constructions in German (Dutch also displays the IPP effect in constructions with verbs like There is significant speaker and dialect variation which we cannot discuss here. A summary of a preliminary questionnaire-based study of these constructions in various German dialects can be found in Wurmbrand (1999b, 2000b, in prep.).

Motivating the lexical/functional

distinction

163

try, dare etc.; cf. (81)a). This is illustrated in (121) (all sentences are grammatical when the infinitive is changed to a participle; (121)e is grammatical since the infinitive and participle are accidentally identical). (121) IPP effect with lexical restructuring predicates a. b. c. d. e. f. g. h. i.

John *Hans *Hans *Hans *Hans Hans *Hans *Hans *Hans *Hans 'John

has to home to go hat nach Hause zugehen hat nach Hause zugehen ist nach Hause zugehen ist nach Hause zugehen hat nach Hause zu gehen hat nach Hause zugehen hat nach Hause zu gehen hat nach Hause zugehen hat nach Hause zugehen has intended/began... to go

VERB-INF beabsichtigen 'intend' beginnen 'begin' gelingen 'manage' mislingen 'fail' vergessen 'forget' vermeiden 'avoid' versäumen 'miss' versuchen 'try' wagen 'dare' home'

j· k. 1. m. n.

John *Hans *Hans *Hans *Hans *Hans 'John

VERB-INF has it her to eat hat es ihr zu essen empfehlen 'recommend' hat es ihr zu essen erlauben 'allow' hat es ihr zu essen gestatten 'permit' hat es ihr zu essen untersagen 'prohibit' hat es ihr zu essen verbieten 'forbid' has recommended/allowed... her to eat it'

To determine whether functional predicates display the IPP effect is slightly more complex since a number of intervening factors are at work that influence the question. First, raising constructions (i.e., infinitives combining with the verbs begin, threaten, used to, seem, promise) cannot be considered here since these constructions cannot occur in compound tenses in German and hence cannot occur in the perfective construction for independent reasons (cf. (122)a; see also section 4.2). Similarly, the conditional modals dürfte 'might' and möchte 'would like to' only show up in their finite forms and cannot be embedded under other auxiliaries or modals (cf. (122)b,c; see Erb 2001 for a study of finite auxiliaries in German). Finally, the future

164

Functional restructuring

modal or auxiliary will cannot be embedded under a perfective auxiliary for semantic reasons.61 (122) Perfective with raising/auxiliary constructions a. *Hans hat schon nach Hause gegangen zu sein John has already to home gone to be geschienen/gescheint/scheinen seem-PARTa/PARTb/INF 'John has seemed to have gone home already' b. *Hans hat/wird nach Hause gehen möchten John has/will to home go would-like 'It has been/will be the case that John would like to go home' c. *Hans hat/wird nach Hause gehen dürften John has/will to home go might 'There will be/has been a chance that John will/might go home' The remaining (semi-)functional constructions that combine with the auxiliary have display the IPP effect. As illustrated in (123), all modal constructions occur with IPP morphology. The ¿»e+infinitive construction while (at least in spoken language) allowing the perfective construction (cf. (123)f) nevertheless blocks the IPP effect. We assume that this is due to the fact that be takes the auxiliary be rather than have. Further evidence for the claim that the IPP phenomenon is only found under the auxiliary have is provided by the semifunctional constructions in (124). While the causative let construction and perception verb constructions (optionally) display the IPP effect, motion verbs which take the auxiliary be prohibit IPP morphology. Furthermore, there is some uncertainty among speakers concerning the question of whether the deontic Aave+infinitive construction can occur in the perfective. Examples such as %Hans hat nach Hause zu gehen gehabt (Lit. 'John has to home to go had'; 'John has had to go home') are considered as quite marked. The IPP effect is clearly blocked in these examples. Since the issue is not settled yet, we have to ignore this construction here.

Motivating the lexical/functional

distinction

165

(123) IPP effect with functional predicates a. b. c. d. e.

John Hans Hans Hans Hans Hans 'John

has/d to home go VERB-INF/%PART hat nach Hause gehen dürfen/%gedurft 'may' hat nach Hause gehen können/%gekonnt 'can' hat nach Hause gehen müssen/Vbgemußt 'must' hätte nach Hause gehen sollen/%gesollt 'shall' 62 hat nach Hause gehen wollen/%gewollt 'want' was allowed/able... to go home'

f. IDer Kuchen ist nicht zu essen gewesen/*sein the cake-NOM is not to eat been/*be 'The cake could not/?was not allowed to be eaten' (124) IPP effect with functional predicates a. b.

John Hans Hans 'John

is eat VERB -PART/*INF ist essen gegangen!*gehen ist essen gekommen/*kommen went/came for dinner'

c. d. e.

John Hans Hans Hans 'John

has the Peter make-music hat den Peter musizieren hat den Peter musizieren hat den Peter musizieren has heard/seen/let Peter make

'go' 'come'

VERB-INF/PART hören/gehört 'hear' lassen/gelassen 'let' sehen/gesehen 'see' music'

It might again be tempting to relate the IPP effect to the presence vs. absence of the infinitival marker. However, as can be seen from the distribution of the IPP effect and the infinitival marker summarized in Table 17, the generalization goes only one way. While it is indeed the case that the IPP phenomenon is only found in infinitival constructions lacking the infinitival marker, it is not the case that all infinitives lacking the infinitival marker can occur with IPP morphology. The crucial exceptions are motion verb constructions, which appear without the infinitival marker, however, nevertheless clearly

62

The modal sollen 'shall/should' is best with a conditional auxiliary and somewhat marked in the indicative perfect. We have no explanation for this effect.

166

Functional restructuring

prohibit the IPP effect. Table 17. IPP effect and infinitival marker in German Predicates

Examples

Lexical restructuring verb try, dare, manage Modal verbs must, can, want Auxiliary verbs is, have, will Raising verbs seem, promise Motion verbs go, come let Causatives Perception verbs see, hear

IPP effect

Infinitival marker

NO YES NO NO NO YES YES

YES NO YES YES NO NO NO

We thus conclude here that it is indeed the nature of the auxiliary together with the lexical/functional distinction that drives the distribution of the IPP effect in German. We have no further insight into the question of why the IPP effect occurs and simply conclude here that it is a post-syntactic morphological property that is defined as "PART => INF/ v.func have".63 Note that this conclusion seems unavoidable in a restrictive theory of syntax, since the IPP effect is invisible to the semantics of the constructions involved—i.e., IPP infinitives are interpreted as participles.64 A summary of the properties discussed so far is provided in Table 18. Assuming that the IPP effect is excluded for independent reasons 63

This rule does not capture per se the possibility of the IPP-effect in modal+DP constructions for some speakers (cf. examples such as (120)d). To accommodate these speakers' judgments either the IPP rule has to be modified or it has to be assumed that these speakers project a covert modal. 64 It is sometimes claimed that the IPP effect is caused by different word orders of the verbal elements. However, a careful investigation of the empirical situation reveals that the opposite is true—namely, that word order is fed by the IPP. In Wurmbrand (1999b, 2000b) it is shown that the order of verbal elements is determined by their morphological status (i.e., whether the verbs involved are participles or infinitives; auxiliaries or modals etc.). Importantly, the possible word orders in IPP constructions are the orders one would expect if the IPP-verb is an infinitive (and not a participle). Thus, word order is determined after the IPP effect has taken place. Since the IPP effect is not visible to the semantic component and hence cannot be viewed as a syntactic operation, it is concluded that verbal reordering is also a post-syntactic phenomenon.

Motivating the lexical/fiinctional distinction

167

with raising verbs, auxiliaries, and motion verbs (either because these elements cannot occur in compound tenses or they do not take the auxiliary have), we find again a clear split between lexical and functional predicates in German: lexical predicates do not display the IPP effect whereas (semi-)functional predicates exhibit the IPP effect. In the next section, we will turn to properties that will motivate a distinction between the functional and semi-functional categories. Table 18. Predicates

Extraposition, IPP effect, and infinitival marker in German Examples

Lexical restructuring verb try, dare, manage Modal verbs must, can, want Auxiliary verbs be, have, will Raising verbs seem, promise Motion verbs go, come Causatives let Perception verbs see, hear

Extraposition

IPP

zu 'to'

YES NO NO NO NO NO NO

NO YES N/A N/A N/A YES YES

YES NO YES YES NO NO NO

3.2.3. Thematic properties In this section, we will discuss the thematic properties of the predicates in Table 15. According to a common view, a thematic relation is established between a predicate and an argument when the predicate imposes selectional or thematic requirements on the syntactic or semantic properties of the argument. In this section, we will show that only a subgroup of the functional predicates can be considered as non-thematic. Other functional predicates (i.e., the class of semifunctional predicates), which clearly pattern with functional constructions regarding the syntactic properties discussed in the previous section, enter into thematic relations with an argument of the clause, and hence pattern with lexical verbs in their thematic properties. A first difference between lexical and semi-functional predicates, on the one hand, and functional predicates, on the other hand, is found regarding the compatibility with embedded weather predicates. As is illustrated in (125) through (130), lexical and semi-functional predicates prohibit weather-ΰ subjects, whereas purely functional

168

Functional

restructuring

predicates allow these non- or semi-thematic arguments. All lexical predicates in Table 15—except begin—are incompatible with weather-iY arguments—both subjects and objects (cf. in (125)). Following the widely accepted view (originally suggested in Perlmutter 1970), we consider begin as ambiguous between a raising and a control verb. Furthermore, we will show that in German, the raising vs. control distinction corresponds to the lexical (control) vs. functional (raising) distinction. Thus, despite the apparent difference between begin and the other lexical verbs, the inclusion of begin among lexical restructuring verbs (as well as functional restructuring verbs) will turn out to be unproblematic. (125) Weather-// subjects with lexical restructuring predicates It/John VERB

(him/it)

to snow

zu schneien 'intend' beabsichtigte a. *Es 'begin' zu schneien begann b. Es (ihm) zu schneien 'recommend' c. *Hans empfahl (ihm) zu schneien 'recommend' *Es empfahl (ihm) zu schneien 'allow' d. *Hans erlaubte 'allow' (ihm) zu schneien *Es erlaubte (ihm) zu schneien 'manage' gelang e. *Es (ihm) zu schneien 'permit' gestattete f. *Es (ihm) zu schneien 'permit' *Hans gestattete (ihm) 'fail' *Es mißlang zu schneien g· 'prohibit' (ihm) zu schneien h. *Es untersagte (ihm) zu schneien 'prohibit' *Hans untersagte 'forbid' *Es verbot (ihm) zu schneien i. 'forbid' *Hans verbot (ihm) zu schneien zu schneien 'forget' vergaß j· *Es zu schneien 'avoid' k. *Es vermied zu schneien 'miss' versäumte 1. *Es zu schneien 'try' versuchte m. *Es zu schneien 'dare' n. *Es wagte 'It intended/managed ... to snow' 'John recommended/allowed ... it to snow'

Motivating the lexical/functional distinction

169

Turning to the functional predicates in Table 15, it can be seen in (126) through (127) that weather-iV subjects are licensed with these predicates (for want and would like to see below).65 (126) Weather-it subjects with functional predicates a. b. c. d. e. f. g· h. i. j·

It VERB (to) snow zu schneien Es begann Es drohte zu schneien Es dürfte schneien Es kann schneien Es muß schneien Es pflegte zu schneien Es schien zu schneien Es soll schneien Es versprach zu schneien Es wird zu schneien 'It began/threatened... to snow'

'begin' 'threaten' 'might' 'can' 'must' 'used to' 'seem' 'shall' 'promise' 'will'

Note in particular, that weather-/? subjects are not only compatible with the epistemic interpretations of modal constructions (i.e., modal statements that express a necessity or possibility relative to some state of knowledge or belief), but are also possible (though more marked and context dependent) with the so-called root or deontic interpretations (i.e., modal statements that express forces like permission, obligation etc. relative to some normative system). Modal constructions that favor a deontic interpretation in German (such as the indicative modal dürfen 'may' in (127)a or the periphrastic modal haben zu 'have to' in (127)b) can occur with weather-/? subjects in appropriate contexts. Similarly, Hackl (1998) has shown that root modals like ability can in (127)c are compatible with weather-/? subjects. Hackl gives the following evidence for the claim that examples such as the one in (127)c indeed involve an ability reading and not only an epistemic interpretation. He notes (following a suggestion by 65

Weather predicates are independently excluded in the fre+infinitive construction (and hence ignored for this test) since be can only combine with passive or unaccusative predicates in this construction.

170

Functional restructuring

S. Iatridou) that epistemic modals are pragmatically odd if it is clear to all participants in the discourse that the proposition in question is true in the actual world. That is, in the context in (128), a modal statement with a typical epistemic modal like dürfte 'might' is pragmatically ill-formed (cf. (128)a). Since no such effect arises with können 'can' in (128)b, it is reasonable to assume that the sentence in this context is interpreted with a root interpretation. (127) Weather-it subjects with root modals a.

Es darf morgen einfach nicht schneien It may tomorrow just not snow 'It just can't snow tomorrow' b. lEs hat morgen einfach zu schneien It has tomorrow just to snow 'It just has to snow tomorrow' c. Es kann hier sogar schneien It can here even snow 'It can even snow here' (128) Epistemic vs. root modals Context:

It is snowing heavily and everybody involved in the discourse is looking out the window

a. #Oh\ Es dürfte hier Oh! It might here 'It might even snow here' b. Ohl Es kann hier Oh! It can here 'It can even snow here'

sogar even

schneien snow

sogar even

schneien snow

Finally, semi-functional predicates, like lexical predicates, prohibit weather-¿i subjects (cf. (129)). Causatives and perception verbs, however, allow weather-/? objects, which is compatible with the standard assumption that these predicates do not assign a theta role to the "object", but rather the surface object is the thematic subject of the embedded predicate (but see section 4.3).

Motivating the lexical/functional

distinction

171

(129) Weather-/r subjects in semi-functional constructions It VERB snow a. *Es geht schneien b. *Es kommt schneien 'It goes/comes snow' It/John-NOM VERB Peter-ACC/it den Peter c. *Es hört es Hans hört den Peter d. *Es lässt Wans lässt es den Peter sah e. *Es es Wans sah 'It heard Peter thunder'; 'John heard it

'go' 'come' thunder/snow 'hear' donnern 'hear' donnern donnern 'let' donnern 'let' schneien 'see' schneien 'see' thunder'

Note that some of the functional predicates in Table 15 are also listed among the class of semi-functional predicates. The predicates that are ambiguous in this sense are the modals können 'can', wollen 'want', and möchte 'would like to' which have been shown be many authors to form a third class of modal constructions—usually referred to as dynamic modals—that cannot be subsumed under the epistemic vs. root/deontic distinction (see for instance Palmer 1986; Brennan 1993; Warner 1993). For our purposes here, it is sufficient to mention that these constructions are somewhat in-between functional and semifunctional constructions in many respects, which can be accounted for by cross-listing these constructions in both categories. As is illustrated in (130), weather-iY subjects are in principle possible with want and would like to, but only in very restricted contexts. However, the fact that there is a clear contrast between (129) and (130) indicates that these predicates nevertheless have to be distinguished from semi-functional predicates. (130) Weather-iY subjects with (semi-)functional predicates a. llEs möchte hier doch bitte bald einmal schneien It would-like-to here but please soon for-once snow Ί [speaker] would like it if it would snow here soon'

172

Functional restructuring

b. ?Es will hier einfach nicht It wants here just not 'It just does not want to snow here'

schneien snow

The contrasts regarding the compatibility of weather-// subjects have received widespread attention in the literature and although there is dispute about the theoretical implementation of the contrasts found in the examples above, most authors agree that the differences noted above have to do with the fact that only some of the predicates in (125) through (130) impose thematic restrictions on their argument(s) which have to be met at some level of representation (though not necessarily in the syntax). Thus, under any account, the illformedness of the examples in (125) and (129) is in some way or another attributed to the fact that weather expressions cannot satisfy the thematic restrictions of the predicates involved. A second property that is attributable to the fact that certain but not all predicates enter into thematic relations with their argument(s) is the (im)possibility of inanimate subjects. As is shown in (131), the lexical predicates in Table 15 (except begin) are incompatible with inanimate subjects and objects—i.e., these predicates impose a thematic animacy or agentivity requirement on their arguments. (131) Inanimate arguments with lexical predicates a. b. c. d. e. f. g· h. i. j· k. 1.

eaten to AUXPASS The cake VERB zu werden 'intend' *Der Kuchen beabsichtigte gegessen Der Kuchen begann gegessen zu werden 'begin' * Der Kuchen empfahl gegessen zu werden 'recomm.' * Der Kuchen erlaubte gegessen zu werden 'allow' gegessen zu werden 'permit' * Der Kuchen gestattete * Der Kuchen untersagte gegessen zu werden 'prohibit' * Der Kuchen verbot gegessen zu werden 'forbid' * Der Kuchen vergaß gegessen zu werden 'forget' * Der Kuchen vermied {es) gegessen zu werden 'avoid' * Der Kuchen versäumte gegessen zu werden 'miss' * Der Kuchen versuchte gegessen zu werden 'try' * Der Kuchen wagte gegessen zu werden 'dare' 'The cake intended/began. .. to be eaten'

Motivating the lexical/functional distinction

173

John VERB the cake-DAT eaten to AUXPASS m. *Hans empfahl dem Kuchen gegessen zu werden n. *Hans erlaubte dem Kuchen gegessen zu werden *Hans gestattete dem Kuchen gegessen zu werden untersagte dem Kuchen gegessen zu werden P· *Hans *Hans dem Kuchen gegessen zu werden verbot q· 'John {recommended to/allowed/permitted/prohibited/forbade} the cake to be eaten' 0 .

The c a k e - D A T VERB (it) eaten r. * Dem Kuchen gelang (es) gegessen s. *Dem Kuchen mißlang (es) gegessen 'The cake managed/failed to be eaten'

to A U X P A S S zu werden zu werden

Similarly, semi-functional predicates pattern again with lexical predicates in that inanimate subjects are excluded (cf. (132)). Inanimate embedded subjects are possible in causative and perception verb contexts, however, as we will see in section 4.3, embedded passive is not allowed in semi-functional constructions. Thus, the examples in (132)a-e are ruled out on independent grounds. However, (132)f-j show that animacy is a requirement for subjects in semifunctional constructions but not for objects. (132) Inanimate arguments with semi-functional predicates The cake a. b. c. d. e.

VERB

eaten

AUXPASS

*Der Kuchen geht gegessen werden *Der Kuchen kommt gegessen werden *Der Kuchen hört gegessen werden *Der Kuchen sieht gegessen werden *Der Kuchen läßt gegessen werden 'The cake goes, comes. .. being eaten'

The boat VERB in-the water float f. *Das Boot geht im Wasser treiben *Das Boot kommt im Wasser treiben g· 'The boat goes/comes float in the water'

'go' 'come' 'hear' 'see' 'let'

'go' 'come'

174

Functional restructuring

h. i. j.

John Hans Hans Hans 'John

VERB the boat in-the water sieht das Boot im Wasser hört das Boot im Wasser läßt das Boot im Wasser sees/hears/lets the boat float in the

float treiben treiben treiben water'

'see' 'hear' 'let'

In contrast, (full) functional predicates do not impose a thematic restriction such as animacy or agentivity on the subjects in these constructions (cf. (133)). Finally, predicates that are classified as either functional or semi-functional predicates allow inanimate subjects, however, again only in very special contexts (cf. (134) which is considered as stilted or archaic for most speakers). (133) Inanimate arguments with functional predicates a. b. c. d. e. f. g· h. i. j· k.

The cake VERB eaten IDer Kuchen begann gegessen IDer Kuchen drohte gegessen Der Kuchen darf gegessen gegessen Der Kuchen dürfte gegessen Der Kuchen hat gegessen Der Kuchen kann Der Kuchen muß gegessen gegessen Der Kuchen soll IDer Kuchen versprach gegessen Der Kuchen wird gegessen 'The cake began/threatened... to be

(to) AUXPASS

zu werden 'begin' zu werden 'threaten' werden 'may' werden 'might' zu werden 'have to' werden 'can' werden 'must' werden 'shall' zu werden 'promise' werden 'will' eaten'

Der Kuchen ist zu essen The cake is to eat 'The cake can/has to be eaten' 1. Der Kuchen pflegte vor dem Essen gegessen zu werden The cake used before dinner eaten to AUXPAS¡ 'The cake used to be eaten before dinner' m. Der Kuchen scheint schon gegessen worden zu sein The cake seems already eaten been tobe 'The cake seems to have been eaten already'

Motivating the lexical/functional distinction

175

(134) Inanimate arguments with (semi-)functional predicates a.

Der Kuchen möchte doch bitte gegessen werden The cake would-like but please eaten AUXPASS 'Would you [addressee] please eat the cake' b. Der Rasen möchte doch bitte gemäht werden The lawn would-like but please mown AUXPASS 'Would you [addressee] please mow the lawn' c. ΊDieser Kuchen will einfach nicht gekauft werden This cake wants simply not bought A U X P A S S 'People just don't want to buy this cake' To conclude, the predicates in Table 15 fall into two groups: predicates that do and predicates that do not impose thematic restrictions on their argument(s). We have seen that this split only partially corresponds to the lexical/functional distinction. Lexical predicates impose thematic restrictions on their arguments. However, a small class of functional predicates—semi-functional predicates (which clearly qualify as functional predicates regarding extraposition and the IPPphenomenon)— pattern with lexical predicates in that they impose lexical restrictions on their subjects. The properties discussed so far are summarized in Table 19. Table 19. Extraposition, IPP effect, and infinitival marker with functional and lexical predicates in German Predicates

Ex.

Position

Extraposition IPP zu 'to' Θ-restrictions

Lexical restructuring Modal verbs Auxiliary verbs Raising verbs Motion verbs Causatives Perception verbs

try must be seem go let see

V Aux/Mod/v Aux/Mod Aux/some V ν ν ν

YES NO NO NO NO NO NO

NO YES N/A N/A N/A YES YES

YES NO YES YES NO NO NO

YES YES/NO NO NO YES YES YES

Assuming the clausal architecture in (109), the distribution of the properties in Table 19 reflects the fact that different properties target different aspects of the structure. Thematic restrictions only occur with verbal heads of the thematic domain (i.e., V and v-heads). Ex-

176

Functional restructuring

traposition and the IPP effect, on the other hand, target the lexical vs. functional distinction (i.e., V-heads vs. all other categories). The distribution of the infinitival marker is not related to any syntactic or semantic property and as discussed in Chapter 2, section 4.3, has thus to be treated as a pure selectional property of the matrix predicate. Before turning to a detailed discussion of the constructions listed in Table 19, we will briefly discuss an alternative approach to restructuring, namely an approach that considers all restructuring predicates as functional categories and show that the approach suggested here is empirically and theoretically superior.

3.3. Restructuring and functional structure (Cinque 2000) In this final subsection, we will briefly discuss Cinque's approach to restructuring which takes all restructuring to be functional. In earlier work (e.g., Cinque 1999), Cinque claims that clauses are built according to a universal hierarchy of functional projections which host adverbs in their specifiers and verbal categories in their head positions. To capture the observation that the order of adverbs in a clause and the order of restructuring predicates are subject to similar restrictions, Cinque (1997a, 1997b, 2000) suggests that all restructuring predicates are functional categories in a more elaborate clausal architecture. Assuming that (all) restructuring verbs are functional categories, Cinque (2000) is led to the following conclusions: i) restructuring verbs cannot take internal arguments; ii) all restructuring verbs are non-thematic (i.e., they are essentially raising verbs); and iii) restructuring is obligatory (but transparency properties such as clitic climbing are optional). In this section, we will show that all three claims are untenable in light of restructuring in German. First, as we have seen in Chapter 2, section 2-3, certain ditransitive verbs (i.e., recommend, allow, forbid) allow long passive (provided the infinitive receives a simultaneous tense interpretation). Relevant examples are given in (135)a,b. Furthermore, the unaccusative verb manage does not take an external nominative argument in German but selects for an internal dative argument. Nevertheless, long object movement (hence restructuring) is possible (cf. (13), re-

Motivating the lexical/functional

distinction

177

peated as (135)c). Thus, the claim that restructuring predicates cannot select internal arguments is clearly falsified by the examples in (135). (135) The availability of internal arguments a.

Dem Kind wurden nur Kekse zu essen erlaubt the child-DAT were only cookies to eat allowed 'They only allowed the child to eat cookies' b. Der Roman wurde ihm zu lesen empfohlen the novel-NOMwas him to read recommended 'They recommended to him to read the novel' c. Iweil mir der Brief auf Anhieb since me-DAT the letter-NOM straightaway zu entziffern gelungen ist to decipher managed is 'since I managed straightaway to decipher the letter' The second consequence of a unified treatment of restructuring verbs as non-thematic functional categories is that according to Cinque, all restructuring verbs are raising verbs. As we have seen in the previous section, however, this claim is problematic for verbs like try, dare, forget, and the semi-functional categories. The examples in (136) illustrate one of the major differences between non-thematic verbs and thematic verbs—the possibility vs. impossibility of quasi arguments and expletive subjects. (136) Raising vs. control a. *There tried to be a unicorn in the garden b. There seems to be a unicorn in the garden c. *It tried to snow d. It seems to be snowing Note that this problem is independent of the question of whether control is analyzed as raising as suggested by Hornstein (1999), since even under this approach, a difference in the thematic properties between "control" verbs and raising verbs is maintained (i.e., ac-

178

Functional

restructuring

cording to Hornstein, verbs like try do establish a thematic relation with the subject, whereas verbs like seem are non-thematic). Furthermore, lexical predicates differ crucially from (what we call) functional predicates regarding the passive properties found in these constructions. As we will discuss in detail in section 4.1.2.2, lexical restructuring verbs allow long passive (i.e., passive of the matrix predicate and long object movement), whereas passive of functional predicates is generally illicit (cf. the contrast between (137)a and (137)b). On the other hand, passive of the embedded predicate (and long object movement) is possible in functional constructions, whereas this operation is prohibited in lexical restructuring constructions (cf. (137)c vs. (137)d). (137) Passive in control vs. raising constructions a.

Der Wagen wurde zu reparieren versucht The car-NOM was to repair tried They tried to repair the car' b. *Der Wagen wurde (zu) reparieren gemußt/müssen must-PART/INF The car-NOM was (to) repair 'They had to repair the car' c. *Der Wagen versuchte repariert zu werden The car-NOM tried repaired t o A U X P A S S 'They tried to repair the car' d. Der Wagen muss repariert werden The car-NOM must repaired A U X P A S S 'The car must be repaired' The contrasts in (137) follow straightforwardly from the assumption that verbs like try establish a thematic relation with their subjects, whereas verbs like must are true raising verbs that do not assign a theta-role to the subject. Since the car cannot fulfill the theta requirement of try and since raising verbs generally block passivization (see section 4.1.2.2), raising is impossible in (137)b and passive is impossible in (137)c. In order to maintain the claim that all restructuring predicates are non-thematic raising predicates, a different explanation is required to account for the contrasts in (136) through

Motivating the lexical/functional distinction

179

(137). In this respect, various authors (including Prinzhorn 1987, 1990, Roberts 1997, Cinque 2000) have followed Zubizarreta (1982) in claiming that lexical and semi-functional restructuring predicates assign adjunct theta roles rather than true theta-roles. In particular, it has been suggested that restructuring predicates while not projecting arguments on their own nevertheless impose thematic restrictions on certain arguments in the clause (and that these adjunct theta roles are not subject to the Theta Criterion). There are two main reasons why we think that this approach is unsatisfactory. First, it is important to note that these adjunct theta roles have to be assigned obligatorily. If adjunct theta roles were optional, examples such as (136)a,c and (137)c would wrongly be predicted to be possible. The obvious question then is what the status of adjunct theta roles is, and in particular, how the difference between true theta roles and adjunct theta roles is implemented in the grammar and how it can be predicted. Since the postulation of adjunct theta roles as suggested in the works above is not motivated by any thematic or syntactic difference between the two types of theta roles (note that thematically, try does not differ in any obvious way from decide, whereas both predicates differ clearly from raising predicates such as seem), it appears that the concept of adjunct theta roles does no more than save an account of restructuring that is built on the assumption that all restructuring verbs are non-thematic raising verbs. Secondly, as we have seen in the previous section, lexical and functional restructuring infinitives show a number of syntactic differences (such as extraposition and the IPP effect) which can be related to a difference in the syntactic nature of these categories. Under the assumption that all restructuring predicates are functional elements, however, these differences would be left unexplained. The final problem with Cinque's account in light of restructuring in German is that it makes exactly the opposite prediction concerning the optionality of restructuring and certain restructuring properties. Let us recapitulate some of the crucial examples (see Chapter 2, section 2.3 for detailed discussion). In (138)a, we see that an object embedded in an infinitive that combines with a passivized (potential) restructuring verb can bear accusative case in German. The conclusion we have drawn is that examples of this sort are instances of non-

180

Functional

restructuring

restructuring infinitives—i.e., they are at least vPs and hence object case is assigned inside the infinitive. The supporting evidence comes from the fact that scrambling (i.e., another restructuring property) is impossible whenever structural case is assigned inside the infinitive (cf. (31)a, repeated as (138)b). Thus, our crucial claim is that restructuring with lexical restructuring verbs is optional and that these predicates can combine with different types of infinitival complements, which then display different syntactic and semantic properties.66 Under Cinque's approach, on the other hand, restructuring is considered as an obligatory result of the universal clausal architecture suggested in his framework (i.e., the functional status of all restructuring predicates which does not provide room for optionality as Cinque argues explicitly), and hence (138)a would have to be analyzed as a restructuring infinitive. Under this assumption, long passive would then have to be seen as an optional operation since it does not apply in (138)a. The problem, however, that this system is faced with at this point is the ungrammaticality of (138)b. Assuming (as Cinque would have to) that (138)a is a restructuring infinitive, nothing seems to prohibit (138)b, since scrambling should be licensed in restructuring infinitives (as we have shown in Chapter 2, section 2.3, scrambling is possible from extraposed restructuring infinitives). (138) Restructuring vs. non-restructuring infinitives a.

dass that 'that b. *dass that 'that

versucht wurde den Traktor tried was the tractor-ACC they tried to repair the tractor' den Traktor versucht wurde the tractor-ACC tried was they tried to repair the tractor'

zu reparieren to repair zu reparieren to repair

The German facts suggest that restructuring is an 'all or nothing' phenomenon. For any given sentence, either all tests targeting a particular restructuring level come out positive or all tests come out 66

See also Chapter 6 for a detailed comparison of the different structures possible with lexical restructuring verbs and a summary of the motivating evidence.

Motivating the lexical/functional distinction

181

negative (see Chapter 5 for a characterization of what counts as tests for a particular restructuring level). This property is predicted if (lexical) restructuring itself is optional, but it seems that it cannot be explained if restructuring is obligatory and each of the diagnostics is optional.67 Thus, the claim that restructuring is obligatory and that any optionality is due to an optional application of the restructuring properties seems to be seriously challenged again by what we have analyzed as lexical restructuring constructions. Assuming, on the other hand, that verbs like try are lexical predicates that can combine with different types of infinitival complements, the correlations between different properties (see also Chapter 6) follow straightforwardly. To conclude, in this section, we have shown that German restructuring constructions motivate a syntactic distinction between lexical and functional predicates. We have argued that lexical predicates are main verbs that establish thematic relations with their arguments, whereas full functional predicates are non-thematic categories. Furthermore, we have seen that there is a class of semi-functional predi67

Part of Cinque's motivation for the (rigid) functional status of restructuring predicates comes from the prohibition against using the same adverb twice in constructions in which restructuring effects are found (cf. i. which is Cinque's (23)b). However, as Cinque notes, the ungrammaticality is only found when transparency facts obtain; when operations such as clitic climbing do not apply, a structure with more than one occurrence of the same adverb is licensed (cf. ii; Cinque's (23)a). i.

ii.

* Maria Mary 'Mary Maria Mary 'Mary

lo vorrebbe già aver già lasciato him-CL would-want already have already left would already want to have already left him* vorrebbe già averlo già lasciato would-want already have-him-CL already left would already want to have already left him'

Assuming that i. is a (functional) restructuring construction, whereas ii. is a nonrestructuring construction, the facts in i. and ii. are as expected. However, under Cinque's account the optionality of restructuring is rejected and hence both of the examples above would be restructuring constructions. The question that arises for Cinque's system is then why there should be a difference between the two examples. Thus, it appears that even for Italian, the claim that restructuring is obligatory is too strong.

182

Functional

restructuring

cates which display both properties of functional categories as well as properties of lexical categories. Assuming a clausal architecture that divides a clause into a lexical vs. a functional domain on thematic grounds provides room for this three-way distinction, since it allows for functional thematic heads such as voice or aspect. In the next section, we will investigate different functional constructions in German in more detail and provide further support for the clausal architecture as suggested in section 2.2.

4. German functional constructions 4.1. Modal constructions The syntax and semantics of modal constructions has received extensive discussion in the literature and we can only peripherally touch on most of the issues under debate. Following the common classification of modal constructions we assume that there are three basic types of modal interpretations: epistemic, root or deontic, and dynamic (see for instance Lyons 1977; Kratzer 1977, 1981, 1991; Palmer 1979, 1983, 1986; Barbiers 1995; Brennan 1993; Warner 1993). The lines between these classes are often not easy to draw since modals are highly context sensitive and most modal constructions can be associated with a number of modal bases and conversational backgrounds (cf. Kratzer 1991). We will not discuss here how the different modal interpretations arise but rather concentrate on the syntactic aspects of modal constructions in German. In essence, we will argue for two crucial claims. First, modal constructions (with the exception of dynamic modals) are raising configurations—i.e., both epistemic and root/deontic modal statements are non-thematic constructions. Second, we will show that a split syntactic modal domain allows us to account for various ordering and co-occurrence restrictions found in modal constructions. The structure we suggest for German modal constructions is depicted in (139) (throughout this section we ignore TP and AgrSP, however, the reader is asked to keep in mind that projections hosting tense and

German functional constructions

183

agreement morphemes are present in the constructions to be discussed). As can be seen in (139), dynamic modals are generated in the voice/aspect head and are hence theta-assigners, however, at the same time also functional categories. The connection between dynamic modal and voice will be further motivated by certain competition effects with other voice properties. Root/deontic and epistemic modals, on the other hand, occur in the lower Mod head or the higher Aux head, respectively. Both positions are non-thematic positions—i.e., we will argue that in neither type of modal construction does the modal establish a thematic relation with an argument (the subject). (139) Modal constructions in German AuxP

root/deontic

Note finally, that we do not assume that there are three (or more) distinct lexical modal categories. Rather, following a Kratzer-style approach to modality, we assume that the actual modal verbs are unambiguous but that different interpretations arise as a result of different semantic computations. Similarly, the ordering in (139) does not mean that a particular modal element can only be inserted in a particular position; rather (most) modals can appear in any of the three positions, however, the different positions will then correspond to different syntactic properties and different interpretations. A question

184

Functional

restructuring

that we will not be able to answer here is why there is a syntactic ordering of modal meanings (in particular between epistemic and deontic interpretations) as in (139). Ultimately, we hope that this ordering (which appears to be quite robust across languages; cf. Cinque 1999, 2000) can be derived from some semantic aspect of these modal meanings. However, at this stage, it seems that this question has to be left open.

4.1.1. Some ordering issues Let us start with a discussion of (im)possible modal and auxiliary combinations. As was noticed by Picallo (1985, 1990) for Catalan, epistemic modals cannot appear in a perfective construction. The same is true for German. While a sentence like (140)a is ambiguous between an epistemic and a root interpretation, the same sentence involving a modal in the scope of the perfective auxiliary have (cf. (140)b) does not allow an epistemic interpretation but can only be interpreted with a root interpretation. The same effect is found when modals are embedded under the future auxiliary—i.e., (140)c can only receive a root interpretation for the modal. (140) Perfective modal constructions a.

Sue muß gerade zu Hause arbeiten Sue must just at home work 'Sue is obliged to work at home right now' 'It must be the case that Sue is working at home right now' b. Sue hat zu Hause arbeiten müssen Sue has at home work must-IPP 'Sue had an obligation to work at home' *'It must have been the case that Sue worked at home' c. Morgen wird er Kartoffel schälen müssen Tomorrow will he potatoes peel must 'Tomorrow, he will be obliged to peel potatoes' *'Tomorrow, it will have to be the case that he will peel potatoes'

German functional constructions

185

Assuming that epistemic modals and auxiliaries compete for the same position in German—the Aux head—this effect is straightforwardly accounted for (but note again that we do not have an answer for the question of why there is an ordering along the lines suggested). A simplified structure is provided in (141). (141) Epistemic modals and auxiliaries AuxP

If, on the other hand, auxiliaries take lower scope than the modal, an epistemic reading becomes possible again. The sentences in (142) which favor epistemic readings (though do not force them) illustrate this point. The auxiliaries have and be can only take scope over the main verbs (eat and go); and they obligatorily appear in the scope of the epistemic modal. The examples in (142) thus show that auxiliaries can appear in either the Mod head or the Aux head. (142) Auxiliaries under modals a.

b.

Moel Moel 'Moel Moel Moel 'Moel

muß die Oliven gegessen must the olives eaten must have eaten the olives' muß nach Hause gegangen must to home gone must have gone home'

haben have sein be

The assumption that epistemic modals are in the highest position of the modal domain predicts further that a clause can involve at most one epistemic modal. This prediction is borne out as shown by the examples in (143). If a modal is embedded under an epistemic modal

186

Functional restructuring

like might ((143)a, the lower modal can only be interpreted as a root modal. Furthermore, since might only allows an epistemic interpretation, it also follows from the structure in (141) that this modal cannot be embedded under any other modal or auxiliary (cf. (143)b,c). (143) Epistemic might a.

Er dürfte zu Hause sein müssen He might at home be must 'He might have to be at home' *'It might be that it must be the case that he is at home' b. *Er wird wieder singen dürften He will again sing might 'It will be the case that he might sing again' c. *Er muß wieder singen dürften He must again sing might 'It must be the case that he might sing again' Finally, the order between root and dynamic modals is supported by the general preference of embedding dynamic modals under root modals and not vice versa (cf. (144)a vs. (144)b,c). It should be pointed out that these kinds of ordering effects are slightly weaker since in many cases, the interpretations cannot be easily distinguished. Furthermore, since most modals can be interpreted as root, epistemic, or dynamic modals, the sentences as such are grammatical. The fact that speakers prefer (144)a over (144)b,c is expected since the modals appear in the correct order (i.e., root over dynamic) under the unmarked interpretation of the modals involved in the former but not the latter. In (144)b,c, the structure is only felicitous when the modal can is used in the more marked root or epistemic interpretation, which is possible but requires a more elaborate context. (144) Root and dynamic modals a.

Er muß bis morgen schwimmen He must by tomorrow swim 'He must be able to swim by tomorrow'

können can

German fiinctional constructions

187

b. lEr kann morgen schwimmen müssen He can tomorrow swim must 'It is possible that he will have to swim tomorrow' *'He is able to have to swim tomorrow' c. lEr kann morgen schwimmen wollen He can tomorrow swim want 'It is possible that he will want to swim tomorrow' 'He is allowed to want to swim tomorrow' *'He is able to want to swim tomorrow' To sum up, the assumption that epistemic modals are in a position which is higher than the position of root modals which in turn is higher than the position of dynamic modals allows us to account for various ordering restrictions among these elements. We would like to emphasize again that it is very likely that the ordering among modal projections/interpretations is caused by a deeper property of grammar related to the interpretation of modal constructions (note for instance that similar markedness effects arise across clauses in English with periphrastic modal constructions where the syntactic ordering is not an issue). It should thus be kept in mind that the syntactic ordering in German is merely a reflection (rather than the cause) of this yet to be discovered semantic or pragmatic property of modal constructions. In the next sections we will discuss further implications of the structure suggested for the thematic properties, the position of the subject, and the passive properties in modal constructions. 4.1.2. Modal verbs must be raising verbs A common view concerning modal constructions is that different modal interpretations correspond to different syntactic structures. Ross (1969) for instance assumes that root modals are transitive (i.e., assign two theta-roles—an internal theta-role for the infinitive and a subject theta-role), whereas epistemic modals are intransitive (i.e., they assign only a theta-role to the infinitive). Similarly, Roberts (1985) following Zubizarreta (1982) argues that epistemic modals do not assign a subject theta-role, but deontic modals assign a(n adjunct)

188

Functional restructuring

theta role to the surface subject. Assuming a (strong) version of the Theta Criterion (as formulated in Chomsky 1981: 36) that specifies a one-to-one relation between arguments and theta-roles, root modals under these premises have to be represented by a structure such as (145)a. Since the modal and the lower verb each assign a subject theta-role, two syntactic arguments have to be present—the surface subject and a silent infinitival subject (e.g., PRO). In a raising structure such as (145)b, on the other hand, only one subject theta-role is assigned, and hence, only one argument is present in syntax. (145) a.

Control ModP

b.

Raising ModP

We will show in this section that the control/raising distinction is unmotivated for epistemic and root modal constructions and that the properties of modal constructions point to the conclusion that root as well as epistemic modals are represented by a raising structure (see also Vanden Wyngaerd (1994) and Bhatt (1998) who reach the same conclusion). Thus, we will argue for the structure in (145)b, which is a simplified version of the structure suggested in (139). 4.1.2.1.

The subject starts out below the modal

The first set of arguments for the structure in (145)b comes from phenomena that indicate that the subject in modal constructions starts out below the modal verb. Under a raising structure such as (145)b, this fact is straightforwardly accounted for. However, under a control structure such as (145)a, these phenomena would be left unexpected. Let us begin with expletive constructions. As has been pointed out by a number of authors, modal constructions are compatible with ex-

German functional constructions

189

pletive subjects (cf. (146)). Importantly, all the examples in (146) involve root interpretations (see also Brennan 1993; McGinnis 1993; Warner 1993: 16 for (146)a; Kulick 1998).68 Since the subject (i.e., the associate of there) appears inside the infinitival complement, these examples strongly favor a raising analysis for root as well as epistemic modal constructions. (146) Expletive subjects in modal constructions a. b. c. d.

There may be singing but no dancing on my premises There can be a party as long as it's not too loud There must be a solution to this problem on my desk, tomorrow morning! There will be no complaints when we go to Aunt Mary's!

The second argument we would like to discuss is drawn from the case of subjects in Icelandic modal constructions. The unmarked case for Icelandic subjects is nominative. However, certain verbs require a non-nominative (quirky case marked) subject: the verb lack can only occur with an accusative subject, and the verb like requires a dative subject (cf. (147); examples (147)-(149) except (148)c are from Thráinsson and Vikner 1995: 60). (147) Icelandic: quirky case a.

b.

Harald / *Haraldur Harold-ACC/ *Harold-NOM 'Harold lacks money' Haraldi / *Haraldur Harold-DAT/ *Harold-NOM 'Harold likes it in Stuttgart'

vantar peninga lacks money líkar vel likes well

í Stuttgart in Stuttgart

When verbs that require quirky case marked subjects are embedded in a control construction, the case of the matrix subject is determined by the higher verb—i.e., the verb the subject is associated with. Thus, Dynamic modals do not license expletive subjects (cf. Brennan 1993). Since we assume that dynamic modals are semi-functional categories which do assign a theta-role to the subject this fact is unproblematic for our analysis.

190

Functional restructuring

if the higher verb is not a quirky case assigner as in (148), the subject is realized with nominative.69 In raising constructions, on the other hand, the subject is only associated with the lower predicate, and hence case is determined by the lower verb (cf. (148)c; if the lower verb is not a quirky case assigner, the subject shows up with nominative). (148) Icelandic: control vs. raising constructions a.

b.

c.

Haraldur / *Harald vonast Harold-NOM / *Harold-ACC hopes aö vanta ekki peninga to lack not money 'Harold hopes not to lack money' Haraldur / *Haraldi vonast Harold-NOM / *Harold-DAT hopes aô líka vel í Stuttgart to like well in Stuttgart 'Harold hopes to like it in Stuttgart' Harald viröist vanta ekki Harold-ACC seems lack not 'Harold seems not to lack money'

til for

til for

peninga money

Icelandic thus provides a way to distinguish between raising and control constructions—if quirky case is retained, the construction involves raising, if the subject shows the case that the higher verb assigns, the construction involves control. As we will see, modal constructions support the first view. When verbs requiring quirky case marked subjects are embedded under a modal (cf. (149)), the subject has to show up with quirky case and nominative is ungrammatical (if the lower verb does not require a quirky case marked subject, the subject again shows up with nominative).

69

Although not visible on PRO, Sigurösson (1991) provides evidence that quirky case is retained on the embedded infinitival subject. He shows that floating quantifiers associated with PRO show up with quirky case (i.e., the case the embedded verb would assign to an overt subject).

German functional constructions

191

(149) Icelandic: epistemic modal constructions and quirky case a.

b.

Harald / *Haraldur vili vanta peninga Harold-ACC/ *Harold-NOM will lack money 'Harold tends to lack money' Haraldi / *Haraldur cetlar Harold-DAT / *Harold-NOM intends aö líka vel í Stuttgart to like well in Stuttgart 'It looks like Harold will like it in Stuttgart'

Thráinsson and Vikner (1995) claim that the examples in (149) allow only epistemic readings. However, we believe that this effect is caused by the unnaturalness of a deontic interpretation in these examples. If the context is constructed in a way that favors a root/deontic reading as in (150), the examples are grammatical and again, only quirky case is possible for the subject (examples from Ólafur P. Jonsson, Ásta Sveinsdóttir p.c.).70 (150) Icelandic: root modal constructions and quirky case a.

b.

Haraldi l*Haraldur veròur aö líka hamborgarar Harold-DAT/*Harold-NOMmust to like hamburgers 'Harold must like hamburgers' (in order to be accepted by his new American in-laws) Umscekjandann veròur aö vanta peninga The-applicant-ACC must to lack money 'The applicant must lack money' (to apply for this grant)

Assuming that quirky case is not assigned structurally but rather determined idiosyncratically by certain predicates, Icelandic provides strong support for the claim that in epistemic as well as root modal constructions, the surface subject is associated with the lower verb at 70

It is noteworthy that some speakers accept a nominative subject with the verb seem in (148)c. However, as indicated in the text, nominative is clearly impossible in modal constructions (for all speakers). Thus, while the. verb seem could be seen either as a control verb or a raising verb (as for instance in Italian), modal constructions unambiguously qualify as raising constructions.

192

Functional restructuring

some point in the derivation. The assignment of quirky case in (150) follows straightforwardly from a raising structure for modal constructions since the surface subject starts out as the subject of the lower predicate and as such can be assigned quirky case. Under a control structure for modal constructions, the Icelandic case pattern (especially the contrast between (148) and (150)) is not expected. The third argument for a low position of the subject in modal constructions comes from the scope properties of subjects in modal constructions. Since May (1977, 1985) it has been known that control and raising structures differ with respect to the following scope property. Raising constructions but not control constructions allow an interpretation in which the subject takes narrow scope with respect to the matrix verb. In examples like (151)a the subject could take higher scope, yielding an interpretation like There is somebody from N.Y., and it is likely that he will win the lottery; or the subject could take lower scope, yielding the interpretation It is likely that somebody from Ν. Y. will win the lottery (the latter is more natural in an unmarked context). Importantly, no such effect is found in control contexts such as (151)b (i.e., an interpretation such as It was hoped that somebody from New York would win the lottery is not available for (151)b). (151) Scope in raising vs. control constructions a. b.

Someone from New York is likely to win the lottery Someone from New York hoped/wanted/decided to win the lottery

As has been argued by many authors, this contrast is due to the availability of a lower position for the subject at LF in raising constructions but not in control constructions. That is, the low scope reading is achieved either by reconstruction or by simply choosing the lower copy of the movement chain (cf. May 1977, 1985; Lebeaux 1995; Fox 1998/2000, 1999; Sauerland 1998a, 1998b; Bobaljik and Wurmbrand 1999, 2001 for arguments for this approach). Simplified structures are given in (152).

German functional constructions

(152) a.

Control

b.

IP

193

Raising IP

SUBÍ^^VP V°

SUBÍ^^VP INF

Vo

PRÖ^^VP

t«,», '•S^BJ

LF:

X

VP

Vo " ^ ^ O B J

V ^^OBJ SUBJ»VERB

vP

LF:

SUBJ»VERB»SUBJ

Turning now to modal constructions, a raising structure for modal constructions as in (145)b or (152)b makes the prediction that—as in seem-type raising structures—two positions should be available in which the subject can be interpreted. The predicted scope ambiguity extends straightforwardly to epistemic modal constructions which are uncontroversial raising contexts. As is illustrated in (153), a modal statement such as (153)a is ambiguous. If the subject is interpreted in the higher position, it takes scope over the modal and we get the interpretation in (153)b. The perhaps pragmatically more natural interpretation, however, is the interpretation in (153)a in which the subject takes lower scope. (153) Epistemic modals and scope a. b. c.

Somebody from New York must have won in the lottery In view of the evidence available it is necessarily the case that somebody from N.Y. won the lottery There is somebody from N.Y. and in view of the evidence available it is necessarily the case that he won the lottery

What is important for the discussion here, however, is that the same ambiguity is found with root/deontic modals. As is illustrated in (154), root modal contexts are also ambiguous—the subject can take scope over or under the modal. Like in seem-type raising constructions, context and knowledge of the world favor different readings. In (154)a, the wide scope reading for the subject in (154)a.ii is less

194

Functional restructuring

marked, since ski races are generally won by one person only. In (154)b, on the other hand, the lower scope reading for the subject in (154)b.i is less marked, since for a country to win the most gold medals does not require that specific racers win the medals. (154) Root/deontic modals and scope a.

Two Austrian skiers must win the next race ( in order for either of them to win the World Cup) i. #It is necessary that two Austrians win the next race ii. There are two Austrians and for each of them it is necessary to win the next b.

An Austrian must win the next race ( in order for Austria to have the most gold medals) i. It is necessary that an Austrian (whoever it is) win the next race ii. #There is an Austrian and it is necessary that he win the next race Under a raising structure for modal constructions, the ambiguity in (154) is readily accounted for. To maintain a control structure as in (145)a or (152)a, it would have to be assumed that modal verbs start out in the VP (an assumption that seems to be motivated neither by the semantic nor the morphosyntactic properties of modal verbs; see also Brennan 1993) and move to a functional position that is higher than the base position of the subject. While this approach could account for the ambiguity in (154), we will see below that it cannot be extended to account for the scope relations inside the infinitive. A further test to distinguish control from raising constructions is provided by the scope relation between the embedded object and the matrix subject. In raising constructions but not in control constructions, the embedded object can take scope over the surface subject. Thus, sentences involving the raising verb seem as in (155)a are ambiguous between a wide scope reading and a narrow scope reading of the universal quantifier. In (155)b, on the other hand, only a narrow scope reading for the embedded object is available.

German functional constructions

195

(155) Scope in raising vs. control constructions a.

b.

Ein Professor scheint jeden Studenten zu betreuen Some professor seems every student to supervise 'Some professor seems to supervise every student' Ein Professor beschloß jeden Studenten zu betreuen Some professor decided every student to supervise 'Some professor decided to supervise every student'

This contrast is expected under the assumption that only shortdistance quantifier raising is possible and that two LF-positions are available for the subject in raising constructions (Fox 1998/2000, 1999).71 In a control construction such as (152)a, the embedded object can take scope over material inside the infinitive, however, it cannot undergo long-distance quantifier raising to a position above the matrix subject. In a raising structure, on the other hand, short distance quantifier raising can target a position that is higher than the lower position of the subject, hence allowing a wide scope reading of the object. Returning to the main question of this section—the scope properties in modal constructions, we find again that modal constructions pattern with raising structures rather than with control structures. That is, in examples like (156)a, both a narrow scope (cf. (156)b) and a wide scope reading (cf. (156)c) for the embedded object are available. Note again that the context strongly favors a root or deontic interpretation in this example.72 71

The situation seems to be more complex in English. Hornstein (1995: 156, 1998) claims that in the examples below, (a) contrasts with (b,c) in that only the former allows a wide scope reading for the universal quantifier (i.e., long-distance QR is impossible in control constructions). However, since the judgements are controversial they do not seem to allow us to draw any firm conclusions. a. b. c. 72

Someone seemed to attend every class Someone persuaded John PRO to attend every class Someone hoped PRO to recite every poem

We only list the two readings that are important for the present discussion. Various other readings that result from scope interactions between the subject and the modal are ignored.

196

Functional restructuring

(156) Scope in root modal constructions a.

Gemäß Universitätsbestimmungen muß mindestens ein Professor jeden Studenten betreuen 'According to university regulations, at least one professor must supervise every student'

b.

University regulations require that there is at least one professor who supervises every student University regulations require that every student is supervised by at least one professor

c.

The contrast between the unambiguous (155)b and the ambiguous (155)a and (156)a is expected under a raising structure for modal constructions. However, it seems less obvious how this contrast could be derived if modal constructions were represented by control structures. We conclude that the assumption that epistemic and root/deontic modal constructions are represented by raising structures (and hence involve two positions for the subject which are accessible at LF) accounts naturally for the scope interactions between a modal and the subject as well as between the subject and an embedded object in modal constructions. In the next subsections, we will discuss passive in modal constructions and the thematic properties of modal verbs, which will provide further support for a raising structure and against a control structure of modal constructions.

4.1.2.2.

Passive in modal constructions

The second argument for a raising structure of modal constructions comes from the passive properties in these constructions. A wellknown property of modal constructions is that modal verbs cannot be passivized (cf. (157)). Note that the prohibition against passive arises under all modal interpretations; that is, not only epistemic modals but also root/deontic modals block passivization (furthermore, we will see below that dynamic modal constructions also prohibit passive when the modal is not used as a main verb).

German functional constructions

197

(157) Passive of modal verbs a. b. c. d. e.

since the caviar eat VERB-PART/INF was *weil der Kaviar essen gedurft/dürfen wurde 'may' *weil der Kaviar essen gekonnt/können wurde 'can' *weil der Kaviar essen gemußt/müssen wurde 'must' *weil der Kaviar essen gesollt!sollen wurde 'shall' *weil der Kaviar essen gewollt/wollen wurde 'want' *'since the caviar was musted/canned to eat' 'since somebody had to, was allowed to... eat the caviar'

We will show that this fact follows straightforwardly from our analysis of modal constructions. To do so, let us first briefly mention the basic properties of passive in German. In German, transitive and intransitive (unergative) predicates can be passivized (cf. (158)a,b respectively), however, unaccusative predicates do not allow passive (cf. (158)c). (158) Passive with transitve and intransitive predicates a.

Er wurde am Tatort gesehen He was at-the crime-scene seen 'He was seen at the crime scene' b. Es wurde getanzt it was danced *'It was danced' c. *Es wurde (rechtzeitig) angekommen It was (on time) arrived *'It was arrived' The generalization about German passive thus is that passive is only possible when the predicate has an underlying external argument. This generalization extends to verbs that combine with infinitival complements: while passive of unergative verbs like try, decide etc. is possible (cf. (159)a), passive of raising (cf. (159)b) and unaccusative verbs (cf. (159)c) is blocked. Since raising verbs do not have an (underlying) external argument and hence by definition are unaccusatives, it is expected that passive is blocked in raising contexts.

198

Functional

restructuring

(159) Passive in infinitival constructions a.

Es wurde zu tanzen versucht/beschlossen It was to dance tried/decided 'It was tried/decided to dance' (=Somebody tried/decided to dance) b. *Es wurde (zu) tanzen geschienen It was (to) dance seemed * 'It was seemed to dance' c. *Es wurde (ihm) zu tanzen gelungen It was (him) to dance managed *'It was managed to dance' Returning to modal constructions, the prohibition against passivization of modal verbs follows directly from the structure in (139) for modal constructions. Since modal verbs do not assign a theta role to the subject, passive in (157) is blocked as it is with unaccusatives and raising verbs in (159)b,c. In order to maintain a control analysis for modal constructions, one would have to assume that modals are unaccusative predicates that project an internal argument which controls the embedded PRO subject (and which ends up as the surface subject). However, this analysis would not only be problematic for the there-insertion cases, the case properties of Icelandic subjects, and the scope properties in modal constructions as discussed in the previous section, but also constitute a serious challenge for the derivation of constructions involving embedded passive and 'long' object movement in modal constructions (see also Vanden Wyngaerd 1994). As is shown in (160), passive under a modal verb—i.e., passive of the main predicate—is possible in modal constructions (expect in case of want which requires a rich context and is quite marked; see also section 3.2.3). In these constructions, the underlying embedded object shows up as the subject of the modal construction. The same situation holds in English. While control contexts such as (161)a block passivization of the embedded object, embedded passive is possible with raising verbs and modals (cf. (161)b,c).

German functional constructions

199

since the caviar not eaten AUXPASSVERB weil der Kaviar nicht gegessen werden darf weil der Kaviar nicht gegessen werden kann weil der Kaviar nicht gegessen werden muß weil der Kaviar nicht gegessen werden soll V.weil der Kaviar nicht gegessen werden will 'since the caviar may/can... be eaten'

'may' 'can' 'must' 'shall' 'want'

(160) Passive under modal verbs a. b. c. d. e.

(161) Embedded passive a. *The biscuits tried/decided to be finished by Paul b. The biscuits seem to have been finished by Paul c. The biscuits may be finished by Paul (Warner 1993) Before we present our analysis, let us first see how a control structure could handle the contrasts in (160) through (161). As noted above, one could in principle assume that modal verbs are unaccusative predicates that take an internal argument which controls an embedded PRO subject. Examples such as (161)c would then be represented as in (162). (162) Passive under modals: control structure VP the i

be finished

PRO

The major problem for the structure in (162), however, is that the surface subject is clearly not selected by the modal (neither as an external nor an internal argument). Thematically, the surface subject is only related to the lower predicate (i.e., it is the theme of finish, and clearly, the biscuits are in no permission or obligation relation).

200

Functional restructuring

Thus, given that the subject is not an argument of the higher predicate, it has to be explained where it comes from (in particular since in general noun phrases cannot just be inserted anywhere). If the biscuits is considered as an argument of the lower predicate, on the other hand, the structure in (162) would constitute a violation of the theta criterion since there are two arguments (the subject and PRO) which share one theta-role. In order to save the structure in (162), one would thus have to give up the theta criterion. This is a move that has been made by many authors (see for instance Chomsky 1981: 139; fn. 14; Hornstein 1999). However, it also has to be pointed out that this would then mean that the original motivation for a control structure for (root) modal constructions disappears. Turning to our account for the contrasts in (160) through (161), we follow the general idea that the ungrammaticality of examples such as (161)a is due to the fact that verbs like try, decide etc. are lexical verbs that establish a thematic relation with an animate external argument (cf. section 3.2.3). If this relation cannot be met as in (161)a, which involves an inanimate subject, the construction is illformed. Since no such effect arises with raising verbs, the assumption that raising verbs do not impose thematic restrictions on the subject (i.e., do not assign a theta role to the subject) is supported. The same reasoning can then be applied to modal constructions. In examples such as (160) and (161)c, the only thematic restrictions imposed on the surface subject are the theta requirements of the embedded predicate; the modal verb does not impose any theta requirements on the subject. The analysis of these examples is then straightforward under a raising structure for modal constructions such as the structure we have suggested in (139) (relevant parts are repeated here as (163)a). The surface subject the caviar starts out as an argument of the embedded predicate. Passivization of the lower predicate eliminates accusative case and hence causes the object to move to the surface subject position where it is assigned nominative case. Since no other arguments are present (the underlying subject is suppressed or expressed as an adjunct due to passive), this movement is consistent with standard locality constraints.

German functional constructions

(163) a

Passive under modals TP/AgrSP

t

b.

201

Passive & dynamic modals TP/AgrSP

eaten

Furthermore, our analysis provides a way to account for the behavior of dynamic modals in passive constructions. As we have seen in (160)e, the dynamic modal want is less felicitous than other modals in embedded passive constructions. This effect can be seen as a result of the fact that the unmarked interpretation of want is a dynamic interpretation (which we will show is incompatible with passive) and that a root interpretation (i.e., a structure such as (163)a) is only marginally available for this modal. Our system offers two ways to account for the incompatibility of dynamic modals and passive. First, since dynamic modals (i.e., modals that are inserted in the v-head) obligatorily assign a theta role to the subject and since passive voice obligatorily suppresses the theta role for the subject, these two elements involve conflicting requirements that cannot be fulfilled simultaneously (cf. (163)b). Second, one could simply assume that dynamic modals compete with voice elements (e.g., the passive auxiliary) for the v° head and that only one element can occur in this position. Under both assumptions, dynamic modals and passive cannot co-occur (both passive of dynamic modals as well as passive under modals) which is the desired result. As a final point we would like to discuss an apparent counter example to our claim that dynamic modals are incompatible with pas-

202

Functional restructuring

sive, which, on closer scrutiny, however, will support our analysis. As pointed out by Erb (2001: 90), passive of dynamic modals is possible in examples such as (164)a. Note, however, that this claim only holds in constructions in which the modal combines with a nominal complement rather than a verbal infinitival complement. The claim that that Aufsätze schreiben 'essays write' in (164)a is a complex noun phrase (note that infinitives can be used as nouns in German)—i.e., a DP complement of the modal rather than a true infinitival construction—is supported by the lack of agreement between the object and the auxiliary. Since the auxiliary in (164)a does not agree with the underlying plural object essays, the surface subject in this passive construction has to be the whole noun phrase "essays write". If the example is changed to an unambiguous modal-infinitive construction (i.e., a construction in which the underlying object agrees with the matrix auxiliary as in (164)b), the result is again clearly ungrammatical. (164) Passive and dynamic modals a. ΊAufsätze schreiben wird heutzutage Essays write A U X P A S S - S G nowadays von keinem Schüler mehr gekonnt by no pupil anymore can-PART 'Nowadays, pupils cannot write essays anymore' b. *Aufsätze schreiben werden heutzutage Essays write AUXPASS-PL nowadays von keinem Schüler mehr gekonnt by no pupil anymore can-PART 'Nowadays, pupils cannot write essays anymore' We thus assume that dynamic modals that combine with a nominal complement can be full main verbs. However, dynamic modals that combine with an infinitive are functional voice elements, which is supported by their syntactic properties (such as extraposition and the IPP effect) as well as by their incompatibility with passive. To conclude, both properties discussed in this section—the prohibition against passivization of modal verbs in general and the possi-

German functional constructions

203

bility of passivizing the embedded object in epistemic and root/deontic modal constructions—support the claim that modals involve raising structures rather than control structures. 4.1.2.3.

Do modals assign theta roles?

The last point we would like to raise is that the claim that root or deontic modal verbs assign a theta role to the subject is empirically inadequate. We summarize here a number of constructions that show that this claim cannot be maintained. First, as discussed in this and the previous section, the possibility of expletive constructions such as (165)a and passive constructions such as (165)b, clearly shows that there is no thematic relation between the modal and the subject. (165) Expletive and passive construction in root modal constructions a. b.

There can be a party as long as it's not too loud The biscuits may be finished by Paul (Warner 1993)

The same point can be made for the constructions in (166). In (166)a (which is from Newmeyer 1975), the modal which is interpreted deontically is clearly in no thematic relation with the subject an opening hand. Furthermore, McGinnis (1993) notes that an example such as (166)b can be used in a situation where a home owner gives instructions to the care taker of his house (i.e., it can be interpreted deontically roughly as It is allowed that icicles hang from the eavestroughs). Finally, the examples in (166)c-g (can) involve an interpretation that following Barbiers (1995) can be called non-directed deontic interpretation. That is, the obligation or permission expressed by the modals does not have to be directed towards the subject of these sentences. This is most clearly the case in (166)c-f which are most naturally interpreted as It must be the case that the traitor dies, that John is home when the murder happens etc. However, even in examples such as (166)g, a non-directed deontic interpretation is possible in the right context (e.g., some Mafia boss wants to get rid of John and hence he orders one of his henchmen to arrange that John leaves town; in this context the Mafia boss could utter (166)g without

204

Functional restructuring

imposing any obligation on John). (166) Thematic relations in root modal constructions a. b. c. d. e. f. g.

An opening hand must contain thirteen points Icicles may hang from the eavestroughs The traitor must die John must be home when the murder happens The old man must fall down the stairs and it must look like an accident Your children may play in the garden but they cannot go into the barn John must leave town

Importantly, in all of the examples above the modality expressed by the different root modals is not applied to the subject. Hence, if it is the case that modals ever assign a theta role to the subject they do so optionally. The contexts in which (intuitively) the subject appears to be in a thematic relation with the modal are modal constructions that involve a directed deontic interpretation (following again Barbiers' terminology). An example like (166)g could be interpreted as an obligation relation imposed on John. The question, however, is whether these relations (obligation, permission etc.) are (syntactic) thetaroles. The availability of non-directed readings shows that roles/functions like 'obligee' or 'permissee' do not have to coincide with a specific syntactic argument in the sentence. In other words, the determination of these roles cannot be seen as an obligatory mapping between theta roles and syntactic arguments but is rather determined by the context in which a modal construction is used. We thus suggest that modals (again epistemic and root/deontic) in fact never assign a theta role to the subject but that the apparent thematic relation between the subject and a (root) modal in certain contexts is purely contextual.

4.1.3. Summary In this section, we have argued for a functional modal domain of the

German functional constructions

205

clause which in German, is split into different sub-domains which correspond to different modal interpretations. We have summarized and provided further arguments for the claim that root and epistemic modal constructions are represented by a raising structure. We have argued that the thematic and syntactic properties (such as case) of the subject in modal constructions are determined solely by the lower verb and that the modal has no influence on the properties of the subject. Furthermore, we have shown that the subject in modal constructions can be interpreted below the modal—i.e., in a position that is only available in a raising structure. 4.2. Raising constructions Besides modal constructions which we have argued are raising constructions (modulo dynamic modals), functional restructuring predicates in German also include the more typical raising predicates in Table 20. Table 20.

Raising predicates in German

Verb

Gloss

Category

beginnen drohen pflegen scheinen versprechen

begin threaten used to seem promise

Aux/lexical restructuring verb Aux/(reduced) non-restructuring verb Aux Aux Aux/(reduced) non-restructuring verb

In this section, we will show that these verbs fall into two groups: pure raising verbs {seem, used to), and verbs that are ambiguous between a raising and a 'control' version (i.e., begin, threaten, promise). We will summarize the properties of these constructions which will provide evidence for the claim that raising verbs are functional Aux elements in German (see (167)), and that the non-raising versions involve lexical predicates which can combine with restructuring or non-restructuring infinitives according to the properties discussed in Chapter 2 (see also Chapter 5 for a refined structure for non-restructuring infinitives). Our basic claim is that raising verbs in

206

Functional restructuring

German are essentially epistemic modals—i.e., they occur in the Aux head (cf. (167)). (167) Raising constructions in German AuxP

ιΘ'^^^Αιιχ'

The immediate consequence of this claim is that raising verbs (like epistemic modals) cannot be embedded under other modals or auxiliaries in German but can co-occur with lower modals or auxiliaries. The correctness of this prediction is illustrated in (168) through (169) (we use the unambiguous raising verb scheinen 'seem' to illustrate this property; the other raising constructions will be discussed below). (168)a,b show that scheinen cannot be embedded under auxiliaries. Note that in this case it is in fact not even clear what the form of the participle of scheinen would be (cf. (168)b). (168)c,d show that scheinen can appear neither under an epistemic nor under a root/deontic modal. 73

73

The raising construction involving the predicate pflegen 'used to' is also ungrammatical under auxiliaries and modal verbs. However, this effect does not provide strong evidence for the ordering between auxiliaries and pflegen since examples such as the ones in (168) with the verb used to would presumably also be excluded independently on semantic grounds (i.e., the inherent generic nature of this construction makes it incompatible with most of the contexts tested in the text).

German functional constructions

207

(168) Embedded scheinen 'seem' a. * Morgen wird er die Stadt zu verlassen scheinen Tomorrow will he the town to leave seem 'He will seem to be leaving town tomorrow' b. *Sie hat zu Hause zu arbeiten gescheint/geschienen She has at home to work seem-PARTa/PARTb 'She has seemed to work at home' c. * Morgen dürfte er die Stadt zu verlassen scheinen Tomorrow might he the town to leave seem 'He might seem to leave town tomorrow' d. *Sie muß zu Hause zu arbeiten scheinen She must at home to work seem 'She must seem to work at home' On the other hand, nothing prevents (non-epistemic) modals and auxiliaries under raising verbs. Some examples are given in (169). (169) Modals and auxiliaries under scheinen a.

b.

c.

d.

e.

Sie schien zu Hause gearbeitet She seemed at home worked 'She seemed to have worked at home' Sie schien zu Hause gewesen She seemed at home been 'She seemed to have been at home' Sie pflegte um 5h zu Hause She used at 5 o'clock at home 'She used to be home by 5' Sie schien zu Hause arbeiten She seemed at home work 'She seemed to have to work at home' Sie pflegte zu Hause arbeiten She used at home work 'She used to be able to work at home'

zu haben to have zu sein to be gewesen zu sein been to be zu müssen to must zu können to can

Finally, as mentioned before, raising verbs do not passivize but can embed a passivized predicate. This is illustrated in (170).

208

Functional restructuring

(170) Passive in raising constructions a. *Der Kaviar wurde zu essen gescheint/geschienen The caviar was to eat seem-PARiyPARTt, *'The caviar was seemed to eat'; 'It seemed that somebody ate the caviar' b. *Der Salat wurde vor dem Essen zu essen gepflegt The salad was before dinner to eat used *'The salad was used to eat before dinner'; 'It used to be the case that one ate the salad before dinner' c. Der Kaviar schien gegessen worden zu sein The caviar seemed eaten been to be 'The caviar seemed to have been eaten' d. Der Salat pflegte vor dem Essen gegessen zu werden The salad used before dinner eaten to A U X P A S S 'The salad used to be eaten before dinner' Let us now turn to the optional raising predicates. We follow the common assumption that the verbs promise, threaten, and begin are ambiguous between raising and control constructions. 74 In case of promise and threaten the two constructions correspond (more or less) to two different readings illustrated in (171). Roughly, the interpretation arising in the control construction is an agentive reading, whereas the interpretation arising in the raising construction is an epistemic reading. In (171), under the control reading, the subject makes an actual promise or utters a threat; under the raising reading, somebody thinks (promise) or fears (threaten) that John might become a good father. (171) Raising vs. control a.

74

Hans verspricht ein guter Vater zu werden John promises a good father to become 'John promises to become a good father'

Since we assume that begin is a lexical restructuring verb which can combine with a subjectless infinitive (see Chapter 4), the ambiguity is technically not a raising vs. control ambiguity but should be understood as an ambiguity between a functional raising verb and a lexical restructuring verb.

German functional constructions

b.

209

Hans drohte ein guter Vater zu werden John threatened a good father to become 'John threatened to become a good father'

Under our analysis, the raising versions of these predicates are generated in Aux, whereas the control versions are main verbs. Evidence for this claim comes from the loss of the raising interpretation when these verbs are embedded under auxiliaries or modals. Examples such as (172)a,b are possible, however, they can only be interpreted with agentive promise and threaten (the effect is even stronger when a dative argument is realized).75 Similarly, (172)c,d in which the potential raising verbs appear under modals prohibit the epistemic interpretation. (172) Embedded promise, threaten a.

b.

c.

d.

Er hat ein guter Vater zu werden versprochen He has a good father to become promised 'He has promised to become a good father' (control only) Er hat ein guter Vater zu werden gedroht He has a good father to become threatened 'He has threatened to become a good father' (control only) Er muß ein guter Vater zu werden versprechen He must a good father to become promise 'He must promise to become a good father' (control only) Er muß ein guter Vater zu werden drohen He must a good father to become threaten 'He must threaten to become a good father' (control only)

Furthermore, the raising interpretation disappears when the infinitive is extraposed. Although the contrasts are subtle, we find that in contexts such as (173) in which the infinitive is extraposed (which we have argued is only possible when the infinitive combines with a lexical verb), only an agentive interpretation is possible (see Prinzhorn 1987,1990 for similar facts). 75

The same effect has been reported for Dutch by Rutten (1991).

210

Functional restructuring

(173) Extraposition with promise and threaten a.

b.

weil er versprach ein guter Vater zu werden since he promised a good father to become 'He promised to become a good father' weil er drohte ein guter Vater zu werden since he threatened a good father to become 'He threatened to become a good father'

The final properties we will examine here are the following two passive properties of optional raising constructions. First, we will show that passive of the embedded predicate and long object movement are possible in promise, threaten, and begin constructions, which will support the claim that these constructions are (functional) raising constructions. Second, we will show that passive of the main predicate is also possible with these predicates, which will support the claim that these constructions are (lexical) control constructions. Importantly, however, we will see that the two structures are not interchangeable but come with the properties typical for lexical vs. functional predicates. The first property—the possibility of embedded passive and long object movement in promise, threaten, and begin constructions—is illustrated in (174) (the examples are grammatical, but maybe not as naturally as with true raising verbs). Thus, (174) constitutes basic evidence for a raising structure of these constructions. Note that this form of object movement has to be distinguished from long passive with lexical predicates (see below); long passive occurs when the matrix predicate is passivized, whereas in raising constructions only the embedded predicate can be passivized. (174) Passive under promise, threaten, and begin a. IDie Stadt versprach zerstört The town promised destroyed 'The town promised to get destroyed' b. IDie Stadt drohte zerstört The town threatened destroyed 'The town threatened to get destroyed'

zu werden to AUXpAss zu werden to AUXpAss

German functional constructions

c. IDie Stadt begann zerstört The town began destroyed 'The town began to get destroyed'

211

zu werden to A U X P A S S

A crucial fact about the passive constructions in (174) is that these examples become ungrammatical when the matrix verbs appear under an auxiliary (cf. (175)) or a modal (cf. (176)). The ungrammaticality of (175) and (176) is expected under our account, since object movement of the form in (174) is only possible in raising constructions (i.e., when the matrix verb does not establish a thematic relation with the subject). However, since raising predicates—as we claim—have to be realized as the highest verbal element (modulo tense), auxiliaries and modal verbs cannot take scope over the matrix predicates in (174) through (176). (175) Raising constructions under auxiliaries a. *Die Stadt hat zerstört zu werden The town has destroyed to AUX pass 'The town (has) promised to get destroyed' b. *Die Stadt hat zerstört zu werden The town has destroyed to AUX pass 'The town (has) threatened to get destroyed' c. *Die Stadt hat zerstört zu werden The town has destroyed to AUX pass "The town has begun to get destroyed'

versprochen promised gedroht threatened begonnen begun

(176) Raising constructions under modal verbs a. *Die Stadt muss zerstört zu werden The town must destroyed toAUX PASS 'The town must promise to get destroyed' b. *Die Stadt kann zerstört zu werden The town can destroyed to AUX pass 'The town can threaten to get destroyed' c. *Die Stadt darf zerstört zu werden The town may destroyed to AUX pass 'The town may begin to get destroyed'

versprechen promise drohen threaten beginnen begin

212

Functional

restructuring

Similarly, we correctly predict that extraposition is prohibited in (174) (which is illustrated in (177)). Since extraposition is only possible when the matrix verb is a lexical verb, and raising verbs (including the raising options of promise, threaten, and begin) are functional elements, these two properties cannot co-occur. (177) Extraposition and raising predicates a. *weil die Stadt versprach zerstört zu werden since the town promised destroyed t o A U X P A S S 'The town promised to get destroyed' b. *weil die Stadt drohte zerstört zu werden since the town threatened destroyed t o A U X P A S S 'The town threatened to get destroyed' c. *weil die Stadt begann zerstört zu werden since the town began destroyed t o A U X p a s s 'The town began to get destroyed' The second property—passive of the main predicate—provides direct evidence for a lexical version of promise, threaten, and begin constructions. As is illustrated in (178), these predicates can be passivized (in contrast to true raising and modal constructions). Importantly, (178) also shows that the constructions are then not subject to the restrictions found for functional predicates. In particular, the infinitives are extraposed in (178) which is only possible when the matrix verb is a lexical predicate. (178) Passive of promise, threaten, and begin a.

b.

weil (ihm) versprochen / gedroht wurde since (him) promised / threatened was den Turm abzureißen the tower-ACC to-tear-down 'since somebody promised/threatened (him) to tear down the tower' Es wurde begonnen den Wagen zu reparieren It was begun the car-ACC to repair 'They began to repair the car'

German functional constructions

213

Furthermore, as expected under our analysis, embedding of the constructions in (178) under a modal verb is possible, since the matrix verbs in (178) are lexical predicates, and hence nothing prevents them from co-occurring with higher functional verbs as in (179). (179) Embedded passivized promise, threaten, and begin a.

b.

weil (ihm) zuerst versprochen / gedroht werden since (him) first promised / threatened A U X P A S S muss den Turm abzureißen must the tower-ACC to-tear-down 'since somebody must first promise/threaten (him) to tear down the tower' Es muss sofort begonnen werden It must immediately begun AUXPASS den Wagen zu reparieren the car-ACC to repair 'They must begin immediately to repair the car'

Finally, passive allows us to distinguish between restructuring and non-restructuring. While impersonal passive constructions are possible with all three verbs (cf. (178)), it is shown in (180) that long passive is only allowed in begin constructions (cf. (180)a), and clearly prohibited in promise and threaten constructions (cf. (180)b). Thus, begin qualifies as a lexical restructuring verb, whereas (lexical) promise and threaten constructions are non-restructuring infinitives (but see Chapter 5 for different grades of non-restructuring). (180) Long passive a.

Der Wagen wurde zu reparieren begonnen The car was to repair begun 'They began to repair the car' b. *weil (ihm) der Turm abzureißen since (him) the tower-NOM to-tear-down versprochen / gedroht wurde promised / threatened was 'since somebody promised (him) to tear down the tower'

214

Functional restructuring

Note that the examples in (180)b are also excluded as raising structures since raising verbs cannot be passivized. The fact that promise and threaten constructions are non-restructuring infinitives is not unexpected in our approach to restructuring, since both predicates allow an independent embedded tense interpretation (and as we will see in Chapter 4, a non-obligatory control interpretation) which would be incompatible with the restructuring configuration. To conclude, we have argued in this section that raising verbs are functional categories in German which are generated in the highest functional head of the modal domain (i.e., what we have labeled Aux). While this claim correctly accounts for the distribution of raising verbs in German, the question of why raising verbs are functional elements in some languages (e.g., German), but lexical elements in other languages (e.g., English where raising verbs do not seem to compete with auxiliaries etc.) has been left open here. We have also shown that the there is good evidence for the claim that promise, threaten, and begin constructions can involve either a functional matrix predicate (i.e., a raising predicate) or a lexical predicate.76 We have seen that in case of promise and threaten the different interpretations correlate with different syntactic properties which can be reduced to the lexical vs. functional distinction suggested here. In particular, under an epistemic interpretation, these predicates cannot be embedded under other verbal elements (except tense), do not allow extraposition of the infinitive, and prohibit passivization of the matrix predicate. Under the agentive interpretation, on the other hand, these predicates can be embedded under other verbal elements, extraposition of the infinitive is possible, and the matrix predicate can passivize, however, long passive is excluded. 76

Begin constructions might in fact allow another option: one could imagine that begin can also be realized as an aspectual head, in which case it would be a nonthematic raising verb but nevertheless allow higher auxiliaries. We have not distinguished between voice and aspect here, but aspectual begin might provide evidence for a separation of these two categories. In particular, examples such as Es hat zu regnen begonnen (Lit. 'It has to rain begun'; 'It (has) began to rain') could be instantiations of this option. The fact that extraposition is dispreferred in these examples might support this idea. However, the empirical situation is not completely clear and further investigation is necessary before drawing firm conclusions.

German functional constructions

215

Similarly, (functional) begin constructions allow embedded passive but can then not be embedded under other modals or auxiliaries. Passive of (lexical) begin predicates is possible, and since begin is compatible with the restructuring configuration, long passive is licensed as well.

4.3. Semi-functional constructions The last set of constructions we will discuss briefly are constructions involving the predicates in Table 21. The purpose of this section is not to provide an in-depth account of these constructions, which for obvious reasons we cannot do here, but rather to point out some crucial properties of these constructions which seem to bear on the lexical vs. functional distinction as well as the issue of restructuring. Table 21.

Semi-functional predicates

Verb

Gloss

Category

gehen kommen hören lassen sehen

go come hear let see

Aux/lexical restructuring verb Aux/(reduced) non-restructuring verb Aux Aux Aux/(reduced) non-restructuring verb

As discussed in section 3.2, semi-functional predicates behave like functional predicates concerning extraposition and the IPP effect, but like lexical categories concerning their thematic properties. The basic claim we would like to make is that the categories in Table 21 are voice/aspect elements, and that the clausal architecture in (181) allows us to account for the intermediate status of these categories. In (pure) structural terms, semi-functional predicates appear in a functional head position and hence qualify as functional categories for all tests targeting this structural lexical/functional distinction (see section 3.2). However, semi-functional predicates are also thematic categories since they are in the lexical domain of a clause (cf. Marantz 1997) which allows them to establish a thematic relation with an argument in their specifier.

216

Functional restructuring

(181) Semi-functional constructions in German

A question that arises concerning the structure in (181) is how the claim that causatives, perception verbs, and motion verbs are voice/aspect elements can be motivated. For causatives, this assumption seems uncontroversial as it is the standard claim in many recent works on vP structure (cf. Pylkkänen 1999 among many others). As for motion verbs, we suggest that these predicates are aspectual in nature and since we have not been distinguishing here between voice and aspect in German (though further refinements might be necessary upon further inspection), the assumption that motion verbs occupy the v-head does not seem to constitute a problem either. Note also that this assumption receives support from motion verbs in English, since English come and go, while combining with bare infinitives, contrast with modal verbs in that they can co-occur with other auxiliaries (such as in John will go/come eat soon). Concerning perception verbs however, the question seems to be valid since it is not clear a priori what these elements have in common with voice or aspect. We tentatively suggest that the reason has to do with the semantic properties of perception verb complements. A common characterization of (at least some of the) perception verb constructions is that the complement is interpreted as a bare (realis) event which is subject to a strict simultaneity requirement (see for

German functional constructions

217

instance Declerck 1981, 1982, 1983; Higginbotham 1983; Mittwoch 1990; Alsina 1992; Safir 1993; Boivin 1998; Felser 1998 for detailed discussions of the semantic properties of these constructions). Since a VP complement can satisfy this requirement, the minimal structure would be as in (181). However, note again that we do not claim that our structure explains or causes the semantic properties we find in (semi-)functional constructions. Rather, we assume that the order of different categories is a mere instantiation of a deeper semantic interaction of the various elements involved. The co-occurrence restrictions we noted (and will note below) simply show that an ordering is in effect and that German groups verbal categories the way we have suggested. As mentioned at various points, a more fundamental understanding of the cause of these orderings is still desirable. In addition to the syntactic and thematic properties discussed in section 3.2 (e.g., extraposition, IPP effect), the structure in (181) allows us to account for three further (sets of) properties: the ordering between functional and semi-functional categories, the complementary distribution between different semi-functional elements, and the prohibition against passive in semi-functional constructions. The first property that results as an immediate consequence of the structure in (181) is that semi-functional predicates can be embedded under modale and auxiliaries (cf. (182)), however, modals and auxiliaries cannot be embedded under semi-functional predicates (cf. (183); some of the examples are also excluded on semantic grounds assuming the semantic restrictions mentioned above). (182) Embedded semi-functional predicates a.

b.

c.

Hans John 'John Hans John 'John Hans John 'John

ist schwimmen gegangen I gekommen is swim gone / come went/came swimming' muß schwimmen gehen / kommen must swim go / come must go/come swimming' hat Maria musizieren sehen / hören / lassen has Mary make-music see / hear / let has seen/heard/let Mary make music'

218

Functional

d.

restructuring

Hans darf die Maria nicht John may the Mary not musizieren sehen / hören / lassen make-music see-INF / hear-INF / let-INF 'John may not see/hear/let Mary make music'

(183) Semi-functional predicates embedding modals and auxilaires a. *Hans ließ den Peter gewonnen haben John let the Peter won have 'John let Peter be the winner' (Lit. 'John let Peter have won') b. *Hans hörte den Peter musizieren müssen John heard the Peter make-music must 'John heard that Peter had to make music' c. *Hans sah den Peter musizieren wollen John saw the Peter make-music want 'John saw that Peter wanted to make music' d. *Hans ging schwimmen müssen John went swim must 'John went to get ordered to swim' (Lit. 'John went to have to swim') e. *Hans kam essen wollen John came eat want 'John came to get a desire to eat' Second, our analysis can account for the incompatibility of the different semi-functional predicates with each other. (184) shows that perception verbs and causatives cannot be embedded under motion verbs. Note again that this restriction only applies when the lower predicate combines with an infinitival complement—i.e., when the matrix predicate has to be analyzed as a voice head according to our analysis and cannot function as a main verb. If the lower predicate is a non-infinitival taking verb that combines with a nominal or prepositional complement, it can be inserted as the head of the verb phrase and hence, embedding under a semi-functional predicate is possible. A minimal pair illustrating this difference is given in (184)a,b.

German functional constructions

219

(184) Motion verbs embedding caustative/perception verbs a. *Hans kommt den Peter muszieren hören John comes the Peter make-music hear 'John is coming to hear Peter make music' b. Hans kommt Musik hören John is coming music hear 'John is coming to hear music' c. *Hans geht Peter muszieren sehen / hören / lassen John goes Peter make-music hear / hear/ / let 'John is going to see/hear/let Peter make music' d. Hans geht ihr die Luft auslassen John goes her the air out-let 'John is going to let the air out of her tires' The examples in (185) illustrate the same point for motion verbs that are embedded under perception verbs and causati ves. The grammatical sentences in (185)a,c do not involve infinitival taking motion verbs, and hence the motion verbs are analyzed as main verbs which are compatible with higher semi-functional predicates. The ungrammatical sentences in (185)b,d, on the other hand, involve infinitival taking motion verbs, and hence the motion verbs compete with the causatives/perception verbs for the same position. (185) Causative/perception verbs embedding motion verbs a.

Hans läßt / sah ihn ins Kino gehen John lets / saw him in-the cinema go 'John is letting/saw him go to the movies b. *Hans läßt / sah ihn einen Turm bauen gehen John lets / saw him a tower build go 'John is letting/saw him go build a tower' c. Hans ließ ihn zum Essen kommen John let him to dinner come 'John let him come to dinner' d. *Hans ließ / sah ihn Muscheln essen kommen John let / saw him mussels eat come 'John let/saw him come eat mussels'

220

Functional restructuring

The third property we would like to mention here is passive. As in the case of dynamic modals, we find an incompatibility of semifunctional predicates with passive (see also Bennis and Hoekstra 1989; Mittwoch 1990; Felser 1998). The prohibition against passive of semi-functional predicates is illustrated in (186) (the addition of a fry-phrase does not alter the grammaticality of the examples). 77 Furthermore, (187) shows that passive under semi-functional predicates is also prohibited. (186) Passive of semi-functional predicates a. *Ein Turm wurde bauen gegangen A tower was build gone 'Somebody went to build a tower' b. *Die Oma wurde besuchen gekommen come The granny was visit 'Somebody came to visit granny' c. *Der Peter wurde den Wagen reparieren gelassen The Peter was the car repair let 'Somebody let Peter repair the car' c'.*Der Wagen wurde reparieren gelassen The car was repair let 'Somebody let somebody repair the car' d. *Die Kapelle wurde den Marsch blasen gehört The band was the march play heard 'Somebody heard the band play the march' d'. * Der Marsch wurde blasen gehört The march was play heard 'Somebody heard somebody play the march' e. *Der Peter wurde den Kuchen essen gesehen The Peter was the cake eat seen 'Somebody saw Peter eat the cake' 77

Examples with intransitive embedded predicates might sound slightly better (e.g., %Es wurde oft schwimmen gegangen [Lit. 'It was often swim gone'; 'People used to go swimming a lot']; %Er wurde schlafen gelassen [Lit. 'He was sleep let; 'They let him sleep']). However, these constructions are quite restricted and could either be seen as lexicalized complex verbs or as nominalized infinitives combining with main verbs.

German functional constructions

221

e'. * Der Kuchen wurde essen gesehen The cake was eat seen 'Somebody saw somebody eat the cake' (187) Passive under semi-functional predicates a. *Hans John 'John b. *Hans John 'John c. *Hans John 'John d. *Hans John 'John

geht / kommt bestraft goes / comes punished is going/coming to get punished' läßt den Peter unterstützt lets the Peter supported is letting Peter get supported' sah den Kuchen gegessen saw the cake eaten saw the cake being eaten' hörte den Marsch geblasen heard the march played heard the march being played'

werden AUXPASS werden AUXPASS werden AUXPASS werden AUXPASS

Returning to one of the main issues of this chapter—the restructuring status of semi-functional predicates—a question that arises concerning some of the semi-functional predicates is what type of complement causatives and perception verbs take. As we will show in the next chapter, lexical restructuring requires a subject-less infinitival predicate. Assuming that causatives and perception verbs involve an embedded subject, one might think that these predicates should not count as restructuring predicates. However, the presence of an embedded "subject" in these constructions does not pose a problem for our approach in at least three respects (though it raises some questions of classification and terminology). Firstly, as pointed out by Guasti (1993, 1996, 1997), the embedded subject in causative constructions is both an argument of the causative predicate and an argument of the embedded predicate, and Guasti suggests that the complement of causative verbs is best represented as a VPcomplement or small clause. If one adopts this approach, the complement of a causative verb would simply be a complex VP without another vP and the problem could be avoided. However, we will not

222

Functional restructuring

commit ourselves to this analysis (though we leave it as an option) since the problem of embedded "subjects" in functional constructions does not seem to arise independently for two other reasons. Much recent work on the structure and meaning of voice phrases has shown that various aspects are involved and expressed by vprojections. In particular, many authors assume that a causative vP and an agentive vP have to be separated, but crucially this split is not a split into two clauses but rather a split along the lines of AgrSP vs. TP (but in the lexical domain of a clause). Enriching the functional structure of the lexical domain along these lines, would then mean that a causative construction is an inherent mono-clausal construction that does not include two "subjects" but rather a causer argument and an external (causee) argument. Finally, a leading idea throughout our study, which we will elaborate on in detail in Chapter 5, is that there are different grades of both restructuring and non-restructuring. Anticipating some of the conclusions we will reach there, not all restructuring criteria or tests target the same structural property. While long passive requires a complement without embedded case, tense or an embedded subject, properties such as pronoun fronting and focus scrambling are possible from vP infinitives and TP infinitives but blocked across CPs. In other words, the presence of an embedded subject blocks certain properties but not others which will lead to a more complex classification of restructuring and non-restructuring infinitives. In Chapter 5, we will show that these classificaitons are not arbitrary assumptions but can be related directly to the nature of the properties involved (such as passive, scrambling etc.). Since passive (and hence long passive, which is sensitive to intervening subjects and in our classification a criterion to distinguish between restructuring and non-restructuring), is prohibited on independent grounds in causative and perception verb constructions (see above), it is not clear that the presence of a "subject" in these constructions should cause a problem for our approach. As can be seen for instance in (188), scrambling is possible across a causee or perceivee for some speakers, however, the general judgement is that this form of scrambling is focus scrambling and that regular scrambling is illicit (the examples are best when the cheese and the car are focused or preceded by a focus element such

German functional constructions

223

as only). Since, as we will argue, focus scrambling is not sensitive to the presence of an embedded subject, the possibility of these examples is as expected.78 (188) Scrambling in causative and perception verb constructions a. %dass Hans den Käse den Peter that John the cheese-ACC the Peter 'that John let Peter eat the cheese' b. %dass Hans den Wagen den Peter that John the car-ACC the Peter 'that John saw Peter repair the car'

essen eat

ließ let

reparieren sah repair saw

Similarly, in Dutch, a famous restructuring property is verb raising and the IPP effect (see also Chapter 2, section 4.2.1). The crucial condition for this criterion is tenselessness (cf. Chapter 5, section 3.1); the presence of a subject does not seem to bear on the wellformedness of this property. Hence, the class of restructuring constructions defined by verb raising plus the IPP effect in Dutch also includes causatives and perception verbs (see the appendix, section 5 for a summary of verb raising predicates in Dutch). To conclude, an analysis that treats motion verbs, causatives, and perception verbs in German as voice/aspect heads allows us to account for various ordering and co-occurrence restrictions found in these constructions, the thematic properties of these constructions and the prohibition against passive, as well as the fact that these verbal elements pattern with functional categories rather than with lexical categories in German regarding their syntactic properties. While our analysis is able to accommodate a range of empirical properties, a deeper understanding of the nature and motivation of the syntactic structures suggested awaits further research. 78

Although we have not been able to determine a robust empirical generalization, it seems that pronoun fronting is less acceptable for many speakers across an embedded accusative causee or perceivee. If this generalization turns out to be correct, it might show that both pronoun movement and embedded subjects target and compete for the matrix vP. The analysis we will present in Chapter 5 might then have to be modified along these lines.

224

Functional restructuring

5. Summary The properties of (semi-)functional constructions in German discussed in this chapter are summarized in Table 22. Table 22.

Properties of (semi-)functional restructuring predicates in German

Verb

Gloss

dürfen haben können müssen sein sollen werden wollen dürfte möchte scheinen pflegen gehen hören kommen lassen sehen beginnen

may have can must be shall/should will want might would like to seem used to go hear come let see begin

Matrix Matrix passive perfect

-

have % have have be have+coná

-

-

-

have

-

-

-

-

-

-

-

-

-

IPP Extraposition Θ-restriceffect of infinitive tions

+

-

-

-

-

-

-

-

-

-

-

-

-

-

+ + N/A + N/A + N/A N/A N/A N/A N/A + N/A + +

-

-

-

± -

±

-

-

-

-

+ + + + + +

+

be have be have have have

-

+

-

-

N/A

-

-

+

+

-

-

drohen

threaten

+

have

-

-

-

N/A

-

-

versprechen

promise

+

have

-

+

+

-

-

N/A

-

-

For comparison, we also list the properties of lexical restructuring infinitives in Table 23. We have argued that the lexical vs. functional distinction is reflected in two syntactic properties: extraposition (which is only possible with complements of lexical predicates) and the IPP effect (which is only found with functional predicates in the scope of have).

Summary Table 23.

225

Properties of lexical restructuring predicates in German

Verb

Gloss

beabsichtigen gelingen mißlingen vergessen vermeiden versäumen versuchen wagen empfehlen erlauben gestatten untersagen verbieten

intend (want) manage fail forget avoid miss, neglect try dare recommend allow permit prohibit forbid

Matrix Matrix passive perfect + -

+

+ + + +

+ + + +

+

have be be have have have have have have have have have have

IPP Extraposition Θ-restriceffect of infinitive tions

+ + + + + +

+ + + +

+ + +

+ + HO] + [10] + + + + + + + + + +

Furthermore, we have shown that the thematic properties of the predicates in Table 22 cannot be reduced to the lexical/functional distinction but that certain predicates have to be classified as thematic functional (or what we have called semi-functional) predicates. We have argued that dynamic modals, motion verbs, causatives, and perception verbs are functional categories syntactically but nevertheless theta assigners semantically. Assuming that these categories are generated in the semi-functional voice or aspect domain allowed us to account for their syntactic properties as well as the fact that these categories establish theta relations with an argument. Finally, we have suggested that the non-thematic functional domain is split into two sub-domains and that epistemic modals and raising verbs are restricted to the highest projection in this domain, whereas root/deontic modals can occur in either position. A deeper motivation for the clausal architecture suggested has not been provided at this point, however, various co-occurrence and ordering restrictions seem to indicate that the approach is on the right track.

Chapter 4 The infinitival subject

1. Introduction In Chapter 2, we argued that lexical restructuring infinitives are tenseless complements that do not involve a structural case position/assigner. These properties were attributed to the lack of tense and structural case features in the infinitive, which—assuming the minimalist view that phrase structure is determined by combining features (rather than by a universal template of projections in which features are inserted)—translates into the lack of functional projections hosting tense and structural case features. An issue we have only touched upon implicitly so far is the question of whether restructuring infinitives include an embedded subject. Note that the lack of structural case only entails the lack of an external argument, if one assumes a strong version of Burzio's generalization. Similarly, the lack of structural case does not necessarily entail the lack of a voice head (cf. for instance Kratzer 1994, 1996 who assumes that there are different voice heads such as active, passive etc.). In this chapter, we will not take any position on these issues but rather seek (theory) independent evidence for the presence or absence of an infinitival (PRO) subject. The main claim we will make is that control infinitives fall into two classes—infinitives that lack a syntactic subject and infinitives that include a syntactic subject—and that the absence vs. presence of a syntactic subject corresponds to the restructuring vs. nonrestructuring distinction. We will thus again conclude that restructuring infinitives are represented as subject-less VPs (cf. (189)a), whereas the (minimal) structure for non-restructuring infinitives is as in (189)b.

Infinitives with/without a syntactic subject

(189) a.

Restructuring infinitive

b. Non-restructuring infinitive

TP

the tractor

227

TP

to repair

the tractor

to repair

In the first part of this chapter, we will provide arguments for the structures in (189). In the second part, we will discuss the semantic properties of subjectless infinitives and suggest a theory of control that will account for the interpretation of infinitival subjects in the absence of a syntactic PRO subject.

2. Infinitives with and without a syntactic subject The question of whether control infinitives include an embedded syntactic (PRO) subject (which goes hand in hand with the question of whether infinitives are clauses or smaller predicates) has been a longstanding issue in both the syntactic as well as the semantic literature. The common approach in the 1980s was that control infinitives are clauses (IPs or CPs) syntactically and propositions semantically.79 One of the main reasons for the clausal nature of control infinitives in these theories had to do with the idea of uniformity of phrase structure. Since certain control infinitives can involve overt 79

Throughout this chapter, we will use the term proposition to refer to a clause/predicate including a subject (i.e., a predicate that does not have an unsaturated subject position).

228

The infinitival subject

CP-material, control infinitives were generally considered to be clauses (whether they involve CP-material or not). The so-called PRO-theorem provided another theory-internal reason for a clausal structure of control infinitives. Since Chomsky's Lectures on Government and Binding, PRO has been assigned a special status—it is both anaphoric and pronominal. To avoid violations of binding theory, PRO had to be ungoverned. A C-projection in control infinitives was then necessary to protect PRO from government by the matrix verb. Finally, assuming that control infinitives are interpreted as propositions (but see below for a different view), the Projection Principle, which states that an argument is present at all levels of representation, requires the presence of a (PRO) subject in the syntactic structure. A unified account for (all) control infinitives, however, has also been challenged in a variety of works on infinitives. On the one hand, with the development of a Case-theoretic account of PRO (Chomsky and Lasnik 1993, Martin 1996) one of the major theoretical motivations for a CP-structure for control infinitives disappears. Many contemporary syntactic approaches assume that complementizerless and non-wh-infìnitives are smaller categories. To take just one approach, Boskovic (1996, 1997) argues that control infinitives are IPs, while maintaining the idea that control infinitives project an embedded PRO subject which is present in the syntactic representation. On the other hand, considering the semantic properties of different infinitival constructions, many works have shown that a uniform treatment of all control infinitives seems untenable. Rochette (1988), for instance, proposes that the size of an infinitive corresponds to its semantic category: CP-infinitives denote propositions, IP-infinitives denote events, and subjectless VP-infinitives denote actions. In standard Lexical Functional Grammar (LFG), control infinitives are considered as subjectless VPs in syntax and propositions in semantics (cf. Bresnan 1982). Similarly, Chierchia (1984a, 1984b) argues that control infinitives are properties—i.e., subjectless predicates—in syntax and semantics and become associated with a subject later on in context by means of a semantic/pragmatic control principle. While our analysis of restructuring will provide further support for Chierchia's claim that certain infinitives (namely restructuring in-

Infinitives with/without a syntactic subject

229

finitives) are subjectless predicates syntactically and semantically (cf. section 2.2), we will also conclude in section 3.2 that the syntax and the semantics of infinitival constructions have their own lives, and that the syntactic properties cannot be fully reduced to the semantics of a given construction.

2.1.

Binding

The main argument for the lack of an infinitival subject in restructuring infinitives and the presence of a subject in non-restructuring infinitives (i.e., the structures in (189)) that we wish to discuss in this section comes from the binding properties of infinitival constructions in German. Let us start with an example such as (190)a which involves an anaphor embedded in the infinitival complement and see how the binding relation is established. Under a clausal account to restructuring (i.e., an account which assumes that restructuring infinitives include an embedded infinitival PRO subject), the anaphor is bound by the embedded subject as illustrated in (190)b. Under our proposal, on the other hand—i.e., under the assumption that restructuring infinitives lack an embedded PRO subject (cf. (189)a)—there is no embedded subject in restructuring infinitives and the embedded anaphor is bound directly by the matrix subject (cf. (190)c). (190) Embedded anaphors: bound directly or indirectly a.

b. c.

weil der Hansh sichh zu rasieren versuchte since the Johnh SELFh to shave tried 'since John tried to shave himself since John [PROh SELFh to shave] tried since Johnh [SELFh to shave] tried

Thus, both the [+PRO] approach in (190)b and the [-PRO] approach in (190)c provide a way of accounting for the coreference relation between the embedded anaphor and the matrix subject in examples such as (190)a. The major difference between the two approaches is whether the anaphor is bound by the matrix subject directly or whether the anaphor is bound by PRO—the relation of coreference

230

The infinitival subject

with the matrix subject thus being established only indirectly. While (empirically) indistinguishable in examples such as (190)a, the two approaches make different predictions in environments where the matrix subject is for some reason absent or unavailable. Under the [+PRO]-approach, the presence or absence of a matrix subject should not affect the binding properties in the embedded infinitive, since the PRO subject would still be available as an antecedent (unless additional assumptions are made). Under the [-PRO]approach, on the other hand, the elimination of the matrix subject would eliminate the only available binder for the anaphor in the infinitive, and the structure should thus become ungrammatical. As we will see, the binding facts in German will support the analysis suggested here (i.e., the [-PRO approach]), since anaphors are indeed illicit in restructuring infinitives when there is no appropriate antecedent in the matrix predicate. However, since anaphors are licensed in non-restructuring infinitives, we will conclude that non-restructuring infinitives project a PRO subject. We will first discuss contexts in which the infinitive is the only internal argument of the matrix predicate. Ditransitive constructions—i.e., contexts in which the selecting verb combines with an infinitive and a dative argument will be discussed in section 2.I.2. 80

2.1.1. Transitive constructions The first fact to be noted is that in German, anaphors are not generally impossible in infinitival complements combining with an impersonal matrix predicate (i.e., a predicate lacking an overt controller). As can be seen in (191), embedded anaphors can occur in impersonal passive constructions (cf. (191)a) and impersonal adjectival constructions (cf. (191)b). Furthermore, embedded anaphors are possible in the impersonal non-restructuring versions of constructions involving a potential restructuring verb (cf. (191)c; recall that infini 80

Ditransitive constructions in which the verb combines with an infinitive and an accusative argument will be ignored since these constructions are generally nonrestructuring infinitives (cf. Sabel 1996) and as predicted, embedded anaphors will always be licensed by the embedded PRO subject.

Infinitives with/without a syntactic subject

231

tives involving a structural accusative object in passive or unaccusative matrix contexts are non-restructuring infinitives). (191) Embedded anaphors in non-restructuring infinitives a.

b.

c.

Es wurde beschlossen It was decided PRO¡ sichi den Fisch mit Streifen vorzustellen PRO¡ SELF) the fish with stripes-ACC to-imagine 'They decided to imagine what the fish would look like with stripes' Es war notwendig It was necessary PRO¡ sich¡ einen Wagen zu kaufen PRO¡ SELFj a car-ACC to buy 'It was necessary to buy oneself a car' Es wurde versucht It was tried PROj sich, den Fisch mit Streifen vorzustellen PROj SELF; the fish with stripes-ACC to-imagine 'They tried to imagine what the fish would look like with stripes'

As indicated in the examples above, we assume that the anaphors in (191) are bound by an embedded PRO subject. Note that it cannot be assumed that the anaphors are bound directly by an implicit argument in the matrix clause in (191). As is shown in (192)b, implicit passive arguments cannot bind anaphors in German.81 The examples in (191) and in particular the contrast between these examples and the ungrammatical (192)b thus provide evidence for the existence of an embedded subject in non-restructuring infinitives (a further argument for this claim will be provided in section 2.1.2).

81

There are some well-known counter examples to this claim. In particular, anaphors bound by an implicit argument seem to improve in generic contexts (at least for some speakers). Since the examples we use are non-generic, the judgements are consistent and this interfering factor does not affect the argument made in the text.

232

The infinitival subject

(192) Implicit arguments a.

b.

Frederikf hat sichf ein Haus gekauft Frederic^ has SELFf a house bought 'Frederic bought himself a house' Ein Haus wurde (*sich) gekauft A house was (*SELF) bought Ά house was bought (*oneself)'

Since the examples in (191) are non-restructuring infinitives (as evidenced by the embedded accusative objects), we correctly predict that infinitival subjects are projected in these examples, and binding of the anaphors is thus straightforward. Turning to restructuring infinitives, it is easy to see that the situation changes drastically. In a clear restructuring configuration (e.g., a construction involving "long" passive) which does not involve a potential matrix antecedent, inherently reflexive predicates such as sich vorstellen 'to imagine/recall the picture of cannot occur in the infinitival complement (cf. (193)a). Furthermore, as is shown in (193)b, benefactive anaphors associated with the implicit subject are impossible (however, the sentence is grammatical without the anaphors). (193) Embedded anaphors in restructuring infinitives a. *weil {sich} der Fisch {sich} vorzustellen since {SELF} thefish-NOM {SELF} to-imagine versucht/begonnen/beabsichtigt wurde tried/begun/intended was 'since they tried/began/intended to recall the image of the fish' b. weil {*sich} der Turm {*sich} zubauen since Ì*SELF} thetower-NOM j*SELF} to build versucht/begonnen/beabsichtigt wurde tried/begun/intended was 'since they tried/began/intended to build themselves the tower'

Infinitives with/without a syntactic subject

233

Assuming that restructuring infinitives lack an embedded (syntactic) subject and implicit arguments cannot bind anaphors in German, the contrasts in (191) and (193) can straightforwardly be accounted for. Since restructuring infinitives do not include an embedded subject there is no possible antecedent for anaphors or reflexive predicates, and unbound anaphors are thus ruled out by the conditions of binding theory..

2.1.2. Ditransitive constructions A further argument for the analysis proposed here comes from infinitival constructions combining with verbs that also select a dative argument. Let us start with an observation about dative arguments in German that will be of importance for the discussion to follow. As has been pointed out by Grewendorf (1984, 1988), anaphors cannot be bound by dative arguments in German (cf. (194)). Whatever the nature of this restriction is, it will allow us to draw certain conclusions about the presence or absence of infinitival subjects. (194) Dative antecedents weil der Hansh derMariam sichu,m since the Johnh-NOM to Marym-DAT S E L F ^ auf dem Photo zeigte in the picture showed 'since John showed Mary himself/*herself in the picture' Consider first the example in (195)a with the (potential) restructuring verb manage. Since the infinitive involves an accusative object which cannot have been case-marked by the matrix predicate (recall that manage is an unaccusative predicate in German), this construction can only be a non-restructuring infinitive. As expected under our account, an anaphor can occur in this context, since non-restructuring infinitives project an embedded syntactic subject which can function as the antecedent for the embedded anaphor. Similarly, the ditransitive constructions in (195)b,c involving potential restructuring verbs allow embedded anaphors which we assume are bound by the em-

234

The infinitival subject

bedded PRO subject.82 (195) a.

b.

c.

Es ist ihm gelungen It is him-DAT managed PRO¡ sich{ einen Turm zu bauen PRO, SELF) [a tower]-ACC to build 'He (has) managed to build himself a tower' Sie hat dem Hansh erlaubt/verboten She has Johnh-DAT allowed/forbidden PROh sichh den Fisch mit Streifen vorzustellen PROh SELFh the fish with stripes-ACC to-imagine 'She allowed/forbade John to imagine what the fish would look like with stripes' Sie hat dem Hansb erlaubt/verboten She has Johnh-DAT allowed/forbidden PROh sichh den Turm anzuschauen PROh SELFh the tower-ACC to-look-at 'She allowed/forbade J. to look at/ investigate the tower'

Importantly, the grammaticality of the examples in (195) (i.e., examples with a dative "controller" for the embedded subject) provides a further argument for the presence of an embedded subject in nonrestructuring contexts. If one were to assume that potential restructuring infinitives are generally subjectless predicates—i.e., if examples such as the ones in (195) would lack an embedded subject, it would not be clear how the anaphors are bound in these examples. Under this view, the dative arguments would be the only potential binders for the embedded anaphors, however, since dative arguments cannot bind anaphors in German (cf. (194)), the embedded anaphors 82

Note that only (195)a is necessarily a non-restructuring infinitive. Concerning the case properties, the examples in (195)b,c could in principle also be restructuring infinitives, since the matrix predicates are potential structural case assigners and hence accusative could be assigned by the matrix predicate. However, as we will see below, the binding properties provide clear evidence for the claim that these infinitives involve embedded PRO subjects—i.e., are non-restructuring infinitives. See also Wurmbrand (2000a) for a way to distinguish between case assignment in the matrix vs. embedded predicate.

Infinitives with/without a syntactic subject

235

would end up without an antecedent, and the sentences should be ungrammatical (contrary to fact). We thus conclude that nonrestructuring infinitives indeed project a syntactic (PRO) subject. Turning to (true) restructuring constructions, we find again a clear contrast between restructuring and non-restructuring infinitives concerning the possibility of embedded anaphors. If examples such as the ones in (195) are changed to (clear) instances of restructuring infinitives, anaphors become impossible. This is illustrated in (196) by the impossibility of reflexive predicates in infinitives involving "long" passive—i.e., restructuring infinitives (the position of the reflexive pronoun has no effect on the ungrammaticality of these examples). Note in particular (196)c which shows that long passive (and hence restructuring) is possible in this context, however, only when the embedded predicate is not reflexive. (196) a. *weil der Fisch dem Hansb sichh since the fish-NOM the Johnh SELFh mit Streifen vorzustellen gelungen ist with stripes imagine managed is 'since John managed to imagine what the fish would look like with stripes' b. *weil der Fisch dem Hansh sichh since the fish-NOM the Johnh SELFh mit Streifen vorzustellen erlaubt wurde with stripes imagine allowed was 'since they allowed John to imagine what the fish would look like with stripes' c. weil ein Turm dem Hansh (*sichh) since a tower-NOM Johnh-DAT (*SELFh) anzuschauen verboten/empfohlen wurde to look at forbidden/recommended was 'since they forbade/recommended (to) John to look at/investigate (*refl.) a (certain) tower' To conclude this section, the distribution of anaphors in German shows the following contrast: non-restructuring infinitives (i.e., infinitives characterized by the presence of a structural case posi-

236

The infinitival subject

tion/assigner in the infinitive) allow embedded anaphors whereas restructuring infinitives prohibit embedded anaphors when the matrix predicate does not include an appropriate binder for the anaphor. We have argued that this contrast follows from the presence vs. absence of infinitival subjects and that neither an account that assumes that infinitives generally involve a syntactic subject, nor an account that assumes that infinitives generally lack a syntactic subject seems to be able to account for these contrasts without further assumptions. In the next section, we will show that the property of subjectless infinitives is independently motivated by the interpretation of these constructions and we will suggest an account of control that captures these generalizations.

2.2. Semantic control properties of restructuring infinitives The assumption that restructuring infinitives are subjectless predicates raises some important questions about the interpretation of these infinitival constructions, which we will turn to in section 3. In this section, we will first provide an empirical summary of the semantic control properties of restructuring infinitives. Since Williams (1980), it has been widely accepted that the control phenomenon subsumes two sub-cases: obligatory and non-obligatory control. What we will see in this section is that restructuring infinitives are obligatory control infinitives. 2.2.1. The obligatory/non-obligatory control distinction While most authors agree that an appropriate account of control has to distinguish between obligatory and non-obligatory control, the actual classification of infinitival constructions as obligatory vs. nonobligatory control infinitives is subject to some debate. To avoid confusion, we will first outline how we will use the obligatory vs. nonobligatory control distinction. The examples in (197) and (198) illustrate a number of control phenomena that have been distinguished in the literature. The example in (197)a is an instance of exhaustive

Infinitives with/without a syntactic subject

237

control (cf. Landau 1999/2000)—i.e., the infinitival subject refers exhaustively to the matrix subject John. Examples such as the one in (197)b from Kawasaki (1993) involve arbitrary control—i.e., the reference of the infinitival subject is left unspecified. The example in (197)c is an instance of long-distance control (cf. Manzini 1983a; Landau 1999/2000)—i.e., control is established across another (more local) potential controller. Examples such as the one in (197)d involve implicit control—i.e., the implicit argument of difficult functions as the controller of the infinitival subject (see Landau 1999/2000 for arguments against an arbitrary control analysis for examples of this sort). The example in (197)e is an instance of split control—i.e., the infinitival subject is controlled jointly by the matrix subject and object. And (197)f represents partial or imperfect control—i.e., the controller denotes a subset of the reference set of the infinitival subject (cf. Martin 1996; Petter 1998; Wurmbrand 1998a, 1998b; Landau 1999/2000).83 (197) a. b. c. d. e. f.

John tried to leave It is dangerous for babies to smoke around them Mary knows that it would help Bill to behave herself in public It was difficult to leave John persuaded Mary to leave together We thought that the chair preferred to gather at 6

Finally, the German examples in (198) illustrate the phenomenon of control shift or variable control which is found in many languages.84 As is illustrated in (198)a, the infinitival subject can be associated with either the matrix subject or object. The availability of variable control depends on the type of predicate (see below), as well as cer83

For many speakers, partial control is a marked interpretation, and there are also speakers who reject this form of control (see also Landau 1999/2000 for a discussion of partial control in other languages and ways to facilitate this interpretation). The judgements and contrasts reported in this chapter are from speakers who readily allow partial control. 84 As has been noted by Comrie (1984), the possibility of control shift is subject to variation (both across and within languages).

238

The infinitival subject

tain thematic and pragmatic properties of the construction (see for instance Farkas 1988; Petter 1998; Landau 1999/2000). As can be seen in (198)b-d for instance, constructions involving the (object control) verb bitten 'ask' allow control shift best when the infinitival predicate is passivized (cf. (198)c) or involves a modal predicate (cf. (198)d)—i.e., when the (surface) infinitival subject bears a nonagentive theta role (a similar effect has been noted in English for promise-constructions). (198) a.

lchx habe ihm¡ angeboten I¡ have him-DAT offered [SUBJyj mich zu erschießen] [SUBJj/j me/myself to shoot] Ί offered him to shoot myself' Ί offered him that he could shoot me' b. Icht habe ihnj gebeten I¡ have him¡-ACC asked [SUBJ??i/j ins Kino zu gehen] [SUB J??i/j in-the cinema to go] Ί asked him to go to the cinema' ??'I asked him that I could go to the cinema' c. Ich¡ habe ihn} gebeten Ij have himj-ACC asked [SUBJi/??j ins Kino gehen zu dürfen] [SUB Jj/rç in-the cinema go to may] ??'I asked him that he would be allowed to go to the cinema' Ί asked him to allow me to go to the cinema' d. Ichj habe ihn} gebeten Ij have him-ACC asked [SUBJj^j erschossen zu werden] [SUBJyq shot to A U X p a s s ] ?'I asked him to allow/arrange that he get shot (by somebody)' Ί asked him to allow/arrange that I get shot (by somebody)'

Infinitives with/without a syntactic subject

239

As far as we are aware, the classification of exhaustive control as in (197)a as obligatory control is uncontroversial. Similarly, arbitrary and long-distance control as in (197)b,c, respectively, are generally considered as the prototypical cases of non-obligatory control. Concerning the examples in (197)d-f and (198), on the other hand, the classification is less straightforward. Like obligatory and unlike nonobligatory control infinitives, partial, split, and variable control constructions require an antecedent for the infinitival subject in the higher predicate (i.e., long distance or arbitrary control are prohibited). However, the antecedent does not necessarily fully coincide with the infinitival subject—it suffices that the antecedent is included in the reference set of the embedded subject. Unlike obligatory and like non-obligatory control, on the other hand, partial, split, and variable control do not involve a unique pre-determined controller but allow different possibilities for the interpretation of the infinitival subject. We will take this latter property to group partial, split, and variable control as part of non-obligatory control. Infinitives that in principle allow more than one interpretation of the embedded subject will be considered as non-obligatory control infinitives, whereas infinitives that—irrespective of the syntactic and pragmatic context—require a particular interpretation of the embedded subject will count as obligatory control infinitives. 85 We thus assume that the

This classification requires some clarification concerning implicit control. As we will see below, implicit control in impersonal iAj-constructions (i.e., constructions involving an exhaustive control predicate) and in examples such as (197)c (cf. Landau 1999/2000) is no different from the control properties of the non-implicit versions of these constructions. That is, the antecedent of the infinitival subject is pre-determined as the implicit argument of the matrix predicate and no other interpretation is possible. Given the implicit nature of the antecedent, there are of course different values that the implicit argument and hence the infinitival subject can receive, however, what is crucial is that there is no interpretation in which the infinitival subject and the implicit argument receive a different interpretation (at the same time). Implicit control in variable, split, or partial control constructions, on the other hand, does not involve a pre-determined antecedent (i.e., the infinitival subject does not have to coincide fully with any particular argument of the matrix predicate). Examples such as Es wurde angeboten das Haus zu verkaufen (Lit. 'It was offered to sell the house) involve two implicit matrix arguments, and a range

240

The infinitival subject

obligatory/non-obligatory control distinction is a basic semantic distinction. Note that by grouping partial, split, and variable control with nonobligatory control, we do by no means claim that non-obligatory control is a homogeneous phenomenon or that all forms of nonobligatory control involve the same control mechanism. As discussed in detail in Landau (1999/2000) there are very important syntactic and semantic differences between the different forms of nonobligatory control. In particular, as mentioned above, partial, variable, and split control differ from arbitrary and long-distance control in whether a local antecedent is required or not. In partial, variable, and split control constructions the embedded subject has to be associated (at least partially) with an antecedent in the matrix predicate (however, the antecedent is not pre-determined and can be variable), whereas no local antecedent is required in arbitrary and long-distance control constructions. Given this property, Landau (1999/2000) argues that exhaustive, partial, split, and variable control are to be characterized as obligatory control and derived by the same control mechanism. According to Landau, the obligatory vs. non-obligatory control distinction is a syntactic rather than a semantic distinction: VP-internal infinitives trigger obligatory control, VP-external infinitives allow non-obligatory control. The exhaustive vs. partial control distinction, on the other hand, is based on a semantic difference: tenseless infinitives (e.g., infinitives combining with implicative predicates) yield exhaustive control, whereas tensed infinitives (desiderative and interrogative infinitives) allow partial control. The main reason for why we consider the obligatory/nonobligatory control distinction a semantic distinction rather than a syntactic distinction is that in a number of contexts, the syntactic configuration and the tense properties are not sufficient to determine which form of control is possible, and additional lexical/semantic specifications are necessary to overrule the syntactic control properof interpretations are possible (e.g., 'X offered Y that Y could sell the house', 'X offered Y that X would sell the house'). We thus assume that implicit control per se is neither obligatory nor non-obligatory control but that the control properties depend on the type of the matrix predicate.

Infinitives with/without a syntactic subject

241

ties. As Landau notes, certain predicates (e.g., silly, kind, polite, irresponsible of) require coreference between the complement of the adjective and the embedded subject, independently of the syntactic configuration these predicates appear in. Similarly, certain infinitival adjuncts (e.g., right-adjoined temporal adjuncts or purpose clauses) are specified for obligatory control although the syntactic properties would be compatible with a non-obligatory control interpretation (Uchiumi, p.c.). Finally, the tense properties do not fully correlate with the exhaustive/partial control distinction. In particular, we will see below that certain tensed infinitives only yield exhaustive control and that not all desiderative predicates allow partial control (e.g., verbs like order, permit which combine with irrealis/tensed infinitives require exhaustive control). Hence, an additional lexical specification of obligatory control is necessary for a number of desiderative predicates that prohibit partial control. While we take these lexical/semantic factors to reflect the basic obligatory/non-obligatory control distinction, we follow Landau in that the syntactic properties are crucial for the determination of nonobligatory control (in our sense). Thus, the difference mentioned above between partial, variable, and split control, on the one hand, and arbitrary and long-distance control, on the other hand, is the result of different syntactic configurations (e.g., whether the infinitive is VP-internal or VP-external) and hence different forms of control. Since our major concern here is the nature of the obligatory vs. nonobligatory control distinction as defined above and the question of where and how this distinction is established, we will not go into detail regarding the technical details of non-obligatory control but refer the reader to Landau (1999/2000). 2.2.2. Restructuring infinitives involve obligatory control Before we turn to the relevance of the obligatory vs. non-obligatory control distinction for restructuring, let us start with an empirical delineation of obligatory vs. non-obligatory control as suggested in the previous section in German. As we have seen above, infinitives combining with verbs like offer or ask allow either subject or object con-

242

The infinitival subject

trol (modulo certain thematic/pragmatic restrictions mentioned). The same situation holds for verbs like propose or promise. As is illustrated in (199)a, infinitives combining with vorschlagen 'propose' allow either subject or object control; infinitives combining with versprechen 'promise' which are generally subject control verbs (cf. (199)b), allow object control for instance when the embedded predicate involves a modal (cf. (199)c). (199) a.

Ich¡ habe ihm} vorgeschlagen I¡ have hinij-DAT proposed [SUBJj/j mich zu erschießen] [SUBJyj me/myself to shoot] Ί proposed to him that I shoot myself Ί proposed to him that he shoot me' b. Ich, habe ihm¡ versprochen I¡ have hinij-DAT promised [SUBJi/??j ins Kino zu gehen] [SUB J;/,?) in-the cinema to go] Ί promised him to go to the cinema' ??'I promised him that he would go to the cinema' c. Ich¡ habe ihm^ versprochen I¡ have hinij-DAT promised [SUBJyij ins Kino gehen zu dürfen] [SUBJ^j in-the cinema go to may] Ί promised him he would be allowed to go to the cinema' ?'I promised him I would be allowed to go to the cinema'

Despite the fact that variable control (control shift) is quite readily available in German, there are constructions that clearly block this form of control. Infinitives combining with verbs like forbid, allow, or recommend, prohibit any kind of flexible interpretation and only allow object control (cf. (200)). (200) a.

Icht habe I¡ have [SUBJ.j/j [SUBJ,¡y:

ihm} verboten/erlaubt/empfohlen hinij-DAT forbidden/allowed/recommended mich zu erschießen] me/myself to shoot]

Infinitives with/without a syntactic subject

243

Ί forbade/allowed/recommended (to) him to shoot me' *'I forbade/allowed/recommended (to) him that I should/could shoot myself b. lieh, habe ihmi verboten/erlaubt/empfohlen I¡ have him-DAT forbidden/allowed/recommended [SUBJ, i / ? j ins Kino gehen zu dürfen] [SUBJ.^j in-the cinema go to may] ?'I forbade/allowed/recommended (to) him to be allowed to go to the cinema' *'I forbade/allowed/recommended (to) him that I should/could be allowed to go the cinema' c. 7/c/ij habe ihmj verboten/erlaubt/empfohlen I¡ have himpDAT forbidden/allowed/recommended [SUBJ.^j gewählt zu werden] [SUBJ.y?j elected to A U X P A S S ] ?'I forbade/allowed/recommended (to) him to get elected' *'I forbade/allowed/recommended (to) him that I should/could get elected' Similarly, only a subgroup of infinitival constructions allows partial control. As has been noticed by Williams (1980), collective predicates such as to meet, to gather, together (i.e., predicates that require a plural subject or associate) can occur in certain infinitival constructions that involve a singular controller in the matrix predicate (cf. his examples in (201)a vs. (201)b). (201) a. * I will try to meet at 6 b. I decided to meet at 6 Looking at a range of infinitival constructions, we find again that some infinitival constructions allow partial control (cf. the examples in (202)), whereas others block this form of interpretation (cf. the examples in (203)). As is illustrated in (202)a, constructions with variable control verbs (such as propose), not only allow subject and object control but also allow partial subject and object control ("k" refers to referents that are provided by the context).

244

The infinitival subject

(202) a.

b.

c.

Der Direktor¡ hat ihm} vorgeschlagen The principali has hinij proposed [SUBJ sich im Schloß zu versammeln] [SUBJ SELF in-the castle to gather] 'The principal proposed to him to gather in the castle' => SUBJi+k (e.g., the teachers), SUBJj+k (e.g., the students), SUBJi+j+k (e.g., the whole school) Eri beschloß/erwog/plante He¡ decided/contemplated/planned [SUBJi+k sich im Schloß zu versammeln] [SUBJi+k SELF in the castle to gather] 'He decided/contemplated/planned to gather in the castle' Hans{ beschloß/erwog/plante John¡ decided/contemplated/planned [SUBJi+k sich gemeinsam zu bewerben] [SUBJi+k SELF together to apply] 'John decided/contemplated/planned to apply together'

In contrast to the examples in (202), the examples in (203) show that infinitives combining with verbs like dare, try, forbid etc. cannot contain collective predicates when the matrix controller is singular—i.e., partial/imperfect control is impossible in these constructions. Thus, variable and partial control split infinitival constructions into two groups which we labeled obligatory vs. non-obligatory control. (203) a. * Der Bürgermeister wagte/versuchte/begann/vergass The mayor dared/tried/began/forgot [sich im Schloß zu versammeln] [SELF in the castle to gather] 'The mayor dared/tried/began/forgot to gather in the castle' b. *Es ist ihm gelungen[sich im Schloß zu versammeln] It is him managed [SELF in the castle to gather] 'He managed to gather in the castle' c. *Der Beamte hat ihm verboten/erlaubt/empfohlen The clerk has him disallowed/allowed/recommended

Infinitives with/without a syntactic subject

245

[sich gemeinsam zu bewerben] [SELF together to apply] 'The clerk disallowed/allowed/recommended (to) him to apply together' Having established that there are two basic classes of control constructions, let us return to lexical restructuring predicates—i.e., predicates that allow long passive in German—and see what type of control is involved in these constructions. As is illustrated in Table 24, there is a striking correlation between restructuring and obligatory control. That is all predicates that allow long passive are obligatory control predicates and prohibit any form of non-obligatory control. Table 24.

Lexical restructuring and obligatory vs. non-obligatory control

Verb

Gloss

Lexical restructuring predicates

Non-obligatory control

Obligatory control

beabsichtigen beginnen gelingen mißlingen vergessen vermeiden versäumen versuchen wagen empfehlen erlauben gestatten untersagen verbieten

intend (want) begin manage fail forget avoid miss, neglect try dare recommend allow permit prohibit forbid

+ + +

-

+ +

-

+

+

-

+

-

% ? + + + +

% + +

-

+ + + + + + + + + + +

Note that the correlation between obligatory control and restructuring is only a one way correlation—i.e., obligatory control does not entail restructuring (however, restructuring entails obligatory control). For instance, verbs like anfangen 'start' or aufhören 'stop' clearly trigger obligatory control, however, due to the fact that these predicates are particle-verb constructions, they do not allow restructuring (see

246

The infinitival subject

Chapter 2, section 3.5). As mentioned before, restructuring is subject to a number of syntactic and semantic conditions that all have to be satisfied. Thus, like with the tense interpretation, obligatory control is a necessary condition for restructuring but it is not a sufficient one. Assuming our analysis is correct and restructuring infinitives lack an embedded subject, the question that arises is how the control effect can be accounted for, or, in other words, how an interpretation can be assigned to the understood embedded subject of a restructuring infinitive. The next section will address this question. 3. What is under control 3.1. Main proposal The numerous works on the obligatory vs. non-obligatory control distinction have identified a number of properties that distinguish the two forms of control (see in particular Williams 1980; Manzini 1983a; Lebeaux 1985; Hornstein 1999; Landau 1999/2000).86 In obligatory control constructions, the antecedent is thematically or grammatically uniquely determined, the antecedent is obligatory, there are strict locality constraints on the relation between the antecedent and the embedded subject, only a sloppy interpretation is available under ellipsis, and only a de se interpretation is possible for the embedded subject. In non-obligatory control constructions, none of these properties has to hold. What the various obligatory control properties have in common is that obligatory control involves a very tight relation (both structurally and semantically) between the antecedent and the infinitival subject. The major aim of obligatory control analyses is thus to derive this close relation between the infinitiSome of the properties are controversial (see Landau 1999/2000 for a critical overview). However, much of the controversy appears to be the result of terminological differences. As we have seen in section 2.2.1, the classification as obligatory vs. non-obligatory control is not always straightforward. Given the various forms of control which all come with their own properties and restrictions, much depends on the classification one uses.

What is under control

247

val subject and the antecedent. Possibly the most common way to do so is by assuming that obligatory control expresses an anaphoric relation (cf. Manzini 1983a; Bouchard 1984; Köster 1984; Borer 1989). A different approach is that obligatory control is a (very local) agreement relation (cf. Landau 1999/2000). Finally, the most recent proposals reduce the effects of obligatory control to the claim that the antecedent and the embedded subject are one and the same element which occurs in two positions—i.e., obligatory control is analyzed on a par with raising (O'Neil 1997; Hornstein 1999; Manzini and Roussou 2000). In contrast, non-obligatory control is subject to looser conditions, which is generally attributed to the pronominal (Manzini 1983a; Bouchard 1984, 1995; Köster 1984; Hornstein 1999) or logophoric (Grinder 1970; Kuno 1975; Lebeaux 1985; Landau 1999/2000) nature of the embedded subject. While syntactic approaches to control are intended to derive the properties of obligatory vs. non-obligatory control constructions, they are generally not aimed at explaining or predicting which predicates can or have to involve which form of control (though see Landau 1999/2000 for an exception). Going back to the contrasts noted in the previous sections, the question that arises for the determination of obligatory vs. non-obligatory control is how it is ensured that variable and partial control are excluded with matrix verbs like order, permit, recommend, try etc., but allowed with verbs like offer, propose, ask, or decide. A different syntax for these constructions would presumably derive the different control properties, however, the question of how we can guarantee that these constructions involve different structures nevertheless remains. In other words, it is not obvious how the syntax knows what type of subject or construction can or has to be chosen (e.g., anaphoric vs. pronominal subject, PRO¡pro vs. trace etc.), or what kind of control mechanism applies. The general solution to questions of this sort is to invoke additional lexical or selectional specifications. To account for the differences in the control behavior of the constructions mentioned above, it seems necessary for any approach to supplement the syntax of control with certain lexical or semantic specifications about the possible control relations. Note that shifting the burden to some unsystematic pragmatic accident (i.e., an account that assumes that there is no dif-

248

The infinitival subject

ference in the lexical properties and the syntax of these constructions) would miss the generalization that these two classes of control in fact correlate with a number of syntactic and semantic properties (as we will show below). Thus, taking verbs like order, permit, or recommend, for instance, the lexical meaning of these predicates has to include the information that the object of these verbs is the sole controller of the embedded subject. On the other hand, there is no pre-determined control relation in constructions with verbs like offer, propose, ask, or decide. This lexical or semantic aspect of the determination of the control relation is the core feature of many semantic approaches to control. Chierchia (1983, 1984a, 1984b) for instance, denies the relevance of syntactic control and suggests instead that (obligatory) control infinitives are properties rather than propositions—i.e., they do not involve a subject. The control relations, according to Chierchia's approach, are entailment relations that are made explicit as part of the meaning of the selecting predicate. More specifically, the subject is "added" later on—i.e., in context—by the application of meaning postulates such as the one in (204) (where q¡ is a context dependent modal operator; (204)b is a rough paraphrase of (204)a). (204) a. b.

try' (P) (x) - > q P (x) (Chierchia 1984a: 34) whenever χ tries to bring about P, then in all the contextually relevant situations (namely those where what χ tries actually succeeds) χ does Ρ

Thus, in case of try, the obligatory control effect arises as a result of the entailment relation built into the meaning of the higher verb. German infinitival constructions provide support both for and against a direct translation of a Chierchia style account into syntactic terms. As we have seen in section 2.1, a certain class of infinitival constructions (restructuring infinitives), which crucially only involves obligatory control infinitives, displays syntactic properties that strongly point to the lack of a syntactic subject. Although these properties indicate that there is a correlation between subjectless infinitives and obligatory control, we will demonstrate below that the correlation is only a one-way correlation and that obligatory control

What is under control

249

does not entail a subjectless infinitive (in syntactic terms). The conclusion will be that the properties of German infinitives provide evidence for the necessity of two distinctions—a semantic distinction between obligatory and non-obligatory control, and a syntactic distinction between subjectless infinitives vs. infinitives with an embedded subject. Although these distinctions show an overlap they cannot be reduced to each other. The proposal that we would like to put forward is that the crucial distinction between obligatory and non-obligatory control is whether the antecedent is determined lexically/semantically or syntactically. In particular, we will assume that non-obligatory control predicates do not include 'instructions' for the interpretation of the embedded subject. An infinitival subject is projected as part of the syntactic structure and the antecedent is determined purely syntactically (by whatever one considers the correct control mechanisms). The meaning of the infinitival construction is then determined strictly compositionally from the output of the syntactic structure. Thus, nonobligatory control constructions are clauses syntactically and propositions semantically. Obligatory control on the other hand, is determined lexically/semantically. In line with semantic approaches to control, we assume that the obligatory control effect is the result of an inherent lexical property—e.g., an entailment relation built into the meaning of the selecting predicate (which identifies the infinitival subject with a pre-specified antecedent). Looking at the syntactic and semantic properties of obligatory control infinitives, however, we will see that in the case of obligatory control infinitives, the syntax/semantics mapping is not perfect. While obligatory control infinitives can be represented as subjectless predicates (at least in German), the lack of an infinitival subject (in syntax) is not obligatory. That is, we will provide evidence for the presence of a syntactic subject in obligatory control constructions. There seem to be two basic ways to accommodate this mismatch. It could be assumed that obligatory control infinitives (but not nonobligatory control infinitives) that include a syntactic subject are translated into semantic properties. This line is taken for instance by Chierchia (1989) who suggests that PRO functions essentially as a

250

The infinitival subject

property abstractor—i.e., a lambda-abstractor that turns the infinitive into a property. Note that it is crucial to restrict property abstraction to obligatory control infinitives. As has been argued extensively by Landau (1999/2000), the lack of coreference between the understood infinitival subject and its antecedent in partial control contexts necessitates the presence of a separate infinitival subject (in syntax and semantics)—i.e., property abstraction would make it impossible to assign a partial control interpretation. The second way to accommodate the syntax/semantics mismatch found in obligatory control constructions is to assume that obligatory control infinitives that include a syntactic subject are indeed translated into semantic propositions, but that these propositions are nevertheless subject to the effects of obligatory control. Thus, in this approach, obligatory control is a lexical property of certain predicates, however, this lexical specification is not tied to the assumption that obligatory control infinitives are semantic properties. We will pursue this option here and assume that obligatory control predicates are defined as functions that can either apply to properties or propositions. Crucially, however, obligatory control is built into the meaning of both definitions. 87 Assuming that obligatory control is determined lexically/semantically, a syntactic subject and the application of (syntactic) control mechanisms is in a sense vacuous in obligatory control constructions, since the antecedent of the infinitival subject is already pre-specified as part of the meaning of an obligatory control predicate. We claim that it is exactly this redundancy that licenses (but does not necessitate) the omission of a syntactic subject in obligatory control constructions. The lexical determination of obligatory control and the (partial) separation of the syntactic and semantic properties has two important effects. First, a structure involving an obligatory control predicate and an infinitival subject is possible; second, such a configuration will nevertheless prohibit non-obligatory control relations, since obligatory control (which is an inherent lexical/semantic requirement 87

Presumably, obligatory control meaning postulates can then not simply be seen as entailment relations as characterized above but rather as identity relations between the understood infinitival subject and a pre-specified antecedent.

What is under control

251

of these predicates) has to be met to yield a well-formed interpretation. In the following sections, we will provide support for this analysis. We have already seen in section 2.1 that properties targeting the syntactic structure (in particular the binding properties of German infinitives) support a syntactic distinction between subjectless infinitives and infinitives with a subject. We have also seen that subjectless infinitives are obligatory control infinitives (cf. the correlation between lexical restructuring contexts and obligatory control). A closer look at the binding properties as well as the interpretation of elided infinitives, however, will show that this correlation is only a weak correlation and that both distinctions—a semantic distinction between obligatory and non-obligatory control as well as a syntactic distinction between subjectless infinitives and infinitives with a syntactic subject—are necessary to account for the properties of infinitival constructions.

3.2. Syntax Φ semantics We have argued for the following correlations between the syntactic and semantic properties of infinitival constructions: infinitives that lack a syntactic subject (i.e., restructuring infinitives) are obligatory control infinitives (or, in Chierchia's terms, semantic properties rather than propositions). We have also assumed that non-obligatory control requires the presence of a syntactic subject. A closer look at the interaction of the syntax and semantics of infinitival constructions, however, shows that the correlation between the lack of a syntactic subject and obligatory control is only a one-way correlation and that the distribution of restructuring and the binding properties in infinitives also raise a serious question for a one-to-one mapping between syntax and semantics. Recall that we have claimed that non-restructuring infinitives (including the non-restructuring versions of constructions with a potential restructuring verb) project an embedded syntactic (PRO) subject. The evidence for this claim came from the binding asymmetry between (191)c and (193)a (repeated here as (205)a,b): while embedded

252

The infinitival subject

anaphors are licensed in non-restructuring infinitives such as (205)a, they are impossible in restructuring constructions lacking an antecedent in the matrix predicate (cf. (205)b). We have argued that this contrast is due to the presence ((205)a) vs. absence ((205)b) of a syntactic antecedent for the anaphor—an embedded syntactic PRO subject as indicated in the examples. Most importantly, however, both examples in (205) are obligatory control constructions. Similarly, the non-restructuring versions of examples involving a potential restructuring verb (i.e., examples such as the ones in (28), repeated as (205)c) can only receive an obligatory control interpretation. (205) a.

Es wurde versucht It was tried PRO¡ sicht den Fisch mit Streifen vorzustellen PROi SELFj the fish with stripes-ACC to-imagine 'They tried to imagine what the fish would look like with stripes' b. *weil {sich} der Fisch {sich} vorzustellen since {SELF} thefish-NOM {SELF} to-imagine versucht wurde tried was 'since they tried to recall the image of the fish' c. dass versucht wurde/*wurden that tried was/*were [PRO den Traktor und den Lastwagen zu reparieren] [PRO [the tractor and the truck]-ACC to repair] 'that they tried to repair the tractor and the truck'

Assuming our analysis is correct, the examples in (205) show that obligatory control infinitives can be represented as clausal complements (i.e., vPs, TPs, or CPs) in the syntax (or more specifically, at the level where binding applies).88 Thus, the presence of a syntactic 88

By clausal complement, we refer to a structure that includes a subject (i.e., at least a vP). Whether these infinitives involve other functional projections such as TP, CP, is irrelevant for the discussion here.

What is under control

253

subject does not entail non-obligatory control (or obligatory control does not entail the lack of a syntactic subject). In other words, while the semantic control properties of an infinitival construction play an important role for the question of whether a PRO subject is projected or not, they are not sufficient to determine whether the infinitive can or cannot include a syntactic subject. The presence or absence of an infinitival subject is ultimately a syntactic issue which is reflected in the distinction between restructuring and non-restructuring infinitives. Although these facts show that there is only a one way correlation between obligatory vs. non-obligatory control and the absence vs. presence of an infinitival subject, they nevertheless seem to support a lexical/semantic approach to obligatory control. Let us see how the obligatory control effect in (205)a,c is accounted for in our approach. Since these examples include a syntactic infinitival subject, the syntactic derivation could in principle yield a non-obligatory control interpretation. However, at the syntax/semantics interface, this interpretation will clash with the inherent semantic properties of the selecting (obligatory control) verbs, and hence the structures would yield contradictory interpretations. Thus, the semantic requirement of obligatory control (i.e., the entailment or identity relation which is built into the lexical meaning of these verbs), will guarantee the correct control properties even in cases where a syntactic subject is present.

3.3.

"It" anaphors

The final point we want to mention in support of our analysis is built on the interpretation of non-overt infinitives that are represented by /i-anaphors, and we will show again that the syntactic structure cannot be fully reduced to the interpretation of infinitives. Chierchia (1984a, 1984b) has noted that //-anaphors replacing an infinitive or a gerund in examples such as (206) display an ambiguity between a strict and a sloppy interpretation.89 In (206), the antecedent of it can In English, //-anaphors can only replace gerunds. As we will see below, this is

254

The infinitival subject

refer to a situation in which the understood subject of the nonexpressed gerund is identical to the subject of the verb that combines with it (i.e., a sloppy interpretation with Mimi as the understood subject of the elided gerund), or to a situation in which the understood subject of the non-expressed gerund is the subject of the verb that combines with the associate or antecedent of it (i.e., a strict interpretation with Ezio as the understood subject of the elided gerund). (206)

Ezio likes fooling around, but I would bet that his wife Mimi doesn 't like it at all a. b.

it: Mimi fooling around it: Ezio fooling around

sloppy reading strict reading

The situation is different when the matrix predicate is changed to a verb like begin. As can be seen in (207) (slightly modified from Chierchia), only a sloppy interpretation is available; a strict reading in which the understood subject of the elided gerund refers back to the subject of the antecedent of it is excluded. (207)

Ezio began playing the violin for fun and Nando began it out of lust a. b.

it: Nando's playing it: *Ezio's playing

sloppy reading * strict reading

The analysis Chierchia (1984a, 1984b) provides to account for the (im)possibility of the strict interpretation of ii-anaphors is built on the assumption that infinitives and gerunds are semantic properties (i.e., subjectless predicates) that become associated with a subject contextually via the application of meaning postulates. Furthermore, Chierchia assumes that iY-anaphors are interpreted essentially like pronouns, i.e., as variables that are either bound by a structural antecedent or receive their value from context. not the case in German where it can stand for an infinitive. Note that we do not make any claims about the structure of gerunds in English; we only use examples with gerunds in this section to illustrate the strict vs. sloppy ambiguity and to lay out Chierchia's analysis.

What is under control

255

Let us first see how the two interpretations are accounted for in Chierchia's theory. For the sloppy interpretation in (206) (repeated in detail in (208)), Chierchia assumes that it represents a property in syntax/semantics which is bound by the property fooling around in the preceding conjunct (cf. (208)a,b). The entailment relation (meaning postulate) built into the meaning of the predicate like then identifies the subject of the embedded property with the subject of the selecting verb. Thus, in both conjuncts, the embedded property is associated with the subject of the verb it is the complement of—Ezio in the first gerund, Mimi in the second gerund (cf. (208)c). (208) Ezio likes fooling around... Mimi doesn't like it at all a. b.

c.

Ezio likes Mimi doesn't like it-association Ezio likes Mimi doesn't like Mimim doesn't like meaning postulate Ezioe likes Mimira doesn't like

[fooling around]... ^property

[fooling around]k.... itk at all [fooling around]k [SUB Je fooling around]k.... [SUBJm fooling around]k

For the strict reading in (206) (repeated in (209)), on the other hand, Chierchia assumes that it stands for a proposition in syntax/semantics. The antecedent for this proposition is determined contextually by associating it with the proposition SUBJ fooling around of the preceding conjunct. As is illustrated in (209)a,b, the first conjunct is associated with a subject via the meaning postulate of like; the resulting proposition is then taken as the antecedent for the iV-anaphor (cf. (209)c). Thus, the antecedent of it in (209) is not a structural entity (i.e., it is neither present in syntax nor in semantics) but rather an object that is created post-semantically. (209) Ezio likes fooling around... Mimi doesn 't like it at all a.

Ezio likes Mimi doesn't like

[fooling around]. itPROposmoN

256

The infinitival subject

b.

c.

meaning postulate Ezioe likes [SUBJe fooling around].... Mimi doesn't like ^proposition it-association Ezioe likes [SUBJe fooling around] k ... Mimi doesn't like itk Mimi doesn't like [SUBJe fooling around]k

Since, according to Chierchia, the predicate like can in principle combine with a proposition, an alternative derivation—a structure in which it refers back to a structural proposition—might be possible for the strict interpretation in examples such as (209). However, as we will see below, the mechanism of associating it with a nonstructural antecedent (i.e., a contextually created proposition) will still be required in Chierchia's system for examples in which the first predicate can only combine with a property. To account for the impossibility of a strict interpretation of itanaphors in ¿>eg/n-contexts, Chierchia (1984a, 1984b) assumes that certain predicates are incompatible with propositional complements. Chierchia points out that syntactically, verbs such as like can combine with a clausal complement or a subjectless VP-complement (cf. (210)a,b); verbs like begin, on the other hand, are only compatible with a VP-complement (cf. (210)c,d).90 Assuming that there is a straightforward mapping between syntax and semantics, we can then expect that like is compatible with both a property and a proposition, whereas begin is only compatible with a property. In other words, if a begin-type predicate combines with an i'r-anaphor, it can only stand for a property and hence will always have to refer back to a property, yielding only a sloppy interpretation. (210) a. b.

like + PROPOSITION John likes Mary's fooling around like + PROPERTY John likes fooling around

(210)d is acceptable under the (irrelevant) interpretation where begin is interpreted as cause.

What is under control

257

c.

begin + PROPERTY John began fooling around d. *begin + PROPOSITION *John began Mary's fooling around Before we discuss Chierchia's analysis, let us first give an overview of the situation in German, where iY-anaphors can replace infinitives. The interpretation of ù-anaphors in German shows two crucial generalizations. First, as is summarized in Table 25, the strict vs. sloppy interpretation of iï-anaphors depends solely on the type of predicate combining with if, the nature of the antecedent of it is irrelevant. Second, we will see that strict interpretations are only possible when the if-anaphor combines with a non-obligatory control predicate (according to our classification in section 2.2.1; the syntactic status of the infinitive (i.e., whether it is a subjectless restructuring complement or a non-restructuring complement with an infinitival subject) will turn out to be irrelevant. Table 25.

Strict/sloppy readings of "it"-anaphors

Antecedent

"it"

strict

sloppy

Example

Non-obligatory control Obligatory control Non-obligatory control Obligatory control

Non-obligatory control Non-obligatory control Obligatory control Obligatory control

+ +

+ + + +

(211)a (21l)b (211 )c (21 l)d

-

Examples illustrating the four possibilities in Table 25 are given in (211). In (21 l)a,b, it combines with non-obligatory control predicates and a strict interpretation is possible (the first paraphrase of the examples in (211) represents the sloppy interpretation, the second paraphrase represents the strict interpretation). As an aside, the two interpretations in (211)a,b correlate with an interesting difference in intonation, which we have no explanation for at this point: under the sloppy interpretation, the main sentence stress falls on the subject (i.e., Peter)·, under the strict interpretation, the main sentence stress falls on the matrix verb (i.e., angekündigt 'announce'). In (211)c,d, on the other hand, it combines with obligatory control predicates and a strict interpretation is impossible.

258

The infinitival subject

(211) a.

Hans beschloß/plante/bot an zu heiraten John decided/planned/offered to get-married nachdem Peter es angekündigt hatte after Peter it announced had 'John decided/planned/offered to get married after Peter had announced that he, Peter, would get married' 'John decided/planned/offered to get married after Peter had announced that John would get married' b. Hans wagte/versuchte/vergaß zu heiraten John dared/tried/forgot to get-married nachdem Peter es angekündigt hatte after Peter it announced had 'John dared/tried/forgot to get married after Peter had announced that he, Peter, would get married' 'John dared/tried/forgot to get married after Peter had announced that John would get married' c. Hans beschloß/plante/bot an zu heiraten John decided/planned/offered to get-married nachdem Peter es gewagt hatte after Peter it dared had 'John decided/planned/offered to get married after Peter had dared to get married' * 'John decided/planned/offered to get married after Peter had dared that John would get married' d. Hans wagte/versuchte/vergaß zu heiraten John dared/tried/forgot to get-married nachdem Peter es versucht hatte after Peter it tried had 'John dared/tried/forgot to get married after Peter had tried to get married' *'John dared/tried/forgot to get married after Peter had tried that John would get married'

The same situation holds for ditransitive constructions. As can be seen in (212)a, non-obligatory control predicates such as offer allow

What is under control

259

a strict and a sloppy interpretation for the iY-anaphor.91 Λ-anaphors combining with an obligatory control predicate such as order (cf. (212)b), on the other hand, can only be interpreted with a sloppy interpretation (i.e., the understood subject of the elided infinitive is obligatorily coreferent with the object of the selecting verb). Finally, (212)c shows that partial control patterns with non-obligatory control—a partial control interpretation is possible for both the strict and the sloppy interpretation of the ¿/-anaphor. As noted in (212)c, obligatory control verbs such as try cannot combine with an /r-anaphor that refers back to a collective predicate when the subject of the obligatory control verb is singular.92 Since obligatory control predicates prohibit a partial control interpretation and only allow a sloppy interpretation in ellipsis contexts, the understood subject of the infinitive replacing the ti-anaphor in (212)c can only be der Direktor 'the principal'. This singular subject, however, is incompatible with the collective predicate (of the elided infinitive) and the example is thus correctly predicted to be ungrammatical. (212) a.

b.

91

Peter hat sich geweigert Peter has SELF refused [fìir die Freunde seiner Mutter Holz zu hacken] [for the friends of-his mother wood to chop] obwohl sie es ihnen schon angeboten hatte although she it them already offered had 'Peter refused to chop wood for the friends of his mother, although she had already offered them that he/she/they (Peter and his mother) would chop wood for them' Peter hat angeboten Peter has offered fìir die Freunde seiner Mutter Holz zu hacken [for the friends of-his mother wood to chop]

The pragmatic context of the first sentence in this example favors subject control of the elided infinitive. However, if the context is modified accordingly, object control is equally available. 92 We thank Idan Landau for pointing us to this prediction.

260

The infinitival subject

nachdem sie es ihnen befohlen hatte after she it them ordered had 'Peter offered to chop wood for the friends of his mother, after she had ordered them to chop wood' *'Peter offered to chop wood for the friends of his mother, after she had ordered them that he (Peter) should chop wood' c. Peter hat versprochen Peter has promised [sich im Schloß zu versammeln] [SELF in-the castle to gather] nachdem der Direktor es beschlossen/*versucht hatte after the principal it decided/*tried had 'Peter promised to gather in the castle after the principal had decided that Peter and his associates would gather in the castle' 'Peter promised to gather in the castle after the principal had decided that he [the principal] and his associates would gather in the castle' *'Peter promised to gather in the castle after the principal had tried that he [the principal] (and his associates) would gather in the castle' The examples in (211) through (212) thus show that the distinction between non-obligatory and obligatory control as suggested here is directly reflected in the possibility vs. impossibility of a strict interpretation of infinitival complements expressed by iY-anaphors. The final question we want to address again is whether this semantic distinction determining the interpretation of it-anaphors is also reflected in the syntactic structures, or in other words, whether obligatory control infinitives (i.e., infinitives characterized as properties semantically) necessarily correspond to subjectless VP-predicates in syntax. As in the case of binding and restructuring (see section 3.2), the answer will be negative, and ¿ί-anaphors will hence provide another piece of evidence against a one-to-one mapping of syntactic and semantic structure. The crucial fact that will allow us to make this point is that /f-anaphors that combine with an obligatory control predicate

What is under control

261

but that at the same time give rise to the assumption that the itanaphor represents a syntactic clause (i.e., vP, TP, or CP) clearly disallow a strict interpretation. Hence, the syntactic structure (whether the infinitive projects a subject or not) seems to be irrelevant in itanaphor contexts—what is crucial is the semantic property of obligatory control. To show that an iY-anaphor represents a clausal structure syntactically is not straightforward and can only be shown indirectly. In examples such as (213)a, the i7-anaphor combines with the obligatory control predicate manage. Note first that the only interpretation it can receive in this example is a sloppy interpretation (i.e., the subject of the elided infinitive is understood to be Mary); a strict interpretation in which the subject of the elided infinitive refers back to Peter is strictly impossible (cf. (213)b). Secondly and most importantly, this example provides (indirect) evidence for the claim that it represents a syntactic clause (i.e., vP, TP, or CP) rather than a subjectless VP. Consider the simplified structure in (213)c—i.e., a structure in which it has been associated with an antecedent. As we have seen in the previous section, dative arguments cannot bind anaphors in German. Assuming that the same binding conditions hold in (213)c (i.e., for the associate of it), the grammaticality of (213) then provides evidence for the presence of an infinitival subject as an antecedent for the embedded anaphor in (213)c. If the infinitive does not include a PRO subject at the level where binding applies (as in (213)d), the anaphor would end up without an antecedent and the structure would be predicted to be ungrammatical (like (205)b). We thus conclude that the correct structure for it in (213)a is the clausal structure in (213)c rather than the VP-structure in (213)d.93

93

Note that if one were to assume that binding takes place after the meaning postulate applies (i.e., (213)d is the syntactic structure for (213) and binding applies to the structure in (213)c which is created contextually), the idea that obligatory control infinitives are subjectless VPs (syntactically) could be maintained. However, this approach would then not be able to account for the ungrammaticality of examples such as (205)b, which would (incorrectly) be predicted to be grammatical. The same problem would arise for the other examples provided in section 2.1 as evidence against a syntactic PRO subject.

262

The infinitival subject

Peter hat beschlossen Peter has decided [sich den Fisch mit Streifen vorzustellen] [SELF [the fish with stripes]-ACC to-imagine] nachdem es der Maria gelungen ist after it the Mary-DAT managed is 'Peter decided to imagine what the fish would look like with stripes after Mary managed to imagine what the fish would look like with stripes' b. Mary-DATM managed [SUBJm S E L F m the fish-ACC to-imagine] *[SUBJp SELFp the fish-ACC to-imagine]

(213) a.

c.

Mary-DATM managed [SUBJM SELFM

INF]

Mary-DATM managed [ I ¿

INF]

SELFM 1

To summarize, an analysis that ties the possibility of a strict interpretation of an i'í-anaphor to the syntactic structure (in particular the presence of a syntactic subject) seems to make the wrong prediction for the interpretation of i7-anaphors in German. Rather, the crucial property that licenses a strict interpretation is the semantic property of obligatory control. Going back to our analysis of obligatory control, we have proposed that obligatory control is compatible with a syntactic subject (as evidenced by the restructuring and binding properties). However, as a result of the inherent lexical/semantic property of obligatory control predicates, which identifies the embedded subject with a pre-specified argument of the obligatory control verb, only control relations that meet this semantic requirement will be interpretable. The same reasoning can be applied to the interpretation of iY-anaphors. Since obligatory control predicates impose as part of their meanings a certain control relation on the embedded subject, this control relation has to be met in the embedded structure—i.e., the infinitive that replaces it. Thus, whether the i7-anaphor stands for an infinitive that is a subjectless VP (i.e., a restructuring infinitive) or a clause (i.e., a non-restructuring infinitive) has no

Conclusion

263

bearing on the control relation when the selecting predicate is an obligatory control predicate. Since the strict interpretation (like long distance control) is incompatible with the semantic requirement of obligatory control, obligatory control predicates combining with an ii-anaphor will thus generally block a strict interpretation of their complements.

4. Conclusion The main claim we have made in this chapter is that there are two basic forms of control—syntactic control and semantic control—which are distinguished semantically. If the interpretation of the understood subject is variable, control is determined syntactically. If the interpretation of the understood subject is fixed, control is determined lexically/semantically. Syntactic control configurations are represented by a syntactic PRO subject; semantic control configurations do not involve a syntactic subject. Table 26 provides an overview of the properties found in infinitival constructions in German that we have discussed in this chapter. In this table, restructuring refers to infinitives without an embedded (PRO) subject; non-restructuring refers to infinitives with an embedded (PRO) subject. Table 26.

Syntactic and semantic properties of infinitival constructions

Control relation

Syntactic structure

Strict Long Embedded anaphors (& reading passive no matrix antecedent)

Obligatory Obligatory Non-obligatory Non-obligatory

Restructuring Non-restructuring Non-restructuring Restructuring

-

+

+ N/A

N/A

+ + N/A

The discussion in this chapter has shown that the properties in Table 26 target different aspects of the syntactic and semantic structure of an infinitival construction. Long passive and binding (principle A) target the syntactic restructuring vs. non-restructuring distinction, whereas the interpretation of ¿/-anaphors targets the semantic

264

The infinitival subject

obligatory vs. non-obligatory control distinction. As is summarized in Table 26, a strict reading for if-anaphors requires a non-obligatory control context. Whether the infinitive is a restructuring or a nonrestructuring infinitive is irrelevant for this property. The opposite situation holds for long passive and binding. Long object movement requires the lack of a syntactic subject and hence it is only possible in restructuring contexts. Embedded anaphors, on the other hand, require the presence of a syntactic subject (in the absence of an appropriate matrix antecedent) and hence they are only possible in nonrestructuring contexts. Whether the infinitive is an obligatory or nonobligatory control infinitive has only an indirect bearing on these two syntactic properties. We have suggested that subjectless infinitives (i.e., restructuring infinitives) are only licensed when the control relation is recoverable from the meaning of the selecting predicate (i.e., in obligatory control contexts). Finally, we have shown that there is a close relation between the syntactic and semantic properties of infinitival subjects (i.e., syntactically subjectless infinitives are necessarily obligatory control infinitives), however, we have also concluded that a full correlation between the presence vs. absence of a syntactic subject and the interpretation of the infinitival subject is not possible.

Chapter 5 Grades of (non-)restructuring

1. Introduction In the final chapter of this study, we will turn to non-restructuring infinitives. The main goal of this chapter is to further elaborate on the restructuring/non-restructuring distinction suggested here, as well as to provide an overview of the structure and properties of constructions that we classify as non-restructuring constructions in German. We will show that the class of non-restructuring infinitives also cannot be viewed as a uniform phenomenon but is distinguished into two subclasses: (full clausal) non-re structuring infinitives vs. reduced non-restructuring infinitives, which systematically differ in a number of syntactic and semantic properties. We will further argue that propositional and factive infinitives are clausal non-restructuring infinitives, which are represented as CP complements, whereas irrealis infinitives are restructuring or reduced non-restructuring infinitives, which are represented as VP complements (restructuring), or TP or vP complements (reduced non-restructuring infinitives). This view of graded (non-) restructuring is an important step to a better understanding of infinitival constructions since it not only avoids conflicting classifications of restructuring and non-restructuring infinitives, but it also provides an empirically adequate characterization of the properties of different classes of infinitival constructions, as well as a straightforward way to account for the possibility vs. impossibility of different operations in different syntactic and semantic contexts. While the particulars of the analysis we suggest are tailored to the properties of German infinitival constructions, it can be applied to other languages provided that language specific properties and phenomena are taken into consideration.

266

Grades of (non-Restructuring

2. Restructuring vs. non-restructuring We will start the discussion of graded (non-)restructuring by presenting further evidence for the restructuring vs. non-restructuring distinction as proposed in this study. We will first show that not all "restructuring" properties define the same class of infinitival constructions. The different properties, however, will then be shown to define coherent semantic classes which will be the first step in motivating a system of graded (non-)restructuring. Let us begin with a short summary of the restructuring phenomenon. In Chapters 2 to 4 we have argued for two classes of restructuring: lexical and functional restructuring. Table 27 summarizes the distribution and properties of these two classes of restructuring. Table 27. Type

German restructuring constructions Structure Properties, distribution

(Im)Possible operations

INF = · no embedded (PRO) subject possible: VP-layer · no embedded structural case • long object movement • no embedded tense • scrambling • no embedded negation • pronoun fronting • obligatory control impossible: • possible with: (strong) im- • ??extraposition of infinitive plicatives, aspectuals, irrealis • relative clause pied piping predicates Functional INF = • thematic properties are depossible: main termined by the embedded • IPP effect predicate predicate (except semi• raising functional predicates) impossible: • possible with: modal, rais- • extraposition of infinitive ing, aspectual, causative, • matrix passive perception, motion verbs • relative clause pied piping Lexical

We have argued that both types of constructions are represented by mono-clausal structures—i.e., structures that involve a single set of clausal functional projections (such as tense, complementizers, and structural case) and only one external argument. The differences between the two classes of restructuring have been reduced to the functional vs. lexical status of the matrix predicate: first, functional

Restructuring vs. non-restructuring

267

predicates, being non-thematic, do not associate with an external theta-role and hence cannot be passivized; second, we have claimed that functional verbs cannot be stranded by extraposition since functional verbs form a prosodie phrase with the main predicate and extraposition of a partial prosodie phrase is impossible. In our system, the major difference between restructuring and non-restructuring infinitives is the possibility vs. impossibility of long object movement—i.e., the absence vs. presence of an embedded structural case position/assigner and an embedded subject. This property distinguishes constructions such as the one in (214)a, which we have labeled lexical restructuring, from constructions such as the ones in (214)b,c, which we consider non-restructuring constructions (but see below). (214) Long passive a.

dass that 'that b. *dass that 'that c. *dass that 'that

der Traktor zu reparieren versucht wurde the tractor-NOM to repair tried was they tried to repair the tractor' der Traktor zu reparieren geplant wurde the tractor-NOM to repair planned was they planned to repair the tractor' der Traktor zu reparieren beschlossen wurde the tractor-NOM to repair decided was they decided to repair the tractor'

The assumption that long passive is one of the crucial restructuring properties, however, is not shared by all researchers (see in particular Grosse 2000). A different operation that is often claimed to be licensed only in restructuring constructions is pronoun fronting (see among many others Haider 1993, Kiss 1995, Grosse 2000). As is illustrated in (215), a pronominal argument originating in the embedded predicate of a /^-infinitive can be moved to the Wackernagel position—i.e., to the left of the matrix subject. Most speakers allow pronoun fronting from extraposed infinitives (cf. (215)b) as well as from infinitives in situ (cf. (215)a). (However, it should be pointed out that there are also speakers who disprefer (215)b. We will return

268

Grades of (non-Restructuring

to this variation in section 3.2.2.) (215) Pronoun fronting in restructuring constructions a.

b.

weil ihn der Hans zu reparieren versuchte since it-ACC the John to repair tried 'since John tried to repair it' weil ihn der Hans versuchte zu reparieren since it-ACC the John tried to repair 'since John tried to repair it'

Looking at the distribution of pronoun fronting in more detail, it can be observed that this operation is possible in a larger range of infinitival constructions than long passive. As is illustrated in (216), pronoun fronting is perfectly wellformed in constructions involving matrix predicates such as plan or decide (with the same speaker variation regarding pronoun fronting in extraposed contexts). The examples in (216) thus contrast with examples such as (214)b,c which block long passive in the same type of infinitival constructions. (216) Pronoun fronting in non-restructuring constructions a.

b.

c.

d.

weil ihn der Hans zu reparieren since it-ACC the John to repair 'since John planned to repair it' weil ihn der Hans zu reparieren since it-ACC the John to repair 'since John decided to repair it' weil ihn der Hans beschloss since it-ACC the John decided 'since John decided to repair it' weil ihn der Hans plante since it-ACC the John planned 'since John planned to repair it'

plante planned beschloss decided zu reparieren to repair zu reparieren to repair

The contrast between (214)b,c and (216) shows that different "transparency" criteria define different classes of infinitival constructions. Mismatches of this sort provide a challenge to the classification of

Restructuring vs. non-restructuring

269

infinitival constructions on the assumption (maintained by most authors) that there are only two types of constructions (restructuring and non-restructuring infinitives). To accommodate contrasts such as the one between (214)b,c and (216) in a two-class system, one has to declare one criterion as the relevant one and ignore the other criterion. Thus, much of the debate in the literature evolves around the question of which criterion is the "right" one, and mismatches are usually left unexplained. We believe, however, that this approach is inadequate and that mismatches of the sort mentioned should not be ignored since they express important generalizations that would otherwise be lost. In what follows, we will show that the contrast between (214)b,c and (216) is indeed not arbitrary but correlates with various other properties, and hence provides important evidence for a more fine-grained classification of infinitival constructions and the approach of graded (non-)restructuring advocated here. Note first, that there is a further "transparency" property that splits infinitival constructions into two groups. As is illustrated in (217), scrambling of an embedded argument to the matrix predicate yields different acceptability for iry-infinitives vs. /?/an-infinitives. As pointed out in Chapter 2, section 2.3.1, some speakers allow scrambling from non-restructuring infinitives such as (30)d (repeated here as (217)a).94 However, in contrast to scrambling from restructuring infinitives (cf. (217)b), scrambling from non-restructuring is marked (or even ungrammatical for some speakers) and requires a focus interpretation which is not necessary in (217)b.95 94

We will continue to refer to plan/decide-type infinitives as non-restructuring infinitives (see section 3.1). However, it is important again not to confuse the terminological question with the real issue at hand. In light of the preceding discussion it is crucial to keep in mind that we do not maintain a simple binary division between restructuring and non-restructuring. Our use of the term "nonrestructuring" is intended to capture the fact that constructions with verbs such as plan or decide systematically differ in behavior from lexical restructuring constructions. As we will see presently, the class of verbs represented by plan/decide displays clearly defined syntactic and semantic properties which distinguish it from what we call true restructuring infinitives. 95

The distinction between focus scrambling and non-focus scrambling is admittedly shaky. However, we believe that the contrasts are sharp when the two forms

270

Grades of (non-)restructuring

(217) Focus vs. non-focus scrambling a. Sodass Hans den Traktor geplant hat zu reparieren that John the tractor-ACC planned has to repair 'that John (has) planned to repair the tractor' b. dass Hans den Traktor versucht hat zu reparieren that John the tractor-ACC tried has to repair 'that John (has) tried to repair the tractor' Thus, so far we have seen that passive and unmarked scrambling are possible from fry-infinitives and blocked from p/an-infinitives, whereas pronoun fronting and focus scrambling are possible in both constructions. The second generalization we would like to point out is that this split is not a special property of the two kinds of infinitives discussed, but in fact extends to a much broader range of infinitival constructions. Let us look closer at the distribution of the four criteria discussed so far. Table 28 provides a (partial) list of predicates that allow pronoun fronting and focus scrambling (a more exhaustive list of predicates and examples illustrating each property is given in the appendix, section 3). The table shows two important generalizations. Firstly, it illustrates that long passive patterns with unmarked (i.e., non-focus) scrambling, whereas pronoun fronting patterns with focus scrambling. Secondly, it shows that the there is a semantic uniformity to the classes defined by these properties: the verbs that allow pronoun fronting and focus scrambling occur in either irrealis or aspectual constructions (see Chapter 2, section 3 for a discussion of the irrealis property). As will be illustrated in section 3.2, propositional and factive infinitives, on the other hand, systematically block all four operations. are clearly distinguished. As we have shown in Wurmbrand (1998b), contexts that strongly disfavor a focus interpretation of the scrambled phrase (e.g., constructions that involve another focused element) are generally judged worse than constructions that allow focus interpretation (by the same speaker). Similarly, there is a robust difference between scrambling to the left vs. scrambling to the right of the matrix subject. The former is clearly preferred over the latter, since the pre-subject position is an inherent focus position.

Restructuring vs. non-restructuring

271

Table 28. The distribution of long passive, scrambling, and pronoun fronting in German Verb

Gloss

Long passive

anbieten anfangen aufhören befehlen • beschließen drohen entscheiden ermöglichen erwägen fortfahren glauben hoffen planen raten versprechen verstehen vorschlagen vorziehen wünschen

offer start stop order decide threaten choose enable contemplate continue mean hope plan advise promise know how propose prefer wish

Non-focus scrambling

Pronoun fronting

Focus scrambling

+ +

+

?

+ + +

+

+ + + + + +

+ + + +

+

+ + + +

+ +

+ + +

+

+ + + +

+ + +

+ +

+

The distribution of long passive, scrambling, and pronoun fronting in German can be summarized as in Table 29 (see also the appendix, section 2-2 for motivation for these classifications including various tests applied to a larger class of constructions). Table 29. The distribution of long passive, scrambling, and pronoun fronting in German (verb classes) Verb

Long passive

Restructuring predicates + Irrealis predicates Propositional predicates Factive predicates

Non-focus scrambling

Pronoun fronting

Focus scrambling

+ -

+ +

+ +

-

-

-

-

-

-

To capture the generalizations in Table 29, an account is called for

272

Grades of (non-)restructuring

that i) allows for a three-way distinction of infinitival constructions (setting aside here the lexical vs. functional distinction of restructuring infinitives), ii) derives the properties illustrated in Table 29, and iii) explains the correlation between the (un)availability of the properties mentioned and the semantic type of the infinitive.96 Although some of the details will require a more refined implementation of the analysis than will be given here, we will outline an account of the generalizations in Table 29 in the next sections, which will allow us to get a basic understanding of the distribution of infinitival constructions in German. We will first discuss constructions that allow pronoun fronting and focus scrambling but block long passive and non-focus scrambling (i.e., infinitives involving the verbs in Table 28). For reasons that will become clear as we proceed, we will call these constructions reduced non-restructuring infinitives. Full nonrestructuring infinitives (i.e., infinitives that block all four properties) will be discussed in section 3.2. Finally, in section 4, we will suggest how the semantic classification of infinitival constructions can be related to the syntactic structure. 3. Graded (non-)restructuring The analysis we suggest is based on the idea of non-templatic, deterministic syntactic structure and graded (non-)restructuring. In short, we assume that non-restructuring infinitives—like restructuring infinitives—come in (at least) two versions: full non-restructuring infinitives and reduced non-restructuring infinitives (a similar approach has been suggested by Rutten 1991). Full non-restructuring infinitives are syntactically represented as CPs, whereas reduced non-restructuring infinitives are either vPs or TPs (the claim that As can be seen in the appendix, Table 40, weak implicatives prohibit pronoun fronting and scrambling. However, a firm conclusion seems premature at this point, since the class of predicates tested is very small and we have not been able to test these predicates with a larger group of speakers (in particular, we have not been able to establish if there are differences correlating with the interpretation). Note also that one of the verbs (sich weigern 'refuse') is an inherently reflexive verb which blocks restructuring independently.

Graded (non-Restructuring

273

certain (control) infinitives lack a CP in German has also been made for instance by Tappe 1984; and Fanselow 1989). Thus, reduced nonrestructuring infinitives are in a sense between restructuring and nonrestructuring infinitives, which in light of their intermediate status with respect to some of the transparency properties is a desired result. The aim of this section is to motivate these structures and to see how they account for the properties of infinitival constructions in German. In section 3.1, we will discuss the differences and similarities between restructuring constructions and reduced non-restructuring constructions. In section 3.2, we then turn to the differences and similarities between full and reduced non-restructuring constructions.

3.1.

Restructuring vs. reduced

non-restructuring

An account for the difference between restructuring constructions and reduced non-restructuring constructions (i.e., the possibility vs. impossibility of long passive) has already been provided. As discussed in detail in Chapter 2, we assume that long object movement and non-focus scrambling are case-driven movements which are subject to standard locality constraints such as "closeness". Thus, passive and non-focus scrambling can only target a case position in the matrix predicate when there is no closer case position in the embedded predicate; embedded case positions such as vP and TP, as well as embedded subjects block passive and non-focus scrambling. The fact that long object movement and non-focus scrambling are only possible from restructuring infinitives and not from reduced non-restructuring infinitives is thus derived from the difference in syntactic size between these categories: the former are "VP" infinitives, whereas the latter are (at least) vP infinitives. We have also seen in Chapter 2 to Chapter 4 that this difference is not arbitrary but (largely) predictable from a combination of the tense, control, and certain structural properties such as the presence of particles. To account for the similarities between restructuring constructions and reduced non-restructuring constructions (i.e., the possibility of pronoun fronting and focus scrambling in both constructions), we follow Müller (1996) who has shown that pronoun movement has to

274

Grades of (non-)restructuring

be treated differently from Α-movement and scrambling. In particular, we assume that pronoun movement to the left of the subject does not target a case position but rather a position in the C-domain. We do not have any new insights concerning the question of why pronouns move to this position, but we simply assume that this movement (i.e., the step of pronoun movement that places the pronoun in Wackernagel position) is triggered by some special "C" or "Wackernagel"-feature, which attracts pronouns to this position. In contrast to passive and scrambling, pronoun movement to Wackernagel position is then not sensitive to embedded case positions but only to embedded C/Wackernagel-positions. Note that this analysis is also compatible with approaches that assume that pronoun movement involves two steps: a step of Α-movement and a step of head movement (cf. for instance Sportiche 1996; Roberts 1997 and references therein). While the first step of pronoun movement would again be blocked by an embedded case position, pronouns can nevertheless escape infinitives with embedded case positions, since the second step is not sensitive to case positions but only to C-positions. Finally, we assume that focus scrambling differs from non-focus scrambling (i.e., object shift-type movement) in that the former is not case-driven but rather triggered by some interpretive property or feature such as focus (situated in the C-domain), and hence this form of scrambling is also not sensitive to embedded case positions but only to intervening focus domains or positions. Our claims are summarized in Table 30. Table 30. Property

Long passive, scrambling, and pronoun fronting Matrix target

Blocking categories

Infinitive

Long object movement case position embedded vP/TP/subject restructuring Non-focus scrambling case position embedded vP/TP/subject restructuring Focus scrambling focus position embedded CP/focus reduced nonrestructuring Pronoun fronting C-domain embedded CP reduced nonrestructuring

In sum, without going into detail about the technicalities of pronoun fronting and focus scrambling, we assume that both properties target

Graded (non-Restructuring

275

the C-domain in German (either overtly or covertly), and hence will be blocked in a familiar way by intermediate C-domains which would represent closer landing sites (cf. (218)a,b).97 Blocking effects of non-restructuring infinitives (i.e., CP-infinitives) will be illustrated in the next section. CP-less infinitives (such as restructuring infinitives and reduced non-restructuring infinitives), on the other hand, are transparent for pronoun movement and focus scrambling, since the matrix C-domain is the closest attractor or landing site for these movement operations (cf. (218)c,d). (218) Pronoun fronting, focus scrambling a. *[Cp C°WACKERNAGEL

NON-RESTRUCTURING

1

b.

* [Cp

C°F0CUS

[CP ...pronoun ...]] X

NON-RESTRUCTURING

I

1

[ C P . . . FOCUS-DP... ] ] Χ 1

C.

[Cp C°WACKERNAGEL

REDUCEDNON-RESTR.

[TP ...pronoun ...]]

d.

[ c p C°FOCUS

REDUCED NON-RESTR. [TP . . . F O C U S - D P . . . ] ]

In the rest of this section, we will provide evidence for this analysis and show that despite the fact that reduced non-restructuring constructions display some transparency properties, these constructions 97

There are two basic ways to implement the claim that focus scrambling targets the C-domain. First, it can be assumed that topic/focus features are situated in C° and that focus phrases move into the C-domain in covert syntax. This claim is supported by the fact that scrambling does not have to occur to assign focus to an embedded phrase. Second, following recent approaches by Rizzi (1997), it can be assumed that the C-domain is split into a number of projections, representing among others topic and focus. Under this approach movement can be overt (if this option is chosen, it has to be assumed that in examples such as (217)a the subject which precedes the scrambled phrase is also moved to the C-domain). This approach would be supported by the fact that scrambling from reduced nonrestructuring infinitives is clearly preferred when the scrambled phrase occurs to the left of the matrix subject. We will not decide between these options here since either approach is compatible with what is crucial for the present analysis, namely the claim that focus scrambling is only sensitive to CPs and not case positions.

276

Grades of (non-)restructuring

also differ significantly from restructuring configurations, and that hence a separation of these two constructions is motivated. The assumption that pronoun fronting and focus scrambling are possible from vP/TP-infinitives and that only restructuring infinitives can be represented by a bare VP-structure makes the prediction that reduced non-restructuring infinitives should not be subject to the restructuring requirements (i.e., obligatory control, simultaneous tense, lack of an embedded subject, and lack of embedded negation)—even in cases in which pronoun fronting or focus scrambling occur. In the rest of this section, we will see that this is indeed the case. First, in contrast to long passive as in (219)a (repeated from (64)), pronoun fronting and focus scrambling are possible in contexts in which the matrix and the embedded predicate involve independent tense interpretations. Examples such as the ones in (219)b,c show that the acceptability of pronoun fronting (cf. (219)b) and focus scrambling (cf. (219)c) is not reduced when the temporal location of the infinitival event is different from the time of the matrix event (scrambling in (219)c is marked for some speakers, however, the presence of the future adverbial does not affect the acceptability). (219) Tense with passive, pronoun fronting, and scrambling a.

dass dem Kind nur Kekse that the child-DAT only cookies (*morgen) zu essen erlaubt wurden (* tomorrow) to eat allowed were 'that the child was only allowed to eat cookies tomorrow' b. dass ihn der Hans that him/it- ACC the John-NOM [(morgen) zu reparieren] beschlossen hatte [(tomorrow) to repair] decided had 'that John had decided to repair it tomorrow' c. Woweil nur den Wagen der Hans since only the car-ACC the John-NOM [(morgen) zu reparieren] beschlossen hatte [(tomorrow) to repair] decided had 'since John had decided to repair only the car tomorrow'

Graded (non-)restructuring

277

Second, pronoun fronting and focus scrambling are compatible with non-obligatory control. As is illustrated in (220)a,b, a partial control interpretation is available despite the fact that pronoun fronting has occurred.98 Similarly, (220)c illustrates that variable and partial control (and various combinations thereof) are possible in reduced nonrestructuring constructions involving focus scrambling. (220) Non-obligatory control with pronoun fronting and scrambling a.

weil sie derHansb since her-ACC the Johnh-NOM PROh+x gemeinsam zu überraschen beschloss PROh+x together to surprise decided 'since John decided that he and somebody else would surprise her together' b. weil ihn die Mariam since him-ACC the Marym-NOM PROm+x gemeinsam aufzusuchen plante PROm+x together to visit planned 'since Mary planned that she and somebody else would pay him a visit together' c. Iweil mit grüner Farbe nur der Hans since with green paint only the John-NOM der Maria vorschlug PRO den Zaun zu streichen the Mary-DAT proposed PRO the fence to paint 'since only John suggested to Mary to paint the fence with green paint'; 'since only John had the idea of painting the fence with green paint' PRO: Mary, John, John+Mary, Mary+others, John+others 98

We only present cases of partial control here since we have not been able to establish the empirical situation of pronoun fronting in variable control contexts. Some of the speakers we consulted reject pronoun fronting in variable control contexts across dative arguments (both across controlling and non-controlling datives). Pronoun fronting across datives that are obligatory controllers, on the other hand, is accepted. However, since the generalizations are not robust and judgements differ significantly in different (pragmatic and discourse) contexts, we cannot draw any conclusion about the nature of this restriction at this point (see also the appendix, section 4.2).

278

Grades of (non-)restructuring

Third, pronoun fronting and focus scrambling (in contrast to long passive and non-focus scrambling) are possible in the presence of an embedded PRO subject. Recall that examples such as (193)-(196) (repeated here as (221)) are ungrammatical since they cannot meet the requirements imposed by restructuring and the requirements imposed by binding at the same time. The restructuring configuration requires the absence of an embedded PRO subject, whereas binding requires the presence of an embedded PRO subject to bind the embedded anaphors in these examples (recall that dative arguments cannot bind anaphors in German). (221) Binding and passive a. *weil {sich} der Fisch {sich} vorzustellen since {SELF} thefish-NOM {SELF} to-imagine versucht/begonnen/beabsichtigt wurde tried/begun/intended was 'since they tried/began/intended to imageine the fish' b. *weil der Fisch dem Hansh sichh since the fish-NOM the Johnh SELFh mit Streifen vorzustellen erlaubt wurde with stripes imagine allowed was 'since they allowed John to imagine what the fish would look like with stripes' In (222), on the other hand, we find that pronoun fronting (i.e., movement of es 'it' or ihn 'him' to the left of the matrix dative argument ihm 'him') does not block embedded anaphors. Since the only possible antecedents for the embedded anaphors in (222) are the embedded PRO subjects, the grammaticality of theses examples provides evidence for an embedded subject. However, since pronoun fronting is possible at the same time in these contexts, we can conclude again that pronoun fronting differs crucially from passive and (non-focus) scrambling in that it is does not require a restructuring configuration.

Graded (non-)restructuring

279

(222) Binding and pronoun fronting a.

b.

weil sie es ihm since she-NOM it-ACC him-DAT [PRO t¡t sich mit Streifen vorzustellen] erlaubte [PRO tit SELF with stripes to-imagine] allowed 'since she allowed him to imagine it (e.g., the picture) with stripes' weil sie es ihm [PRO tit sich anzuschauen] verbot since she it him [PRO t¡, SELF to-look-at] forbade 'since she forbade him to look at it'

The same situation holds for focus scrambling. As is shown in (223)a (cf. Sabel 1996: 103), scrambling of a PP (which can only be an instance of focus scrambling) has no effect on the grammaticality of a sentence involving an embedded anaphor that cannot be bound by an argument in the matrix predicate. The only available antecedent for the embedded anaphor is again an embedded PRO subject. Similarly, the example in (223)b, which involves scrambling of an accusative argument is acceptable when the scrambled phrase is heavily focused. The less clear status of this example can be attributed to the potential ambiguity of this example and the fact that it is only grammatical under one of the two structures: if a speaker attempts to analyze the example as a non-focus scrambling configuration, it would be ruled out since either binding or the restructuring configuration would not be met. However, the facts are clear when focus scrambling is ensured, and importantly, binding is then not affected by the occurrence of focus scrambling. (223) Binding and focus scrambling a.

weil since PROh PROh 'since in this

in diesem Spiegel niemand der wilden Hildeh in this mirror nobody the wild Hildeh-DAT sichh tSCR anzuschauen erlaubte SELFh tSCR to-look-at allowed nobody allowed the crazy Hilde to look at herself mirror' (cf. Sabel 1996: 103)

280

Grades of (non-Restructuring

b. %weil sie nur den Fisch dem Hansb since she only the fish the Johnh-DAT PROh sichb tSCR mit Streifen vorzustellen erlaubte PROh SELFh tscR with stripes to-imagine allowed 'since she only allowed John to imagine what THE FISH would look like with stripes' Finally fourth, pronoun fronting and focus scrambling (in contrast to long passive and non-focus scrambling) are compatible with embedded negation. As we have shown in Chapter 2, section 4.4, examples involving long passive such as (91) ((91)a is repeated here as (224)a), prohibit an interpretation with embedded negation. By comparing (224)a to the examples in (2?4)b-d, however, it can be observed that embedded negation is available in examples in which pronoun fronting has taken place. All of the examples in (224)b-d are in principle ambiguous, however the different scope interpretations require different intonations and certain interpretations are less felicitous than others for pragmatic reasons. (224) Negation with passive and pronoun fronting a.

weil dem Hans [der Spinat nicht zu essen] since the John-DAT [the spinach-NOM not to eat] erlaubt wurde allowed was 'since John was not allowed to eat the spinach' *'since John was allowed not to eat the spinach' b. weil ihn der Hans nicht mehr since him-ACC theJohn-NOM not anymore zu bevorzugen beschlossen hatte to privilege decided had #'since John didn't decide anymore to privilege him' 'since John decided not to privilege him anymore' c. weil es dem Hans nicht zu essen gelungen ist since it-ACC the John-DAT not to eat managed is 'since John didn't manage to eat it' 'since John managed not to eat it'

Graded (non-Restructuring

d.

281

weil es ihr der Hans since it-ACC her-DAT theJohn-NOM nicht zu essen erlaubt hatte not to eat allowed had 'since John didn't allow her to eat it' 'since John allowed her not to eat it'

The same contrast can be found regarding scrambling vs. focus scrambling. To begin with, compare examples involving long passive, where embedded negation was blocked (cf. (224)a), with examples such as in (225) involving scrambling. As expected, scrambling is marked and is only possible under à focus interpretation (the examples can again be improved by adding nur 'only' to the scrambled phrase). (225) Negation and scrambling a. %weil die Maria den Kuchen versuchte nicht zu essen since the Mary the cake-ACC tried not to eat 'since Mary tried not to the cake' b. %weil ihm die Maria den Kuchen since him-DAT the Mary the cake-ACC erlaubt hat nicht zu essen allowed has not to eat 'since Mary allowed him not to eat the cake' Furthermore, consider the examples in (226) which display an interesting contrast that will follow from our distinction of scrambling vs. focus scrambling. While (226)a,b allow both matrix and embedded negation, the most natural interpretations in (226)c,d are interpretations in which negation takes embedded scope. Let us discuss the interpretations and structures of these examples in more detail. Assume first that the infinitives are reduced non-restructuring infinitives, i.e., infinitives that license embedded negation. In this case, both the embedded object and the negation are inside the infinitival complements and hence negation takes low scope. The examples in (226)a,c express that Mary made an explicit attempt or decision to achieve her (negative) goal which is 'not to help her son' (e.g., to force him to

282

Grades of (non-Restructuring

grow up and to learn to handle his problems without her help).99 Similarly, the examples in (226)b,d, convey that Mary explicitly allowed or suggested the 'non-eating of the cake'. (226) Negation and scrambling a.

weil die Maria ihrem Sohn nicht zu helfen versuchte since the Mary her son-DAT not to help tried 'since Mary didn't try to help her son' 'since Mary tried not to help her son' b. weil ihm die Maria den Kuchen since him-DAT the Mary the cake-ACC nicht zu essen erlaubt hat not to eat allowed has 'since Mary didn't allow him to eat the cake' 'since Mary allowed him not to eat the cake' c. weil die Maria ihrem Sohn nicht zu helfen beschloß since the Mary her son-DAT not to help decided %'since Mary didn't decide to help her son' 'since Mary decided not to help her son' d. weil ihm die Maria den Kuchen since him-DAT the Mary the cake-ACC nicht zu essen vorschlagen hat not to eat proposed has %'since Mary didn't propose to him to eat the cake' 'since Mary proposed to him not to eat the cake' To interpret these sentences with matrix negation, on the other hand, the examples have to involve scrambling. (Assuming that negation does not move covertly but starts out in the predicate it takes scope over, the embedded object has to have undergone scrambling since it occurs to the left of negation.) The examples in (226)a,b are potential restructuring constructions, and hence scrambling is permitted. The interpretation in which negation takes matrix scope is thus licensed in 99

Since in many cases (e.g., íry-infínitives), an interpretation with embedded negation entails the interpretation with matrix negation, the two readings can often only be distinguished by differences in the presuppositions involved.

Graded (non-Restructuring

283

the first two examples. The constructions in (226)c,d, however, cannot involve a restructuring configuration (cf. the impossibility of long passive in these contexts), and the only form of scrambling possible in these examples is focus scrambling. The intuition that wide scope of the negation is less natural in (226)c,d than in (226)a,b can then again be reduced to the fact that some speakers disfavor (or even disallow) this form of scrambling. Speakers that allow this form of scrambling, require a richer context (in particular a focus interpretation) to get the matrix negation reading. To sum up the distribution of "restructuring" criteria in German, we have shown that not all properties define the same class of infinitival constructions. We have discussed four criteria and argued that they split infinitival constructions into two classes: a class of infinitives that allow long passive and scrambling, and a class of infinitives that block long passive and scrambling but allow pronoun fronting and focus scrambling. We have suggested that these mismatches should not be ignored but rather provide a crucial piece of evidence for the notion of graded (non-)restructuring: the first class of infinitives are restructuring infinitives, which are subject to a range of restructuring requirements (such as the lack of tense etc.; see Chapter 2 to Chapter 4), whereas the second class of infinitives are reduced non-restructuring infinitives which are not subject to restructuring requirements but are only sensitive to the presence vs. absence of a C-domain in the infinitive. An interesting question one might ask at this point is whether reduced non-restructuring infinitives also comprise two separate classes (i.e., TP-infinitives vs. vP-infinitives). A distinction between two classes of reduced non-restructuring infinitives would be motivated if two subgroups can be singled out that are systematically distinguished by a particular property or operation and if this distinction correlates with the tense properties of the constructions. Since so far, we have not found properties or criteria that target this difference in German (i.e., we have not found operations that are only possible in tenseless/tensed reduced non-restructuring infinitives), we have subsumed both types of constructions under reduced non-restructuring infinitives. In the rest of this section, we will briefly discuss infiniti-

284

Grades of (non-)restructuring

val constructions in Dutch which will provide some initial evidence for a split of reduced non-restructuring infinitives along the lines suggested, and hence further support for the view of graded (non-)restructuring. As we have argued in Chapter 2, section 2.1, the lack of long passive in Dutch indicates that true restructuring (i.e., constructions involving lexical VP complements) is unavailable in this language and that Dutch infinitival complements are at least vPs. Furthermore, in Chapter 2, section 4.2, we have illustrated that Dutch displays certain "transparency" properties—namely, verb raising plus the IPP effect and remnant extraposition. Interestingly, however, these two properties do not define the same class of constructions. As can be seen in the appendix, section 5, verb raising and the IPP-effect are only found with predicates that are inherently incompatible with tensed infinitival complements (thus the class defined by verb raising/IPP is by and large the same as the class of restructuring predicates in German). Remnant extraposition, on the other hand, is possible in a larger class of constructions than verb raising (e.g., verbs such as besluiten 'decide', beloven 'promise', voorstellen 'propose' allow remnant extraposition but not the IPP effect). As far as we have been able to determine, the predicates that are compatible with remnant extraposition include irrealis, implicative, and aspectual predicates but not propositional or facti ve predicates (cf. Rutten 1991: 78 for a list of remnant extraposition constructions).100 We therefore propose that both verb raising/IPP and remnant extraposition define reduced non-restructuring infinitives in Dutch, but that the two properties target different grades of reduced non-restructuring: verb raising re100

Rutten (1991) also includes the verbs geloven 'believe', menen 'think, mean' which presumably can receive a propositional interpretation. However, all the examples provided by Rutten seem to involve the irrealis interpretation of these predicates. As shown in the appendix for German, many verbs are ambiguous, and generalizations can only be made when the properties are tested in contexts that clearly involve a particular interpretation (see for instance the contrasts regarding pronoun fronting with irrealis vs. propositional fear/think/mean/hope in (248) and (267)). Thus, while the possibility of remnant extraposition with 'think' etc. in Dutch might challenge the generalization suggested in the text, it does not a priori show that it is incorrect.

Graded (non-Restructuring

285

quires a TP-less infinitive, whereas remnant extraposition is possible from TP-infinitives but blocked from CP-infinitives. Although a detailed survey and comparison of these two properties with a larger class of constructions is still outstanding and the different predicates in Dutch have not been tested regarding their semantic properties, this short survey of verb raising/IPP vs. remnant extraposition shows that a separation of different classes of (reduced non-restructuring) infinitives seems to be necessary. Assuming the account suggested is on the right track, Dutch provides further support for the view of graded (non-)restructuring and in particular for a finer grained classification of the class of reduced non-restructuring infinitives. To conclude, the main purpose of the analysis provided in this section was to account for the properties and distribution of German infinitival constructions. We have argued that not all properties define the same class of infinitival constructions. In particular, we have shown that it is necessary to distinguish pronoun fronting and focus scrambling from long passive and non-focus scrambling, and that only the latter properties define true restructuring (long passive and non-focus scrambling are only possible in restructuring configurations, whereas pronoun fronting and focus scrambling are possible in restructuring and reduced non-restructuring infinitives). Despite the language specific flavor of the analysis provided, the approach taken here has direct consequences for the classification of infinitival constructions and the nature of restructuring in general. The investigation of German infinitival constructions as well as the short discussion of Dutch infinitival constructions have shown that a simple restructuring/non-restructuring distinction is insufficient and that the view of graded (non)-restructuring provides an empirically more adequate way of characterizing infinitival constructions in these languages. Under this view, conflicting classifications of different restructuring properties are not surprising (but are in fact expected) and can be accounted for by a careful characterization of the properties employed. Thus we would expect that the distinction between different grades of (non-)restructuring can be detected in one form or another in other languages as well; however, an exhaustive crosslinguistic study of the distribution of different restructuring properties is obviously beyond the scope of this book.

286

Grades of (non-Restructuring

3.2.

Non-restructuring

In this section, we will discuss full non-restructuring infinitives and provide further evidence for our claim that full non-restructuring infinitives differ from reduced non-restructuring infinitives in that the former but not the latter involve an embedded C-domain. We will introduce two further properties—extraposition/intraposition and relative clause pied piping—and argue that these properties require the presence of a CP. Crucially, these properties will again only be found in full non-restructuring infinitives and be blocked in restructuring and reduced non-restructuring infinitives. Before we turn to these properties, let us recapitulate the relevant tests showing that (full) non-restructuring infinitives indeed prohibit all operations possible in restructuring and reduced non-restructuring infinitives. As we will argue in section 4.2, propositional and factive complements are unambiguously (full) non-restructuring infinitives. While examples such as (227)a involving an infinitival complement to a factive verb such as regret are possible in German, they become ungrammatical when long passive (cf. (227)b), pronoun fronting (cf. (227)c), or (focus and non-focus) scrambling (cf. (227)d) take place (a more exhaustive list of examples illustrating these properties with a range of non-restructuring infinitives can be found in the appendix, section 3). (227) Non-restructuring infinitives a.

dass Hans bedauerte den Traktor repariert zu haben that John regretted the tractor-ACC repaired to have 'that John regretted that he had repaired the tractor' b. *dassder Traktor repariert zu haben bedauert wurde that the tractor-NOMto repair to have regretted was 'that they regretted that they had repaired the tractor' c. *dass ihn Hans repariert zu haben bedauerte that it-ACC John repaired to have regretted 'that John regretted that he had repaired it' d. *dass Hans nur den Traktor bedauert hat that John only the tractor-ACC regretted has

Graded (non-)restructuring

287

repariert zu haben repaired to have 'that John {only} regretted having repaired {only} the tractor' Assuming as we do here that non-restructuring infinitives such as the ones in (227) project CP-complements and that the properties mentioned above are subject to clause boundedness (i.e., cannot occur across a CP-boundary), the ungrammaticality is straightforwardly accounted for. A question arising in this respect is why these types of infinitival constructions can only be represented by CP-complements whereas other infinitival constructions allow smaller complements. We will postpone this question until section 4.2 and first turn to our final argument for graded (non-)restructuring which will be based on two further properties (relative clause pied piping and extraposition vs. intraposition) that we will argue differentiate between CPinfinitives and smaller complements. 3.2.1. Relative clause pied piping As has been pointed out by many authors, restructuring infinitives differ from (full) non-restructuring infinitives in that the latter but not the former allow an operation called relative clause pied piping (see van Riemsdijk 1984, 1985; von Stechow 1984; Haider 1993; Kiss 1995; Grosse 2000). This phenomenon is illustrated in (228)a: if an element embedded in the infinitival complement is relativized, the whole infinitival complement can be pied piped to SpecCP. The structure we suggest for examples such as (228)a is given in (228)b. (228) Relative clause pied piping a.

...derRoman [den schon gelesen zuhaben ] .. .the novel [ that already read to have ]INF der Hans bedauerte the John-NOM regretted ... 'the novel that John regretted having read already'

288

Grades of (non-)restructuring

b.

...the novel

tlNF-CP

V"

V regretted

read

Importantly, relative clause pied piping is impossible in restructuring infinitives involving modal or raising verbs (cf. (229)). (229) Relative clause pied piping with functional restructuring a. *... der Roman [ den lesen ] der Hans muß .the novel [ that read ] the John-NOM must . 'the novel that John must read' b. *...der Roman [den zu lesen] der Hans schien .the novel [ that to read ] the John-NOM seemed . 'the novel that John seemed to be reading' To account for the contrast between (228) and (229) we make the following assumption: w/i-features are located in C° and have to be checked via specifier-head agreement with a wh-element. Furthermore, we assume that in German, successive cyclic movement of a w/i-element 'transfers' the w/i-features to all intermediate C-heads. Thus, in (228)b, movement of the relative pronoun that marks the embedded C° (and hence the embedded CP) as a wh-CP. Assuming the embedded CP inherits w/i-features from the relative pronoun in its specifier, either the relative pronoun or the whole embedded CP can then move to the higher CP to check the w/i-features of the higher C°. Crucially, however, wA-feature transfer or inheritance only

Graded ( non-Restructuring

289

occurs in potential w/z-landing sites—i.e., SpecCP. From these assumptions it follows that only infinitives that involve an embedded CP allow relative clause pied piping. Since functional restructuring predicates such as the ones in (229) do not have the option of embedding a CP-complement, these constructions will generally prohibit relative clause pied piping (except restructuring verbs such as begin or promise which are ambiguous between a functional and a lexical restructuring verb; see below). Let us now turn to lexical restructuring and reduced nonrestructuring constructions. Recall that we have noted at several times in this study that restructuring is not obligatory, but only an option along with other possible structures. These options, however, are not arbitrary but have strictly defined syntactic and semantic consequences. That is, if a certain configuration is chosen (as evidenced by operations such as long passive for restructuring and pronoun fronting for reduced non-restructuring) the particular construction is subject to all restrictions defined for that configuration. If for instance an infinitive involves long passive, the infinitive has to be a "VP"-complement and all properties associated with TP, vP, and embedded subject are prohibited or missing (see also Chapter 6 for a detailed exemplification and summary). In the remainder of this section, we will provide the final argument for this approach. The analysis of relative clause pied piping suggested above constitutes an interesting testing ground for the compatibility of various restructuring properties and hence the structure of the infinitival complements. As we have argued in the previous section, pronoun fronting and focus scrambling require the absence of an infinitival CP. Similarly, restructuring properties such as long passive and nonfocus scrambling require, among other things, the lack of an infinitival CP. Assuming that relative clause pied piping requires the presence of an infinitival CP, the obvious prediction is that relative clause pied piping can never co-occur with any of the above properties or operations. We will see momentarily that this prediction is borne out. The first fact to note is that both potential lexical restructuring infinitives and potential reduced non-restructuring infinitives permit relative clause pied piping (cf. (230)). The grammaticality of (230) leads to the conclusion that potential lexical restructuring and poten-

290

Grades of (non-Restructuring

tial reduced non-restructuring constructions can be represented by a full clausal structure—i.e., as full non-restructuring infinitives. (230) Relative clause pied piping in potential restructuring and reduced non-restructuring contexts a.

b.

...der Roman [den zu lesen] der Hans versuchte ...thenovel [ that toread ] theJohn-NOM tried ... 'the novel that John tried to read' ...der Roman [den zu lesen] der Hans plante ...the novel [that toread ] the John-NOM planned ... 'the novel that John planned to read'

If, however, examples such as the ones above are constructed in a way that guarantees a restructuring or reduced non-restructuring configuration, the results are clearly illformed. The examples in (231)a,a' which do not involve any restructuring properties are full non-restructuring infinitives and hence allow relative clause pied piping. The examples in (231)b-d, on the other hand, illustrate the incompatibility of relative clause pied piping with the restructuring properties long passive (cf. (231)b), scrambling (focus or non-focus scrambling; cf. (231)c,c'), and pronoun fronting (cf. (231)d,d'). (231) Relative clause pied piping and restructuring properties a.

...derRoman [den ihr zugeben ] ...the novel [that her-DAT to give ] der Hans schon oft versucht hatte the John-NOM already often tried had ...'the novel that John had tried to give to her already several times' a'. ...der Mann [dem einen Roman zugeben] ...the man [that anovel-ACC to give ] nur der Hans versucht hatte only the John-NOM tried had ... 'the man that only John had tried to give a novel' b. *...der Mann [ dem zugeben] ...the man [that to give ]

Graded ( non-Restructuring

291

nur ein Roman versucht wurde only a novel-NOM tried was ... 'the man that they (imp.) tried to give only a novel' c. *... der Roman [ den zugeben] ...the novel [ that to give ] der Hans nur der Maria versucht theJohn-NOM only the Mary-DAT tried 'the novel that John tried to give only to Mary' c'. *... der Mann [ dem zugeben] ...the man [ that to give ] der Hans nur einen Roman versucht theJohn-NOM only a novel-ACC tried 'the man that John tried to give only a novel' d. *... der Roman [ den zugeben] ...the novel [ that to give ] ihr nur der Hans versucht her-DAT only the John-NOM tried 'the novel that only John tried to give to her' d'. *.. .der Mann [ dem zu geben ] ...the man [ that to give ] ihn nur der Hans versucht him/it-ACC only the John-NOM tried 'the man that only John tried to give it to'

hatte had

hatte had

hatte had

hatte had

Since none of the operations possible in restructuring or reduced nonrestructuring configurations are possible when relative clause pied piping takes place we can conclude that relative clause pied piping indeed targets a full clausal structure which then however precludes any restructuring property. 3.2.2. Extraposition vs. intraposition The last property we will discuss here is the position of the infinitival complement with respect to the matrix verb since this property is tied in intricate ways to the restructuring/non-restructuring distinction. As discussed in Chapter 3 section 3.2.1, functional restructuring infini-

292

Grades of (non-)restructuring

tives in contrast to (potential) lexical restructuring infinitives have to occur in pre-verbal (i.e., intraposed) position and cannot undergo extraposition. The relevant data are repeated here as (232). (232) Extraposition/intraposition in functional vs. lexical contexts a.

weil der Hans der Maria helfen muß since the John-NOM the Mary-DAT help must 'since John must help Mary' b. *weil der Hans muß der Maria helfen since the John-NOM must the Mary-DAT help 'since John must help Mary' c. weil der Hans der Maria zu helfen versuchte since the John-NOM the Mary-DAT to help tried 'since John tried to help Mary' d. weil der Hans versuchte der Maria zu helfen since the John-NOM tried the Mary-DAT to help 'since John tried to help Mary' The distribution of actual restructuring infinitives and reduced nonrestructuring infinitives is somewhat more complex. Note first that both constructions are well-formed in intraposed position as is illustrated in (233) ((233)a involves pronoun fronting, (233)b involves scrambling, and (233)c involves long passive). (233) Intraposition with restructuring and reduced non-restructuring a.

b.

c.

dass that 'that dass that 'that dass that 'that

ihr der Hans ther zu helfen beschlossen her the John-NOM ther to help decided John decided to help her' dem Hans Maria zu helfen beschlossen the John-DAT Mary to help decided Mary decided to help John' der Wagen zu reparieren versucht wurde thecar-NOM to repair tried was they tried to repair the car'

hat has hat has

Concerning extraposition, however, the situation is less clear. While

Graded (non-)restructuring

293

some researchers (including the author here) clearly accept extraposed reduced non-restructuring infinitives such as (234)a,b (see also Bayer and Kornfilt 1990 and Sabel 1994/1996), some authors consider these examples as ungrammatical (cf. Kiss 1995, Grosse 2000). Furthermore, comparing (234)a,b to the non-extraposed versions of these examples usually yields a preference for the latter. Similarly, examples with long passive show a clear decrease in acceptability when the infinitive is extraposed as in (234)c. (234) Extraposition with restructuring and reduced non-restructuring a. %dass ihr der Hans beschlossen hat ther zu helfen that her the John-NOM decided has ther to help 'that John decided to help her' b. %dass dem Hans Maria beschlossen hat zu helfen that the John-DAT Mary decided has to help 'that Mary decided to help John' c. Ί Idass der Wagen versucht wurde zu reparieren that thecar-NOM tried was to repair 'that they tried to repair the car' Finally, turning to non-restructuring infinitives, we find a similar uncertainty. While extraposition is generally accepted (cf. (235)a), intraposition of a non-restructuring infinitive is marked and again impossible for some speakers (cf. (235)b). (235) Extraposition vs. intraposition in non-restructuring infinitives a.

weil der Hans zutiefst bedauerte since the John-NOM deeply regretted der Maria nicht geholfen zuhaben the Mary-DAT not helped to have 'since John regretted it deeply that he didn't help Mary' b. %weil der Hans [ der Maria nicht since the John-NOM [ the Mary-DAT not geholfen zuhaben] zutiefst bedauerte helped to have ] deeply regretted 'since John regretted it deeply that he didn't help Mary'

294

Grades of (non-)restructuring

We will not take one or the other intuition concerning the question of extraposition vs. intraposition as superior but rather aim for an explanation that allows us to encode the variation found in (234) to (235). Looking at the distribution of extraposition and intraposition of infinitival complements (and its variation) as summarized in Table 31, a clear generalization emerges: the 'bigger' a complement (assuming our system of graded (non-)restructuring), the more likely it is to extrapose; the 'smaller' a complement, the more likely it is to occur in intraposed position. Table 31.

Extraposition and Intraposition

Context

Extraposition

Intraposition

Functional restructuring Lexical restructuring Reduced non-restructuring Non-restructuring

*

OK OK OK OK-* (marked)

?? OK-* (marked) OK

In light of the distribution in Table 31, we will assume that the question of extraposition vs. intraposition is indeed a question of prosodie or syntactic markedness rather than an issue of 'hard' syntactic constraints. In particular, we suggest that the position of a non-nominal complement is determined by the following markedness constraints (assuming that there is a direct correspondence between syntactic structure and prosodie structure, it does not seem to matter whether these constraints are ultimately defined over syntactic or prosodie categories): i) the unmarked position of a full clause (i.e., a CP) is post-verbal, which in prosodie terms would correspond to a constraint against embedding a category of the highest type on the prosodie hierarchy (e.g., an utterance or intonational phrase following Selkirk 1984, 1986) inside another prosodie phrase; ii) the unmarked position of a non-clausal category is its base position, which prosodically could be seen as a reluctance against re-ordering smaller prosodie units (e.g., prosodie phrases). Thus, in the unmarked case, nonrestructuring infinitives undergo extraposition, whereas restructuring and reduced non-restructuring infinitives remain in situ.101 101

An interesting question we have not been able to test yet is whether speakers

Graded ( non-Restructuring

295

To account for the difference found between extraposition of restructuring infinitives (which is close to unacceptable) vs. extraposition of reduced non-restructuring infinitives (which is possible but marked or dispreferred) we suggest tentatively that restructuring infinitives can be reanalyzed as part of the prosodie phrase of the matrix predicate. That is, like functional restructuring constructions, lexical restructuring constructions can represent one single prosodie domain, whereas reduced non-restructuring infinitives constitute an independent prosodie domain. This claim is confirmed by the intuition that (as mentioned at various points in this study) pre-verbal non-restructuring infinitives but not pre-verbal restructuring infinitives require an intonational break before and after the infinitival complement. An illustration of this contrast is provided by the reduced non-restructuring examples in (236)a,b vs. the restructuring examples in (236)c,d. To ensure that the infinitive is a reduced non-restructuring infinitive, we use examples involving pronoun fronting in infinitival constructions with embedded tense or embedded negation. Pronoun fronting is excluded in full non-restructuring infinitives and embedded tense and negation are excluded in restructuring infinitives—hence the constructions can only be reduced non-restructuring infinitives. Importantly, in both cases in (236)a,b, the infinitive has to be set off from the matrix predicate by an intonational break (indicated by the brackets). If on the other hand, negation is interpreted as matrix negation—i.e., the infinitive can be a restructuring infinitive—no intonational pause is necessary (cf. (236)c). (236) Intonation in intraposed infinitival constructions a.

dass ihr Hans [ morgen zu helfen ] beschloß that her John [ tomorrow to help ] decided 'that John decided to help her tomorrow'

who prefer intraposition of reduced non-restructuring infinitives recognize a difference between TP infinitives vs. vP-infinitives (i.e., whether these speakers for instance allow extraposition of TP-infinitives more readily than extraposition of vPinfinitives).

296

Grades of (non-Restructuring

b.

dass that 'that c. dass that 'that *'that d. dass that 'that

ihr Hans [ nicht zu helfen ] versuchte her John [ not to help ] tried John tried not to help her' ihr Hans nicht zu helfen versuchte her John not to help tried John didn't try to help her' John tried not to help her' (*if no intonational break) der Wagen zu reparieren versucht wurde thecar-NOM to repair tried was they tried to repair the car'

Likewise, in examples with long passive such as (236)d (i.e., clear cases of restructuring infinitives), again no intonational break is required. Although the contrasts in (236) support the view that lexical restructuring infinitives do not constitute independent prosodie domains, further empirical experiments are necessary to test this hypothesis in a broader range of constructions and with a larger set of speakers. We thus leave this analysis as preliminary. To conclude, we have seen that the question of extraposition/intraposition can be seen as another criterion that targets the distinction between full clauses (i.e., CPs) vs. smaller complements: extraposition of full non-restructuring infinitives (i.e., CPcomplements) is preferred or required by some speakers; whereas extraposition of smaller complements (i.e., reduced non-restructuring infinitives and restructuring infinitives) is marked or impossible. A summary of the properties targeting the presence/absence of a Cdomain is provided in Table 32. Table 32.

Properties targeting the presence/absence of CP

Property

Embedded CP

No embedded CP

Focus scrambling Pronoun fronting Intraposition of infinitive Extraposition of infinitive Relative clause pied piping

impossible impossible marked or impossible possible possible

possible possible possible possible, marked impossible

Semantic classification

297

4. Semantic classification of infinitives In this final section, we will discuss the influence of the semantic category of an infinitival construction on its syntactic structure. Let us first recapitulate the four classes of infinitival constructions we have distinguished in German. The smallest category are restructuring infinitives. We have shown in Chapter 2 that these infinitives lack CPs, TPs, and vPs, and that semantically, restructuring infinitives express (tenseless) actions or events (cf. Cremers 1983; Rochette 1988, 1990), and are properties rather than propositions (in Chierchia's terms). We have also pointed out that the semantic properties of a particular infinitival construction alone are not sufficient to characterize an infinitival construction (see in particular Chapter 2, section 3.5, and Chapter 4, section 3.2). While the lack of tense and the semantic status of the infinitive as a property rather than a proposition are necessary conditions for restructuring they are not sufficient conditions. Whether an infinitive is a restructuring infinitive or a non-restructuring infinitive is a question that ultimately both the syntax and the semantics have to decide. Turning to non-restructuring infinitives, we have argued that the class of non-restructuring infinitives should also be split into two subgroups: reduced non-restructuring infinitives and full nonrestructuring infinitives. As mentioned in section 1 of this chapter, the former includes irrealis constructions, whereas the latter includes factive and propositional constructions. In this section, we will suggest a possible account to derive this correlation between the syntactic structure and the semantic category of infinitival constructions. We will start with a comparison of the properties of the verb forget which shows an interesting ambiguity for some German speakers—forget can be interpreted as an implicative verb or a factive verb. Crucially, restructuring properties are only found under the first interpretation. In section 4.2, we will then turn to an account of factive and propositional infinitives, which will show that the presence of a CP is crucial for the syntax and semantics of these constructions.

298

Grades of (non-Restructuring

4.1. Two ways to forget Let us begin with an illustration of the two readings of forget. The first and most typical reading of forget is an implicative reading illustrated in (237). The verb forget has a negative implicative reading, which means that the truth of the sentence involving forget implies the falsity of the embedded predicate; if, on the other hand, the sentence is negated, the truth of the embedded infinitive is implied (cf. Karttunen 1971). Thus, to forget ρ implies ->p, whereas not to forget ρ implies p. Note that the implication in (237)b might appear too strong. As has been proposed by many authors, the meaning of implicative forget is essentially negation plus a presupposition that there is some obligation to perform the event denoted by the embedded predicate (cf. Karttunen 1971). That is, (237)a can be paraphrased as John was supposed to water the flowers but he didn 't do it. Under this assumption, it is not surprising that the implication in (237)b for instance might be considered as not valid since the presupposition involved with the verb forget can be cancelled in certain contexts (e.g., as in John didn't forget to water the flowers, he simply didn't want to). (237) Implicative forget a.

b.

weil Hans vergaß die Blumen zu gießen since John forgot the flowers to water 'since John forgot to water the flowers' =» he didn't water the flowers weil Hans nicht vergaß die Blumen zu gießen since John not forgot the flowers to water 'since John didn't forget to water the flowers' =» he watered the flowers

The second reading of forget (which is available for some speakers) is a factive reading.102 As illustrated in (238), under a factive interpretation, the truth of the complement is presupposed whether the 102 We have noticed that not all speakers allow this reading. The examples and judgements to follow are from speakers who permit a factive forget.

Semantic classification

299

matrix predicate is positive or negative (cf. Kiparsky and Kiparsky 1970, 1971). (238) Fictive forget a.

b.

dass Hans vergaß die Blumen schon gegossenzu haben that John forgot the flowers already watered to have 'that John forgot having watered the flowers already' => he watered the flowers dass Hans nicht vergaß that John not forgot die Blumen schon gegossen zu haben the flowers already watered to have 'that John didn't forget having watered the flowers already' => he watered the flowers

Thus, the truth conditions of the embedded predicate do not change when the matrix verb is negated (in contrast to the implicative reading in (237)). Comparing the properties of these two versions of forget, we will see that the implicative construction can be represented by a restructuring configuration, whereas the factive construction only allows a representation as a non-restructuring configuration. The examples in (240) to (243) illustrate the possibility of restructuring properties with implicative forget (i.e., the a.-examples) and the impossibility of restructuring properties with factive forget (i.e., the b.-examples): (239) involves pronoun fronting, (240) involves long passive, and (241) involves scrambling. (239) Pronoun fronting a.

dass that 'that b. *dass that 'that

ihn Hans zu reparieren vergaß it-ACC John to repair forgot John forgot to repair it' ihn Hans schon repariert zu haben vergaß it-ACCJohn already repaired to have forgot John forgot having repaired it already'

300

Grades of (non-Restructuring

(240) Long passive a.

dass der Traktor und der Lastwagen that [the tractor and the truck]-NOM zu reparieren vergessen wurden to repair forgotten were 'that they forgot to repair the tractor and the truck' b. *dass der Traktor und der Lastwagen that [the tractor and the truck]-NOM schon repariert zu haben vergessen wurden already repaired to have forgotten were 'that they forgot having repaired the tractor and the truck already ' (241) Scrambling a. %dass Hans den Traktor vergessen hat zu reparieren that John the tractor-ACC forgotten has to repair 'that John (has) forgotten to repair the tractor' a', dass Hans den Traktor nicht zu reparieren vergaß that John the tractor-ACC not to repair forgot 'that John didn't forget to repair the tractor' 'that John forgot not to repair the tractor' b. *dass Hans nur den Traktor vergessen hat that John only the tractor-ACC forgotten has schon repariert zu haben already repaired to have 'that John {only} forgot having repaired {only} the tractor already' b \ dass Hans [den Traktor nicht repariert that John [ the tractor-ACC not repaired zuhaben] vergessen hat to have ] forgotten has 'that John forgot having not repaired the tractor' *'that John did not forget having repaired the tractor' Furthermore (242) shows that an embedded tense is only available with factive forget and that implicative forget requires a simultane-

Semantic classification

301

ous interpretation; and finally (243) illustrates that factive (but not implicative) forget is compatible with a non-obligatory control interpretation. (242) Tense a.

Er vergaß (*morgen) die Blumen zu gießen He forgot (*tomorrow) the flowers to water 'He forgot to water the flowers (*tomorrow)' b. Er vergaß die Blumen schon gestern gegossen zu haben He forgot the flowers already yest. watered to have 'He forgot that he watered the flowers already yesterday' *'He forgot to have watered the flowers yesterday' (*if implicative) b'. Er vergaß morgen die Blumen gießen zu müssen He forgot tomorrow the flowers water to must 'He forgot that he will have to water the flowers tomorrow' (243) Non-obligatory control a. *Der Direktor vergaß sich im Schloß zu versammeln The principal forgot SELF in-the castle to gather 'The principal forgot to gather in the castle' b. Der Direktor vergaß sich schon oft The principal forgot SELF already often im Schloß versammelt zu haben in-the castle gathered to have 'The principal forgot having gathered in the castle already a few times' b'. Der Direktor vergaß sich morgen The principal forgot SELF tomorrow im Schloß versammeln zu müssen in-the castle gather to must 'The principal forgot that he + his colleagues will have to gather in the castle tomorrow' To sum up, the ambiguity of forget shows that the restructuring/non-

302

Grades of (non-)restructuring

restructuring distinction is sensitive to the semantic category of an infinitival construction. Whereas implicati ves are compatible with a restructuring configuration (e.g., they involve tenseless infinitives, obligatory control etc.), facti ve complements are tensed and allow non-obligatory control and are hence not compatible with a restructuring configuration.103 4.2. F active (and propositional) infinitives The final issue we will address is the question of why factive (and propositional) constructions generally prohibit restructuring properties and why these constructions in contrast to irrealis infinitives are always represented by non-restructuring infinitives. While our account of factive infinitives will be based on common assumptions about the syntax and semantics of these constructions, the account of propositional infinitives will be somewhat more tentative and require further elaboration. However, since the generalization that propositional infinitives do not show restructuring properties is not particular to German but holds throughout the languages showing any kind of restructuring effects, there is reason to assume that our suggestion—despite its preliminary character—is on the right track. Let us start with a few remarks about the syntax and semantics of factive complements. As was first noted by Kiparsky and Kiparsky (1970, 1971), finite factive complements in English are special in that they show island effects for adjunct extraction but not for object extraction (cf. (244)a vs. (244)b). (244) Extraction from factive complements a. lWhox do you regret that John met t¡ ? b. *Whyi do you regret that John met Mary t¡?

As is shown in the appendix, section 1, there seems to be a split between strong implicatives such as manage or forget and weak implicatives such as refuse or pretend. The latter behave more like factives with respect to the restructuring properties. At this point, we have no explanation for this fact and the generalization also has to be tested against a broader range of data.

Semantic classification

303

This property distinguishes factive complements from prepositional complements—the latter do not show any island effects; thus both object and adjunct extraction are possible in (245). (245) Extraction from prepositional complements a. b.

Who, do you think that John met t¡ ? Why, do you think that John met Mary t¡?

Various solutions have been proposed to this puzzle. Kiparsky and Kiparsky (1970, 1971) argue that factive clauses are embedded in a noun phrase of the form [the fact that...]. As has been noted by various authors (cf. Cattell 1978, Hegarty 1991 and others), however, this raises the question of why object extraction shows only a mild deviance (in contrast to regular complex NP constraint violations). If factive complements were part of a complex noun phrase, it would be expected that object extraction should result in a subjacency violation like in other cases of extraction from complex NPs. In more recent approaches, factive complements are thus considered as clauses rather than as noun phrases. To account for the island character of factive complements, various proposals about the structure and semantics of factive clauses have been made. Although the different approaches to factivity differ significantly in the way factive structures are represented and how the phenomenon is implemented technically and theoretically, there is also consensus among the various analyses that factive complements are syntactically and semantically opaque domains. Setting aside the actual mechanism that is employed to 'close off or isolate factive clauses, the assumption that remains is that factive complements do not enter into any direct relations with the matrix predicate (such as selection; cf. Cinque 1990b, or theta-binding; see below), but rather saturate or discharge all relations inside the complement. We will not discuss any of the factivity approaches in detail but only mention two approaches that are important for the present purposes (i.e., the question of why factive complements cannot be restructuring infinitives). Melvold (1991) for instance assumes that factive complements involve a factive operator that is inserted in

304

Grades of (non-)restructuring

SpecCP at LF. The operator in SpecCP then competes with the intermediate adjunct trace that is necessary to γ-mark the original trace of the adjunct (following Lasnik and Saito 1984, Melvold assumes that γ-marking of objects takes place at S-structure, while γ-marking of adjunct traces can only take place at LF). Assuming an account along these lines, the question of why factive complements can only be (full) non-restructuring infinitives follows directly from the necessity of a factive operator in SpecCP and our assumption about the structure of restructuring vs. non-restructuring infinitives: since restructuring infinitives and reduced non-restructuring infinitives lack a C-domain, factive operators cannot be inserted in these constructions, and hence restructuring infinitives and reduced non-restructuring infinitives cannot be interpreted as factive complements. Similarly, Hegarty (1991) proposes that factive complements involve an obligatory complementizer that binds the event position of the embedded clause. Hegarty's system is aimed at accounting for the syntactic as well as semantic properties of factive complements. The crucial semantic property of factive constructions is the notion of semantic factivity. As mentioned before, factive complements are described as presupposing the truth of the embedded complement, or as involving the presupposition that the embedded event in fact occurred. This is illustrated again in (246): regret that ρ presupposes p. Importantly, the presupposition is kept under negation (cf. (246)b). (246) Factive presupposition generation a. b.

John => John =>

regrets that he went to Turkestan he went to Turkestan doesn 't regret that he went to Turkestan he went to Turkestan

According to Hegarty, a factive presupposition can only be generated when the complement does not involve an unsaturated event or argument position. To saturate the embedded event argument, an operator is thus necessary in the infinitival complement (which in Hegarty's approach is a factive complementizer) that binds the open argument positions in the infinitive. Furthermore, Hegarty assumes that

Semantic classification

305

adverbial w/i-elements have an event theta-position that has to be construed with the event position of the modified clause by thetaidentification. However, since the event position of an embedded factive complement is discharged by the complementizer, the whadverbial cannot be construed with the embedded clause. Returning to our question of why restructuring is prohibited with factive complements, Hegarty's analysis also provides a straightforward answer: factivity presupposition requires the presence of a factive operator in the C-domain, whereas restructuring and reduced non-restructuring, on the other hand, require (among other things) the absence of a C-domain (according to our analysis). Thus, again the only structures that are compatible with a factive interpretation are CP-infinitives—i.e., full non-restructuring infinitives. The final generalization we would like to address is the difference between irrealis and propositional infinitives (see also Chapter 2, section 3.1) with respect to restructuring. As mentioned before, irrealis infinitives can be represented as restructuring infinitives (provided that all properties associated with the restructuring configuration are met), reduced non-restructuring infinitives, or full nonrestructuring infinitives. Propositional constructions, on the other hand, lack the first two options and can only involve a nonrestructuring configuration—i.e., a CP-infinitive. To give an example, the verb fear is ambiguous between an irrealis interpretation and a propositional interpretation. Under the propositional interpretation, truth/falsity predication is possible (cf. (247)a) whereas it is blocked under the irrealis interpretation (cf. (247)b; see also the appendix, section 2 for further irrealis/propositional tests and examples). (247) Propositional vs. irrealis fear b.

Hans befürchtete der Maña mißfallen zu haben John feared the Mary-DAT displeased to have was auch stimmte which also was-true 'John feared that he had displeased Mary, which was true' =» it is true that John feared he had displeased Mary => it is true that John had displeased Mary

306

Grades of (non-)restructuring

b.

Hans befürchtete der Maña zumißfallen John feared the Mary-DAT to displease was auch stimmte / *stimmen wird which also was-true / *be-true will 'John feared that he would displease Mary, which was/will be true' => it is true that John feared that he would displease M. =}> it is true that John displeased/will displease Mary

Interestingly, however, pronoun fronting (i.e., a criterion that requires a restructuring or reduced non-restructuring infinitive) is only possible in the irrealis construction (cf. (248)a) and blocked in the prepositional construction (cf. (248)b; a full list of examples testing restructuring properties with a range of irrealis vs. propositional predicates can be found in the appendix, section 3). (248) Pronoun fronting with /ear-infinitives a.

weil ihr Hans zu mißfallen befürchtete since her-DAT John to displease feared 'since John feared to displease Mary' b. *weil ihr Hans mißfallen zu haben befürchtete since her-DAT John displeased to have feared 'since John feared that he had displeased Mary' Thus, the semantic type of an infinitival construction is again a crucial factor for the determination of the syntactic structure. To accommodate this generalization we will assume that propositionality like factivity is a property of the C-domain (see Cremers 1983 for a similar claim), whereas irrealis is an aspectual property which is encoded (if at all syntactically) in an aspect phrase lower than TP. Assuming that a propositional infinitive requires the presence of a CP, it then follows again from our claim that reduced non-restructuring infinitives and restructuring infinitives lack a C-domain that propositional infinitives can only be non-restructuring infinitives. As mentioned above, this account of the syntax and semantics of propositionality is preliminary and requires further elaboration and motivation, but we also believe that it provides a first step towards a better

Semantic classification

307

understanding of the structure and meaning of infinitival constructions. Note also that our approach is incompatible with recent approaches by Pesetsky and Torrego (2001) and Landau (1999/2000), which crucially rely on the assumption that the irrealis property is represented by features in C. In light of the distribution of reduced vs. full non-restructuring infinitives, this view cannot be adopted (at least not for German), since a number of properties (i.e., the possibility of pronoun fronting and focus scrambling, the impossibility of relative clause pied piping, and the marked nature of extraposition) suggest that irrealis infinitives in German lack a C-domain. The distinction between irrealis and propositional infinitives, on the one hand, and the distinction between the various grades of restructuring, on the other hand, thus provide an interesting empirical testing ground for further research in the area of the syntax and semantics of infinitival constructions.

Chapter 6 Summary and conclusion

One of the main claims we have argued for in this study is that control or raising infinitives do not form uniform classes but fall into a number of subclasses that are characterized by different syntactic and semantic properties. Since the tests that distinguish between different types of infinitival constructions are to a large extent language specific, we have concentrated on the properties of infinitival constructions in one language—German—and suggested that infinitival constructions fall into four basic classes: functional restructuring infinitives, lexical restructuring infinitives, reduced non-restructuring infinitives, and full non-restructuring infinitives. While we assume that this system of graded (non-)restructuring is a general crosslinguistic property of infinitival constructions, the actual classification of infinitival constructions may somewhat vary from language to language (subject to language specific properties). The distribution and properties of the four classes of infinitival constructions in German are summarized in Table 33. The table presents an overview of the four types of infinitival constructions, their syntactic and semantic distribution, the structures suggested, the main syntactic and semantic properties, as well as the operations that are allowed or disallowed in each type of construction. As the table shows, a simple restructuring/non-restructuring system is insufficient to capture the distribution of infinitival constructions in German. Rather, we have argued that at least two types of restructuring (lexical vs. functional restructuring), as well as two grades of nonrestructuring (full vs. reduced non-restructuring) have to be distinguished. Note that the labels we have assigned to these classes (i.e., restructuring, non-restructuring etc.) are immaterial; what is essential, however, is what these labels stand for. Since the different types of

Summary and conclusion

309

constructions are characterized by clearly defined properties, these infinitival classes refer to true linguistic entities which cannot be ignored and which any account of infinitival constructions should capture. Table 33. German infinitival constructions (repeated) Type

Structure Properties, distribution

(Im)Possible operations

Restructuring: Lexical INF = • no embedded (PRO) subject possible: VP-layer • no embedded structural case • long object movement • no embedded tense • scrambling • no embedded negation • pronoun fronting • obligatory control impossible: • possible with: (strong) im- • ??extraposition of infinitive plicatives, aspectuals, irrealis • relative clause pied piping predicates Functional INF = • thematic properties are de- possible: main termined by the embedded • IPP effect predicate predicate (except semi• raising functional predicates) impossible: • possible with: modal, rais- • extraposition of infinitive • matrix passive ing, aspectual, causative, perception, motion verbs • relative clause pied piping Non-restructuring: INF = Reduced possible: • embedded (PRO) subject • pronoun fronting vP or TP • embedded structural case * possible with: implicatives, • focus scrambling aspectuals, irrealis predicates • %extraposition of infinitive impossible: possible: • embedded tense • long object movement • (non-focus) scrambling • embedded negation • non-obligatory control • relative clause pied piping INF = Clausal • embedded (PRO) subject possible: (filli) • relative clause pied piping CP • embedded structural case • embedded tense • extraposition of infinitive * possible with: all lexical impossible: • long object movement predicates • obligatory with: preposi• scrambling • pronoun fronting tional, factive predicates • %intraposition of infinitive

310

Summary and Conclusion

The main purpose of this section is to provide a summary of the major points of our analysis. Table 34 gives an overview of the tests and criteria we have used and developed to distinguish between the different classes of infinitival constructions. The table also lists the crucial properties that these tests target and the conditions and constraints these criteria are subject to. Below, we will then selectively illustrate some of the major properties in detail and highlight the main findings and advantages of our system. Table 34.

Properties of German infinitival constructions

FR: LR: RNR: NR:

Functional restructuring Lexical restructuring Reduced non-restructuring Full clausal non-restructuring

Properties (possible)

Conditions

Long object movement Non-focus scrambling Non-obligatory control Independent embedded tense contribution Focus scrambling Pronoun fronting Relative clause pied piping Intraposition of infinitive Extraposition of infinitive

restructuring configuration + restructuring configuration + •restructuring configuration •restructuring configuration -

+ +

absence of CP absence of CP presence of CP absence of CP (markedness) functional vs. lexical; presence of CP (markedness) functional vs. lexical

IPP effect

FR

LR

RNR NR -

-

-

-

-

+ +

+ +

+ +

+ +

+ +

-

-

-

+

+

+

%

-

??

%

+

+

-

-

-

-

-

+

As is evident from Table 34, a system that treats all restructuring criteria (such as scrambling, long passive, pronoun fronting) alike is seriously challenged by the conflicting classifications these criteria define. In light of the distribution in Table 34, we have therefore concluded that restructuring properties are not simply diagnostics that classify infinitival constructions into two subgroups (such as restructuring vs. non-restructuring infinitives), but rather, restructuring properties are taken to provide direct evidence for a particular syntactic and/or semantic property (e.g., the lack of embedded structural case, the lack of an independent tense interpretation etc.). The re-

Summary and conclusion

311

suiting picture is a view of graded (non-)restructuring—i.e., the claim that German displays a variety of different types of infinitival constructions which can be distinguished by their syntactic and semantic properties. The advantage of this view of graded (non-)restructuring is that it is not only explained but also predicted that different restructuring properties define different classes of infinitival constructions, since different operations are sensitive to different structural conditions, positions, and/or properties. Let us now recapitulate some of the major properties of infinitival constructions in German as illustrated in Table 34. We have first suggested that the first four properties in Table 34 follow from the restructuring configuration—i.e., the presence (non-restructuring) vs. absence (restructuring) of vP, TP, and an external argument. Nonfocus scrambling and long object movement (by virtue of being casedriven movements) are only possible in the absence of infinitival case positions and of an infinitival subject—i.e., essentially when the restructuring configuration is met. Non-obligatory control and an independent tense interpretation, on the other hand, require the presence of an embedded syntactic subject and an embedded TP, respectively, and are hence only possible in non-restructuring infinitives. The next four properties in Table 34 follow from the presence (full non-restructuring) vs. absence (restructuring and reduced nonrestructuring) of an infinitival C-projection. We have argued that pronoun fronting and focus scrambling target the C-domain and are hence blocked across an embedded CP. Similarly, for many speakers, the pre-verbal position is only available for infinitival complements that lack a CP (however, this condition was shown to be a markedness condition rather than an inviolable syntactic constraint). Relative clause pied piping, on the other hand, requires the presence of an embedded C-domain. Finally, the last two properties target (among other things) the lexical vs. functional distinction. The IPP effect in German is only found with functional predicates and extraposition cannot strand functional verbs (in addition, many speakers only allow extraposition of CPs—at least in the unmarked case). One of the major contributions of this study is that although infinitival constructions can be represented by a variety of structures

312

Summary and Conclusion

(at least in German), these structures are not arbitrary but have unavoidable and clearly defined structural and semantic consequences which we will summarize below. In this respect, it is very important to note again that the tables above have to be read as involving the following conditional: a particular infinitival construction is subject to the properties associated with a construction of type X as listed in the table (e.g., restructuring, non-restructuring etc.), only if this particular construction unambiguously qualifies as an infinitive of type X. Thus, for instance, we do not claim that by virtue of being a potential lexical restructuring construction, fry-constructions always have the properties listed for lexical restructuring constructions. In other words, by simply looking at a sentence like (249)a, for instance, little can be said about the properties of infinitival constructions. As is illustrated in (249)b-d, in our account, the sentence is three-way s ambiguous: it can be represented as a lexical restructuring configuration ((249)b), a reduced non-restructuring configuration ((249)c), or a full clausal non-restructuring configuration ((249)d). (249) Ambiguity of ^-constructions a.

b. c. d.

weil Hans den Wagen zu reparieren since John thecar-ACC to repair 'since John tried to repair the car' weil Hans [ den Wagen zu reparieren since John [ the car-ACC to repair weil Hans [ den Wagen zu reparieren since John [ the car-ACC to repair weil Hans [ den Wagen zu reparieren since John [ the car-ACC to repair

versuchte tried ]Vp ]VP ]vP/TP ] VP/TP ]CP ÏCP

versuchte tried versuchte tried versuchte tried

However, as the following summary of our analysis will show, these options are not arbitrary or simply suggested for convenience, but they can be tested empirically by investigating the properties of each construction in light of the criteria we have developed. Since this is a very crucial methodological point that has led to a significant amount of misunderstanding in the literature, we will illustrate it here once again in some detail.

Summary and conclusion

313

Starting with the first option, if we choose the lexical restructuring option in (249)b, the construction is subject to the restructuring configuration. That is, the infinitive allows long passive (cf. (250)a), prohibits an independent tense interpretation (cf. (250)b which would in principle be compatible with embedded future tense), cannot involve embedded negation (cf. (250)c), and lacks an infinitival subject that could function as a binder for embedded anaphors (cf. (250)d). Note again that to test these properties, one has to make sure that the construction under consideration is indeed (249)b. Thus, our crucial claim is that the properties associated with lexical restructuring can only be verified if the tests are applied to structures that cannot be analyzed as (249)c or (249)d. The examples in (250) control for this effect since they all involve long passive which—as we have argued extensively—requires a structure as in (249)b and is incompatible with the structures in (249)c,d. (250) Lexical restructuring a.

dass der Traktor zu reparieren versucht wurde that the tractor-NOM to repair tried was 'that they tried to repair the tractor' b. dass dem Kind nur Kekse that the child-DAT only cookies (*morgen) zu essen erlaubt wurden (*tomorrow) to eat allowed were 'that the child was only allowed to eat cookies tomorrow' c. weil [der Kuchen nicht zu essen] versucht wurde since [thecake-NOM not to eat] tried was 'since they didn't try to eat the cake' *'since they tried not to eat the cake' d. weil {*sich} der Wagen {*sich} since {*SELF} thecar-NOM {*SELF} zu reparieren versucht wurde to repair tried was 'since they tried to repair themselves the car' Turning to the second option—the option of reduced non-

314

Summary and Conclusion

restructuring as in (249)c—we have claimed that reduced nonrestructuring infinitives permit pronoun fronting and focus scrambling, however, in contrast to restructuring infinitives, these constructions are not subject to the restructuring configuration. Reduced non-restructuring infinitives (as evidenced by pronoun fronting) are compatible with an independent embedded tense (provided that the matrix predicate is compatible with a complement with independent tense (cf. (251a)), allow embedded negation (cf. (251)b), and involve an infinitival PRO subject that functions as a binder for an embedded anaphor (cf. (251)c). However, crucially, reduced non-restructuring infinitives do not project an embedded CP and hence these contexts block relative clause pied piping (cf. (251)d), which would require an infinitival C-domain. (251) Reduced non-restructuring a.

weil ihn der Hans since him/it-ACC the John-NOM [(morgen) zu reparieren] beschlossen hatte [(tomorrow) to repair] decided had 'since John had decided to repair it tomorrow' b. weil ihn der Hans nicht mehr since him-ACC the John-NOM not anymore zu bevorzugen beschlossen hatte to privilege decided had #'since John didn't decide anymore to privilege him' 'since John decided not to privilege him anymore' c. weil sie es ihm [PRO tit sich anzuschauen] verbot since she it him [PRO tit SELF to-look-at] forbade 'since she forbade him to look at it' d. *... der Roman [den zugeben] ...the novel [ that to give ] ihr nur der Hans versucht hatte her-DAT only the John-NOM tried had 'the novel that only John tried to give to her' Finally, the sentence in (249)a can also be represented by a full clausal structure as in (249)d. If this non-restructuring option is cho-

Summary and conclusion

315

sen, relative clause pied piping is licensed (cf. (252)a), however, at the same time scrambling (cf. (252)b), pronoun fronting (cf. (252)c=(251)d), and long passive (cf. (252)d) are then prohibited. (252) Non-restructuring a.

...derRoman [ den ihr zugeben ] ...the novel [ that her-DAT to give ] der Hans schon oft versucht hatte the John-NOM already often tried had ... 'the novel that John had tried to give to her already several times' b. *.. .der Roman [ den zu geben ] ...the novel [ that to give ] der Hans nur der Maria versucht hatte the John-NOM only the Mary-DAT tried had 'the novel that John tried to give only to Mary' c. * ...der Roman [ den zugeben] ...the novel [ that to give ] ihr nur der Hans versucht hatte her-DAT only the John-NOM tried had 'the novel that only John tried to give to her' d. *...der Mann [ dem zugeben] ...the man [ that to give ] nur ein Roman versucht wurde only a novel-NOM tried had ... 'the man that they (imp.) tried to give only a novel to' To conclude, the distribution of infinitival constructions in German is subject to a range of syntactic and semantic conditions that interact in non-trivial ways with each other and with the structures and properties of the different types of infinitival constructions. These interactions lead to the conclusion that a simple binary classification as plus vs. minus restructuring is empirically inadequate, and thus any theory reducing the data to such a binary division will not be sufficient in accounting for the range and variety of properties (see Table 33) that have been identified and systematized in this study. We have attempted to show here that there is a systematicity to the data when

316

Summary and Conclusion

the entire range of infinitival complementation constructions is examined. In presenting these data, the generalizations, and a theoretical account, we hope to have demonstrated that a number of interesting and robust generalizations in the syntax and semantics of these constructions can be captured naturally by an independently motivated syntax and the recognition of a more finely graded approach to the (non)-restructuring dimension.

Appendix

1. Overview The appendix provides a summary and reference source for the data used in this study. The judgements are based on previous works by other authors, consultation of up to 20 German speakers from various German speaking areas, and the author's own intuitions. Section 2 summarizes the predicates in German that were tested for this study. Only verbs that combine with infinitival complements and that do not select accusative arguments in addition to the infinitive are included (constructions with matrix accusatives generally involve clausal infinitives prohibiting transparency effects; cf. Sabel 1994/1996; Haider 1993; Grosse 2000). Section 3 classifies the predicates as irrealis, propositional, or factive predicates according to the tests in Table 35. Table 35. Criteria for semantic classification Type of construction

Irrealis

Embedded gestern 'yesterday' * * Embedded perfective * Truth/falsity predication * "The fact that* Implicit control OK Embedded modal #

Propositional

Factive

OK OK OK

OK OK

*

OK ?? OK

?? OK

*

Section 4 summarizes the generalizations of graded (non-)restructuring. The properties tested are given in Table 36. Finally, section 5 gives a list of restructuring predicates in German, Dutch, Italian, Spanish, and Japanese, and summarizes some of the previous restructuring/non-restructuring classifications by other authors. Table 36. (Non-)Restructuring properties Operation

Restructuring Reduced non-restructuring Non-restructuring

Long object movement Non-focus scrambling Pronoun movement Focus scrambling

+ + + +

+ +

-

318

Appendix

2. Verb classification Table 37 summarizes the German verbs tested in this study and their semantic classification. Verbs that are listed in more than one category are ambiguous. Note that the classification is not absolute but intended to represent the verbs' unmarked interpretations; some verbs allow re-classification in appropriate contexts. Table 37. Verb classes Class Aspectual

Verb

anfangen beginnen Irrealis ?ablehnen beabsichtigen befürchten drohen entscheiden ermöglichen gestatten hoffen raten verbieten versprechen vorschlagen ?sich weigern Implicative gelingen (strong) vergessen wagen Implicative ablehnen (weak) vermeiden sich weigern angeben Factive bedauern gestehen vergessen Propositional angeben annehmen behaupten erklären erzählen glauben ?leugnen vorgeben

Gloss

Verb

Gloss

start begin refuse, reject intend fear threaten choose enable permit hope advise forbid promise propose refuse manage forget dare refuse, reject avoid refuse declare, report regret confess forget declare, report assume claim declare tell believe deny pretend

aufhören fortfahren anbieten befehlen beschließen empfehlen erlauben erwägen glauben planen untersagen verlernen versuchen vorziehen wünschen mißlingen versäumen vermeiden aufgeben vorgeben

stop continue offer order decide recommend allow contemplate mean plan prohibit unlearn try prefer wish (=order) fail miss, neglect avoid give up pretend

ankündigen erwähnen leugnen zugeben ankündigen befürchten bezweifeln erwähnen gestehen hoffen vermuten zugeben

announce mention deny admit announce fear doubt mention confess hope suspect admit

Distinguishing irrealis, propositional, factive

319

3. Distinguishing irrealis, propositional, factive The examples in (253)-(262) below involve the verbs as listed in Table 38 (i.e., the order of the verbs is kept constant in the examples and the numbering corresponds to the letters in the first column). Table 38. Verbs used in examples (253)-(262) Exx Irrealis a b c d e f g h i j k 1 m η 0 Ρ q r s t u V w χ y ζ

3.1. —

ablehnen anbieten beabsichtigen befehlen befürchten beschließen drohen empfehlen entscheiden erlauben ermöglichen erwägen gestatten glauben hoffen planen raten sich weigern untersagen verbieten verlernen versprechen versuchen vorschlagen vorziehen wünschen

Irrealis

Gloss

Factive/propositional Gloss

refuse, reject offer intend order fear decide threaten recommend choose allow enable contemplate permit mean hope plan advise refuse prohibit forbid unlearn promise try propose prefer wish

angeben ankündigen annehmen bedauern befürchten behaupten bezweifeln erklären erwähnen erzählen gestehen glauben hoffen leugnen vergessen vermuten vorgeben zugeben

declare, report announce assume regret fear claim doubt declare mention tell confess believe hope deny forget suspect pretend admit

properties

Embedded perfective is impossible in irrealis constructions Befürchten 'fear', glauben 'believe, mean', and hoffen 'hope' are compatible with irrealis and non-irrealis interpretations and hence allow embedded perfective Perfective is marginally possible in some of the other examples when a non-irrealis (e.g., factive) interpretation is coerced

320

Appendix

(253)

a. b. c. d. e. f. g. h. i. j. k. 1. m. n. o. p. q. r. s. t. u. v. w. χ. y. z.

-»

(254)

??Hans lehnt(e) es ab (gestern) gewonnen zu haben ??Hans bietet/bot (der Maria) an (gestern) gewonnen zu haben *Hans beabsichtigt(e) (gestern) gewonnen zu haben *Hans befiehlt/befahl der Maria (gestern) gewonnen zu haben Hans beftirchtet(e) (gestern) nicht gewonnen zu haben ??Hans beschloß (gestern) gewonnen zu haben *Hans droht(e) (der Maria) (gestern) gewonnen zu haben *Hans empfiehlt/empfahl (der Maria) (gestern) gewonnen zu haben ??Hans hat entschieden (gestern) gewonnen zu haben *Hans erlaubt(e) der Maria (gestern) gewonnen zu haben *Hans hat der Maria ermöglicht (gestern) gewonnen zu haben *Hans erwog (gestern) gewonnen zu haben *Hans gestattet(e) der Maria (gestern) gewonnen zu haben Hans glaubt(e) (gestern) gewonnen zu haben ?Hans hofft(e) (gestern) gewonnen zu haben *Hans plante (gestern) gewonnen zu haben *Hans riet der Maria (gestern) gewonnen zu haben *Hans untersagt(e) der Maria (gestern) gewonnen zu haben *Hans verbietet/verbot der Maria (gestern) gewonnen zu haben *Hans hat verlernt (gestern) gewonnen zu haben ??Hans verspricht/versprach (der Maria) (gestern) gewonnen zu haben *Hans versucht(e) (gestern) gewonnen zu haben *Hans schlägt/schlug (der Maria) vor (gestern) gewonnen zu haben ??Hans zieht/zog es vor (gestern) gewonnen zu haben ??Hans weigert(e) sich (gestern) gewonnen zu haben *Hans wiinscht(e) (gestern) gewonnen zu haben

Past future interpretation is possible in irrealis constructions (except with versuchen 'try', verlernen 'unlearn* which require a simultaneous temporal orientation of the matrix and embedded event) Only future adverbials in past future interpretation (impossible with embedded gestern 'yesterday') a. b. c. d. e. f.

Hans hatte es (vor einer Woche) abgelehnt *gestern/°Kam nächsten Tag abzureisen Hans hatte (der Maria) (vor einer Woche) angeboten *gestern/°Kam nächsten Tag abzureisen Hans hatte (vor einer Woche) beabsichtigt *gestern/?am nächsten Tag abzureisen Hans hatte der Maria (vor einer Woche) befohlen *gestern/°Kam nächsten Tag abzureisen Hans hatte (vor einer Woche) befürchtet *gestern/°Kam nächsten Tag zu verlieren Hans hatte (vor einer Woche) beschlossen *gestern/°Kam nächsten Tag abzureisen

Distinguishing irrealis, propositional, /active g.

K

321

Hans hatte (der Maria) (vor einer Woche) gedroht *gestern/° am nächsten Tag abzureisen h. Hans hatte der Maria (vor einer Woche) empfohlen *gestern/°Kam nächsten Tag abzureisen i. Hans hatte (vor einer Woche) entschieden *gestern/°Kam nächsten Tag abzureisen j. Hans hatte der Maria (vor einer Woche) erlaubt *gestern/°Kam nächsten Tag abzureisen k. Hans hatte der Maria (vor einer Woche) ermöglicht *gestern/°Kam nächsten Tag abzureisen 1. Hans hatte (vor einer Woche) erwogen *gestem/°Kam nächsten Tag abzureisen m. Hans hatte der Maria (vor einer Woche) gestattet *gestem/°Kam nächsten Tag abzureisen n. ?Hans hatte (vor einer Woche) geglaubt *gestem/°Kam nächsten Tag zu gewinnen o. Hans hatte (vor einer Woche) gehofft *gestem/°Kam nächsten Tag zu gewinnen p. Hans hatte (vor einer Woche) geplant *gestern/°Kam nächsten Tag abzureisen q. Hans hatte der Maria (vor einer Woche) geraten *gestern/°Kam nächsten Tag abzureisen r. Hans hatte der Maria (vor einer Woche) untersagt *gestem/°Kam nächsten Tag abzureisen s. Hans hatte der Maria (vor einer Woche) verboten *gestern/°Kam nächsten Tag abzureisen t. *Hans hatte verlernt gestern/am nächsten Tag zu schwimmen u. Hans hatte (der Maria) (vor einer Woche) versprochen *gestern/°Kam nächsten Tag abzureisen v. *Hans hat versucht gestern/am nächsten Tag abzureisen w. Hans hatte (der Maria) (vor einer Woche) vorgeschlagen *gestern/°Kam nächsten Tag abzureisen χ. ?Hans hatte es (vor einer Woche) vorgezogen *gestern/°Kam nächsten Tag abzureisen y. ?Hans hatte sich (vor einer Woche) geweigert *gestern/°Kam nächsten Tag abzureisen z. ?Hans hatte (vor einer Woche) gewünscht *gestern/°Kam nächsten Tag abzureisen Matrix passive/implicit control is possible (but marked to impossible for verbs that also allow a non-irrealis interpretation) (255)

a. ??Es wurde abgelehnt den Vorschlag in Betracht zu ziehen b. Es wurde (ihnen) angeboten den Peter zu besuchen c. Es wurde beabsichtigt den Peter zu besuchen d. Es wurde (ihnen) befohlen den Peter zu besuchen

322

Appendix e. f. g. h. i. j. k. 1. m. n. o. p. q. r. s. t. u. v. w. χ. y. z.

??Es wurde befürchtet den Bus zu versäumen Es wurde beschlossen den Peter zu besuchen Es wurde (ihnen) gedroht den Strom abzustellen Es wurde (ihnen) empfohlen den Wagen zu reparieren Es wurde entschieden den Peter zu besuchen Es wurde (ihnen) erlaubt den Peter zu besuchen Es wurde ihnen ermöglicht den Peter zu besuchen Es wurde erwogen den Peter zu besuchen Es wurde (ihnen) gestattet den Peter zu besuchen ??Es wurde geglaubt das Spiel noch gewinnen zu können ??Es wurde gehofft den Peter besuchen zu können Es wurde geplant den Peter zu besuchen Es wurde ihnen geraten den Peter zu besuchen weigern: no passive Es wurde ihnen untersagt den Peter zu besuchen Es wurde (ihnen) verboten den Peter zu besuchen ?Es wurde immer mehr verlernt gute Arbeiten zu schreiben Es wurde (ihnen) versprochen den Peter zu besuchen Es wurde versucht den Peter zu besuchen Es wurde (ihnen) vorgeschlagen den Peter zu besuchen ??Es wurde vorgezogen den Bus zu nehmen Es wird gewünscht keine Abfälle zu hinterlassen

Embedded modals are marked pragmatically or redundant (requires coersed interpretation); in contexts that allow a factive or propositional interpretation modals are unmarked (256)

a. b. c. d. e. f. g· h. i. j· k. 1. m. n. o. P· q·

#Hans #Hans #Hans #Hans Hans #Hans ?Hans #Hans #Hans #Hans #Hans #Hans #Hans Hans Hans #Hans #Hans

Distinguishing irrealis, propositioned, f active r. s. t. u. v. w. χ. y. z.

3.2.

323

#Hans untersagte der Maria verreisen zu dürfen/wollen/müssen #Hans verbot der Maria verreisen zu dürfen/wollen/müssen #Hans hat verlernt verreisen zu dtlrfen/wollen/müssen ?Hans versprach (der Maria) verreisen zu dürfen/wollen/müssen #Hans versuchte verreisen zu dürfen/wollen/müssen #Hans schlug (der Maria) vor verreisen zu dürfen/wollen/müssen #Hans zog es vor verreisen zu dürfen/wollen/müssen #Hans weigerte sich verreisen zu dürfen/wollen/müssen ?Hans wünschte verreisen zu dürfen

F active, propositional Embedded perfective and embedded gestern 'yesterday' are possible (marked for verbs that also allow an irrealis interpretation)

(257)

a. Hans b. Hans c. Hans d. Hans e. Hans f. Hans Hans g· h. Hans i. Hans Hans j· k. Hans 1. Hans m. ?Hans n. Hans o. Hans Hans P· Hans q· r. Hans Matrix passive/implicit control is marginal to impossible (it is OK with vorgeben 'pretend' which can also be an implicative); the examples might have different degrees of unacceptability; we simply mark them with ??

(258)

a. b. c. d. e. f.

??Es wurde angegeben einen Mord begangen zu haben ??Es wurde (dem Gericht) ankündigt einen Mord begangen zu haben ??Es wurde angenommen einen Mord begangen zu haben ??Es wurde bedauert einen Mord begangen zu haben ??Es wurde befürchtet einen Mord begangen zu haben ??Es wurde behauptet einen Mord begangen zu haben

324

Appendix g. h. i. j. k. 1. m. n. 0. p. q. r.

??Es wurde bezweifelt einen Mord begangen zu haben ??Es wurde (dem Gericht) erklärt einen Mord begangen zu haben ??Es wurde erwähnt einen Mord begangen zu haben ??Es wurde (dem Gericht) erzählt einen Mord begangen zu haben ??Es wurde (dem Gericht) gestaneinen einen Mord begangen zu haben ??Es wurde geglaubt einen Mord begangen zu haben ??Es wurde gehofft keinen Mord begangen zu haben ??Es wurde geleugnet einen Mord begangen zu haben ??Es wurde vergessen einen Mord begangen zu haben ??Es wurde vermutet einen Mord begangen zu haben ?Es wurde vorgegeben einen Mord begangen zu haben ??Es wurde zugegeben einen Mord begangen zu haben

Embedded modals are unmarked (259)

a. ?Hans hat angegeben (nicht) verreisen zu wollen/müssen/dürfen b. Hans hat angekündigt (nicht) verreisen zu ?wollen/m(issen/dürfen c. Hans hat angenommen (nicht) verreisen zu müssen/dürfen d. Hans hat bedauert nicht (nicht) verreisen zu müssen/dürfen e. Hans hat befürchtet (nicht) verreisen zu müssen/dürfen f. Hans hat behauptet (nicht) verreisen zu wollen/müssen//dürfen g. Hans hat bezweifelt verreisen zu dürfen h. Hans hat erklärt (nicht) verreisen zu wollen/dürfen 1. Hans hat erwähnt (nicht) verreisen zu wollen/müssen/dürfen j. Hans hat erzählt (nicht) verreisen zu wollen/müssen/dürfen k. Hans hat gestanden (nicht) verreisen zu wollen/dürfen 1. Hans hat geglaubt (nicht) verreisen zu müssen/dürfen m. ?Hans hat gehofft (nicht) verreisen zu müssen/dürfen n. Hans hat geleugnet verreisen zu wollen/müssen o. Hans hat vergessen (nicht) verreisen zu wollen/müssen/dürfen p. Hans hat vermutet nicht verreisen zu dürfen q. Hans hat vorgegeben (nicht) verreisen zu wollen/dürfen r. Hans hat zugegeben (nicht) verreisen zu wollen/dürfen The the fact that construction is possible with factives, impossible with propositional predicates

(260)

a. Hans hat die Tatsache den Mord begangen zu haben angegeben b. Hans hat die Tatsache den Mord begangen zu haben angekündigt c. *Hans hat die Tatsache den Mord begangen zu haben angenommen ^ [assume interpretation; OK under accept interpretation] d. Hans hat die Tatsache den Mord begangen zu haben bedauert e. *Hans hat die Tatsache den Mord begangen zu haben befürchtet f. *Hans hat die Tatsache den Mord begangen zu haben behauptet g. ??Hans hat die Tatsache den Mord begangen zu haben bezweifelt

Distinguishing irrealis, propositional, f active h. i. j. k. 1. m. n. 0. p. q. r.

325

*Hans hat die Tatsache den Mord begangen zu haben erklärt Hans hat die Tatsache den Mord begangen zu haben erwähnt *Hans hat die Tatsache den Mord begangen zu haben erzählt Hans hat die Tatsache den Mord begangen zu haben gestanden ??Hans hat die Tatsache den Mord begangen zu haben geglaubt *Hans hat die Tatsache den Mord begangen zu haben gehofft Hans hat die Tatsache den Mord begangen zu haben geleugnet Hans hat die Tatsache den Mord begangen zu haben vergessen *Hans hat die Tatsache den Mord begangen zu haben vermutet *Hans hat die Tatsache den Mord begangen zu haben vorgegeben Hans hat die Tatsache den Mord begangen zu haben zugegeben

Truth/falsity predication is possible with propositional constructions, impossible with all others (261)

[]: was auch stimmte 'which is true' modifies the infinitive a. b. c. d. e. f. g.

'Hans gab an [den Mord begangen zu haben was auch stimmte] Hans kündigte an [den Mord begangen zu haben was auch stimmte] Hans nahm an [im Lotto gewonnen zu haben was auch stimmte] #Hans bedauert [den Mord begangen zu haben was auch stimmte] Hans befürchtete [den Bus verpasst zu haben was auch stimmte] Hans behauptete [im Lotto gewonnen zu haben was auch stimmte] Hans bezweifelte im Lotto gewonnen zu haben was auch stimmte: Hans hat nicht im Lotto gewonnen was aber stimmte: ?Hans hat im Lotto gewonnen h. Hans erklärte [im Lotto gewonnen zu haben was auch stimmte] 1. 'Hans erwähnte [im Lotto gewonnen zu haben was auch stimmte] j. Hans erzählte [im Lotto gewonnen zu haben was auch stimmte] k. 'Hans gestand [den Mord begangen zu haben was auch stimmte] 1. Hans glaubte [im Lotto gewonnen zu haben was auch stimmte] m. ?Hans hoffte [im Lotto gewonnen zu haben was auch stimmte] n. ?Hans leugnete den Mord begangen zu haben was auch stimmte: Hans hat den Mord nicht begangen was aber stimmte: ?Hans hat den Mord begangen o. #Hans vergaß [im Lotto gewonnen zu haben was auch stimmte] p. Hans vermutete [im Lotto gewonnen zu haben was auch stimmte] q. Hans gab vor [im Lotto gewonnen zu haben was auch stimmte] r. ?Hans gab zu [im Lotto gewonnen zu haben was auch stimmte] -»

?

We have found only two true factive (infinitival taking) verbs—bedauern 'regret', vergessen 'forget' (see Table 39) Generalization: all verbs ambiguous between propositional and factive are verbs of saying (see (262))

326

Appendix

Table 39. Prepositional and factive Verb

"thefact that" "which is true" Verb of saying Classification

declare/report announce assume regret claim fear doubt declare mention tell confess believe hope deny forget suspect pretend admit

+ +

+ +

-

+/? +

+

-

-

-

+

+

-

+

-

??

+

-

+

+ -

+ ?? -

+ + -

+

+/? ?+ +/? +

? ?

-

+ +

-

-

+ + +

+ -

+ -

+ +

ambiguous ambiguous propositional factive propositional propositional propositional propositional ambiguous propositional ambiguous propositional propositional ambiguous factive propositional propositional ambiguous

Verbs of "saying": possible in 'He lied when he...' contexts (262)

a. b. c. d. e. f. g. h. i. j. k. 1. m. n. o. p. q. r.

Hans log als er angab den Mord begangen zu haben Hans log als er ankündigte den Mord begangen zu haben #Hans log als er annahm den Mord begangen zu haben #Hans log als er bedauerte den Mord begangen zu haben #Hans log als er befürchtete den Mord begangen zu haben Hans log als er behauptete den Mord begangen zu haben #Hans log als er bezweifelte den Mord begangen zu haben Hans log als er erklärte den Mord begangen zu haben Hans log als er erwähnte den Mord begangen zu haben Hans log als er erzählte den Mord begangen zu haben Hans log als er gestand den Mord begangen zu haben #Hans log als er glaubte den Mord begangen zu haben #Hans log als er hoffte den Mord begangen zu haben Hans log als er leugnete den Mord begangen zu haben #Hans log als er vergaß den Mord begangen zu haben #Hans log als er vermutete den Mord begangen zu haben Hans log als er vorgab den Mord begangen zu haben Hans log als er zugab den Mord begangen zu haben

(Non-)Restructuring properties

327

4. (Non-)Restructuring properties Lexical predicates were tested for four properties (summarized in Table 40). The examples illustrating the properties are listed below (following the numbering as in the first column). Table 40. Long object movement (PASS), non-focus scrambling (OS), pronoun fronting (PRON), focus scrambling (SCR) Exx

Verb

Gloss

a b c d e f g h i j k 1 m η A Β C D E

beabsichtigen beginnen empfehlen erlauben gelingen gestatten mißlingen untersagen verbieten vergessen vermeiden versäumen versuchen wagen anbieten anfangen aufhören befehlen befürchten

intend begin recommend allow manage permit fail prohibit forbid forget avoid miss/neglect try dare offer start stop order fear

F G H I J Κ L

beschließen drohen entscheiden ermöglichen erwägen fortfahren glauben

M

hoffen

Ν 0 Ρ Q R S Τ

planen raten verlernen versprechen vorschlagen vorziehen wünschen

Class

irrealis aspectual irrealis irrealis strong implicative irrealis strong implicative irrealis irrealis strong implicative strong implicative strong implicative irrealis strong implicative irrealis aspectual aspectual irrealis irrealis propositional irrealis decide threaten irrealis irrealis choose enable irrealis contemplate irrealis aspectual continue mean irrealis propositional hope irrealis propositional plan irrealis irrealis advise unlearn modal/irrealis promise irrealis irrealis propose irrealis prefer wish /order irrealis

PASS OS +

+

+

+ +

+

PRON SCR + +

+

+ N/A ΗΝ/Α +

+ + + +

+ +

+ +

+ +

+ +

+ + + +

-

-

+ +

% +

% +

+ % +

%

+

+ + + + +

+ +

+ + + + + + +

+ + + + +

+

+ + + + +

+ +

+ +

-

+ +

-

-

+

-

-

+ + + +

-

?

-

-

-

-

_

_

-

-

-

-

-

-

-

_

-

-

-

-

??

-

-

-

-

-

-

-

-

-

+ +

+

+

+

+ + + + + + +

+ +

+ + + +

+

328

Appendix

EXX VERB

GLOSS

i ii iii iv

ablehnen reject/refuse aufgeben give up vorgeben pretend sich weigern refuse V angeben declare ankündigen announce vi annehmen assume vii vili bedauern regret behaupten claim ix bezweifeln doubt χ xi erklären declare xii erwähnen mention erzählen tell xiii xiv gestehen confess glauben believe XV leugnen deny xvi xvii vergessen forget suspect xviii vermuten zugeben admit xix a-n: A-T: i-xix:

4.1.

CLASS weak implicative weak implicative weak implicative weak implicative propositional/factive propositional/factive propositional factive propositional propositional propositional propositional/factive propositional propositional/factive propositional factive/?propositional factive propositional propositional/factive

Pass OS

PRON SCR ??

*

??

*

?? -

-

-

-

-

-

-

-

-

-

-

-

-

Lexical restructuring Reduced non-restructuring (Full) Non-restructuring

Passive

General comment regarding the data: It has been noted by several authors that long passive is somewhat 'marked' and hence should not be considered as a productive syntactic property (see for instance Grosse 2000). We disagree with this claim on the following grounds. It is well-known that passive in general, and in particular impersonal passive, has to be licensed pragmatically. That is, passive requires an appropriate context and out-of-the-blue passive sentences (such as Es wurde getanzt 'It was danced') are less felicitous for most speakers. Long passive not only involves impersonal passive (in the sense that there is no thematic relation between the embedded object and the verb undergoing passive) but also a complex VPstructure (i.e., it involves two predicates with their own event structures etc.). The fact that long passive is more felicitous in some contexts than in others (involving the same verb) is then not surprising since long passive constructions involve an increased pragmatic complexity, which requires a richer context and a good pragmatic motivation for this operation (see also Wurmbrand 1998b for some discussion of the pragmatic salience requirement). To avoid this potential problem, we will provide more than one example per verb for cases that are controversial or show speaker variation.

(Non-)Restructuring properties

329

Lexical restructuring: long object movement is possible a.

b. c.

d.

e. f.

g.

h.

i. j. k. 1.

?weil der Wagen zu reparieren beabsichtigt wurde ?weil die Autos zu reparieren beabsichtigt wurden weil dieser Plan schon seit langem zu realisieren beabsichtigt wurde (Grosse 2000:71) Zu reparieren beabsichtigt wurden/*wurde nur orange Autos Zu reparieren beabsichtigt wurde nur der VW/*den VW •Morgen zu reparieren beabsichtigt wurde nur der VW *Zu reparieren beabsichtigt wurde morgen nur der VW *Zu reparieren beabsichtigt wurde nur der VW morgen weil der Wagen zu reparieren begonnen wurde weil die Autos zu reparieren begonnen wurden weil ihm dieser Turm (*morgen) zu besichtigen empfohlen wurde weil ihm diese Türme (*morgen) zu besichtigen empfohlen wurden weil der Artikel ihm zu lesen empfohlen wurde (Säbel 1996: 205) weil ihm der Roman (*morgen) zu lesen erlaubt wurde weil ihm fünf Romane (*morgen) zu lesen erlaubt wurden Zu essen erlaubt wurden ihnen nur die Kekse •Gestern zu essen erlaubt wurden ihnen nur die Kekse Zu essen erlaubt wurden ihnen gestern nur die Kekse •Morgen zu essen erlaubt wurden ihnen nur die Kekse *Zu essen erlaubt wurden ihnen morgen nur die Kekse weil ihm der Roman zu lesen gelungen ist weil ihm fünf Romane zu lesen gelungen sind %weil ihm der Brief zu lesen gestattet wurde %weil ihm die Briefe zu lesen gestattet wurden ?Zu essen gestattet wurden ihnen nur die Kekse ?Zu besuchen gestattet wurden ihnen nur wenige Länder ?weil ihm der Brief zu entziffern mißlungen ist ?weil ihm die Briefe zu entziffern mißlungen sind Zu entziffern mißlungen sind/*ist ihm nur ägyptische Schriften Zu entziffern mißlungen ist ihm nur der Brief/*den Brief %Zu besuchen untersagt wurden ihnen alle Länder Europas %Zu essen untersagt wurden den Kindern alle Pilze, die sie finden % Viele Länder wurden ihnen zu besuchen untersagt weil ihm der Roman (*morgen) zu lesen verboten wurde weil ihm fünf Romane (*morgen) zu lesen verboten wurden weil der Wagen zu reparieren vergessen wurde weil die Autos zu reparieren vergessen wurden %Zu reparieren vermieden wurden alle alten Autos ?Zu essen vermieden werden konnten nur die Pilze weil der Direktor zu verständigen versäumt wurde weil die Direktoren zu verständigen versäumt wurden

330

Appendix

m. n.

Zu schließen versäumt wurden nur die Kellerfenster Zu schließen versäumt wurde nur der Tresor weil der Wagen (*morgen) zu reparieren versucht wurde weil die Autos (*morgen) zu reparieren versucht wurden weil der Wagen (*morgen) nicht zu reparieren gewagt wurde weil die Autos (*morgen) nicht zu reparieren gewagt wurden

Reduced non-restructuring: long object movement is impossible (264)

A. *weil (ihnen) der Wagen zu reparieren angeboten wurde *weil (ihnen) die Autos zu reparieren angeboten wurden *Zu bezahlen angeboten wurden (ihm) seine Rechnungen *Zu reparieren angeboten wurden (ihm) nur die orangen Autos *Zu besuchen angeboten wurden ihm alle seine Verwandten B. *weil der Wagen zu reparieren angefangen wurde *weil die Autos zu reparieren angefangen wurden ??Zu reparieren angefangen wurden nur orange Autos *Zu essen angefangen wurden nur die Vorspeisen C. *weil der Wagen zu reparieren aufgehört wurde *weil die Autos zu reparieren aufgehört wurden *Zu reparieren aufgehört wurden nur orange Autos D. *weil ihm der Wagen zu reparieren befohlen wurde •weil die Autos zu reparieren befohlen wurden ??Zu reparieren befohlen wurden ihnen nur orange Autos ??Zu essen befohlen wurden ihm die vergifteten Pilze E. *weil der Wagen nicht reparieren zu können befürchtet wurde *weil die Autos nicht reparieren zu können befürchtet wurden *Nicht reparieren zu können befürchtet wurden nur orange Autos *Zu reparieren befürchtet wurden nur orange Autos F. *weil der Wagen zu reparieren beschlossen wurde *weil die Autos zu reparieren beschlossen wurden *Zu reparieren beschlossen wurden nur orange Autos G. *weil (ihm) der Wagen zu reparieren gedroht wurde *weil (ihm) die Autos zu reparieren gedroht wurden *Zu reparieren gedroht wurden (ihm) nur orange Autos *Zu vernachlässigen gedroht wurden (ihm) seine Verwandten [Es wurde (ihm) gedroht dass man seine Verwandten vernachlässigen würde] *Zu schicken gedroht wurden (ihm) schlechte Haushaltshilfen •Besuchen zu müssen gedroht wurden ihm alle Verwandten H. *weil der Wagen zu reparieren entschieden wurde *weil die Autos zu reparieren entschieden wurden *Zu reparieren entschieden wurden nur orange Autos I. *weil (ihm) der Wagen zu reparieren ermöglicht wurde

CNon-)Restructuring properties

J.

K.

L.

M.

N.

O.

P.

Q.

331

*weil (ihm) die Autos zu reparieren ermöglicht wurden ??Zu reparieren ermöglicht wurden (ihm) nur orange Autos ??Zu besuchen ermöglicht wurden ihnen nun alle Länder Europas *weil der Wagen zu reparieren erwogen wurde *weil die Autos zu reparieren erwogen wurden *Zu reparieren erwogen wurden nur orange Autos *weil der Wagen zu reparieren fortgefahren wurde *weil die Autos zu reparieren fortgefahren wurden •Zu reparieren fortgefahren wurden nur orange Autos •weil der Wagen zu reparieren geglaubt wurde •weil die Autos zu reparieren geglaubt wurden *Zu reparieren geglaubt wurden nur orange Autos *weil der Wagen zu reparieren gehofft wurde *weil die Autos zu reparieren gehofft wurden *Zu reparieren gehofft wurden nur orange Autos •Reparieren zu können gehofft wurden nur orange Autos *weil der Wagen zu reparieren geplant wurde *weil die Autos zu reparieren geplant wurden *Zu reparieren geplant wurden nur orange Autos *weil (ihm) der Wagen zu reparieren geraten wurde *weil (ihm) die Autos zu reparieren geraten wurden ??Zu reparieren geraten wurden (ihm) nur orange Autos ??weil der Wagen zu reparieren verlernt wurde ??weil die Autos zu reparieren verlernt wurden ?Zu reparieren verlernt wurden nur orange Autos *Zu besuchen versprochen wurden (ihm) seine Verwandten oft

[Es wurde (ihm) versprochen dass man seine Verwandten oft besuchen würde] *Zu schicken versprochen wurden (ihm) Haushaltshilfen *Zu lösen versprochen wurden (ihm) alle Hausaufgaben •Besuchen zu dürfen versprochen wurden ihm nur Verwandten •Reparieren zu dürfen versprochen wurden ihnen nur orange Autos R. *weil (ihm) der Wagen zu reparieren vorgeschlagen wurde •weil (ihm) die Autos zu reparieren vorgeschlagen wurden •Zu besuchen vorgeschlagen wurden ihm alle seine Verwandten •Reparieren zu dürfen vorgeschlagen wurden ihm die Autos •Zu essen vorgeschlagen wurden ihm die Kekse S. *weil der Wagen zu reparieren vorgezogen wurde •weil die Autos zu reparieren vorgezogen wurden •Zu reparieren vorgezogen wurden nur orange Autos T. •weil der Wagen zu reparieren gewünscht wurde •weil die Autos zu reparieren gewünscht wurden •Zu reparieren gewünscht wurden nur orange Autos

332

Appendix Non-restructuring: long object movement is impossible

(265)

4.2.

i. ii. ii. iv. v.

*weil der Wagen zu reparieren abgelehnt wurde *weil der Wagen zu reparieren aufgegeben wurde *weil der Wagen zu reparieren vorgegeben wurde sich weigern 'refuse': no passive *weil der Wagen zu reparieren angegeben wurde *weil der Wagen repariert zu haben angegeben wurde vi. *weil der Wagen zu reparieren angekündigt wurde *weil der Wagen repariert zu haben angekündigt wurde vii. *weil der Wagen zu reparieren angenommen wurde *weil der Wagen repariert zu haben angenommen wurde viii. *weil der Wagen zu reparieren bedauert wurde *weil der Wagen repariert zu haben bedauert wurde xi. *weil der Wagen zu reparieren behauptet wurde *weil der Wagen repariert zu haben behauptet wurde χ. *weil der Wagen zu reparieren bezweifelt wurde •weil der Wagen repariert zu haben bezweifelt wurde xi. *weil der Wagen zu reparieren erklärt wurde *weil der Wagen repariert zu haben erklärt wurde xii. *weil der Wagen zu reparieren erwähnt wurde •weil der Wagen repariert zu haben erwähnt wurde xiii. *weil der Wagen zu reparieren erzählt wurde *weil der Wagen repariert zu haben erzählt wurde xiv. *weil der Wagen zu reparieren gestanden wurde *weil der Wagen repariert zu haben gestanden wurde XV. *weil der Wagen zu reparieren vorgegeben geglaubt wurde *weil der Wagen repariert zu haben vorgegeben wurde xvi. *weil der Wagen zu reparieren geleugnet wurde *weil der Wagen repariert zu haben geleugnet wurde xvii. *weil der Wagen zu reparieren vergessen wurde *weil der Wagen repariert zu haben vergessen wurde xviii. *weil der Wagen zu reparieren vermutet wurde *weil der Wagen repariert zu haben vermutet wurde xix. *weil der Wagen zu reparieren zugegeben wurde *weil der Wagen repariert zu haben zugegeben wurde

Pronoun fronting

General comment regarding the data: Pronoun fronting is highly sensitive to "diversity" and "complexity" effects. Acceptability of pronoun fronting is reduced (to different degrees) i) when the matrix predicate is ditransitive, ii) with the increase of animate arguments (both pronominal and non-pronominal) in the sentence; iii) when pronouns marked for a particular case co-occur with other arguments (in the

(Non-)Restructuring properties

333

sentence) with the same case (again both pronominal and non-pronominal, but stronger effects are found with other pronominal arguments), and iv) with certain morphological complexities which we have not been able to fully systematize yet (e.g., the co-occurrence of pronouns with the same gender but different animacy values is sometimes more marked than the co-occurrence of pronouns with different gender). The strength and degree of diversity and complexity effects differs from speaker to speaker. -»

Pronoun fronting is possible from restructuring infinitives

(266)

a. b. c. d.

e. f. g. h. i. j. k. 1. m. n.

weil ihr der Hans (?morgen) zu helfen beabsichtigte als ihr der Hans (*morgen) zu helfen begann weil es ihr der Hans (?morgen) zu essen empfahl weil es/ihn ihr der Hans (?morgen) nicht zu betreten erlaubte es: das Zimmer, ihn: den Saal weil ihn uns jemand nicht auszukosten erlaubte (Haider 1993: 307) weil es dem Hans (*morgen) nicht zu öffnen gelungen ist weil es/?ihn ihr der Hans (?morgen) nicht zu betreten gestattete weil es dem Hans (*morgen) zu entziffern mißlungen ist weil es/?ihn ihr der Hans (?morgen) zu essen untersagte weil es/?ihn ihr der Hans (?morgen) zu essen verboten hatte weil ihr der Hans (*morgen) zu helfen vergaß weil ihr der Hans (*morgen) zu helfen vermied weil ihr der Hans (*morgen) zu helfen versäumte weil ihr der Hans (*morgen) zu helfen versuchte weil ihr der Hans (??morgen) nicht zu helfen wagte

Pronoun fronting is possible from irrealis and aspectual infinitives (modulo complexity and diversity effects, which might be stronger here for some speakers since additional complexity is given by the richer structure of these types of infinitival constructions) In non-obligatory control contexts, pronoun fronting across a dative argument is degraded for most speakers10^ Note that there appears to be some speaker variation for ditransitive constructions that allow variable control. Grosse (2000) reports that anbieten 'offer' allows pronoun fronting only with subject control, whereas vorschlagen 'propose' allows pronoun fronting only with object control (but in general, these verbs are ambiguous between subject and object control). As the judgements in the text show, we have not been able to find native speakers who confirm these intuitions (except for the general impossibility/marginality of pronoun fronting in ditransitive nonobligatory control contexts). We have noticed, however, that judgements vary significantly in different discourse contexts. Since at this stage, the facts are rather shaky, no clear generalizations can be stated, and the question of whether the

334

Appendix

(267)

Α.

cf.

B. C.

D. E.

F. G.

H.

I. J.

weil ihn der Hans (morgen) zu besuchen angeboten hatte weil ihr der Hans (morgen) zu helfen angeboten hatte intended interpretation: 'he offered to help her' ?weil es der Hans zu vergessen angeboten hatte ??weil es ihr der Hans zu vergessen angeboten hatte %weil es Hans ihnen zu reparieren anbot (OK for Grosse 2000: 71) intended interpretation: 'he offered them that he would repair it' the sentence is OK if: 'he offered to repair it for them' ??weil es ihnen Hans zu reparieren anbot ??weil er ihn ihr (?morgen) zu besuchen angeboten hatte ??weil er ihn ihr¡ filr sie¡ zu restaurieren anbot = weil ere ihr¡ anbot PROe ihn (den Turm) für sie¡ zu restaurieren ??weil er ihn ihr¡ für sichj zu restaurieren anbot = weil er ihr¡ anbot PRO¡ ihn (den Turm) für sich¡ zu restaurieren weil ihr der Hans auf die Nerven zu gehen anfing weil ihn der Hans zu reparieren anfing als ihn die Maria zu unterstützen aufhörte begann er zu trinken als ihr der Hans endlich Vorschriften zu machen aufhörte war es schon zu spät weil er ihr schon vor Jahren gute Ratschläge zu geben aufgehört hatte weil ihn die Maria zu bemitleiden aufhörte als er sich wieder einmal daneben benahm weil es ihr der Hans (?morgen) zu lesen befohlen hatte es=das Buch ?weil ihn ihr der Hans (?morgen) zu besuchen befohlen hatte weil ihr Hans zu mißfallen befürchtete ?weil ihr Hans nicht widerstehen zu können befürchtete *weil ihn der Hans nicht zu mögen befürchtete *weil ihn der Hans beleidigt zu haben befürchtete weil ihr der Hans (?morgen) zu helfen beschloß weil sie der Hans (?morgen) zu besuchen beschloß weil ihn der Hans (morgen) zu verprügeln gedroht hat weil es der Hans (morgen) der Polizei zu melden gedroht hat ??weil es ihnen der Hans der Polizei zu melden gedroht hat •weil sie/ihn ihnen der Hans zu besucht gedroht hat weil ihr der Hans eine Freude zu machen entschied ?weil ihnen der Hans doch zu helfen entschied ?weil ihn der Hans wieder einzustellen entschied weil es ihr der Hans sich zu kaufen ermöglichte weil ihn ihr der Hans zu besuchen ermöglichte weil ihn der Hans zur Rede zu stellen erwog weil es der Hans zu melden erwog (Grosse 2000: 69)

variation is due to complexity/diversity effects or some syntactic/semantic property has to be left open.

(Non-)Restructuring properties

335

weil ihr der Hans einen Brief zu schreiben erwog weil ihr der Hans den Sachverhalt zu erklären fortfuhr weil ihn der Hans zu demütigen fortfuhr als ihr der Hans Vorschriften zu machen fortfuhr, verließ sie ihn L. weil ihr der Hans ständig Vorschriften machen zu müssen glaubte weil ihn der Hans belehren zu müssen glaubte ?weil ihn der Hans zu erkennen glaubte weil ihr der Hans damit helfen zu können glaubte *weil ihr der Hans damit geholfen zu haben glaubte M. weil ihr der Hans (morgen) helfen zu können hoffte weil sie der Hans (morgen) wiederzusehen hoffte weil es der Hans (morgen) wieder ansehen zu dürfen hoffte ?weil ihn der Hans (?morgen) zu treffen hoffte *weil ihn der Hans nicht beleidigt zu haben hoffte N. weil ihr der Hans (morgen) zu helfen plante K.

O.

P. Q.

R.

S.

T.

weil ihn der Hans (morgen) zu besuchen plante weil ihn ihr der Hans zu vergessen riet weil es ihr der Hans gut zu würzen riet dass ihn mir jemand zu konsultieren geraten hat (Haider 1993: 307) weil ihr der Hans Komplimente zu machen verlernt hatte ?weil ihm der Hans heimlich Kirschen zu stehlen verlernt hatte weil ihr der Hans (morgen) zu helfen versprach weil ihn der Hans (morgen) zu besuchen versprach weil er es ihr (morgen) zu erledigen versprach dass es ihr jemand zu lesen versprochen hat (Haider 1993: 307) ??weil er ihn ihr (morgen) zu besuchen versprach ??weil er es ihr zu vergessen versprach weil es der Hans zu vergessen vorschlug weil ihr der Hans zu helfen vorschlug ?weil es der Hans für die Maria zu besorgen vorschlug ??weil er es ihr zu vergessen vorschlug ??weil er ihn ihr zu besuchen vorschlug ??weil er ihn ihr¡ für sie¡ zu restaurieren vorschlug = weil ere ihr¡ vorschlug PROe ihn (den Turm) fllr sie¡ zu restaurieren ??weil er ihn ihr¡ für sich¡ zu restaurieren vorschlug = weil er ihr¡ vorschlug PRO¡ ihn (den Turm) für sich¡ zu restaurieren weil ihr der Hans zuzuhören vorzog weil er ihr vorzog zuzuhören weil ?sie/es der Hans zu vergessen vorzog weil ihr der Hans die schlechte Nachricht nach dem Abendessen mitzuteilen vorzog ?weil ihr Hans bei den Mathematikaufgaben zu helfen vorzog weil ihn der Hans (sofort/morgen) zu sehen wünschte only if wish = order!

336

Appendix Weak implicatives allow pronoun fronting only very marginally

(268)

i.

ii.

iii.

iv.

??weil ihr der Hans zu helfen ablehnte *weil ihr der Hans morgen zu helfen entschieden abgelehnt hat ??weil ihn der Hans zu besuchen ablehnt *weil ihn der Hans morgen zu besuchen ablehnte ??weil ihr der Hans gute Ratschläge zu geben schon vor langem aufgeben hatte ??weil ihn der Hans zu überzeugen zu versuchen letztendlich aufgab ??weil sie der Hans schon vor Jahren zu bemitleiden aufgegeben hatte ??weil ihn der Hans nicht zu mögen vorgab ??weil ihr der Hans nicht helfen zu können vorgab *weil ihn der Hans gefeuert zu haben vorgab *weil ihn der Hans {sich} zu besuchen {sich} geweigert hat *weil ihn der Hans sich geweigert hat zu besuchen •weil es der Hans sich geweigert hat zu essen *weil es sich der Hans geweigert hat zu essen

Propositional and factive complements block pronoun fronting (269)

v.

vi.

vii.

viii.

ix.

χ.

xi.

xii.

*weil ihn die Maria nicht zu kennen angab *weil ihr der Hans nicht helfen zu können angab *weil ihn der Hans nicht gesehen zu haben angab *weil ihn die Maria sehr zu mögen ankündigte *weil ihn der Hans (dem Komitee) feuern zu müssen ankündigte *weil sie der Hans geheiratet zu haben ankündigte *weil ihn die Maria nicht zu mögen annahm *weil ihr der Hans helfen zu können annahm *weil ihn der Hans beleidigt zu haben annahm *weil sie der Hans zu mögen bedauerte *weil ihr der Hans nicht helfen zu können bedauerte *weil ihn der Hans gefeuert zu haben bedauerte *weil sie der Hans zu mögen behauptete *weil ihr der Hans helfen zu wollen behauptete *weil ihn der Hans gefeuert zu haben behauptete *weil ihn der Hans zu mögen bezweifelte *weil ihr der Hans helfen zu können bezweifelte *weil ihn der Hans beleidigt zu haben bezweifelte *weil ihn der Hans zu mögen erklärte *weil ihr der Hans helfen zu wollen erklärte *weil ihn der Hans eingestellt zu haben erklärte *weil ihn der Hans zu mögen erwähnte *weil ihr der Hans helfen zu wollen erwähnte

(Non-)Restructuring properties

337

*weil ihn der Hans eingestellt zu haben erwähnte *weil ihn der Hans zu mögen erzählte *weil ihr der Hans helfen zu wollen erzählte *weil ihn der Hans eingestellt zu haben erzählte xiv. *weil ihn der Hans (dem Peter) zu mögen gestand *weil ihn der Hans (dem Peter) nicht einstellen zu können gestand *weil ihn der Hans (dem Peter) eingestellt zu haben gestand XV. *weil ihn der Hans eingestellt zu haben glaubte *weil ihn der Hans schon einmal gesehen zu haben glaubte xvi. *weil ihn der Hans zu mögen leugnete *weil ihr der Hans helfen zu wollen leugnete •weil ihn der Hans eingestellt zu haben leugnete xvii. *weil ihn der Hans nicht zu mögen vergaß •weil ihr der Hans nicht helfen zu können vergaß *weil ihn der Hans beleidigt zu haben vergaß xviii. *weil ihn der Hans nicht zu mögen vermutete •weil ihr der Hans nicht helfen zu können vermutete *weil ihn der Hans beleidigt zu haben vermutete ??weil ihr der Hans morgen wieder über den Weg zu laufen vermutete xix. *weil ihn der Hans nicht zu mögen zugab *weil ihr der Hans nicht helfen zu wollen zugab •weil ihn der Hans eingestellt zu haben zugab xiii.

Scrambling Non-focus scrambling is possible from restructuring infinitives The embedded adverbials marked with % are impossible under the nonfocus scrambling interpretation, but become possible when the scrambled phrase is (heavily) focused (i.e., under focus scrambling) a. weil der Hans den Ball beabsichtigte(%morgen) der Maria zu geben b. weil der Hans den Zaun begonnen hat (*morgen) zu streichen c. weil ihr der Hans den Zaun empfahl (%morgen) zu streichen d. weil ihr der Hans den Zaun nicht erlaubte (%morgen) zu streichen e. non-focus scrambling cannot be demonstrated f. weil ihr der Hans das Büro nicht gestattete (%morgen) zu betreten g. non-focus scrambling cannot be demonstrated h. %weil ihr der Hans das Büro untersagte (%morgen) zu betreten i. weil ihr der Hans das Büro verbot (%morgen) zu betreten j. weil der Hans den Zaun vergessen hat (*morgen) zu streichen k. weil der Hans den Zaun vermieden hat (*morgen) zu streichen 1. weil der Hans den Zaun versäumt hat (*morgen) zu streichen m. weil der Hans den Zaun versucht hat (*morgen) zu streichen n. weil der Hans den Zaun gewagt hat (*morgen) zu streichen

338

Appendix Non-focus scrambling is impossible from restructuring infinitives (without focus, most speakers find the examples (271) marginal to ungrammatical); the examples improve significantly when the scrambled phrase is focused (or nur 'only' is added) %: to be understood as "*" or "??" in neutral conntexts, and as "OK/?" in focus contexts

(271)

A. %weil der Hans der Maria den Zaun anbot (morgen) zu streichen weil nur im Spiegel der Hans der Maria anbot sich zu betrachten B. ?weil der Hans der Maria anfing auf die Nerven zu gehen %weil der Hans den Zaun anfing zu reparieren weil mit grüner Farbe der Hans den Zaun zu streichen anfing C. %als der Hans seinen Sohn aufgehört hatte zu unterstützen D. %weil es ihr den Zaun befohlen hat (morgen) zu streichen ?weil nur im Spiegel der Hans der Maria befahl sich zu betrachten E. %weil der Hans der Maria befürchtete zu mißfallen %weil Hans der Maria befürchtete nicht widerstehen zu können ?weil in diesem Anzug der Hans der Maria zu mißfallen befürchtete *weil in diesem Anzug der Hans der Maria mißfallen zu haben befürchtete *weil der Hans den neuen Boss befürchtete nicht zu mögen *weil der Hans den neuen Boss befürchtete beleidigt zu haben F. %weil der Hans den Zaun beschloß (?morgen) zu streichen weil mit grüner Farbe der Hans beschloß nur den Zaun zu streichen G. %weil der Hans (??ihr) den Peter gedroht hat (morgen) einzuladen ?weil mit grüner Farbe der Hans drohte den Zaun zu streichen H. %weil der Hans den Zaun entschied (morgen) zu streichen weil mit grüner Farbe der Hans entschied nur den Zaun zu streichen I. %weil ihr der Hans den Zaun ermöglichte zu streichen ?weil nur mit grüner Farbe er ihr ermöglichte den Zaun zu streichen J. %weil der Hans den Peter erwog (?morgen) zur Rede zu stellen weil mit grüner Farbe der Hans erwog den Zaun zu streichen K. %weil der Hans den Zaun fortfuhr zu streichen weil mit grüner Farbe der Hans den Zaun zu streichen fortfuhr L. %weil der Hans der Maria ständig glaubte Vorschriften machen zu müssen %weil der Hans den Peter glaubte belehren zu müssen *weil der Hans den Zaun glaubte schon gestrichen zu haben M. %weil er der Maria hoffte helfen zu können weil mit grüner Farbe nur der Hans hoffte den Zaun streichen zu können %weil er die Frau hoffte (morgen) wiederzusehen *weil er den neuen Boss hoffte nicht beleidigt zu haben N. %weil der Hans den Zaun plante (morgen) zu streichen

(Non-)Restructuring properties

339

weil mit grüner Farbe nur der Hans plante den Zaun zu streichen O. %weil er ihr den Zaun riet (morgen) zu streichen weil mit grüner Farbe ihr nur der Hans riet den Zaun zu streichen P. %weil der Hans der Maria verlernt hatte Komplimente zu machen Q. %weil der Hans (??ihr) den Zaun versprach (morgen) zu streichen weil mit grüner Farbe ihr nur der Hans versprochen hatte den Zaun zu streichen R. %weil der Hans ihr den Zaun vorschlug zu streichen weil mit grüner Farbe ihr nur der Hans vorschlug den Z. zu streichen S. %weil der Hans der Maria vorzog zuzuhören weil mit grüner Farbe nur der Hans vorzog den Zaun zu streichen T. %weil der Hans den Peter wünschte (sofort/morgen) zu sehen weil mit grüner Farbe der Hans wünschte den Zaun zu streichen ?weil in seinem Büro der Hans wünschte den Peter sofort zu sehen Weak implicative, facti ve, and propositional complements do not allow scrambling (neither focus nor non-focus) (272)

i. ii. iii.

iv. v.

vi.

vii.

viii.

ix.

105

*weil nur den Zaun der Hans ablehnte zu streichen 105 •weil nur mit grüner Farbe der Hans ablehnte den Zaun zu streichen *weil nur den Zaun der Hans aufgab zu streichen *weil nur mit grüner Farbe der Hans aufgab den Zaun zu streichen *weil nur mit grüner Farbe der Hans vorgab den Zaun zu streichen •weil nur mit grüner Farbe der Hans vorgab den Zaun streichen zu wollen/gestrichen zu haben •weil nur mit grüner Farbe sich der Hans weigerte den Zaun zu streichen •weil nur diesen Mann der Hans angab zu kennen (see fn. 105) •weil {nur }mit grüner Farbe der Hans angab {nur }den Zaun gestrichen zu haben *weil nur diesen Mann der Hans ankündigte nicht zu kennen •weil {nur} mit grüner Farbe der Hans ankündigte {nur} den Zaun gestrichen zu haben •weil nur diesen Mann der Hans annahm nicht zu kennen •weil {nur} mit grüner Farbe der Hans annahm {nur} den Zaun gestrichen zu haben •weil nur diesen Mann der Hans bedauerte nicht zu kennen •weil {nur} mit grüner Farbe der Hans bedauerte {nur} den Zaun gestrichen zu haben •weil nur diesen Mann der Hans behauptete nicht zu kennen

Some of the examples could appear grammatical when the scrambled object is analyzed as a matrix argument and the infinitive as an adjunct (such as in 'He refused the fence for painting').

Appendix *weil {nur} mit grüner Farbe der Hans behauptete {nur} den Zaun gestrichen zu haben χ. *weil nur diesen Mann der Hans bezweifelte zu kennen •weil {nur} mit grüner Farbe der Hans bezweifelte {nur} den Zaun gestrichen zu haben xi. *weil nur diesen Mann der Hans erklärte nicht zu kennen •weil {nur} mit grüner Farbe der Hans erklärte {nur} den Zaun gestrichen zu haben xii. *weil nur diesen Mann der Hans erwähnte gut zu kennen *weil {nur} mit grüner Farbe der Hans erwähnte {nur} den Zaun gestrichen zu haben xiii. *weil nur diesen Mann der Hans erzählte gut zu kennen *weil {nur} mit grüner Farbe der Hans erzählte {nur} den Zaun gestrichen zu haben xiv. *weil nur diesen Mann der Hans gestand gut zu kennen •weil {nur} mit grüner Farbe der Hans gestand {nur} den Zaun gestrichen zu haben XV. *weil {nur} mit grüner Farbe der Hans glaubte {nur} den Zaun gestrichen zu haben xvi. *weil nur diesen Mann der Hans leugnete zu kennen •weil {nur} mit grüner Farbe der Hans leugnete {nur} den Zaun gestrichen zu haben xvii. •weil nur diesen Mann der Hans vergaß gut zu kennen *weil {nur} mit grüner Farbe der Hans vergaß {nur} den Zaun gestrichen zu haben xviii. *weil nur diesen Mann der Hans vermutete gut zu kennen *weil {nur} mit grüner Farbe der Hans vermutete {nur} den Zaun gestrichen zu haben xix. *weil nur diesen Mann der Hans zugab gut zu kennen •weil nur mit grüner Farbe der Hans zugab den Zaun gestrichen zu haben

Verb classification in different languages

341

5. Verb classifications in different languages This section provides a (preliminary) summary of the restructuring/nonrestructuring classification of various predicates in German, Dutch, Italian, Spanish, and Japanese. The data are based on the classifications found in the literature (see sections 5.1 through 5.11 for a selective summary of the classifications of verbs as (non-)restructuring verbs by different authors), but have been supplemented with data collected from consultations with native speakers of these languages. We would like to thank (for German) Christine Czinglar, Martin Hackl, SBär Martin Haiden, Conny Krause, Winnie Lechner, Renate Musan, Friedrich Neubarth, Wilfried öller, Claudia Plaimauer, Joachim Säbel, Ulrich Sauerland, Hubert Truckenbrodt, Martina Wiltschko, and the 100 volunteers from various German organizations; (for Spanish) Calixto Aguero-Bautista, Karlos ArreguiUrbina, Elena Benedicto, Olga Fernandez-Soriano, and Isabel Oltra-Massuet; (for Italina) Sveva Besana, Caterina Donati, Michela Ippolito, and Paola Sica; (for Dutch) Sjef Barbiers, Edith Kaan, Jeannette Schaeffer, Fleur Veraart, and many Dutch speakers from the Euro Members (MIT) list; and for Japanese, Kazuko Yatsushiro and Mikinari Matsuoka. Table 41 summarizes the (non-)restructuring status of the predicates tested. Predicates that either do not combine with infinitival constructions (such as re-do or fail in many languages), or predicates that involve an additional definable restriction that influences their restructuring properties (such as the fact that particle verbs in Dutch and German block restructuring) are marked as N/A. 1 0 6 Predicates marked with "±" either show speaker variation or have more than one corresponding verb in that language and not all of these predicates behave alike concerning the restructuring/non-restructuring status. Note that the table represents selected properties. The criteria used are: long object movement—i.e., object raising from passivized infinitives in functional constructions, long passive/unaccusative movement in lexical constructions (German); verb raising plus IPP (Dutch); clitic climbing (Italian, Spanish); and lack of embedded tense marking (Japanese). Technically, some of these tests do not distinguish between "restructuring" and "non-restructuring" (in our terminology), but only between full and reduced non-restructuring. We abstract away from this issue here (see Chapter 5). A comparison of the predicates in Table 41 with other restructuring properties (e.g., remnant extraposition in Dutch, object preposing in Italian and Spanish etc.) is intended for future research.

In Dutch, easy-adjectives are excluded on independent grounds since adjectives do not undergo verb raising. We thank a reviewer for pointing us to this fact.

342

Appendix

Table 41. Restructuring predicates in five languages Predicates

German

Dutch

Italian

Spanish

Japanese

can, may must, need want other (semi) modals causatives begin come, go (be) used to easy continue start finish, stop be about to re-do know how return intend (-want, mean) forget try manage/succeed fail dare seem promise, threaten allow, permit forbid recommend refuse, reject appear order decide, chose plan

+ +

+

+ +

+ +

+ +

+

+ + +

+ + +

+ + +

+

+ +

+ + + + N/A N/A N/A N/A N/A N/A N/A +

+ + + N/A N/A N/A N/A N/A N/A

+

+ + +

+ + + +

+

+

+

+

+ +

+ + +

+ +

+

Ν/Α N/A

+

wish offer propose prepositional factive

+

+

+ + + N/A + + +

+

N/A +

N/A +

± ±

+ ±

+ +

+

±

±

+ +

η-

+ +

Ν/Α +

+

+

± ±

+

-

-

+

-

-

-

-

-

-

-

-

-

-

+

-

-

-

N/A

+

-

-

-

-

±

-

-

-

-

-

-

-

-

-

-

-

'

+ +

±

±

±

N/A

N/A

Η-

±

+ ±

-

+ + N/A

-

-

-

-

±

-

-

Verb classification in different languages

5.1. I.

II. ΙΠ.

5.2. I.

II. ΙΠ.

5.3. I.

II. ΙΠ.

5.4. I.

II.

Rizzi

343

(1976,1982)

Restructuring Verbs a. modals: b. aspectuals: c. motion verbs: d. others: Variation: Non-Restructuring Verbs:

Italian want, must, can start, finish, be about to, continue come, go, return know how, easy-adjectives try, seem all others

Napoli (1981) Restructuring Verbs a. modals: b. aspectuals: c. motion verbs: d. others: Variation:

Italian want, must, can, have to start, be about to, continue come, go, return, send know how, succeed dare, intend, learn, used to, finish, seem, forget, try Non-Restructuring Verbs: all others

Zagona (1982), Aissen and Perlmutter Restructuring Verbs a. modals: b. aspectuals: c. motion verbs: d. others: Variation: Non-Restructuring Verbs:

(1976,1983)

Spanish want, ought/should, can, have just start, finish, stop, keep on, continue, re-do come, go, return tend, try, easy-adjectives permit, order seem, must (epist.), insist, dream, decide, avoid, suggest, ask, say, affirm

Picallo (1985) Restructuring Verbs a. modals: b. aspectuals: c. motion verbs: Non-Restructuring Verbs:

5.5.

Burzio (1986)

I.

Restructuring Verbs a. modals:

Catalan want, must, can/may, have to start, finish, used to come, go, return seem

Italian want, must, can

344

II.

5.6. I.

II.

5.7. I.

II.

Appendix b. aspectuals: c. motion verbs: d. others: e. raising: Non-Restructuring Verbs:

Fanselow (1989) Restructuring Verbs a. modals: b. others: c. raising: Non-Restructuring Verbs:

German want, must, can, should... try, dare, promise, let, allow seem ask, regret, force, tell, claim

Rutten (1991) Restructuring Verbs (verb raising and IPP verbs) Dutch want, must, can, have to, need, be, do, will,... modals: a. begin, stay, used to aspectuals: b. come, go, walk, sit, stand, lie (down) motion verbs: c. dare, help, learn, intend, try, refuse, know how others: d. seem, appear, threaten, promise raising: e. Non-Restructuring Verbs: all others

5.8.

Haider (1993)

I.

Restructuring Verbs a. modals: b. aspectuals: d. others: e. raising:

5.9.

Sabel (1994/1996)

I.

Restructuring Verbs a. modals: b. aspectuals: c. others:

107

start, be about to, continue, finish 107 come, go know how, easy-adjectives seem return, all others

German want, must, can, should, be allowed start, continue know how, easy-adjectives, try seem, threaten, promise

German want, must, can,... start try, forget, dare, hope, help, permit, promise,

Burzio excludes the verb finire 'finish' from the list of RVs since it does not show all restructuring properties (Burzio 1986:385, fn. 4 and 5). While clitic climbing and object preposing are possible, auxiliary switch is not found with this verb. However, since this is due to some independent factor (namely, the obligatory presence of the infinitival marker di), I include the verb finish here.

Verb classification in different languages II.

345

choose, intend, recommend Non-Restructuring Verbs: claim, refuse, fear, hesitate, accuse, assure, encourage, teach

5.10. Cinque (1997) I.

II.

Restructuring Verbs a. modals: b. aspectuals: c. motion verbs: d. others: Non-restructuring Verbs:

Italian want, must, can start, finish, be about to, continue, come, go, return, send know how, try, dare, tend, manage, resume all others

5.11. Grosse (2000) Note: Grosse considers pronoun fronting and the intraposed position of the infinitival complement as the major restructuring criteria and ignores passive and full DPscrambling. As discussed in Chapter 5, pronoun fronting and intraposition cannot be taken to distinguish between restructuring and non-restructuring, but rather distinguish full non-restructuring from restructuring and reduced non-restructuring. Grosse's class of "restructuring verbs" hence refers to a different class of verbs (in particular, it captures what we have classified as restructuring constructions and reduced non-restructuring constructions). German want, must, can... begin intend, decide, threaten, manage, hope, fail, plan, refrain, forget, avoid, neglect/miss, promise, try, dare, wish (=order), recommend, allow, enable, advise, disallow, forbid, propose try, forget, begin, intend, miss/neglect, allow, Passive possible: forbid, recommend Non-restructuring Verbs: ACC-verbs, reject, beg, announce, give up, stop, regret, fear, claim, doubt, ask, threaten, choose, declare, think about, continue, confess, think, deny, unlearn, assume, tell, propose, prefer, wish, hesitate, admit, offer Restructuring Verbs a. modals: b. aspectuals: others: d.

I\ II.

References

Aissen, Judith and David Perlmutter 1976 Clause reduction in Spanish. In: Henry Thompson, Kenneth Whistler, Vicki Edge, Jeri Jaeger, Ronya Javkin, Miriam Petruck, Christopher Smeall and Robert D. Van Valin Jr. (eds.), Proceedings of the Second Annual Meeting of the Berkeley Linguistic Society (BLS 2), 1-30. Berkeley, CA: Berkeley Linguistic Society. 1983 Clause reduction in Spanish. In: David Perlmutter (ed.) Studies in Relational Grammar 1, 360-403. Chicago: The University of Chicago Press. Akmajian, Adrian, Susan Steele and Tomas Wasow 1979 The category AUX in Universal Grammar. Linguistic Inquiry 10: 1-64. Alsina, Alex 1992 On the argument structure of causatives. Linguistic Inquiry 23: 517-555. Baker, Mark 1988 Incorporation: A theory of grammatical function Changing. Chicago: The University of Chicago Press. Barbiers, Sjef 1995 The syntax of interpretation. Ph.D. Dissertation, University of Leiden, The Netherlands. [Distributed by: Holland Academic Graphics (HIL), The Hague.] Bayer, Josef and Jacqueline Kornfilt 1990 Restructuring effects in German. In: Elisabet Engdahl, Mike Reape, Martin Mellor and Richard Cooper (eds.), Parametric variation in Germanic and Romance: Center of Cognitive Science Working Papers, 21-42. Edinburgh: ICCS, University of Edinburgh. Bech, Gunnar 1955 Studien zum deutschen Verbum infinitum. Tübingen: Max Niemeyer Verlag. Belletti, Adriana 1990 Generalized verb movement: Aspects of verb syntax. Turin: Rosenberg & Sellier. Bennis, Hans and Teun Hoekstra 1989 Why Kaatje was not heard sing a song. In: Dany Jaspers, Wim Klooster, Yvan Putseys and Pieter Seuren (eds.), Sentential complementation and the lexicon: Studies in honour of Wim de Geest, 21-40. Dordrecht: Foris.

References

347

Besten, Hans den and Jean Rutten 1989 On verb raising, extraposition and free word order in Dutch. In: Dany Jaspers, Wim Klooster, Yvan Putseys and Pieter Seuren (eds.), Sentential complementation and the lexicon: Studies in honour of Wim de Geest, 41-56. Dordrecht: Foris. Bhatt, Rajesh 1998 Obligation and possession. In: Heidi Harley (ed.), Papers from the UPenn/MIT roundtable on argument structure and aspect: MIT Working Papers in Linguistics 32: 21-40. Bobaljik, Jonathan David 1995 Morphosyntax: The syntax of verbal inflection. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] 1998 Floating quantifiers: Handle with care. Glot International 3:3-10. Bobaljik, Jonathan David and Höskuldur Thráinsson 1998 Two heads aren't always better than one. Syntax 1: 37-71. Bobaljik, Jonathan and Susi Wurmbrand 1999 Modals, raising & A-reconstruction. Talk given at Leiden University, The Netherlands, and University of Salzburg, Austria [http://web.mit.edu/susi/www/Leiden.pdf]. 2001 Reconstruction and A-traces: A reply to Lasnik (1999). Ms. Montréal, Québec: McGill University. Boivin, Marie Claude 1998 Complementation and interpretation: The concrete and imaginative readings of visual perception verbs. In: Uli Sauerland and Orin Perçus (eds.), The interpretive tract: MIT Working Papers in Linguistics 25: 103-123. Bok-Bennema, Reineke and Brigitte Kampers-Manhe 1994 Transparency effects in the Romance languages. In: Michael Mazzola (ed.) Issues and theory in Romance linguistics: Selected papers from the Linguistic Symposium on Romance Languages XXIII, 199-217. Washington, DC: Georgetown University Press. Borer, Hagit 1989 Anaphoric AGR. In: Osvaldo Jaeggli and Kenneth Safir (eds.), The null subject parameter, 69-109. Dordrecht: Kluwer Academic Publishers. Boskovic, Zeljko 1995 Principles of economy in nonfinite complementation. Ph.D. Dissertation, University of Connecticut, Storrs, CT. [Published as Boskovic 1997.] 1996 Selection and the categorial status of Infinitival Complements. Natural Language and Linguistic Theory 14: 269-304. 1997 The syntax of nonfinite complementation: An economy approach. Cambridge, Mass: ΜΓΓ Press.

348

References

Bouchard, Denis 1984 On the content of empty categories. Dordrecht: Foris. Brennan, Virginia 1993 Root and epistemic modal auxiliary verbs in English. Ph.D. Dissertation, University of Massachusetts, Amherst. [Distributed by: Graduate Linguistic Student Association, Amherst, Mass.] Bresnan, Joan 1971 Sentence stress and transformations. Language 47: 257-297. 1982 Control and complementation. In: Joan Bresnan (ed.) The mental representation of grammatical relations, 282-390. Cambridge, Mass: MIT Press. Brody, Michael 1995 Lexico-logical form: A radically minimalist theory. Cambridge, Mass: ΜΓΓ Press. Broekhuis, Hans 1992 Chain-government: Issues in Dutch syntax. Ph.D. Dissertation, University of Amsterdam, The Netherlands. [Distributed by: Holland Academic Graphics (HIL), The Hague.] Broekhuis, Hans, Hans den Besten, Kees Hoekstra and Jean Rutten 1995 Infinitival complementation in Dutch: On remnant extraposition. The Linguistic Review 12: 93-122. Browning, Marguerite 1987 Null operator constructions. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] Burzio, Luigi 1986 Italian Syntax. Dordrecht: Reidel. Cattell, R. 1978 On the source of interrogative adverbs. Language 54: 61-77. Chierchia, Gennaro 1983 Outline of a semantic theory of (obligatory) control. In: Michael Barlow, David P. Flickinger and Michael T. Westcoat (eds.), Proceedings of the 2nd West Coast Conference on Formal Linguistics (WCCFL 2), 19-31. Stanford, CA: Stanford Linguistics Association. 1984 Anaphoric properties of infinitives and gerunds. In: Mark Cobler, Susannah MacKaye and Michael T. Wescoat (eds.), Proceedings of the 3rd West Coast Conference on Formal Linguistics ( WCCFL 3), 28-39. Stanford, CA: Stanford Linguistics Association. 1984b Topics in the syntax and semantics of infinitives and gerunds. Ph.D. Dissertation, University of Massachusetts, Amherst. [Distributed by: Graduate Linguistic Student Association, Amherst, Mass.]

References 1989

349

Anaphora and attitudes de se. In: Renate Bartsch, J. van Benthem and P. van Emde Boas (eds.), Semantics and contextual expression, 1-32. Dordrecht: Foris. Chomsky, Noam 1957 Syntactic structures. The Hague: Mouton. 1980 On binding. Linguistic Inquiry 11: 1-46. 1981 Lectures on Government and Binding. Dordrecht: Foris. 1982 Some concepts and consequences of the theory of Government and Binding. Cambridge, Mass: MIT Press. 1986a Barriers. Cambridge, Mass: MIT Press. 1986b Knowledge of language. New York: Praeger. 1995 The Minimalist Program. Cambridge, Mass: MIT Press. Chomsky, Noam and Howard Lasnik 1993 Principles and parameters theory. In: Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), Syntax: An international handbook of contemporary research, 506569. Berlin: Walter de Gruyter. 1995 Principles and parameters theory. In: Noam Chomsky (ed.) The Minimalist Program, 13-127. Cambridge, Mass: MIT Press. Cinque, Guglielmo 1990a Two classes of intransitive adjectives in Italian. In: Günther Grewendorf and Wolfgang Sternefeld (eds.), Scrambling and barriers, 261-349. Amsterdam/Philadelphia: John Benjamins. 1990b Types of A '-dependencies. Cambridge, Mass: MIT Press. 1997a The interaction of passive, causative, and 'restructuring' in Romance. Ms. Venice: University of Venice. 1997b Restructuring and the order of aspectual and root modal heads. Ms. Venice: University of Venice. 1998 On clitic climbing and other transparency effects. Talk given at MIT, Cambridge, Mass. 1999 Adverbs and functional heads: A cross-linguistic perspective. Oxford, UK: Oxford University Press. 2000 "Restructuring" and functional structure. Ms. Venice: University of Venice. Comrie, Bernard 1984 Subject and object control: Syntax, semantics and pragmatics. Berkeley Linguistic Society, 450-464. Cremers, C. 1983 On two types of infinitival complementation. In: Frank Heny and Barry Richards (eds.), Linguistic categories: Auxiliaries and related puzzles, 169-221.1.) Dordrecht: Reidel. Declerck, Renaat 1981 On the role of progressive aspect in nonfinite perception verb complements. Glossa 15: 83-114.

350 1982

References

The triple origin of participial perception verb complements. Linguistic Analysis 10: 1-26. 1983 On the passive of infinitival perception verb complements. Journal of English Linguistics 16: Dikken, Marcel den 1996 The minimal links of verb (projection) raising. In: Werner Abraham, Samuel David Epstein, Höskuldur Thráinsson and JanWouter C. Zwart (eds.), Minimal ideas: Syntactic studies in the minimalist framework, 67-96. (Linguistik Aktuell 12.) Amsterdam/Philadelphia: John Benjamins Publishing Company. DiSciullo, Anna Maria and Edwin Williams 1987 On the definition of word. Cambridge, Mass: MIT Press. Erb, Marie Christine 2001 Finite auxiliaries in German. Ph.D. Dissertation, University of Tilburg, The Netherlands. Evers, Arnold 1975a The Guillotine principle. In: A. Kraak (ed.) Linguistics in the Netherlands 1972-1973, 147. Amsterdam: Von Gorcu. 1975b The transformational cycle of Dutch and German. Ph.D. Dissertation, University of Utrecht, The Netherlands. 1986 Clause union in French and German. Groninger Arbeiten zur germanistischen Linguistik 28: 170-201. 1988 Non-finite verb forms and subject theta assignment. In: Martin Everaert, Arnold Evers, Riny Huybregts and Mieke Trommelen (eds.), Morphology and modularity: In honour of Henk Schultink, 105-128. Dordrecht: Foris. 1990 The infinitival prefix "zu" as INFL. In: Günther Grewendorf and Wolfgang Sternefeld (eds.), Scrambling and barriers, 217-238. Amsterdam/Philadelphia: John Benjamins. Fanselow, Gisbert 1989 Coherent infinitives in German. In: Christa Bhatt, Elisabeth Löbel and Claudia Schmidt (eds.), Syntactic phrase structure phenomena in noun phrases and sentences, 1-16. Amsterdam: John Benjamins. Farkas, Donka 1988 On obligatory control. Linguistics and Philosophy 11: 27-58. Felser, Claudia 1998 Perception and control: a minimalist analysis of English direct perception complements. Journal of Linguistics 34: 351-385. Fiengo, Robert 1974 Semantic Conditions on Surface Structure. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.]

References Fox, Daniel 1998

351

Economy and semantic interpretation: A study of scope and variable binding. Ph.D. Dissertation, MIT, Cambridge, Mass. [Published as Fox 2000.] 1999 Reconstruction, binding theory, and the interpretation of chains. Linguistic Inquiry 30: 157-196. 2000 Economy and semantic interpretation. (Linguistic Inquiry Monograph 35.) Cambridge, Mass: ΜΓΓ Press/MITWPL. Gonçalves, Anabela 1998 On restructuring constructions in European Portuguese. In: Tina Cambier-Langeveld, Aníkó Lipták and Michael Redford (eds.), Proceedings ofConSOLEó, 75-88. Leiden: SOLE. Goodall, Grant 1987 Parallel structures in syntax. Cambridge/New York: Cambridge University Press. 1991 Wanna-contraction as restructuring. In: Carol Georgopolous and Roberta Ishihara (eds.), Interdisciplinary approaches to language: Essays in honor of S.-Y. Kuroda, 239-254. Dordrecht: Kluwer Academic Publishers. Grewendorf, Günther 1984 Reflexivierungsregeln im Deutschen. Deutsche Sprache 1: 14-30. 1987 Kohärenz und Restrukturierung. In: B. Asbach-Schnitker and J. Roggenhofer (eds.), Neuere Forschungen zur Wortbildung und Histographie der Linguistik, 123-144. Tübingen: Narr. 1988 Aspekte der deutschen Syntax. Tübingen: Narr. Grewendorf, Günther and Joachim Säbel 1994 Long scrambling and incorporation. Linguistic Inquiry 25: 263308. Grimshaw, Jane 1991 Extended projection. Ms. Waltham, Mass: Brandeis University. Grinder, John T. 1970 Super equi-NP deletion. Papers from the 6th regional meeting: Chicago Linguistic Society (CLS 6), 297-317. Groat, Erich and John O'Neil 1996 Spell-out at the LF interface. In: Werner Abraham, Samuel David Epstein, Höskuldur Thráinsson and C. Jan-Wouter Zwart (eds.), Minimal ideas: Syntactic studies in the minimalist framework, 113139. (Linguistik Aktuell Amsterdam/Philadelphia: John Benjamins Publishing Company. Grosse, Julia 2000 Kontrollprobleme bei Infinitiven: Eine Auseinandersetzung mit "Infinitives" von Susanne Wurmbrand. M.A. Thesis, University of Tübingen, Germany.

352

References

Guasti, Maria Teresa 1992 Incorporation, excorporation and lexical properties of causative heads. The Linguistic Review 8: 209-232. 1993 Causatives and perception verbs: a comparative approach. Turin: Rosenberg and Sellier. 1996 Semantic restrictions on Romance causatives and the incorporation approach. Linguistic Inquiry 27: 294-313. 1997 Romance causative. In: Liliane Haegeman (ed.) The new comparative syntax, 124-144. London: Longmans Linguistics Library. Guéron, Jacqueline and Teun Hoekstra 1988 T-Chains and the constituent structure of auxiliaries. In: Anna Cardinaletti, Giuliana Giusti and Guglielmo Cinque (eds.), Constituent structure, 35-100. Dordrecht: Foris. Hackl, Martin 1998 On the semantics of "ability attributions". Ms. Cambridge, Mass: MIT. Haegeman, Liliane 1995 The syntax of negation. Cambridge, UK: Cambridge University Press. 1996 Finite V-movement in embedded clauses in West Flemish. South African Journal of Linguistics 29: 69-104. Haegeman, Liliane and Henk van Riemsdijk 1986 Verb projection raising, scope, and the typology of rules affecting verbs. Linguistic Inquiry 17: 417-466. Haegeman, Liliane and Raffaella Zanuttini 1991 Negative heads and the NEG criterion. The Linguistic Review 8: 233-251. Haider, Hubert 1986a Deutsche Syntax - generativ. Habilitationsschrift, University of Vienna, Austria. 1986b Fehlende Argumente: vom Passiv zu kohärenten Infinitiven. Linguistische Berichte 101: 3-33. 1986c Nicht-sententiale Infinitive. Groninger Arbeiten zur germanistischen Linguistik 28: 73-114. 1991a Fakultativ kohärente Infinitivkonstruktionen. Arbeitspapiere des Sonderforschungsbereichs 340. 1991b PRO-bleme? In: Gisbert Fanselow and Sascha Felix (eds.), Strukturen und Merkmale syntaktischer Kategorien, 121-143. (Studien zur deutschen Grammatik 39.) Tübingen: Narr. 1993 Deutsche Syntax Generativ. Tübingen: Narr. Harley, Heidi 1995 Subjects and clause structure. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.]

References

353

Hegarty, Michael 1991 Adjunct extraction and chain configurations. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] Higginbotham, James 1983 The logic of perceptual reports: an extensional alternative to situation semantics. The Journal of Philosophy 80: 100-127. 1985 On semantics. Linguistic Inquiry 16: 547-593. Hinterhölzl, Roland 1997 A VP-based approach to verb raising. In: K. Kusumoto (ed.), Proceedings of the North Eastern Linguistics Society Annual Meeting (NELS 27), 187-201. Amherst, Mass.: Graduate Linguistic Student Association. 1998 The syntax of IPP constructions and the structure of the lower middlefield in Westgermanic. In: Alexis Dimitriadis, Hikyoung Lee, Christine Moisset and Alexander Williams (eds.), Proceedings of the 22nd annual Penn Linguistics Colloquium: UPenn Working Papers in Linguistics 5.1: 59-70. 1999 Restructuring infinitives and the theory of complementation. Ph.D. Dissertation, University of Southern California, Los Angeles, CA. Hoeksema, Jacob 1988 A constraint on governors in the West Germanic verb cluster. In: Martin Everaert, Arnold Evers, Riny Huybregts and Mieke Trommelen (eds.), Morphology and modularity: In honour of Henk Schultink, 147-161. Dordrecht: Foris. Hoekstra, Teun 1984 Government and the distribution of sentential complementation in Dutch. In: Wim de Geest and Yvan Putsey (eds.), Sentential complementation: Proceedings of the international conference held at UFSAL, Brussels (June 1983), 105-116. Dordrecht: Foris Publications. Höhle, Tilman 1978 Lexikalistische Syntax: Die Aktiv-Passiv-Relation und andere Infinitkonstruktionen im Deutschen. (Linguistische Arbeiten 67.) Tübingen: Max Niemeyer Verlag. Hoji, Hajime 1985 Logical form constraints and configurational structures in Japanese. Ph.D. Dissertation, University of Washington, Seattle, WA. Hornstein, Norbert 1995 Logical form: From GB to Minimalism. Oxford, UK; Cambridge, Mass: Blackwell. 1998 Movement and chains. Syntax 1: 99-127. 1999 Movement and control. Linguistic Inquiry 30: 69-96.

354

References

Iatridou, Sabine 1990 About Agr(P). Linguistic Inquiry 21: 551-577. 1994 Comments on the paper by Roberts. In: David Lightfoot and Norbert Hornstein (eds.), Verb movement, 243-260. Cambridge, UK: Cambridge University Press. Jackendoff, Ray 1972 Semantic interpretation in generative grammar. Cambridge, Mass: MIT Press. 1977 X-bar-syntax: A study of phrase structure. Cambridge, Mass: MIT Press. Johnson, Kyle 1990 On the syntax of inflectional paradigms. Ms. Madison, Wisconsin: University of Wisconsin. Kageyama, Taro 1993 Bunpo to Gokeisei [Grammar and Word Formation]. Kasukabe: Hituzi Syobo. Karttunen, Lauri 1971 Implicative verbs. Language 47: 340-358. Kawasaki, Noriko 1993 Control and arbitrary interpretation in English. Ph.D. Dissertation, University of Massachusetts, Amherst. Kayne, Richard 1989 Null subjects and clitic climbing. In: Osvaldo Jaeggli and Kenneth Safir (eds.), The null subject parameter, 239-261. Dordrecht: Kluwer Academic Publishers. 1990 Romance clitics and PRO. In: Juli A. Carter, Rose-Marie Dechaine, William Philip and Tim D. Sherer (eds.), Proceedings of the North Eastern Linguistics Society Annual Meeting (NELS 20), 255-302. Amherst, Mass: Graduate Linguistic Student Association. 1991 Romance clitics, verb movement, and PRO. Linguistic Inquiry 22: 647-686. 1992 Italian negative imperatives and clitic climbing. In: L. Tasmowsky and A. Zribi-Hertz (eds.), Hommages à Nicolas Ruwet, 300-312. Gent: Communication and Cognition. 1998 Overt vs. covert movement. Syntax 1: 128-191. Kiparsky, Paul and Carol Kiparsky 1970 Fact. In: M Bierwisch and K.E. Heidolph (eds.), Progress in linguistics, 143-173. The Hague: Mouton. 1971 Fact. In: D.D. Steinberg and L.A. Jakobovits (eds.), Semantics: An interdisciplinary reader in philosophy, linguistics and psychology, Cambridge, UK: Cambridge University Press.

References Kiss, Tibor 1995

355

Infinite ¡Complementation: Neue Studien zum deutschen Verbum infinitum. (Linguistische Arbeiten 333.) Tübingen: Max Niemeyer Verlag. Klein, Wolfgang 1994 Time in language. London and New York: Routledge. Koizumi, Masatoshi 1995 Phrase structure in minimalist syntax. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] Koopman, Hilda and Anna Szabolcsi 2000 Verbal complexes. (Current Studies in Linguistics 34.) Cambridge, Mass: MIT Press. Koster, Jan 1984 On binding and control. Linguistic Inquiry 15: 417-459. 1987 Domains and dynasties: The radical autonomy of syntax. Dordrecht: Foris. Kratzer, Angelika 1977 What must and can must and can mean. Linguistics and Philosophy 1: 337-355. 1981 The notional category of modality. In: Hans-Jürgen Eikmeyer and Hannes Rieser (eds.), Words, worlds and C=contexts: New approaches in world semantics, 38-74. (Research in Text Theory/Untersuchungen zur Texttheorie 6.) Berlin: Walter de Gruyter. 1991 Modality. In: Arnim von Stechow and Dieter Wunderlich (eds.), Semantik/Semantics. Ein internationales Handbuch der zeitgenössischen Forschung. An Internatonial Handbook of Contemporary Research., 639-650. New York/Berlin: Walter de Gruyter. 1994 On external arguments. Functional projections: University of Massachusetts Occasional Papers 17: 103-130. 1996 Severing the external argument from its verb. In: Johan Rooryck and Laurie Zaring (eds.), Phrase structure and the lexicon, 109137. (Studies in Natural Language & Linguistic Theory 33.) Dordrecht: Kluwer Academic Publishers. Kroch, Anthony and Beatrice Santorini 1991 The derived constituent structure of the West Germanic verbraising construction. In: Robert Freidin (ed.) Principles and Parameters in Comparative Grammar, 269-338. Cambridge, Mass: MIT Press. Kulick, Seth 1998 Generalized transformations and restructuring in Romance. Ms. Philadelphia: University of Pennsylvania.

356

References

Kuno, Susumu 1973 The structure of the Japanese language. (Current studies in linguistics series 3.) Cambridge, Mass: MIT Press. 1975 Super-equi-NP deletion is a pseudo-transformation. Proceedings of the Fifth Annual Meeting of the North Eastern Linguistic Society (NELS 5), 29-44. Amherst, Mass: Graduate Linguistic Student Association. Landau, Idan 1999 Elements of control. Ph.D. Dissertation, MIT, Cambridge, Mass. [Published as Landau 2000.] 2000 Elements of control: Structure and meaning in infinitival constructions. (Studies in Natural Language and Linguistic Theory 51.) Dordrecht/Boston/London: Kluwer Academic Publishers. Larson, Richard K. 1988 On the double object construction. Linguistic Inquiry 19: 335-391. Lasnik, Howard (ed.) 1989 Essays on anaphora. (Studies in Natural Language and Linguistic Theory 16.) Dordrecht: Kluwer Academic Publishers. Lasnik, Howard and Robert Fiengo 1974 Complement object deletion. Linguistic Inquiry 5: 535-571. Lasnik, Howard and Mamoru Saito 1984 On the nature of proper government. Linguistic Inquiry 15: 235289. Lebeaux, David 1985 Locality and anaphoric binding. The Linguistic Review 4: 343-363. 1995 Where does binding theory apply? University of Maryland Working Papers in Linguistics 3: 63-88. Li, Yafei 1990 Conditions on Xo movement. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] Luján, Marta 1980 Clitic promotion and mood in Spanish verbal complements. Linguistics 18: 381-484. Lyons, John 1977 Semantics. Cambridge, UK: Cambridge University Press. Manzini, Rita 1982 Italian prepositions before infinitives. In: Alec Marantz and Tim Stowell (eds.), Papers in syntax: MIT Working Papers in Linguistics 4: 115-155. 1983a On control and control theory. Linguistic Inquiry 14: 412-446. 1983b Restructuring and reanalysis. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.]

References

357

Manzini, Rita and Anna Roussou 2000 A minimalist theory of Α-movement and control. Lingua 110: 409447. Marantz, Alec 1993 Implications of asymmetries in double object constructions. In: Sam A. Mchombo (ed.) Theoretical aspects of Bantu grammar, 113-150. (CSLI Lecture Notes 381.) Stanford, CA: CSLI Publications. 1995 'Cat* as a phrasal idiom. Ms. Cambridge, Mass: MIT. 1997 No escape from syntax: Don't try morphological analysis in the privacy of your own lexicon. UPenn Working Papers in Linguistics 4.2: 201-225. Martin, Roger Andrew 1992 On the distribution and case features of PRO. Ms. Storrs, CT: University of Connecticut. 1996 A minimalist theory of PRO and control. Ph.D. Dissertation, University of Connecticut, Storrs, CT. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] May, Robert 1977 The Grammar of Quantification. Ph.D. Dissertation, MIT, Cambridge, Mass. 1985 Logical Form: Its Structure and Derivation. Cambridge, Mass: ΜΓΓ Press. McGinnis, Martha Jo 1993 The deontic/epistemic distinction in English modals: a compositional analysis. M.A. Thesis, University of Toronto, Ontario. Melvold, Janis 1991 Factivity and definiteness. In: Lisa Cheng and Hamida Demirdache (eds.), More papers on Wh-movement: MIT Working Papers in Linguistics 15: 97-117. Meurers, Walt Detmar 2000 Lexical generalizations in the syntax of German non-finite constructions. Ph.D. Dissertation, University of Tübingen, Germany. Mittwoch, Anita 1990 On the distribution of bare infinitive complements in English. Journal of Linguistics 26: 103-131. Miyagawa, Shigeru 1987 Restructuring in Japanese. In: Takashi Imai and Mamoru Saito (eds.), Issues in Japanese linguistics, 273-300. Dordrecht: Foris. Montalbetti, Mario, Mamoru Saito and Lisa Travis 1982 Three ways to get tough. Papers from the 18th regional meeting: Chicago Linguistic Society (CLS18), 348-366.

358

References

Moore, John 1990

1994

Spanish clause reduction with downstairs cliticization. In: Katarzyna Dziwirek, Patrick Farrell and Errapel Mejias-Bikandi (eds.), Grammatical relations: a cross-theoretical perspective, 319334. Stanford, CA: Center for Study of Language and Information. Romance cliticization and relativized minimality. Linguistic Inquiry 25: 335-344.

Müller, Gereon 1996 A constraint on remnant movement. Natural Language and Linguistic Theory 14: 355-407. Napoli, Donna 1981 Semantic interpretation vs. lexical governance: clitic climbing in Italian. Language 841-887. Neeleman, Ad 1994 Complex predicates. Ph.D. Dissertation, Utrecht University, The Netherlands. [Distributed by: LEd (OTS Dissertation Series), Utrecht, The Netherlands.] Nemoto, Naoko 1993 Chains and case positions: A study from scrambling in Japanese. Ph.D. Dissertation, University of Connecticut, Storrs, CT. [Distributed by: ΜΓΓ Working Papers in Linguistics, Cambridge, Mass.] Newmeyer, Frederick J. 1975 English aspectual verbs. The Hague: Mouton. Nishigauchi, Taisuke 1993 Long distance passive. In: Nobuko Hasegawa (ed.) Japanese syntax in comparative grammar, 79-114. Tokyo: Kuroshio Shuppan. O'Neil, John H. 1997 Means of control: deriving the properties of PRO in the Minimalist Program. Ph.D. Dissertation, Harvard University, Cambridge, Mass. Ouhalla, Jamal 1991 Functional categories and parametric variation. London and New York: Routledge. Palmer, Frank R. 1979 Why auxiliaries are not main verbs. Lingua 47: 1-25. 1983 Modality and the English modals. London and New York: Longman. 1986 Mood and modality. Cambridge, UK: Cambridge University Press. Perlmutter, David 1970 The two verbs 'begin'. In: Roderick Jacobs and Peter S. Rosenbaum (eds.), English transformational grammar, 107-119. Waltham, Mass: Ginn & Company.

References

359

Pesetsky, David 1992 Zero syntax II: An essay on infinitives. Ms. Cambridge, Mass: MIT. 1997 Optimality Theory and syntax: Movement and pronunciation. In: Diana Archangeli and Terence D. Langendoen (eds.), Optimality Theory: an overview, 134-170. Maiden, Mass: Blackwell. Pesetsky, David and Esther Torrego 2001 T-to-C movement: causes and consequences. In: Michael Kenstowicz (ed.) Ken Hale: A life in language. Cambridge, Mass: ΜΓΓ Press. Petter, Marga 1998 Getting PRO under control. Ph.D. Dissertation, Netherlands Graduate School of Linguistics (LOT), Amsterdam, The Netherlands. [Distributed by: Holland Academic Graphics (HIL), The Hague.] Picallo, Carme 1985 Opaque domains. Ph.D. Dissertation, CUNY, New York. 1990 Modal verbs in Catalan. Natural Language and Linguistic Theory 8: 285-312. Pollock, Jean-Y ves 1989 Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry 20: 365-424. Prinzhorn, Martin 1987 T-Struktur, S-Struktur und Lexikon. M.A. Thesis, University of Vienna, Austria. 1990 Head movement and scrambling domains. In: Günther Grewendorf and Wolfgang Sternefeld (eds.), Scrambling and barriers, 199-215. Amsterdam/Philadelphia: John Benjamins. Pylkkänen, Liina 1999 Causation and external arguments. In: Liina Pylkkänen, Angeliek van Hout and Heidi Harley (eds.), Papers from the UPenn/MIT roundtable on the lexicon: MIT Working Papers in Linguistics 35: 21-40. Radford, Andrew 1977 Italian Syntax: Transformational and Relational Grammar. Cambridge, UK: Cambridge University Press. Raposo, Eduardo 1987 Case theory and Infl-to-Comp: The inflected infinitive in European Portuguese. Linguistic Inquiry 18: 85-109. Reinhart, Tanya and Eric Reuland 1993 Reflexivity. Linguistic Inquiry 24: 657-720.

360

References

Reuland, Eric 1981

On extraposition of complement clauses. In: Victoria Burke and James Pustejovsky (eds.), Proceedings of the North Eastern Linguistics Society Annual Meeting (NELS 11), 296-318. Amherst, Mass: Graduate Linguistic Student Association. Riemsdijk, Henk van 1984 On pied-piped infinitives in German relative clauses. In: Jindrich Toman (ed.) Studies in German Grammar, 165-192. Dordrecht: Foris. 1985 Zum Rattenfängereffekt bei Infinitiven in deutschen Relativsätzen. In: Werner Abraham (ed.) Erklärende Syntax des Deutschen, 7599. Tübingen: Narr. Categorial feature magnetism: The endocentricity and distribution 1998a of projections. Journal of Comparative Germanic Linguistics 2: 148. 1998b Head movement and adjacency. Natural Language and Linguistic Theory 16: 633-678. Rizzi, Luigi Ristrutturazione. Rivista di grammatica generativa 1. 1976 1978 A restructuring rule in Italian syntax. In: Samuel Jay Keyser (ed.) Recent transformational studies in European languages, 113-158. Cambridge, Mass: MIT Press. 1982 Issues in Italian syntax. Dordrecht: Foris. 1990 Relativized minimality. Cambridge, Mass: MIT Press. 1993 Some notes on Romance cliticization. Paper presented at the ESF workshop on clitics., Durham, Great Britain. 1997 The fine structure of the left periphery. In: Liliane Haegeman (ed.) Elements of grammar: Handbook of generative syntax, 281-337. Dordrecht: Kluwer. Roberts, Ian Agreement parameters and the development of English auxiliaries. 1985 Natural Language and Linguistic Theory 3: 21-58. Excorporation and minimality. Linguistic Inquiry 22: 209-218. 1991 Restructuring, pronoun movement and head-movement in Old 1993 French. Ms. Bangor: University of Wales. 1994 Two types of head movement in Romance. In: David Lightfoot and Norbert Hornstein (eds.), Verb movement, 207-242. Cambridge, UK: Cambridge University Press. 1997 Restructuring, head movement, and locality. Linguistic Inquiry 28: 423-460. Rochette, Anne 1988 Semantic and syntactic aspects of Romance sentential complementation. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.]

References 1990

1999

361

On the restructuring classes of verbs in Romance. In: Anna Maria Di Sciullo and Anne Rochette (eds.), Binding in Romance: Essays in honour of Judith McA'Nulty., 96-128. Ottawa: Canadian Linguistic Association. The selection properties of aspectual verbs. In: Kyle Johnson and Ian Roberts (eds.), Beyond Principles and Parameters: Essays in memory of Osvaldo Jaeggli, 145-165. Dordrecht: Kluwer Academic Publishers.

Rooryck, Johan 1994 Against optional movement for clitic climbing. In: Michael Mazzola (ed.) Issues and theory in Romance linguistics: Selected papers from the Linguistic Symposium on Romance Languages XXIII, 417-443. Washington, DC: Georgetown University Press. Rosen, Sara Thomas 1989 Argument structure and complex predicates. Ph.D. Dissertation, Brandeis University, Waltham, Mass. 1990 Restructuring verbs are light verbs. In: Aaron Halpern (ed.), Proceedings of the 9th West Coast Conference on Formal Linguistics (WCCFL 9), 477-491. Stanford, CA: CSLI Publications. Rosengren, Inger 1992 Zum Problem der kohärenten Verben im Deutschen. In: Peter Suchsland (ed.) Biologische und soziale Grundlagen der Sprachen, 265-296. (Linguistische Arbeiten 280.) Tübingen: Max Niemeyer Verlag. Ross, John Robert 1967 Constraints on variables in syntax. Ph.D. Dissertation, MIT, Cambridge, Mass. 1969 Auxiliaries as main verbs. In: W. Todd (ed.) Studies in philosophical linguistics, 1. Evanston, IL: Great Expectations Press. Rowe, Janet 1994 English modals and the functional-lexical distinction. M.A. Thesis, University of Toronto, Ontario. Rutten, Jean 1991 Infinitival complements and auxiliaries. Ph.D. Dissertation, University of Amsterdam, The Netherlands. Sabel, Joachim 1994 Restrukturierung und Lokalität: Universelle Beschränkungen für Wortstellungsvariationen. Ph.D. Dissertation, University of Frankfurt, Germany. [Published as Sabel 1996.] 1996 Restrukturierung und Lokalität: Universelle Beschränkungen für Wortstellungsvariationen. (Studia Grammatica 42.) Berlin: Akademie Verlag.

362

References

Safir, Ken 1993

Perception, selection, and structural economy. Natural Language Semantics 2: 47-70. Saito, Mamoru and Hiroto Hoshi 1998 Control in Complex Predicates. Ms. Nanzan and London: Nanzan University and SOAS. 2000 Japanese light verb construction and the Minimalist Program. In: Roger Martin, David Michaels and Juan Uriagereka (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 261295. Cambridge, Mass: MIT Press. Sauerland, Uli 1998a The meaning of chains. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] 1998b Scope reconstruction without reconstruction. In: Kimary Shahin, Susan Blake and Eun-Sook Kim (eds.), Proceedings of the 17th West Coast Conference on Formal Linguistics (WCCFL 17), 582596. Stanford, CA: CSLI Publications. Selkirk, Elisabeth 1984 Phonology and syntax. Cambridge, Mass: MIT Press. 1986 On derived domains in sentence phonology. Phonology 3: 371-405. Shibatani, Masayoshi 1978 Nihongo no Bunseki [Analysis of Japanese]. Tokyo: Taishukan Shoten. Sigurösson, Halldór Ármann 1991 Icelandic case-marked PRO and the licensing of lexical arguments. Natural Language and Linguistic Theory 9: 327-364. Sportiche, Dominique 1996 Clitic constructions. In: Johan Rooryck and Laurie Zaring (eds.), Phrase structure and the lexicon, 213-276. (Studies in Natural Language & Linguistic Theory 33.) Dordrecht: Kluwer Academic Publishers. Stechow, Arnim von 1984 Gunnar Bech's government and binding theory: GB's GB theory. Linguistics 22: 225-241. 1990 Status government and coherence in German. In: Günther Grewendorf and Wolfgang Sternefeld (eds.), Scrambling and barriers, 143-198. Amsterdam/Philadelphia: John Benjamins. Stechow, Arnim von and Wolfgang Sternefeld 1988 Bausteine syntaktischen Wissens: Ein Lehrbuch der modernen generativen Grammatik. Opladen: Westdeutscher Verlag.

References

363

Sternefeld, Wolfgang 1990 Scrambling and minimality. In: Günther Grewendorf and Wolfgang Sternefeld (eds.), Scrambling and barriers, 239-257. Amsterdam/Philadelphia: John Benjamins. Stowell, Tim 1981 The origins of phrase structure. Ph.D. Dissertation, MIT, Cambridge, Mass. 1982 The tense of infinitives. Linguistic Inquiry 13: 561-570. 1996 The phrase structure of tense. In: Johan Rooryck and Laurie Zaring (eds.), Phrase structure and the lexicon, 277-291. (Studies in Natural Language & Linguistic Theory 33.) Dordrecht: Kluwer Academic Publishers. Strozer, Judith Reina 1976 Clitics in Spanish. Ph.D. Dissertation, UCLA, Los Angeles, CA. 1981 An alternative to restructuring in Romance syntax. Papers in Romance (PRom) 3: 177-184. Tada, Hiroaki 1992 Nominative objects in Japanese. Journal of Japanese Linguistics 14:91-108. 1993 A/A-bar partition in derivation. Ph.D. Dissertation, MIT, Cambridge, Mass. Tappe, Hans Thilo 1984 On infinitival clauses without COMP. In: Wim de Geest and Yvan Putsey (eds.), Sentential complementation: Proceedings of the international conference held at UFSAL, Brussels (June 1983), 227237. Dordrecht: Foris Publications. Terzi, Arhonto 1996 Clitic climbing from finite clauses and tense raising. Probus 8: 273-295. Thráinsson, Höskuldur 1996 On the (non)-universality of functional projections. In: Werner Abraham, Samuel David Epstein, Höskuldur Thráinsson and C. Jan-Wouter Zwart (eds.), Minimal ideas: Syntactic studies in the minimalist framework, 253-281. (Linguistik Aktuell 12.) Amsterdam/Philadelphia: John Benjamins Publishing Company. Thráinsson, Höskuldur and Sten Vikner 1995 Modals and double modals in the Scandinavian languages. Working Papers in Scandinavian Syntax 55: 51-88. Travis, Lisa deMena 1984 Parameters and effects of word order variation. Ph.D. Dissertation, MIT, Cambridge, Mass. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.]

364 1994

References

Event phrase and a theory of functional categories. In: Päivi Koskinen (ed.), Proceedings of the 1994 Annual Conference of the Canadian Linguistic Association (CLA), 559-570. Toronto, Ontario: Toronto Working Papers in Linguistics. 2000 The L-syntax/S-syntax boundary: evidence from Austronesian. In: Ileana Paul, Vivianne Phillips and Lisa deMena Travis (eds.), Formal issues in Austronesian syntax, 167-194. (Studies in Natural Language and Linguistic Theory 49.) Dordrecht/Boston/London: Kluwer Academic Publishers. Truckenbrodt, Hubert 1995 Extraposition from NP and Prosodie Structure. In: Jill N. Beckman (ed.), Proceedings of the North Eastern Linguistics Society Annual Meeting (NELS 25), 503-517. Amherst, Mass: Graduate Linguistic Student Association. Warner, Anthony 1993 English Auxiliaries. Cambridge, UK: Cambridge University Press. Watanabe, Akira 1993 The role of equidistance in restructuring verbs: Italian vs. French. Proceedings of the Eastern State Conference on Linguistics (ESCOL '93), 360-371. Williams, Edwin 1980 Predication. Linguistic Inquiry 11: 203-338. Wurmbrand, Susi 1994 The German leicht construction. Wiener Linguistische Gazette 4850: 87-100. 1997 Restructuring infinitives. In: Tina Cambier-Langeveld, Joäo Costa, Rob Goedemans and Ruben van de Vijver (eds.), Proceedings of ConSOLES, 277-292. Leiden: SOLE. 1998a Downsizing infinitives. In: Uli Sauerland and Orin Perçus (eds.), The interpretive tract: MIT Working Papers in Linguistics 25: 141175. 1998b Infinitives. Ph.D. Dissertation, MIT, Cambridge, Mass. 1999a Modal verbs must be raising verbs. In: Sonya Bird, Andrew Carnie, Jason D. Haugen and Peter Norquest (eds.), Proceedings of the 18th West Coast Conference on Formal Linguistics (WCCFL 18), 599-612. Somerville, Mass: Cascadilla Press. 1999b Where do verbs cluster: West Germanic perspectives. Paper presented at the Workshop on Verb Clusters, Ôttevény, Hungary [http://web.mit.edu/susi/www/Hungary.pdf], 2000a Α-movement in German: AGREE and ATTRACT. Colloquium talk Harvard University, Cambridge, Mass [http://web.mit.edu/susi/ www/Harvard.pdf].

References 2000b

365

Verb clusters: variation at the right periphery. Paper presented at the workshop on Syntactic Microvariation, Meertens Instituut, Amsterdam [http://web.mit.edu/susi/www/SAND.pdf]. 2001 Back to the future. Snipppets 3: 15-16. to appear AGREE: The other VP-internal subject hypothesis. In: Karine Megerdoomian and Leora A. Bar-el (eds.), Proceedings of the 20th West Coast Conference on Formal Linguistics (WCCFL 20). Somerville, Mass: Cascadilla Press. in prep. Verb raising, verb clustering, and restructuring. Case entry to be included in The Syntax Companion (SynCom). Oxford: Blackwell. Wyngaerd, Guido J. Vanden 1994 PRO-legomena: Distribution and reference of infinitival subjects. (Linguistic Models 19.) Berlin/New York: Mouton de Gruyter. Yatsushiro, Kazuko 1999 Case-licensing and VP structure. Ph.D. Dissertation, University of Connecticut, Storrs, CT. [Distributed by: MIT Working Papers in Linguistics, Cambridge, Mass.] Zagona, Karen 1982 Government and proper government of verbal projections. Ph.D. Dissertation, University of Washington, Seattle, WA. Zanuttini, Raffaella 1991 Syntactic properties of sentential negation: A comparative study of Romance languages. Ph.D. Dissertation, University of Pennsylvania, Philadelphia. 1996 On the relevance of tense for sentential negation. In: Adriana Belletti and Luigi Rizzi (eds.), Parameters and functional heads: Essays in comparative syntax, 181-207. (Oxford studies in comparative syntax New York/Oxford, UK: Oxford University Press. Zubizarreta, Maria Luisa 1982 On the relationship of the lexicon to syntax. Ph.D. Dissertation, MIT, Cambridge, Mass.

Index

adjunct thêta roles, 179 argument structure merger, 147 aspectual infinitives, 96-98, 208-14 Authors Aissen and Perlmutter, 6, 8,12, 343 Akmajian, Steele and Wasow, 139 Alsina, 217 Baker, 12, 14, 124,128 Barbiers, 182,203, 204 Bayer and Komfilt, 40, 293 Bech, 5,113 Belletti, 115, 124, 126 Bennis and Hoekstra, 102,103, 126, 220 Bhatt, 188 Bobaljik, 122, 124,14(M3 Bobaljik and Thráinsson, 140-43 Bobaljik and Wurmbrand, 192 Boivin, 79,217 Bok-Bennema and KampersManhe, 13, 15,123, 128 Borer, 247 Boskovic, 63, 69,70, 112, 228 Bouchard, 247 Brennan, 171, 182, 189,194 Bresnan, 29, 228 Brody, 122 Broekhuis, 15, 23, 24,120 Broekhuis, den Besten, Hoekstra, and Rutten, 11, 102,103,120 Browning, 28 Burzio, 14, 113, 226, 343 Cattell, 303

Chierchia, 228, 248, 249,251, 253-63, 297 Chomsky, 12,18,28,55, 139, 188, 200, 228 Chomsky and Lasnik, 28, 69, 111,228 Cinque, 2,8, 11,15, 28,31,78, 130,131, 134, 135, 138, 142, 146, 176-82, 184, 303, 345 Comrie, 237 Cremers, 11,79,100, 297, 306 Declerck, 217 den Besten and Rutten, 103,120 den Dikken, 20, 22-25 DiSciullo and Williams, 8, 11, 12,20 Erb, 92,163,202 Evers, 5, 12, 13, 97,120-21 Fanselow, 8, 11,273, 344 Farkas, 238 Felser, 79, 217, 220 Fiengo, 139 Fox, 192, 195 Gonçalves, 13, 15,122 Goodall, 12 Grewendorf, 12,233 Grewendorf and Sabel, 8, 13, 14, 60, 122 Grimshaw, 160 Grinder, 247 Groat and O'Neil, 122 Grosse, 267,287,293, 317, 328, 329, 333, 334, 335, 345 Guasti, 13,15,79,123-24, 221 Guéron and Hoekstra, 15,126 Hackl, 169

Index Haegeman, 32, 79, 116 Haegeman and van Riemsdijk, 12, 125 Haegeman and Zanuttini, 115 Haider, 11, 13, 15, 26,29,44, 115,267,287,317, 333, 335, 344 Harley, 18,143,155 Hegarty, 303-5 Higginbotham, 147, 217 Hinterhölzl, 14 Hoeksema, 97 Hoekstra, 12,102 Höhle, 19 Hoji, 51 Hornstein, 177,178, 195, 200, 246, 247 Iatridou, 129, 141, 170 Jackendoff, 139 Johnson, 141 Kageyama, 89 Karttunen, 79, 153, 298 Kawasaki, 237 Kayne, 8, 13, 32, 101,104, 105, 116, 123,124, 126, 129, 131 Kiparsky and Kiparsky, 299, 302, 303 Kiss, 8, 15, 267, 287, 293 Klein, 72 Koizumi, 33, 34,49,53, 89 Koopman and Szabolcsi, 14 Köster, 23, 247 Kratzer, 18,182,183, 226 Kroch and Santorini, 103 Kulick, 189 Kuno, 33, 247 Landau, 237,238, 239, 240, 241, 246, 247, 250, 307 Larson, 18 Lasnik, 55 Lasnik and Fiengo, 29 Lasnik and Saito, 304

367

Lebeaux, 192, 246, 247 Li, 11, 15 Luján, 8, 130 Lyons, 182 Manzini, 12, 113, 237, 246,247 Manzini and Roussou, 247 Marantz, 18,108, 143, 215 Martin, 63,69, 70, 112,228, 237 May, 192 McGinnis, 140, 189,203 Melvold, 303, 304 Meurers, 8 Mittwoch, 217, 220 Miyagawa, 54, 88 Montalbetti, Saito and Travis, 28,29, 32 Moore, 11 Müller, 273 Napoli, 8,15,108, 130,146-55, 343 Neeleman, 97 Nemoto, 34, 87 Newmeyer, 203 Nishigauchi, 89 O'Neil, 247 Ouhalla, 141 Palmer, 171,182 Perlmutter, 158, 168 Pesetsky, 63, 64, 69, 79,122, 153 Pesetsky and Torrego, 307 Petter, 237,238 Picallo, 8, 11, 15,150,184, 343 Pollock, 115, 124,125 Prinzhorn, 13, 15, 179, 209 Pylkkänen, 216 Radford, 129 Raposo, 70 Reinhart and Reuland, 55 Reuland, 103 Rizzi, 6,12,13, 29, 32, 79, 104, 105,113,121-22,132, 159,

368

Index 275, 343 Roberts, 8, 13, 15, 29, 32, 105, 122, 123, 126, 127-29,131, 133, 146, 148,179, 187, 274 Rochette, 8, 11,15,100, 113, 146-55, 228,297 Rooryck, 11, 13, 15,105-7, 123 Rosen, 8,11,15, 113, 146-55 Rosengren, 8,11, 15 Ross, 121,187 Rowe, 140 Rutten, 8, 11, 13,15,97,102, 103, 120, 209, 272, 284, 344 Sabel, 8, 13, 14,40,60,61, 115, 123,127, 128,230, 279, 293, 317, 329, 344 Safir, 217 Saito and Hoshi, 34,49 Sauerland, 192 Selkirk, 294 Shibatani, 89 Sigurösson, 190 Sportiche, 274 Sternefeld, 8, 13, 14,15,122 Stowell, 62-63, 69-70, 91,99, 112 Strozer, 8, 11,15,61 Tada, 47,49 Tappe, 11,273 Terzi, 13, 123,128 Thráinsson, 140-43 Thráinsson and Vikner, 189, 191 Travis, 18, 114, 124, 128, 143, 155 Truckenbrodt, 160 van Riemsdijk, 97-98, 144, 155, 287 Vanden Wyngaerd, 188,198 von Stechow, 12, 287 von Stechow and Sternefeld, 12

Warner, 171, 182, 189, 199, 203 Watanabe, 131 Williams, 236, 243,246 Wurmbrand, 8, 15, 29, 66, 92, 237, 270, 328 Yatsushiro, 50, 53 Zagona, 8, 11, 15, 343 Zanuttini, 115, 116 Zubizarreta, 179,187 auxiliary-infinitive construction, 29-31, 37-38 binding, 229-36, 278-80 Burzio's Generalization, 18, 55 Catalan, 184, 343 causative construction, 123-24, 215-23 control arbitrary, 237 exhaustive, 237, 239,241-46, 249-51,253 implicit, 230-33,237, 239, 240, 321, 323 long-distance, 237 meaning postulates, 248, 254-57,261 obligatory vs. non-obligatory, 241, 245, 246-51,260 partial, 237, 245,247, 258-60, 277 PRO-less approach, 228, 229, 232-33, 235-36,248-51, 254-57 split, 237 strict vs. sloppy interpretation, 253-63 variable, 237, 2 4 1 ^ 3 , 247, 259, 277, 333 control vs. raising, 167-75, 176-82, 208-14. See also modal constructions copy theory of movement, 122

Index dative antecedents, 233 de se interpretation, 246 Doubly Filled Comp filter, 109, 128 Dutch infinitival complementizer, 102-4 IPP effect, 102-4, 162, 223, 283-85, 341 particle constructions, 97-98 remnant extraposition, 102-4, 283-85 restructuring classification, 6-7, 223, 283-85, 341-43, 344 tense interpretation, 79,223, 283-85 verb raising, 5, 20,22-25, 97-98,102-4,120-21, 125, 283-85, 341 easy-to-please construction, 27-29, 32, 37-38,79 event structure, 146-55, 304 excorporation, 106, 107, 123,129 extraposition, 40-42, 156-61, 167, 175, 209-10, 212, 224-25, 291-96, 308-16. See also remnant extraposition factive infinitives, 80, 286-87, 298-307, 317, 323-26 focus scrambling, 40-42, 116-17, 222, 270-83, 308-16,317, 327, 337-40 French, 32 functional restructuring predicates, 137, 155,224

369

idiomatic wA-infinitives, 107-8 implicative infinitives, 79,153-54, 298-302, 336 implicit arguments, 230-33 indexical modifier, 65-68,74,77 infinitival marker, 109-15 and case, 111-12 and negation, 114 and tense, 112-13 distribution, 110, 161, 166, 167, 175 intensionality, 75-76 intraposition, 39-40, 291-96 IPP effect, 162-67, 175, 224-25. See also Dutch IPP effect irrealis vs. propositional, 62-71, 284, 305-7, 317 control vs. raising/ECM, 69-70 temporal modification, 65-68, 319-23 truth/falsity predication, 64-65, 325 Italian causative construction, 79, 123-24 control vs. raising, 191 dislocation, 132-35 event modification, 148-52 infinitival complementizers, 104 long object movement, 31-33 negation, 129-32 restructuring classification, 6-7, 341-43, 345 short verb movement, 126 verb raising, 121-22 vfA-specifiers, 105-7 /f-anaphors, 253-63

Guillotine principle, 5, 121 Head Movement Constraint, 124, 128 Icelandic, 189-92

Japanese affixal verbs, 89 binding, 54-56 nominative objects. See stative construction

370

Index restructuring classification, 6-7, 91,92-94, 341-43 scope, 49-53 stative construction, 33-35, 46-51 tense morphology, 85-91, 92-94, 341

Lexical Functional Grammar, 228 lexical restructuring predicates, 16, 155, 225, 327 lexical/functional distinction, 17, 144,155-82, 308 lexicalism, 20 light verbs, 147 long passive, 19-27, 31-32, 36, 57, 82,84,96, 118,133, 178, 213,267,271,273-81, 286-93,299-300, 308-16, 317, 327-32 modal constructions case of subject, 189-92 control vs. raising, 187-204 co-occurrence restrictions, 139, 184-87 dynamic modals, 158,171, 182-83,186-87,189,196, 201-2, 205 non-directed deontic interpretation, 203 passive, 196-202 scope, 192-96 motion verb construction, 215-23 negation, 115-18, 129-32, 283 Null Case, 69, 111,228 optionality of restructuring, 35-56, 179-81,231,233-35, 289-90, 312-16 perception verb construction, 79,

215-23 presupposition generation, 304 Projection Principle, 12,228 pronoun fronting, 223, 267-81, 286-87,289-92, 295, 299-300, 306, 308-16, 317, 327, 332-37, 345 propositional infinitives, 302-7, 323-26. See also irrealis vs. propositional prosodie restrictions, 160,294-96 PRO-theorem, 228 quirky case, 189-92 raising constructions, 205-15 raising vs. control. See control vs. raising reanalysis, 12, 125 reconstruction, 192-96 reduced non-restructuring, 41, 105, 115,131,272-96, 308-16, 345 relative clause pied piping, 290-91, 308-16 remnant extraposition, 102-4. See also Dutch remnant extraposition remnant topicalization, 43-46, 134 restructuring approaches auxiliary, light verb approaches, 15, 146-55 bi-clausal approaches, 9,11-14, 57-62, 108,119 head movement approaches, 12-14,59-60. See also verb raising mono-clausal approaches, 9, 10-11 reduced clausal approaches, 11 structure changing approaches, 12 topicalization, remnant

Index movement approaches, 14, 60-61 restructuring configuration, 4,94, 99-101, 108,119, 155,214, 275-84, 308-16 scrambling, 40-46, 116-17, 222, 270-83, 286-93,299-300, 308-16, 317, 327, 337-40 Spanish long passive, 31-32 negation, 130 restructuring classification, 6-7, 341-43 wA-specifiers, 105-7 special meaning, 143 split IP parameter, 140-43 structure pruning principle, 5, 12, 121 tense, infinitival. See also irrealis vs. propositional deficient, 15, 74-76,79-85, 126-27,216 irrealis, 62-63, 70-71, 72-78, 319-23

371

past future interpretation, 66-68 simultaneity. See deficient tense temporal modification, 65-68, 73-74, 76-78, 276, 323 thematic deficiency, 15, 146-55 Theta Criterion, 179,188 Third construction, 103. See also remnant extraposition Underspecification, 147 verb classifications, 6-7, 91, 93, 245,271, 318, 326, 327, 341-43. See also irrealis vs. propositional verb projection raising, 22-25,125 verb raising, 5,14,118-35. See also Dutch, Italian verb raising verbs of saying, 326 VP-ellipsis, 114 Wackernagel, 267,274, 275 weaker-predicates, 167-72 West Flemish, 22, 125 West Germanic, 102, 125, 162 w/i-extraction, 302-5