Studies in Comparative Germanic Syntax: Proceedings from the 15th Workshop on Comparative Germanic Syntax

Studies in Comparative Germanic Syntax Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (L...

Author: C. Jan-Wouter Zwart and Werner Abraham

125 downloads 1534 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Studies in Comparative Germanic Syntax

Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.

Series Editor Werner Abraham University of Vienna

Advisory Editorial Board Guglielmo Cinque (University of Venice) Günther Grewendorf (J.W. Goethe-University, Frankfurt) Liliane Haegeman (University of Lille, France) Hubert Haider (University of Salzburg) Christer Platzack (University of Lund) Ian Roberts (Cambridge University) Ken Safir (Rutgers University, New Brunswick NJ) Lisa deMena Travis (McGill University) Sten Vikner (University of Aarhus) C. Jan-Wouter Zwart (University of Groningen)

Volume 53 Studies in Comparative Germanic Syntax: Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000) Edited by C. Jan-Wouter Zwart and Werner Abraham

Studies in Comparative Germanic Syntax Proceedings from the 15th Workshop on Comparative Germanic Syntax Edited by

C. Jan-Wouter Zwart University of Groningen

Werner Abraham University of Vienna

John Benjamins Publishing Company Amsterdam/Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Studies in comparative Germanic syntax : proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000) / edited by C. Jan-Wouter Zwart and Werner Abraham. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 53) Papers from an international conference held at the University of Groningen, Groningen, Netherlands, May 26–27, 2000. Includes bibliographical references and index. 1. German language--Syntax--Congresses. 2. Germanic languages--Syntax-Congresses. 3. English language--Syntax--Congresses. I. Workshop on Comparative Germanic Syntax (15th : 2000 : Groningen, Netherlands) II. Zwart, C. Jan-Wouter. III. Abraham, Werner. IV. Linguistik aktuell ; Bd. 53. PD361.S78 2002 435-dc21 isbn 90 272 2774 8 (Eur.) / 1 58811 268 3 (US) (Hb; alk. paper)

2002074756

© 2002 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

SUBJECT "Linguistik Aktuell/Linguistics Today, Volume 53"

KEYWORDS ""

SIZE HEIGHT "220"

WIDTH "150"

VOFFSET "4">

Table of contents

Introduction C. Jan-Wouter Zwart List of contributors

vii

xiii

Subordination Wh-movement and integrated parenthetical constructions Marga Reis

3

Van as a marker of dissociation: Microvariation in Dutch Jeroen van Craenenbroeck

41

Expletive subjects in subject relative clauses Line Mikkelsen

69

Syntactic versus semantic control Susi Wurmbrand

93

Movement and Morphology Parametric variation and scrambling in English Roland Hinterhölzl

131

V2 and Holmberg’s Generalization Øystein Nilsen

151

The distribution of declarative verb second in Germanic Olaf Koeneman

175

A verb’s gotta do what a verb’s gotta do! On Scandinavian infinitivals and the AGR parameter Øystein Alexander Vangsnes

203

< /R/TREARGET E FF

"ale"> "hae"> "hoe"> "mer"> "ste"> "zim"> "ref"> "si"> "toc">

vi

Table of contents

On the correlation between morphology and syntax: The case of V-to-I Artemis Alexiadou and Gisbert Fanselow

219

Language Variation and Change Observations on the loss of Verb Second in the history of English Eric Haeberli A structure-based analysis of morphosyntactic regularities in language contact Eric Hoekstra

245

273

Syntax and Semantics Swiping in Germanic Jason Merchant

289

The ambiguity of weak reflexive pronouns in English and German Markus Steinbach

317

‘Binominal each-constructions’ (BECs) in German and English Malte Zimmermann

343

References

373

Name index

393

Subject index

399

Introduction C. Jan-Wouter Zwart University of Groningen

The papers appearing in this volume were presented at the 15th Comparative Germanic Syntax Workshop, organized at the University of Groningen on May 26–27, 2000. The history of the Comparative Germanic Syntax Workshop until 1995 has been recorded in Thráinsson et al. (1996: vii). Since then, the twelfth edition of the workshop has been held at the Free University of Amsterdam (January 1997), the thirteenth at Cornell University in Ithaca (July 1997), the fourteenth at the University of Lund (1999), and the sixteenth at McGill University in Montréal (2001). At the moment of writing, preparations are under way for the seventeenth instalment, at the University of Reykjavík (2002). This is the first proceedings volume to appear since 1997, when a selection of papers from the 1995 Brussels workshop was published in Linguistics 35, 4–6. What transpired in the other workshops has been covered in conference reports which appeared in Glot International and the Germanic Generative Syntax Newsletter.1 Papers from the sixteenth workshop are scheduled to appear in the Journal of Comparative Germanic Linguistics. These proceedings are organized in four chapters. The first chapter deals with issues of sentence embedding in the Germanic languages. We are pleased to present a comprehensive treatment by Marga Reis of her analysis of constructions of the type in (1). (1) a.

Was sollte Max glaubst du lesen? what should Max believe-2sg you read-inf ‘What do you think Max should read?’ b. Was glaubst du sollte Max lesen? what believe-2sg you should Max read-inf ‘What do you think Max should read?’

viii Introduction

c.

Was glaubst du welches Buch ich lesen sollte? what believe-2sg you which book I read-inf should ‘Which book do you think I should read?’

Reis argues compellingly that the much debated construction in (1b) does not involve extraction out of a verb second clause but parenthesis of the part glaubst du ‘do you think’ in preverbal position. Using a battery of diagnostic tests, Reis then considers the question whether the German partial wh-movement constructions illustrated in (1c) involve the syntax of parenthesis or the syntax of extraction and concludes that they share properties of both. Jeroen van Craenenbroeck discusses the properties of the complementizer van ‘of’ in Brabant Dutch spoken in Belgium. Van induces a spatial dissociation reading when used as a preposition, and an aspectual/evidential dissociation reading when used as a complementizer. The author shows that in other dialects of (southern) Dutch, van has grammaticalized into a semantically neutral complementizer, a situation comparable to Romance de/di. The article adopts a Kaynean analysis of prepositional complementizers as being merged not with a complement clause but with a constituent containing the matrix verb (cf. Kayne 1999). The complementizer system of Danish relative clauses is discussed by Line Mikkelsen. She considers the status of the complementizers som, at, and der within a minimal structure of the clause involving a single CP- and IP-level. Mikkelsen argues that som is an invariant operator moving to SpecCP through SpecIP, and that der is an expletive in SpecIP. Both som and der satisfy the Extended Projection Principle in Danish relative constructions. At is a complementizer in C. The analysis explains the fixed order of elements, as well as various other observations about Danish relative clauses. The appearance of various types of complementizers has given rise to analyses involving multiple layers of CP in the past (Vikner 1991, Hoekstra 1993a). Mikkelsen’s contribution is an interesting attempt to account for these phenomena from a more parsimonious view on phrase structure. The chapter on subordination is concluded by Susi Wurmbrand’s article on restructuring and control in Germanic. Constructions of obligatory control in German typically give rise to restructuring, evidenced by verb clustering and object shift into the matrix clause. Wurmbrand argues that under these circumstances, the interpretation of the understood embedded subject is determined as part of the meaning of the infinitive-taking verb, and that no syntactic PRO subject needs to be projected. Hence, only in constructions of nonobligatory control is the projection of a PRO subject required.

Introduction

The second chapter of the book deals with issues of movement and morphology in the Germanic languages. The first two articles present revisions of well-studied phenomena, adverb/object placement in English and verb second in Scandinavian, exploiting the process of collective movement (pied piping) also proposed in Koopman and Szabolcsi (2000) and Koster (2000). Roland Hinterhölzl argues for an analysis of the order of objects and adverbs in English that involves both VP-movement and covert object shift. Øystein Nilsen tackles the problem of violations of the verb second constraint in Mainland Scandinavian involving focus particles; his solution requires that verb movement be regarded as phrasal movement, suggesting that parametric variation in this domain resides in the size of the constituent moved. A more traditional approach to verb second is defended in Olaf Koeneman’s article on the distribution of declarative verb second in the Germanic languages. Koeneman, adopting a flexible approach to phrase structure, argues that verb second serves the purpose of projecting the relevant features (agreement, tense) at the clausal level. Projection of the tense features could be performed by the complementizer as well, yielding absence of verb movement in embedded clauses in certain languages, but languages with rich agreement need to project these features independently, yielding embedded verb movement. Koeneman’s analysis touches on an issue which continues to spark controversy among syntacticians, namely the question whether movement is determined by ‘richness’ of the morphology of the moved element or not. The two remaining articles in this chapter cast doubt on the hypothesis that richness of morphology matters. Øystein Alexander Vangsnes emphasizes the parallel behavior of finite and nonfinite verbs in Mainland Scandinavian languages. Vangsnes does relate movement of infinitivals to ‘agreement’, but the presence of the Agreement head is not a function of morphology, but of semantics, where AgrS is taken to yield an event interpretation independent of temporal anchoring. Artemis Alexiadou and Gisbert Fanselow adress the issue of the correlation of movement and richness of morphology head on, arguing that such a correlation is neither conceptually nor empirically justifiable. Their article concludes with a sketch of tendencies in language change which conspire to yield the impression that in some language families, such as Germanic, the connection between morphology and movement is real. With this historical dimension, the latter article straddles the boundaries between the second and the third chapter, which presents two articles on historical Germanic syntax. Eric Haeberli’s article records in detail the loss of

ix

x

Introduction

subject–verb inversion in English, and shows that inversion in clauses with a fronted non-operator and a full noun phrase subject is already quite frequent in Old English. Eric Hoekstra considers the question of how various feature checking mechanisms fare in language contact situations, and argues on the basis of Frisian contact dialects that imperfect language acquisition involves the projection of contextual checking mechanisms from the first language into the second language, provided that the inherent checking mechanisms from the second language allow this. The fourth and final chapter presents three contributions on the syntaxsemantics interface from the perspective of Germanic. Jason Merchant dusts off a little studied subcase of sluicing involving prepositions, illustrated in (2), which he calls swiping: (2) Peter went to the movies, but I don’t know who with

His analysis has interesting and potentially far reaching consequences, as it leads to the conclusion that head movement is a PF-phenomenon. This is because swiping involves head movement of a wh-element to a preposition, and sluicing is derived by deletion at PF. The article tracks the occurrence of swiping across Germanic. Markus Steinbach argues for a unified analysis of middles and related constructions in English and German, focusing on the morphosyntactic properties of weak reflexive pronouns, which are present in German, overtly, and in English, covertly. Crucially, weak reflexive pronouns can, but need not, be linked to a semantic argument of the verb. The fact that the weak reflexive is covert in English explains the ambiguity of sentences like Peter shaves (i.e. himself or in general). Finally, Malte Zimmermann proposes a unified analysis of binominal each constructions in German and English, of which the similarities and differences are listed. In Zimmermann’s analysis, the binominal element is the predicate of a prepositional Small Clause, which may be reanalysed with the verb into a complex predicate in German, but not in English. Zimmermann shows that the syntactic and semantic properties of binominal each may be derived without recourse to LF-movement of the each-phrase. The editors regret that not all papers presented at the 15th CGSW were submitted for publication in these proceedings. Not included are “(Null) pronouns and (null) agreement”, by Marcel den Dikken, “V2 languages”, by Anders Holmberg, “Toward an understanding of Condition C effects”, by Richard Kayne, and “Symmetry and antisymmetry in Germanic syntax”, by Jan Koster.

Introduction

Finally, we would like to thank Susanne Voorn for invaluable editorial assistance.

Note 1. These conference reports are accessible through the history of CGSW web page at http:// www.let.rug.nl/~zwart/cgsw/history.htm

xi

List of contributors

Artemis Alexiadou Institute of Linguistics University of Potsdam Postfach 601553 14415 Potsdam Germany Jeroen van Craenenbroeck Faculty of Arts Leiden University P. O. Box 9515 2300 RA Leiden The Netherlands Gisbert Fanselow Institute of Linguistics University of Potsdam Postfach 601553 14415 Potsdam Germany Eric Haeberli School of Linguistics and Applied Language Studies Room 185 Faculty of Letters and Social Sciences University of Reading Whiteknights Reading RG6 6AA United Kingdom

Roland Hinterhölzl Institut für deutsche Sprache und Linguistik Humboldt University Unter den Linden 6 10099 Berlin Germany Eric Hoekstra Frisian Academy P. O. Box 54 8900 AB Leeuwarden The Netherlands Olaf Koeneman Faculty of Arts University of Groningen P. O. Box 716 9700 AS Groningen The Netherlands Jason Merchant Department of Linguistics University of Chicago 1010 East 59th Street Chicago IL 60637 United States

xiv List of contributors

Line Mikkelsen Department of Linguistics University of California at Santa Cruz Stevenson C. 1156 High Street Santa Cruz CA 95064 United States Øystein Nilsen Utrecht Institute of Linguistics OTS Utrecht University Trans 10 3512 JK Utrecht The Netherlands Marga Reis Deutsches Seminar University of Tübingen Wilhelmstrasse 50 D-72074 Tübingen Germany

Markus Steinbach Deutsches Institut FB 13 Johannes-Gutenberg-Universität 55099 Mainz Germany Øystein Alexander Vangsnes Department of Scandinavian Studies University of Tromsø N 9037 Tromsø Norway Susi Wurmbrand Department of Linguistics McGill University 1001 Sherbrooke Street West Montréal Québec H3A 1G5 Canada Malte Zimmermann Faculty of Arts University of Amsterdam Spuistraat 210 1012 VT Amsterdam

Subordination

Wh-movement and integrated parenthetical constructions* Marga Reis University of Tübingen

1.

The problem

This article is about the relation between two kinds of complex ±wh-constructions that cooccur in practically all Germanic languages: ±wh-constructions involving wh-movement from complements ( = extraction constructions, ECs) on the one hand, and ±wh-constructions containing ‘Integrated Parentheticals’ ( = integrated parenthetical constructions, IPCs) on the other. Pertinent examples from German (G) and English (E) are given in (1)–(2). (1) a.

EC (−wh):

b. EC (+wh): (2) a.

G E G E

IPC (−wh): G E a¢. G E b. IPC (+wh): G E b¢. G E

Dieses Buch glaube ich, daß Max lesen sollte. This book I think (that) Max should read. Was glaubst du, daß Max lesen sollte? What do you think (that) Max should read? Dieses Buch sollte Max lesen, glaube ich. This book Max should read, I think. Dieses Buch sollte Max glaube ich lesen. This book I think Max should read. Was sollte Max lesen, glaubst du? What should he read, do you think? Was sollte Max glaubst du, lesen? What do you think should he read?

What ECs and IPCs have in common is that their component clauses are connected by an s-selectional relationship: in one clause there is a verb of saying, thinking, believing requiring a propositional argument, which is perceived as satisfied by the other clause. As a consequence, these constructions are very close in meaning and in use. In many contexts, ECs like (1a) and IPCs

4

Marga Reis

like (2a,a´) can be used to perform the same assertion, likewise ECs like (1b) and IPCs like (2b,b¢) can be used to ask the same question. Nonetheless, there seem to be clear differences in structure, roughly exemplified for (1b), (2b¢) in (3). (3) a.

EC structure for (1b) [Whati do you think [t¢i (that) he should read ti]] [Wasi glaubst du [t¢i daß er ti lesen sollte]] b. IPC structure for (2b¢) [Whati [do you think] should he read ti] [Wasi sollte er [glaubst du] ti lesen]

While ECs come about by extraction of a ±wh-phrase from the complement into the matrix clause, IPCs look more like structures where one component clause is parenthetically inserted into the other, be it in medial or peripheral position. Moreover, wh-movement clearly and centrally belongs to sentence grammar, whereas parenthetical insertion has features — for example its being root clause bound — that seem to relegate it to discourse grammar. So at first glance it seems neither surprising nor unjustified that in the many syntactic studies on wh-movement constructions, IPCs have hardly been mentioned let alone systematically been taken into account. In this article, I should like to show that this neglect of IPCs is harmful, even if you are only interested in the syntax of wh-movement constructions. In doing this I will mainly report on work I have done in the last few years on parenthetical and wh-movement constructions in German. I will proceed as follows. In Section 2, the defining properties of integrated parenthetical constructions will be sketched, this will be done by comparing them with their unintegrated parenthetical next of kin. In Section 3, IPCs like (2a,b) will be systematically compared with their EC counterparts (1a,a´,b,b´); the result is a battery of structural and interpretive differences that can be used as a diagnostics for IPC vs. EC status and allow some speculations on the structure of IPCs. I will then turn to two German wh-constructions in which according to standard wisdom long wh-movement is crucially involved: (i) so-called extractions from V2-clauses like (4) (= ‘EV2-constructions’), (ii) partial wh-movement constructions like (5), to which was-IP-counterparts also exist (6).1 (4) a.

Dieses Buch glaube ich sollte Max lesen. ‘This book I think Max should read.’ b. Was glaubst du sollte Max lesen? ‘What do you think Max should read?’

Wh-movement and integrated parenthetical constructions

(5) Was glaubst du, welches Buch Max lesen sollte? what believe you which book Max read should (6) a.

Welches Buch sollte Max lesen, was glaubst du? ‘Which book should Max read, what do you think?’ b. Was glaubst du, welches Buch sollte Max lesen? ‘What do you think (:) which book should Max read?’

Armed with the diagnostics developed before, I will show with respect to EV2-constructions that (G1) holds (Section 4), and with respect to partial wh-movement constructions that (G2) holds (Section 5). (G1) (G2)

German EV2-constructions are not ECs but prefinite IPCs. German partial wh-movement constructions have salient EC features and salient IP-features (too many to be dismissed as accidental).

A number of theoretical and comparative implications of these results will be addressed as we go along. Section 6 contains a short summary.

2. Delimiting ‘Integrated’ Parenthetical Constructions (IPCs)2 2.1 was-cases: Unintegrated vs. integrated constructions In delimiting IPCs, let us begin with was-cases where the distinction between unintegrated and integrated cases shows up most clearly. 2.1.1 Unintegrated was-constructions In the typical case, these are so-called ‘sequential questions’ as in (7), i.e. paratactic sequences consisting of a was-clause and a further interrogative clause (‘+wh-clause’), usually a wh-interrogative as in (7a) or a yes/no interrogative as in (7b).3 (7) a.

Was glaubst DU (\): Wann ist GOEthe geboren? what believe you when is Goethe born ‘What’s your guess: What’s Goethe’s birthday?’ b. Wird CLEment gewählt? (\) Was GLAUBst du? ‘Will Clement be elected? What do you believe?’

Their defining formal feature is that the clauses in question are ‘unintegrated’, that is form completely autonomous domains vis-à-vis each other: syntactically, both clauses are full-fledged interrogatives, prosodically, each clause has a

5

6

Marga Reis

Focus-Background Structure (FBS) and an intonation contour of its own (which is, in the case of the was-clauses, either colon intonation or the falling intonation characteristic for wh-interrogatives). Interpretively, the two clauses are understood to be in an ‘indirect dependency’ relation à la Dayal (1994 and later work): the set of possible answers to the was-clause is restricted by the set of possible answers to the related +wh-clause, (in other words, the +wh-clause functions as the restriction on the propositional variable represented by was). Sequential questions are restricted to root position. As a consequence, both clauses making up the sequence have the syntactic form of main clauses and exhibit all the pragmatic signs of root clause status: they have independent and to some degree variable ‘interrogative’ (actually, ‘erotetic’) force (yes/no-/wh- in (7b); yes/no-/deliberative was-clause in (8a)), and they may contain modal particles (8a,b).4 (8) a.

Wird Clement gewählt? Was wohl Angela glaubt? becomes Clement elected what MP Angela believes ‘Will Clement be elected? What does Angela think, I wonder?’ b. Was glaubst du denn nun: Wird Clement gewählt oder nicht? what believe you MP now becomes Clement elected or not ‘Now, what do you think: will Clement be elected or won’t he?’

Moreover, was-clauses in unintegrated constructions may be syntactically complex as in (9a,b), allow more than the typical predicates of thinking, believing and saying (9a,c), and may even contain multiple wh-phrases (9d). (9) a.

Wann ist Goethe geboren? Was scheint Dir richtig zu sein? when is Goethe born what seems you-dat right to be ‘What’s Goethe’s birthday? Which date seems right to you?’ b. Wann ist Goethe geboren? Was glaubst du, daß Peter glaubt? ‘What’s Goethe’s birthday? What do you think that Peter thinks?’ c. Wer gewinnt das Rennen, Karl oder Tim? Was fändest Du who wins the race, Karl or Tim what found-subj you besser? better ‘Who will win the race, Karl or Tim? What would you prefer?’ d. Wann ist Adorno geboren? Was glaubt wer in dieser Runde? when is Adorno born what believes who in this round ‘What’s Adorno’s birthday? Who believes what in this group?’

Note finally that there are also marginal cases where the was-clause is inserted into the related wh-clause, without however losing its unintegrated features.


(10) WoHIN (\), was glaubst DU (\), ist er gegangen? where-to what believe you is he gone ‘What do you think: Where did he go?’

The findings about unintegrated was-cases are summarized in (11). (11) Major properties of unintegrated was-constructions. (i) paratactic combination of was-clause + related interrogative clause (of any type) (ii) occurrence in root position only (iii) position of was-clause: initial, final (??medial) (iv) interpretive relation: related +wh-clause acts as restriction on was-clause (v) mutual prosodic independence (intonation break, independent FBS) (vi) mutual pragmatic independence (w.r.t. force, pragmatic modifiers (MPs), etc.) (i.e. syntactic, prosodic, pragmatic root clause characteristics in both clauses)

2.1.2 Integrated was-constructions Let us now turn to integrated was-constructions. Typical examples are given in (12).5 (13) lists their major properties. (12) a.

Wohin ist er geGANgen, was glaubst du? where-to is he gone what believe you a¢. Was glaubst du, wohin ist er geGANgen? what believe you where-to is he gone a≤. ??Wohin was glaubst du, ist er geGANgen? where-to what believe you is he gone ‘Where do you think did he go?’ b. ?Wird er morgen KOMmen, was glaubst du? will he tomorrow come what believe you b¢. Was glaubst du, wird er morgen KOMmen? what believe you will he tomorrow come b≤.??Wird er was glaubst du, morgen KOMmen? will he what believe you tomorrow come ‘Will he come tomorrow do you think?’

7

8

Marga Reis

(13) Major properties of integrated was-constructions ( = was-IP-constructions/was-IPCs) (i) paratactic combination of was-clause + related interrogative clause (of any type) (ii) occurrence in root position only (iii) position of was-clause: initial [best], final, (?medial) (iv) interpretive relation: related +wh-clause acts as restriction on was-clause (v) prosodic autonomy of was-clause much reduced: a. no really explicit comma intonation b. integration into the FBS of the related +wh-clause c. no focus exponent/main accent in the was-clause (vi) pragmatic autonomy of was-clause much reduced (w.r.t. force, pragmatic modifiers, etc.) =(v,vi) fi prosodic+pragmatic root clause characteristics confined to the related +wh-clause fi status distinction: was-clause = integrated parenthetical clause (IP): was-IP related +wh-clause = the main clause of the IP-construction, hosting the IP. (vii) further restrictions on the was-IP: a. specific selectional restrictions on IP-predicates spelled out in (16) b. no syntactic complexity/only (restricted) IP iteration c. subject restrictions and other idiosyncrasies (see Section 3.1)

Obviously, there is no syntactic surface difference vis-à-vis their unintegrated counterparts; cf. (11i–iii)/(13i–iii). Likewise, the relevant interpretive relation between was and the related wh-clause is the same; cf. (11iv)/(13iv). There are, however, major prosodic differences. Unlike unintegrated was-constructions, parenthetical was-constructions are prosodically integrated in the sense of (13v) as illustrated in (14), (14) a. *Was glaubst DU, wohin ist er gegangen? what believe you where-to is he gone b. *Wohin ist er gegangen, WAS glaubt WER? where-to is he gone, what believes who

which has the effects summed up in (13vi) as loss of pragmatic autonomy of the was-clause.6 First, due to (vib,c), the was-clause is always less prominent than


the related wh-clause, that is, its communicative weight is roughly equivalent to the adverbial in sentences like Wohin ist er Deiner Meinung nach gegangen (‘In your opinion, where did he go?’) rather than to the was-clause in unintegrated constructions. Second, while the was-clause forces (via was) the sentence mood of the whole to be semantically interrogative, it has no communicative force of its own: the illocutionary force of the host clause always prevails (thus ja/nein ‘yes/no’ are clearly felicitous answers to all questions in (12b)). Likewise modal particles and other communicative modifiers may appear in the host clause only:7 (15) a. *Was glaubst du denn/eigentlich, wohin ist er gegangen? what believe you MP/MP where-to is he gone b. *Wohin was glaubst du denn/eigentlich, ist er gegangen? where-to what believe you MP/MP is he gone

What this amounts to, in effect, is that the function of the was-clause is reduced to putting the host clause proposition into the respective attitudinal perspective of the was-clause subject. In other words, as spelled out below (13vi), the related +wh-clause is the main clause in integrated constructions, hosting the was-clause as an integrated parenthetical, a ‘was-IP’. The parenthetical use-value just described is the primary, but not the only diagnostic property of IPs correlating with prosodic integration. Further correlating features are listed in (13viia–c), of which I will just illustrate (13viia,b). Concerning (13viia), the selectional restrictions on was-IP-predicates are stated in (16). (16) Was-IP-predicates: (i) always select a propositional argument, which is lexically specified as a finite −wh-sentential argument in structural object position (ii) include (nonnegative/unnegated) verbs of saying, thinking, believing (iii) do not include – preference predicates – (strong) factive predicates – negative/negated predicates – adjectival predicates in general

Thus, conforming to (16ii), the verbs appearing in was-IPs are just simple verbs of saying, thinking, believing (sagen ‘say’; glauben, denken, meinen ‘think’, ‘believe’; schätzen ‘guess’). All other verbs, in particular those listed in (16iii), are disallowed; cf. (17)–(19).

9

10

Marga Reis

(17) preference predicates besser/das beste sein ‘be better/best’, jem. lieber sein ‘be preferable for sb.’, optative wollen/wünschen/möchte ‘wish’, vorziehen ‘prefer’, etc. a. *Was ist besser, wohin geht er zu Fuß? what is better where-to goes he on foot b. *Soll er zu Fuß gehen oder nicht, was ist besser? should he on foot go or not what is better c. *Was wollte/wünschte Hans, wohin wäre er zu Fuß gegangen? what wished Hans where-to were-subj he on foot gone (18) (strong) factive predicates bedauern ‘regret’, berücksichtigen ‘take into account’, sich entsetzen ‘be appalled’, jem. zürnen ‘be angry with sb.’, etc. a. *Was bedauerte sie, wohin ging Hans? what regretted she where-to went Hans b. *Ging Hans zu Fuß, was berücksichtigte sie? went Hans on foot what took-into-account she (19) negated predicates by nicht ‘not’, keineswegs ‘by no means’, kein- ‘nobody’, etc. negative predicates bezweifeln ‘doubt’, verbieten ‘forbid’, vergessen ‘forget’, verheimlichen ‘keep (it) a secret’, etc. a. *Was glaubt keiner, mit wem ist sie verheiratet? what believes nobody to whom is she married b. *Mit wem ist sie verheiratet, was bezweifelst du? to whom is she married what doubt you

In addition, was-IP-predicates are subject to the general categorial restriction under (16iii): all adjectival predicates seem to be excluded, even if they have the right semantics; cf. (20). (20) *Was ist klar /für dich wahrscheinlich, wohin geht er zu Fuß? what is clear for you likely where goes he on foot

It is easy to check that at least some of these restrictions do not hold for unintegrated was-constructions; cf. for example (9a,c) above. As for (13viib), (21) illustrates that was-IPs tend to be syntactically simple. (21) *Was glaubt sie (daß) er meint, was wird er morgen tun? what believes she (that he thinks what will he tomorrow do

In particular, was-IPs containing a finite complement clause are unacceptable.


All we meet with are cases like (22), which however will turn out to be combinations of was-IPs with a further type of IPs: V1-IPs. (22) Welchen Auftrag wird er erst morgen erledigen, was meint er which job will he only tomorrow do what thinks he glaubst du? believe you ‘Which job does he think you believe, will he postpone till tomorrow?’

Again, unintegrated was-constructions did allow complex was-clauses; cf. (9b,c) above. 2.2 Other integrated parenthetical constructions There are three further types of IPCs: constructions involving wie-IPs (23), so-IPs (24), and the IPs already illustrated in (2), which have verb-initial order, and will henceforth be called VIPs (25). Wie sie glaubt/sagt, muß in der Partei vieles anders werden. as she believes/says must in the party much other become b. In der Partei muß, wie sie glaubt/sagt, vieles anders werden. c. In der Partei muß vieles anders werden, wie sie glaubt/sagt. ‘As she believes/says, a lot must change in the party.’

(23) a.

In der Partei, so glaubt/sagt sie, muß vieles anders werden. in the party so believes/says she must much other become b. In der Partei muß vieles anders werden, so glaubt/sagt sie. ‘A lot must change in the party, so she believes/says.’

(24) a.

(25) a. a¢. b. b¢. c. c¢.

(−wh): Dieses Buch sollte Max lesen, glaube ich/sagt er. ‘This book Max should read, I think/he says.’ Dieses Buch sollte Max glaube ich/sagt er, lesen. ‘This book I think/he says Max should read.’ (+wh): Was sollte Max lesen, glaubst du/sagt er? ‘What should Max read, do you think/does he say?’ Was sollte Max glaubst du/sagtest du, lesen? ‘What do you think/did you say should Max read?’ (+whyes/no): Soll Max diesen Roman lesen, glaubst du/sagt er? ?Soll Max glaubst du/sagt er, diesen Roman lesen? ‘Should Max read this book do you think/does he say?’

Note that these IPCs may have two readings: (i) a reported speech reading (= ‘subject-oriented’ reading according to Reinhart 1983a), requiring a verbum

11

12

Marga Reis

dicendi reading of the IP-predicate; (ii) a true integrated reading (= Reinhart’s ‘speaker-oriented’ reading), requiring a content reading of the IP-predicate. For my purposes, the only interesting reading is (ii), which is shared by the extraction counterparts of IPCs (and, remarkably, also the only one present in was-IPCs); as a consequence, the examples in the following will be restricted to verbs like glauben where a verbum dicendi reading is hard to get.8 Note, moreover, that all these IPCs have unintegrated counterparts (that is the so-/wie-/V1-clause is prosodically set off from the related clause); with so-/wie-constructions this may even be the preferred option.9 It is easy to check that these cases behave just like unintegrated was-constructions in all relevant respects as listed in (11). In turn, the integrated versions share the characteristic properties of integrated was-constructions. In showing this, I will concentrate on VIP-constructions, which come closest to normal ECs like (1) and are crucial for the argument in Section 4.10 To be sure, there are some differences beween VIPs and was-IPs. VIPs appear in declarative and interrogative host clauses alike, they must not express their object argument (which seems to force them more or less to be integrated), and they have different placement preferences. Otherwise, however, their behavior seems to be completely parallel: (26) Major properties of integrated V1 parenthetical constructions ( = VIP-constructions). (i) paratactic combination of V1-clause + related declarative or interrogative clause (of any type) (ii) occurrence in root position only (iii) position of V1-clause: medial, final (as to pre-final position see Section 4.) (iv) interpretive relation: the related clause specifies the object argument of the IP-verb unexpressed in the VIP (identifying it in −wh-IPCs, restricting it in +wh-IPCs, i.e. +wh-VIPCs and was-IPCs have the same interpretation, see Steinbach 1999b) (v) prosodic autonomy of VIP much reduced: a. no really explicit comma intonation b. integration into the FBS of the related clause c. no focus exponent/main accent in the VIP (vi) pragmatic autonomy of VIP-clause much reduced (w.r.t. force, pragmatic modifiers, etc.)


=(v,vi)fi

prosodic and pragmatic root clause characteristics are confined to the related ±wh-clause fi status distinction: V1-clause = integrated parenthetical clause (IP) = VIP related ±wh-clause = the main clause of the IPC, hosting the VIP. (vii) further restrictions on VIPs: a. the selectional restrictions on IP-predicates spelled out in (16) b. no syntactic complexity/only (restricted) IP iteration c. subject restrictions and other idiosyncrasies (see Section 3.)

For reasons of space, I will confine myself to just illustrating the most salient signs of IP-status. (26va–c) and its pragmatic consequences are illustrated in (27)–(28); that (26viia,b) hold, is shown by (29) and (30) respectively. (27) a. *Wohin ist er — glaubst du — geGANgen? b. *Wohin ist er glaubst DU geGANgen? c. *Wohin ist er glaubst DU gegangen? where-to is he believe you gone

(— = comma intonation) (no comma intonation) (no comma intonation)

(28) a. *Wohin ist er gegangen, glaubst du denn/eigentlich? where-to is he gone, believe you MP/MP b. *Nach Prag ist er gegangen, glaubt doch/eigentlich Peter. to Prague is he gone, believes MP/MP Peter preference predicates *Wohin geht er ist besser, zu Fuß? where-to goes he is better on foot *Dahin wäre sie zu Fuß gegangen, wollte/wünschte Hans. there-to were-subj she on foot gone wished Hans b. (strong) factive predicates *Hans ging bedauerte sie/berücksichtigte sie, zu Fuß. Hans went regretted she/took-into-account she on foot c. negated/negative predicates *Mit wem ist sie glaubt keiner/bezweifelst du, seit 1980 with whom is she believes nobody/doubt you since 1980 verheiratet? married d. adjectival restriction *Dahin ist klar/für mich wahrscheinlich, geht er zu Fuß. there-to is clear/for me likely goes he on foot

(29) a.

13

14

Marga Reis

(30) a. *Was wird er glaubt sie (daß) er meint, morgen tun? what will he believes she (that he thinks tomorrow do b. Diesen Auftrag wird er erst morgen erledigen meint er this job will he not-before tomorrow do thinks he glaubt sie. believes she ‘This job he thinks she believes, will he postpone till tomorrow.’

There is one apparent difference concerning (26viia). Was-IPs do not tolerate predicates containing es or full object NPs, which are fine or much better in VIPs; cf. (31) vs. (32). (31) *Was scheint es / hat sie das Gefühl, what seems it has she the feeling wo wird er sich endgültig niederlassen? where will he himself finally down-settle Wo wird er scheint es / ?hat sie das Gefühl, sich endgültig ? where will he seems it has she the feeling himself finally niederlassen? down-settle ‘Where does it seem/does she feel, will he settle down for good?’ b. Dort wird er scheint es / ?hat sie das Gefühl, sich endgültig there will he seems it ?has she the feeling himself finally niederlassen. down-settle ‘There it seems/she feels, he will settle down for good.’

(32) a.

This, however, is clearly due to the presence of was, whose occurrence must be licensed by the argument grid of the IP-predicate — which is not the case in (31); cf. the unacceptability of bare questions like *Was scheint es?/*Was hat er das Gefühl? (See also 3.2 below.) In sum, ceteris paribus there are clear parallels between was-IPs and VIPs, which, as could be easily shown, also extend to so-IPs and wie-IPs. This suggests strongly that IPCs be treated as a unitary phenomenon. The correlating (bundles of) grammatical and pragmatic properties define IPs/IPCs as a grammatical class and are distinctive vis-à-vis their unintegrated counterparts.


3. Integrated parenthetical constructions vs. extraction constructions 3.1 Diagnostic differences Let us now quickly compare the relevant properties of IPCs and ECs. For obvious reasons (see Sections 4, 5), comparison will be confined to VIP-constructions vs. ECs involving daß-clauses, matching the EC matrix with the IP clause and the EC complement with the IP host clause (the former being the ‘selecting clause’, the latter the ‘selected clause’). This given, let us see how ECs behave with respect to the characteristic IPC profile (26). The overlap turns out to be slight. ECs share with IPCs just the interpretive property (iv) and the integrational prosodic properties (va,b). Everything else is different; cf. the list of EC properties (33) and the examples (34)–(39) illustrating the respective differences (i)–(viib). (33) Major properties of extraction constructions ( = w…daß-constructions): (i) embedding construction (the selected clause always has embedded form) (34) (ii) occurrence in root and arbitrary embedded positions (35) (iii) position of selecting clause: fixed (in particular: no insertion into selected clause) (iv) the selected clause specifies the object argument of the verb in the selecting clause (v) prosodic properties a. no really explicit comma intonation b. integration into one FBS c. focus exponent/main accent may be in the selecting clause (36) (vi) if in root position, the selecting clause is the main clause of ECs (i.e. the carrier of illocutionary force, pragmatic modifiers, etc.) (37) (vii) w.r.t. further restrictions: a. restrictions on EC bridge predicates different from/weaker than restrictions on IP predicates (ECs admit factive, negative, preference, adjectival bridge predicates) (38) b. the structure embedding the selected clause may be infinitely complex (‘unboundedness of movement’) (39) c. no comparable subject restrictions, etc. (40)–(42) (34) Wohin glaubt Sue, daß Tom gegangen ist? where-to believes Sue that Tom gone is ‘Where does Sue think that Tom went?’

15

16

Marga Reis

(35) Peter weiß, was Fritz denkt … wohin Sue glaubt, daß Tom Peter knows what Fritz thinks where-to Sue believes that Tom gegangen ist. gone is ‘Peter knows what Fritz thinks … where Sue thinks that Tom went.’ (36) a.

Wohin glaubt nur SIE, daß er gegangen ist? where-to believes only she that he gone is ‘Where does only she believe that he went?’ b. Nach Prag glaubt sie, daß nur Fritz gegangen ist. to Prague believes she that only Fritz gone is ‘To Prague she believes that nobody went but Fritz.’

(37) a.

Wohin glaubt sie denn/eigentlich, daß er gegangen ist? where-to believes she MP/MP that he gone is ‘Where does she believe he went?’ b. *Wohin glaubt sie daß er denn/eigentlich gegangen ist? where-to believes she that he MP/MP gone is

(38) a. ?Wen begrüßt Fritz / ist Fritz dankbar, daß seine Tochter whom welcomes Fritz is Fritz thankful that his daughter heiratet? marries ‘Who is Fritz pleased/thankful that his daughter will get married to?’ b. ?Wen kann Fritz nicht glauben / bezweifelt er, daß seine Tochter whom can F. not believe doubts he that his daughter heiratet? marries ‘Who can’t Fritz believe/does he doubt that his daughter will get married to?’ c. Wen möchte / will Fritz, daß seine Tochter heiratet? whom would like wants Fritz that his daughter marries ‘Who would Fritz want his daughter to get married to?’ c¢. Wohin würde Fritz vorziehen / wäre es besser, daß (/wenn) where-to would Fritz prefer were-subj it better that (/if sie zu Fuß ginge? she on foot went-subj ‘Where would Fritz prefer/it be better for her to walk?’ ? d. Wen ist klar / wahrscheinlich, daß sie heiratet? whom is clear likely that she marries ‘Who is it clear/likely that she will get married to?’


(39) Wohin glaubt Sue, daß Paul wünscht, daß Uli annimmt, daß … daß where-to believes Sue that Paul wants that Uli assumes that that Tom geht. Tom goes ‘Where does Sue believe that Paul wants Uli to assume that … that Tom will go?’

Let me now turn to illustrating the distinctive behavior of IPCs vs. ECs with respect to properties (26/33viic), which has been postponed until now. Subject restrictions. ECs have no subject restrictions on their matrix clauses. By constrast, IP clauses are choosy; in particular, if hosted by interrogative clauses, they (i.e. VIPs as well as was-IPs) abhor first person subjects: (40) a.

EC:

b. VIP: c. was-IP:

[Na rat mal:] [dP guess MP:] ‘[Well guess:] [Na rat mal:] [Na rat mal:] [Na rat mal:]

Wen glaub ich, daß sie heiraten wird? Whom believe I that she marry will? Who do I believe (that) she will marry?’ *Wen wird sie glaub ich heiraten? *Wen wird sie heiraten, glaub ich? *Wen wird sie heiraten, was glaub ich?

Rhetorical readings of wh-questions. While ECs allow them without any problems, IPCs do not; this is again true for VIPs and was-IPs alike: (41) A: Ich weiß nicht, was ich tun soll. Ich muß mal Paul fragen. ‘I don’t know what to do. I think, I’ll ask Paul.’ B: a. EC: Was (schon) wird der sagen, daß du tun sollst — beten und arbeiten, das sagt er immer. what (MP-rhet.) will he say that you do shall — pray and work, that says he always ‘Come on, what he will say, is obvious — pray and work, that’s what he always says.’ b. VIP: *Was (schon) sollst du wird der sagen, tun — … *Was (schon) sollst du tun, wird der sagen — … c. was-IP:*Was wird der sagen, was (schon) sollst du tun — …

Differences in relative predicate scope. There are many more differences between ECs and IPCs along those lines, which I cannot dwell on here. (See Reis 2000b, where an attempt is made to also account for these differences on the basis of the distinctive properties (26i–vi) vs. (33i–vi).) So let us just take note of a further difference that also throws some light on the syntax of IPCs.

17

18

Marga Reis

In embedding constructions, relative predicate scope is determined by the embedding relation (42a). ECs are normal embedding constructions, so the prediction is that they behave the same way, and they do (42b). (42) Predicate scope in embedding constructions. a. … wenn sie glaubt, (daß) alle meinen, daß das Problem gelöst ist if she believes (that all think that the problem solved is a¢. … wenn sie glaubt, alle meinen, das Problem ist gelöst if she believes all think the problem is solved ‘…if she believes (that) everyone thinks (that) the problem is solved.’ [ 1 [ 2 [ 3 ]]] relative predicate scope: – 1 > [2 > [3]] * 2 > [1 > [3]] b. Das Problemi glaubt sie, ti daß alle meinen daß ti jetzt gelöst ist. the problem believes she that all think that now solved is ‘The problem, she believes that everyone thinks is now solved.’ [ 1 [ 2 [ 3 ]]] relative predicate scope: – 1 > [2 > [3]] * 2 > [1 > [3]]

But this prediction is not true for iterated IP-constructions, which are possible in case VIPs are involved — see above (21)–(22) —, no matter where the IPs are located: (43) Iterative IPCs. a. Das Problem wurde damals glaubt sie meinen alle gut gelöst. the problem was then believes she think all well solved b. Das Problem wurde damals gut gelöst, glaubt sie meinen alle. the problem was then well solved believes she think all ‘The problem was, she believes everybody thinks, solved well at the time.’ ‘The problem was, everybody thinks she believes, solved well at the time.’ relative predicate scope: – 1 > [2 > [3]] – 2 > [1 > [3]]

(43) shows that the second IP may have scope over the first one, which is incompatible (i) with (43) having the same hierarchical structure as (41)–(42), (ii) with linearity being the factor determining relative predicate scope in IPCs. This suggests that iterated IPs do not form one complex parenthetical à la (44a), but rather are inserted into their host clause one by one, in other words, that


every simple IP is directly related to its host clause, no matter whether or not it already contains another IP as it does in the iterated cases (44b,b¢). (44) Structural options for iterated VIPs. a. Das Problem wurde damals [glaubt sie meinen alle] gut gelöst. b. Das Problem wurde damals [glaubt sie] meinen alle gut gelöst. b¢. Das Problem wurde damals glaubt sie [meinen alle] gut gelöst.

This way, sentences containing iterated IPs are always structurally ambiguous, and this ambiguity translates directly into a scope ambiguity, unless a nonstructural factor resolving it — for example subjunctive in one of the IPs signalling dependency — intervenes.11 These data then seem to provide evidence that IPC formation is not just paratactic concatenation of clauses as we know it from sentence grammar, but an insertion operation of one complete syntactic object into another one operating on a different level (where hierarchical structure can be neither built up nor respected).12 Be this as it may, the iterative data clearly provide a further diagnostic difference between ECs and IPCs. 3.2 Some remarks on the structure of IPCs (vs. ECs) Before going on let me add a few remarks on two questions: (i) What is the structure of IPs and IPCs in German? (ii) Do the facts about IPs and IPCs vs. ECs as stated for German hold in general, in particular for the other Germanic languages? In view of what follows in Sections 4, 5, I will concentrate on VIP(C)s and was-IP(C)s. Concerning (i), the central fact to be accounted for is the perceived argument satisfaction relation between the IP-predicate and the host clause. For was-IPs this is straightforward: the relation is mediated by a canonically licensed IP-argument — was — that is in turn specified by the host clause. But VIPs must not contain overt object elements to which the host clause could be suitably related. This leaves in principle three options: (a) the host clause originates in fact in structural object position, i.e. as a subordinate to the VIP; (b) there are non-overt mediating VIP arguments; (c) (V)IPCs represent cases of non-canonical argument licensing. Option (a) — which inevitably leads to deriving the paratactic structure of IPCs by some kind of Slifting analysis à la Ross (1973), plus subsequent ‘Niching’

19

20

Marga Reis

in the case of medial IP positions — can be rejected out of hand. There are three strong arguments against it: –

The analysis covers at best VIPCs involving V2-host clauses, but not the many cases of VIPCs involving +wh-host clauses, nor occurrences of VIPs in assertional V-final clauses (i.e. pragmatic root clauses) like (45) (cf. Reis 1995: 64), for neither clause type can be selected by IP-predicates. (45) a.

Peter, der glaube ich gestern Geburtstag hatte, … Peter who believe I yesterday birthday had ‘Peter whose birthday was yesterday I think, …’ b. [Er erreicht viel,] obwohl er wenig arbeitet, scheint mir … [he achieves much although he little works seems me-dat ‘[He achieves a lot,] although he doesn’t work much it seems to me.’ c. [Er ist ernstlich krank,] weshalb er glaubt sie bald zurücktritt. [he is seriously ill why he believes she soon back-steps ‘[He is seriously ill,] for which reason he will resign soon, she thinks.’

–

–

Even in the V2-case, there are no clear binding data supporting an initial subordinate structure, and a number of clear indications against it.13 In particular, starting from such a structure the differences between IPCs and ECs (with which bona fide embedding V2-structures side) with respect to predicate and other restrictions as listed in (26vii) would be completely unexpected. Since these restrictions are shared by was-/so-/wie-IPs, VIPs should also share with them a common structural basis with which to correlate these restrictions, but since for was-/so-/wie-IPCs option (a) above is out of the question, (a) should not be an option for VIPCs either. Moreover, clearly, none of the other structural idiosyncrasies (root status, the iterative behavior described in Section 3.1) can be accounted for on this basis. If so, it seems better to take IPC structures for what they overtly are: paratactic insertion structures that come about by inserting one complete syntactic object into the other at some postsyntactic level, possible ‘niches’ being left and right periphery, and internal XP boundaries.14

While disposing of option (a) involved clarifying the external syntax of VIPs, in evaluating the remaining options (b)–(c) the internal syntax of VIPs comes in: (b) implies that VIPs are V2-structures with a covert preverbal element (46a); (c) implies that VIPs are, or could be, pure V1-structures (46b), and vice versa. (46) a.

V2-structure of VIP [Ø2 [glaubt1 [er t2 t1]]]

b.

V1-structure of VIP [glaubt1 [er t1]]


Neither alternative can be ruled out a priori, for German allows preverbal ‘topic drop’ in V2-structures (47a) as well as declarative V1-structures (47b). (47) a.

[Wie wäre es mit einem Drink? —] [how was-subj it with a drink Würde ich nicht ablehnen. would I not refuse ‘[How about a drink? —] I wouldn’t be opposed.’ b. Glaubt der Kerl mir einfach nicht. believes the guy me-dat simply not ‘The guy just doesn’t believe me.’

Moreover, there are enough cases of noncanonical argument licensing on record (cf. Reis 1995, 1996a, 1997) to make (c) a reasonable option. Either analysis, however, faces problems arising from the fact that VIPs occur in +wh- and −wh-host clauses alike, with a corresponding difference in meaning. As pointed out by Steinbach (1999b: 14ff.), VIPs in +wh clauses have to be interpreted just like was-IPs that is a restrictive relation to the host clause obtains, whereas with −wh-clauses the relation between VIP and host is one of coreference. In terms of option (b) this implies that the putative silent object element corresponds to a propositional variable in +wh-contexts, but to an anaphoric pronoun in −wh-contexts, — a silent was ‘what’ vs. das ‘that’ so to speak. But if so, ‘topic drop’ can at best be appealed to in the declarative −wh-cases (though whether the pragmatic conditions on topic drop are met remains questionable); for interrogative cases it is a mere stipulation, supported by no syntactic evidence whatever. In this sense, option (b) is more or less ad hoc. In this regard, option (c), viz. V1-structure of VIPs, is of course non-ad hoc. But if VIPs are the same self-contained V1-clauses throughout, there is no way to even represent the semantic diversity of +wh- vs. −wh-VIPCs. (Just marking VIP heads in +wh-clauses with +wh and in −wh-clauses with −wh will not help, for +wh-V1-clauses correspond to yes/no-interrogatives and not to +wh-interrogatives as semantically required). The only way I can think of to circumvent this problem and still maintain (c) is assuming VIPs to be ‘truncated’ structures without mood and topic/focus projections of their own which in order to be licensed must be hosted by a full-fledged CP structure having these projections. This way, the transparency of identical VIPs vis-à-vis the mood and topic/focus structures of the host might be made to follow quite naturally. In view of Rizzi’s (1997) split C-system, this looks like a promising way to go, but there is no space

21

22

Marga Reis

to pursue it here. Still, although the question of (V)IP structure thus remains open, there can be no doubt that the structure of IPCs and ECs are drastically different throughout their derivation. Turning now to the comparative question (ii), the present state of IP-research does not allow any definitive answers. Equivalents to was-IPs in other Germanic languages are practically nowhere mentioned in the literature, but judging from acquisition studies (Thornton 1990, van Kampen 1997) and informations by native speakers, they exist at least in Dutch and English. Their properties are so far unknown. Equivalents to VIPs are much better researched; nonetheless, there are next to no comprehensive systematic studies of VIPs or IPs in general for most Germanic languages, let alone studies in a comparative vein (a rare exception is Holmberg 1983). Moreover, the existing literature, which for English is considerable, is hard to evaluate, since the central parameters (prosodic integration, speaker oriented vs. subject oriented reading) are not always carefully controlled.15 To be sure, even when just comparing German VIPs with their English equivalents, there are many suggestive parallels (as can easily be seen from checking through Ross 1973, Reinhart 1983a or Espinal 1991), but the differences seem, at first glance, to be also considerable. In English, the class of IP predicates seems to be much wider (including elements like be sure, know, wonder, promise, not doubt; cf. Bolinger 1968: 5ff., Ross 1973: 136f.), there may be complex IPs, including even extraction structures (see inter alia Ross 1973: 135), there is the ‘negative concord’ phenomenon, which has no German parallel (48), likewise variation in VIP-form depending on the ±wh-status of the host clause (he thinks — does he think, etc.), and the insertion sites — pace Espinal (1991: 733f.) — seem to be parallel to sentence adverbial positions (which in German they definitely are not; cf. Reis 1995: 74ff.), etc. (48) a. E b. G

Peter won’t come, I don’t think/*I think. Peter wird nicht kommen, glaube ich/*glaube ich nicht.

But given the research situation just mentioned, the complexity of the data to be comparatively checked, plus our present ignorance of possible parametric variation regarding IP structures, any conclusion drawn from isolated parallels and differences would seem rather premature. So let me conclude this section by (a) stating the obvious. The grammar of IPs is in need of much more systematic and comparative study; and (b) insisting on the non-obvious. Even if IP-behavior in other Germanic languages, say English, were ultimately found to be significantly different from IP-behavior in German, this would not


necessarily impair the description of German IP(C)s given above. Rather, it might show that the cake between IP and EC types of complex ±wh-constructions is differently cut in different languages, with (V)IPCs in English being perhaps more EC like than in German. In view of what we will see in the following sections, this is by no means out of the question.

4. The analysis of German EV2-constructions16 4.1 The case for (G1) German EV2-constructions (EV2Cs) like (4), here repeated as (49), have standardly been given an ‘extraction analysis’ like (50a) since Thiersch (1978), which has figured in quite a few theoretical discussions and comparative issues. On my view, this analysis is wrong, see (G1), and should be replaced by an analysis of EV2Cs like (50b), that is as constructions containing a VIP in prefinite position. (49) Dieses Buch glaube ich sollte Max lesen. — Was glaubst du sollte Max lesen? ‘This book I think Max should read.’ ‘What do you think Max should read?’ (50) a.

[Dieses B. glaube ich [t¢ sollte M. t lesen]] [Was glaubst du [t¢ sollte M. t lesen]] b. [Dieses B. [glaube ich] sollte M. t lesen] [Was [glaubst du] sollte M. t lesen]

— —

The way to argue the case is, of course, to systematically compare EV2-constructions with bona fide (i.e. postfinite and final) VIP-constructions on the one hand and bona fide extraction constructions (i.e. ECs involving daß-complements) on the other. If this is done, the standard arguments pro extraction analysis — based on iterative constructions, binding data, subjunctive data, structural object relation between IP and related ±wh-clause — immediately fall apart: the alleged extraction properties of EV2-constructions are either shared by bona fide VIPCs or not shared by bona fide ECs or both.17 More interesting than demolishing the traditional argumentation, which was probably too sketchy to begin with, is going through the systematic checklist of distinctive properties worked out above. Do EV2Cs conform to the VIPC profile (26) or to the EC profile (33)? Let me quickly run through some salient diagnostic data.

23

24

Marga Reis

–Occurrence in ‘root position’ [(26ii) vs. (33ii)]. If we embed (49a,b) as in (51), the result is always ungrammatical. Apparently, EV2Cs are confined to root position as predicted by the IPC-analysis. (51) a. *Peter fragt sich, was du glaubst, soll Max lesen. Peter asks himself what you believe shall Max read b. *Dieses Buch meint Peter, daß ich glaube, soll Max lesen. this book thinks Peter that I believe shall Max read

–Distribution of stress and focus-related material [(26vc) vs. (33vc)]. By and large, EV2-constructions do not allow stress or focus on the alleged matrix clause (52), which, by and large, also confirms the IPC-analysis. (52) ??Wen glaubt (nur) HANS wird der Chef entlassen? ‘Who does (only) Hans believe that the boss is going to fire?’

–Distribution of ‘main clause specific material’ [(26vi) vs. (33vi)]. As pointed out above, certain lexical items, notably modal particles, speaker-oriented expressive elements, discourse connectives, are functionally restricted to main clauses, thus providing a clear diagnostics for ‘main clause’ status. Again the data by and large confirm the IPC analysis; cf. the EV2Cs (53a,b), in which denn (a question-specific modal particle), verdammt noch mal (a pejorative expression ‘damn it’), übrigens (a discourse connective ‘by the way’) happily occur in the related +wh-clause, but not (so) happily in the alleged matrix clause. Wohin glaubt sie, ist er denn / verdammt noch mal/ übrigens where-to believes she is he MP / damn it / by the way gefahren? gone ‘Where does she think he went, damn it/by the way?’ ?? b. Wohin glaubt sie denn/verdammt noch mal/übrigens, ist er gefahren?

(53) a.

–Predicate restrictions [(26viia) vs. (33viia)]. We saw in 3.1 that the class of bridge predicates licensing bona fide ECs is much larger than the class of IP-predicates. Since putative matrix predicates for EV2Cs must also license V2-complements, which negated predicates, and strong factives do not, the class of distinctive test cases is small, but not empty, for preference predicates are bridge predicates (see above, Section 3.1, (33viia)) as well as V2-predicates (54), but not IP-predicates. Since using them in EV2Cs is impossible (55), the IPC-analysis is again confirmed.


(54) a.

Es ist besser, sie geht zu Fuß dorthin. it is better she goes on foot there-to ‘It’s better she walk there.’ b. Er wollte/wünschte, sie ginge dorthin zu Fuß. he wished/wished she went-subj there-to on foot ‘He wished she’d walk there.’

(55) a. *Dorthin ist (es) besser, geht sie zu Fuß. there-to is (it better goes she on foot b. *Wohin wollte /wünschte er, ginge sie zu Fuß? where-to wished/wished he went she on foot

–Further restrictions [(26viic) vs. (33viic)]. The first person subject restriction operative in interrogative IPCs seems to also hold in +wh-EV2Cs (56), although the respective cases do not seem quite as bad as with postfinite VIPs or was-IPs. Likewise, +wh-EV2Cs are subject to the rhetorical reading restriction, this time unequivocally (57). (56) [Na rat mal:] ??Wen glaub ich, wird sie heiraten? [dP guess ??MP whom believe I will she marry (57) A: Ich weiß nicht, was ich tun soll. Ich muß mal Paul fragen. A: ‘I don’t know what to do. I think, I’ll ask Paul’. B: *Was (schon) wird der sagen, sollst du tun — beten und arbeiten, what (MP-rhet. will he say shall you do pray and work das sagt er immer. that says he always

Finally, the data from relative predicate scope in iterated constructions also confirm the IPC-analysis, for iterated EV2Cs do not exhibit the unambiguous scope relation (58) typical of bona fide ECs, but the scope ambiguities typical of bona fide IPCs (59). (58) [das Problemi glaubt sie [ti≤ meinen alle [ti¢ ist ti gelöst]]] [the problem believes she think all is solved. ‘The problem, she believes everybody thinks is solved.’ (59) Das Problem glaubt sie meinen alle, ist gelöst the problem believes-ind she think-ind all is solved relative predicate scope: – 1 > [2 > [3]] – 2 > [1 > [3]]

Rounding up the argument pro (G1), let me just mention two pieces of comparative evidence from closely related languages that also have EV2Cs (for details see Reis 1996a).

25

26

Marga Reis

The first is provided by Bernese Swiss German, which allows resumptive pronouns in long +wh-movement constructions but not in short ones (Penner and Bader 1981: 80). EV2Cs pattern with the latter, which is incompatible with an EC-analysis but predicted by an IPC-analysis. (Note that Bernese Swiss German has VIPCs just like Standard German.) The second is provided by Standard Dutch, for which the EV2 construction is standardly possible, although the occurrence of embedded V2-clauses, in particular outside reported speech contexts, is severely restricted in general, and for many speakers using EV2Cs apparently impossible. Since Dutch liberally uses VIPCs (see Corver 1994), these facts constitute a further obvious argument against an EC-analysis and for an IPC-analysis of EV2Cs. 4.2 Consequences and discussion Let me assume the IP-analysis of EV2Cs just defended is correct. What does it buy us in addition to the descriptive gain? Several things. First, the discrepancy concerning so-called extraction dialects disappears. As is well-known, extraction from daß-clauses is restricted to certain German dialects and subject to strong idiolectal variation, whereas EV2Cs are current everywhere. Under an EC-analysis of EV2Cs, this discrepancy is totally unexpected, but totally expectable under the IP-analysis, for VIPCs are also current everywhere. Second, by getting rid of V2-extraction, we get rid of an extraction instance that has extremely idiosyncratic properties; cf. (60).18 (60) Problematic properties of putative extraction from V2-clauses. (i) Initial gap restriction: regardless of the base position of the movee, extraction leaves a gap in the initial position of the V2-clause (cf. (a)). (ii) V2 route restriction: extraction may occur via V2-clauses and into V2- clauses only (cf. (b)–(d)). a. *Was glaubst du, er sollte lesen? vs. –Was glaubst du, __ sollte er lesen. what believe you he should he read b. Was sagt Peter meint Karl soll sie lesen? [Was sagt Peter [t≤ meint Karl [t¢ soll sie t lesen]]] c. *Was du immer glaubst, sollte er lesen. (V-final matrix clause) what you always think should he read d. *Was glaubst du, daß sie gesagt hat, sollte er lesen what think you that she said has should he read (V-final intermediate clause)


These properties have been notoriously hard to account for; none of the many ingenious attempts has really succeeded. If we adopt the IPC-analysis, they simply disappear as problems for extraction theory. Note that no descriptive problem is left either. All the data motivating (60i–ii) can be handled by appealing to the well-formedness vs. ill-formedness of either the putative host clauses, or the putative VIPs, or both. Third, there is a welcome comparative consequence. If there is no extraction from V2-clauses in German, the Germanic facts allow, without exception, a far-reaching descriptive generalization: clauses with ‘main clause’ word order are strict islands for extraction.19 Thus, (G1) seems all to the good; yet so far it has not really been accepted. Among the possible reasons for this, there is one worth discussing. One might claim that all my arguments in favor of (G1) are correct, but that they do not rule out that extraction from V2-clauses as envisioned by Thiersch (1978) et al. also exists, evidence being the fact that in the case of EV2Cs the IP-restrictions against main clause specific material, against focus, against first person subjects in +wh-contexts, seemed to be blurred, or much less strict for many than in the case of bona fide (V)IPs; cf. (52)–(56) above. I do not dispute these data, but would just like to point out that the clear IPC-features of EV2Cs outnumber these EC-traits by far, and that there are core features among them. Thus, even if in some dialects the restrictions on EV2Cs I just called ‘blurred’ did not hold at all, EV2Cs would still not be an extraction construction pure and simple in these dialects, but at best a ‘blended’ construction, whose IP-traits would also have to be accounted for. Since a convincing account of EC-IPC ‘blends’ is quite hard to give — as we will see in Section 5 — it might seem better after all to treat the rather slight EC-traits of EV2Cs as mere analogical transgressions (a device already appealed to by Chomsky 1970:193f.) of the basic IPC pattern which the formal and interpretive closeness of prefinite IPCs to ECs gives rise to.

5. Partial wh-movement constructions in German20 5.1 The case for (G2) Undoubtedly, was…w-constructions — see (5) or (61) — have a number of salient EC properties, i.e. features which clearly separate them from the was-IP-construction. They are listed in (62), with illustrations provided by (61) and (63)–(69) below.

27

28

Marga Reis

(61) Was glaubst du, wohin er gegangen ist? what believe you where-to he gone is ‘Where do you think that he went?’ (62) Salient EC properties of partial wh-movement constructions. (i) the was-clause is always initial, the related wh-clause is a bona fide subordinate clause; (61) (ii) the related wh-clause must contain a (wh-moved) +wh-phrase (hence must not be an ob-interrogative); (63) (iii) the was…w-construction may be embedded; (64) (iv) the was-clause, i.e. the matrix clause may be iterated; (65) (v) the was-clause may contain main clause specific material and bear main stress, the related wh-clause may not contain main clause specific material; (66) (vi) the was-clause may contain first person subjects; rhetorical readings are allowed; (67)–(68) (vii) the was-clause may contain more complex verbs of saying, thinking, believing. (69) (63) *Was glaubst du, ob er nach Paderborn gegangen ist? what believe you whether he to Paderborn gone is (64) Hans weiß, was Peter glaubt, wieviel das kostet. Hans knows what Peter believes how-much this costs ‘Hans knows how much Peter believes that this costs.’ (65) Was glaubst du, was Peter sagt, was … denkt, wieviel das kostet? what believe you what Peter says what … thinks how-much this costs ‘How much do you believe that Peter says that … thinks that this costs.’ Was glaubst DU denn/eigentlich, wohin er gegangen ist? what believe you MP/MP where-to he gone is ‘Where do you believe that he went?’ b. *Was glaubst du, wohin er denn/eigentlich gegangen ist? what believe you where-to he MP/MP gone is

(66) a.

(67) [Na rat mal:] Was glaube ich (wohl), wen sie besuchen wird? [well guess MP what believe I (MP whom she visit will ‘[Well, guess:] Who do I believe that she is going to visit?’ (68) A: [Was glaubst du, was Paul tun wird? —] [what believe you what Paul do will?


B: Was schon werd ich glauben, was Paul tun wird — weinen und what MP-rhet. will I believe what Paul do will weep and beten wie immer. pray as always ‘[What do you think Paul will do? —]. Come on, what I believe, is obvious — Paul will cry and pray as usual.’ (69) Was behauptet/ ?argwöhnt er, wieviel das kostet? what claims / ?suspects he how-much this costs ‘How much does he claim that this costs?’ kostet das? (cf. Was ??behauptet/*argwöhnt er, wieviel what claims /*suspects he how-much costs this)

According to the ‘partial wh-movement analysis’ (alias ‘direct dependency approach’), (62) reflects the true nature of was…w-constructions: they are held to be ECs just like the interrogative w…daß-constructions described in 3.1, the main difference being that long wh-movement of the relevant wh-phrase into its scope position happens on LF. This analysis entails that the overt movement of this wh-phrase into the embedded Comp position is ‘partial wh-movement’, and that initial was is a kind of wh-expletive functioning as the scope marker for this wh-phrase; as such it is base-generated in the relevant A¢-position.21 Clearly, if analyzed this way the properties (62i–vii) of was…w-constructions all fall into place. The crucial point, however, is that this is only half the truth, for was…w-constructions are also like was-IP constructions and unlike ECs in important respects. The most important parallel concerns predicate restrictions. It is well-known (see already von Stechow & Sternefeld 1988: 356f.) that the was…w-construction differs from the w…daß-construction with respect to admissible bridge predicates. As shown in Reis (1996b: 271ff., see also Reis 2000a), the generalization covering all the differences is (70). (70) Only predicates belonging to the IP-predicate classes appearing in was-IPs may also appear as bridge predicates in was…w-constructions.22

In other words, bridge predicates in was…w-constructions are subject to the semantic restrictions on was-IPs spelled out in (16), which disallow negated/ negative predicates (71), preference predicates (72), strong factive predicates (73), adjectival predicates (74). (71) a. *Was glaubst du nicht, mit wem Hans sich dort treffen wird? what believe you not with whom Hans himself there meet will

29

30

Marga Reis

b. *Was bezweifelst du, wen sie heiraten möchte? what doubt you whom she marry wants (72) a. *Was möchte/will Fritz, wen seine Tochter heiratet? what wants/wishes Fritz whom his daughter marries b. ?*Was würde Fritz vorziehen, wohin seine Tochter geht? what would Fritz prefer where-to his daughter goes (73) a. *Was hat er bedauert/berücksichtigt, wen Hans eingeladen what has he regretted/taken-into-account whom Hans invited hat? has b. *Was fand er entsetzlich/gut, wen Hans eingeladen hatte? what found he terrible/good whom Hans invited had (74) *Was ist klar/bekannt, wen seine Tochter heiraten will? what is clear/well-known whom his daughter marry will

In addition, the same structural restriction we observed for was-IPs (cf. (31)–(32) above) is operative in was…w-constructions as well (75a), again in contrast to normal ECs (75b): complex predicates involving es or full object NPs are disallowed. (75) a. *Was scheint es/hat sie das Gefühl, what seems it/has she the feeling wo er sich niederlassen wird? where he himself down-settle will b. Wo scheint es/hat sie das Gefühl, where seems it/has she the feeling daß er sich niederlassen wird? that he himself down-settle will ‘Where does it seem/does she feel that he will settle down?’

Obviously then, initial was in was…w-constructions, even if it should be no more than a scope-marking expletive, must be simultaneously licensed as a possible argument of the matrix verb, just as if it were part of a was-IP clause. In sum, the parallel between was…w- and was-IP constructions regarding possible was-clause predicates is practically perfect. Note that no independent explanation rendering these parallels accidental is readily available. The semantic restrictions on was…w-constructions do not reduce to ‘weak island’ phenomena (this would not cover the exclusion of preference/volitional and adjectival predicates), nor the structural restrictions to mere ‘CNPC/subject


island’ phenomena, as claimed by Müller (1997: 259ff.). Moreover, an ‘indirect dependency’ approach, by itself, would not automatically account for the semantic restrictions either. If so, the restrictions on was…w bridge predicates must be stated as in (70). Further (halfway) parallels between was…w- and was-IP-constructions are provided by two notorious cases of idiolectal variation. For many people, although perhaps not the majority, was…w-constructions like (76) containing a second wh-phrase in the matrix clause are out.23 (76) a. (%)*Was hat Peter wann gesagt, wieviel das kostet? what has Peter when said how-much this costs b. (%)*Was glaubt wer, wann Peter kommt? what believes who when Peter comes

Now, was-IPs may not contain additional wh-phrases either, which follows from (26vc): multiple wh-phrases are almost invariably, if not inherently focused, and this is what IPs shouldn’t be. If so, the division of judgments on (76) can be related to the alternative points of comparison considered here. For speakers accepting (76) the was…w-construction has one more EC-feature; for speakers rejecting it, it has one more was-IP feature. Judgments on complex was…w-constructions containing intervening daß are similarly divided.24 (77) (%)*Was glaubt Peter, daß Franz jetzt meint, wohin Elke what believes Peter that Franz now thinks where-to Elke gegangen ist? gone is

Recall, however, that was-IPs tend to be syntactically simple (26viic); in particular they may not contain finite dependent clauses. Hence, not accepting (77) reflects an IP-feature of was…w-constructions. Again, the alternative judgments may be related to the alternative points of comparison. For speakers accepting (77), was…w-constructions are more in line with w…daß-constructions, for which a similar ban on complexity is unmotivated. Finally, let us take note of an interpretive difference between was…w-constructions and ECs originally pointed out by Herburger (1994) (reported in Dayal 2000). A question involving extraction leaves it open whether the speaker accepts the presupposition behind the embedded question, while the was…w-construction implies that the speaker is committed to it. In other words, (78a) is compatible with assuming that Mary spoke with nobody,

31

32

Marga Reis

whereas (78b) is not. This explains why only (78a), but not (78b), is an appropriate reaction in the context (79). (78) a.

Mit wem glaubt Karl, daß Maria gesprochen hat? with whom believes Karl that Maria spoken has b. Was glaubt Karl, mit wem Maria gesprochen hat? what believes Karl with whom Maria spoken has

(79) Speaker A: Wir wissen ja beide, daß Maria mit niemand gesprochen hat. Aber Karl glaubt ja ihren Lügen. ‘We both know that Mary talked to nobody. But Karl believes MP her lies.’ Speaker B: –(78a) #(78b)

The point is that this distinctive property of was…w-constructions vis-à-vis ECs is again shared by was-IP-constructions, likewise by VIPCs for that matter. (80a,b) would be just as inappropriate in context (79) as a question in was…w-form. This shows that this interpretive property is again a was-IP feature of was…w-constructions. (80) a.

Was glaubt Karl, mit wem hat Maria gesprochen? what believes Karl with whom has Maria spoken b. Mit wem hat Maria gesprochen, glaubt Karl? with whom has Maria spoken believes Karl ‘With whom does Karl believe did Maria speak?’

In sum, (G2) seems to be correct: was…w-constructions have salient EC-features and salient was-IP-features — both too salient to be dismissed as accidental. 5.2 Consequences and discussion How do we account for the double nature of was…w-constructions captured in (G2)? The way I described the situation, the immediate suspicion is that was-IPCs and was…w-constructions are historically related. In Reis (1996b) I explored this possibility and showed that one can indeed think up a perfectly plausible historical scenario for how was-IPCs could have become reanalyzed as extraction constructions (with partial wh-movement on S-structure plus further wh-movement at LF). A particularly attractive feature of this scenario is that the most puzzling synchronic idiosyncrasies seemingly fall into place. Thus, the lack of was…w-constructions involving ob-clauses, see (62ii), could be traced back


to the lack of a reanalyzable source (unlike +wh-main clauses, yes/no main clauses are structurally too different from their subordinate counterparts — V1vs. ob-clauses — to be reanalyzed as subordinate clauses). And ‘partial wh-movement’ as well as so-called ‘anti-locality’, i.e. the failure of the scopemarker was to appear in simple questions (*was ist wer gekommen [what-iswho-come], intended to mean ‘Who came?’), could be simply taken to be features of the historical source structure that (still) persist. The same could be assumed for the striking IP-features described in 5.1, with the idiolectal variation illustrated in (76)–(77) reflecting typical intermediate stages of still ongoing change. Apparently, this scenario was too convincing, for subsequently ‘historical reanalysis’ was taken as the core of my proposal for coming to terms with the double nature of was…w-constructions (for example by von Stechow 1996 or Sternefeld 2000), and used in different ways. Dayal (1996, 2000) used it as an auxiliary theory in favor of an ‘indirect dependency’ approach, whereas Lutz (2001) used it as an argument for the ‘direct dependency’ approach: was..wconstructions were was-IP-constructions once and thus once amenable to the indirect dependency treatment, but since they are different now, they must be ECs, albeit of a special kind. However, the point I wanted and still want to make (see Reis 1996b: 281ff., much elaborated in Reis 2000a: 391ff.) was quite different, and I take the opportunity to repeat it here in a much condensed form. First, no matter how good a diachronic account we can envision, the IP-features of was…w-constructions act as live restrictions on the present was…w-constructions, and must be accounted for as such. Second, the IP-features seem to be stable, i.e. there are no signs of drift whatever going further into the direction of ECs, as we should expect if this were the target of linguistic change. Third, at least some of the IP-features in question, in particular the selectional restrictions, are core features. Consequently, not only the EC-features but also the IP-features must be covered by a systematic synchronic account of was…wconstructions, and cannot be relegated to some negligible periphery. But in trying to give such an account, and that is the second point I wanted and still want to make, we inevitably reach the conclusion that the present analyses of complex ±wh-constructions must be drastically revised in order to permit one. What is required, clearly, is an account that reflects the fact that was…w-constructions are systematically related by their respective EC- vs. IPproperties to two types of constructions: not only to normal wh-movement constructions but also to the entire range of ±wh-IP-constructions. The current approaches fail to do so, and since the ECs and IPCs have completely disjoint

33

34

Marga Reis

analyses at present, this failure is systematic. In the direct dependency approach, which treats was…w-constructions basically as ECs, only the parallels to other instances of ECs are systematically brought out; the IP features of was…w-constructions appear as mere idiosyncrasies. But an indirect dependency approach, even in versions that allow for variable degrees of subordination (see Dayal 2000), fares no better, for besides missing out on salient EC-features, it still assimilates was…w-constructions just to sequential questions (= ‘unintegrated parentheticals’, see 2.1), thus also missing out on the crucial integrated parenthetical properties shared by was…w-constructions and IP-constructions. So we need a third approach, which presupposes that the analyses of ECs and IPCs be revised so as to be compatible in the right way, and this — I submit — can only be done by daring to go unorthodox ways. In order to add substance to this claim, let me briefly address and dispose of two more orthodox proposals to account for (some of) the IP-features of was…w-constructions, the only ones I know or can think of that at first glance look promising. The first crucially involves the so-called w-copy construction as illustrated in (81), which shares certain IP-features of was…w-constructions, in particular the peculiar semantic restrictions on bridge predicates (82). If so, one might suggest that the different restrictions on bridge predicates in was…w- and w-copy-constructions vs. normal ECs are related to the fact that only in the former cases the bridge predicates select a complement introduced by a wh-phrase that is bound by a higher wh-phrase (the ‘w…w-property’).25 (81) Wen glaubst du, wen er getroffen hat? who believe you who he met has ‘Who do you believe he met?’ (82) a. *Wen glaubst du nicht, wen sie liebt? whom believe you not whom she loves b. *Wen bedauert Peter /möchte Peter /ist besser /ist klar, wen whom regrets Peter/wants Peter/is better/is clear whom Petra heiratet? Petra marries

But on closer inspection, this is no answer to our prayers. To begin with, it is doubtful that the w…w-property has the same systematic status in was…w- and w-copy constructions, for the latter also sides with w…daß constructions in significant respects (i.e. behaves as if the second wh-phrase were a substitute of daß).26 However, even if it had the same status, the proposal would just cover


the selectional parallels between those two constructions (at the cost of a nonlocal subcategorization frame), but miss out on the crucial fact that these — and others — are shared by the entire array of IP-constructions. For one thing, some IP-parallels (ban on multiple wh-phrases, negation in matrix clauses) cannot easily be conceived of as selectional restrictions at all. More importantly, the w…w-approach cannot be extended to even cover the parallel predicate restrictions in IPCs. As for was-IPCs, their w…w-property is an anaphoric, not a selectional one. As for all the other IPCs (VIPCs, so-/wie-IPCs), which also share these restrictions (see 2.2), they do not share a w…w-property of any kind at all. Hence, the IP-features of was…w-constructions (or of w-copy constructions for that matter) cannot be systematically accounted for on this basis. The second attempt is the optimality theoretic approach by Müller (1997), who treats was…w-constructions as cases of direct dependency competing with w…daß-constructions and accounts for the differing predicate restrictions in the following way. The additional restrictions on was…w-constructions are interpreted as heightened sensitivity to weak island and CNPC-/subject island restrictions (1997: 259), and deduced from the fact that w…daß-constructions involving these islands violate BAR-CON against crossing barriers just once, whereas was…w-constructions (due to interpreting was-insertion also as chainformation, and BAR-CON as a derivational constraint also holding at LF) violate it twice. Thus, everything else being equal, the w…daß-construction blocks the was…w-construction in these contexts, thus accounting for the differences observed (1997: 285ff.). Clearly, this account depends on a number of premisses one might take issue with. But even if they are granted, it misses the mark. As already pointed out (see 5.1), the predicate restrictions in question do not reduce to weak island and CNPC/subject island phenomena. Since the additional restrictions include verbs like wollen/möchten that are perfect bridge verbs for ECs, BAR-CON — the only likely candidate for causing the respective difference given Müller’s basic EC-approach to was…w-constructions — cannot be appealed to at all. Hence, the predicate data identified above as IP-parallels cannot be accounted for in this approach, let alone the fact that they are IP-parallels, for this would require that was…w-constructions be also in a candidate set together with IPCs. Given Müller’s assumption of was…w-constructions as being basically ECs plus the complete disjunction of present EC and IPC analyses, this is ruled out a priori, for ECs and IPCs will never have the same LF, which is Müller’s defining criterion for candidate sets.

35

36

Marga Reis

If so, the optimality theoretic attempt brings us right back to the conclusion already stated above. In order to do justice to the double allegiance of was…wconstructions, we have to assimilate the analysis of IPCs and ECs in a way that makes them comparable and thus IP- and EC-features compatible. In Reis (2000a) I argued that trying to restate the successful historical analysis in systematic synchronic terms offers such a way, albeit an unorthodox one, for (putting insights by Dabrowska 1997, Sternefeld 2000 and Dayal 2000 to my own use) analogy rather than syntactic wh-movement plays the major role in accounting for complex wh-extraction constructions, and clause linkage, in particular subordination, is treated as a multi-faceted notion, allowing different linking analyses to coexist. Since the argumentation is rather involved, I will not repeat it here but just refer the reader to Reis (2000a). But let me stress again how important it is to strive for a systematic account, orthodox or not, of was…w-constructions that reflects their systematic relationship to both extraction and IP constructions. Just shrugging off the data presented above that force such an account does not make them go away. Perhaps a final consideration will help to accept this goal as something worth attaining. With the rise of the notion of competition in syntax, the notion of reference set or candidate set has become prominent and played an increasingly important role.27 While none of the different definitions proposed for them is uncontroversial, it is easy to see that under the two major defining notions that are theory-independent — a) sameness of numeration (Chomsky 1991, 1994), b) sameness of (truth-functional) meaning (Fox 1995) — practically all ±wh-IPCs stand a good chance of ending up together with some ±wh-movement construction in the same reference sets. Moreover, there is some substantive indication that these construction types do not lead mutually independent lives. VIPCs and ECs seem to usually coexist, subtypes of VIPCs and ECs appear as cross-linguistic equivalents, prefinite VIPCs and was…wconstructions seem to restrict the operating space of w…daß-constructions (Müller 1997: 279f.), and true cases of V2-extraction appear only where prefinite VIPCs are impossible (Reis 1996a: 76ff.).28 If so, the analyses of IPCs and ECs should reflect the close relationship between the two to which the existence of structurally ‘mixed’ types as the was…w-construction (and the w-copy construction for that matter) also testifies.


6. Concluding remarks What I hope to have shown in this article are three things. i.IP-constructions are a topic worthy of the syntactician’s attention, and, it should be added, also an exciting topic of comparative research, — by itself as well as in relation to ±wh-movement constructions. ii.Constructions in which the defining properties of syntactically differing phenomena blend, seem to be a synchronic fact of linguistic life, and should be taken seriously as such. iii.Despite many subtle pragmatic differences, it turned out that the same or related effects can be achieved by wh-movement constructions and IP-constructions alike, which is also the likely precondition for structural blending and competition between these two types of constructions. This suggests strongly that in studying so-called long-distance wh-phenomena not only extraction constructions but also IP constructions should be systematically taken into account, and perhaps even taken as bona fide members of the same reference set(s), be it in systematic, or acquisitional or diachronic studies of wh-dependencies.

Notes * Thanks to Werner Abraham and Jan-Wouter Zwart for inviting me to present my work on wh-movement and integrated parenthetical constructions at the 15th Comparative Germanic Syntax Workshop in Groningen. Thanks also to the Workshop audience for their comments, which have led to a number of changes in exposition and content. 1. Please note that ‘IP’ (integrated parenthetical) is used to refer to the integrated parenthetical expression or clause, whereas ‘IPC’ is used to refer to the entire construction containing the IP. 2. Sections 2 and 3 are based on the more comprehensive treatments of was-IPCs and Verbfirst-IPCs in Reis (1995, 1996a, 2000a). Section 3.2 also draws on Reis (2000b). 3. (\) indicates the intonation break between the clauses; the difference between the more level ‘colon’ intonation (as in (7a)) and the intonation contour in (7b) (more of a rise contour, due to the first clause being a yes/no clause) is disregarded. As usual, capital letters mark the syllable bearing main stress, which in turn marks the ‘focus exponent’. 4. Abbreviations used in the glosses: MP = modal particle, dP = discourse particle, subj = subjunctive. 5. There is considerable variation in judgments of was-IPCs involving wh- vs. yes/no-clauses as well as initial vs. medial vs. final IPCs. All ratings in the text represent my own judgments.

37

38

Marga Reis

6. Interestingly, the interpretive relation noted in (11iv)/(13iv) is also affected. While in unintegrated cases the restrictive relation between the was-clause and the related +wh-clause can still be cancelled in certain contexts, albeit with effort, this is impossible in the integrated counterparts; see Reis (2000b). 7. The judgments in (15) presuppose the prosodic properties described in (13v). For the grammatical counterparts to the examples in (15), see (66) below. 8. Verbum dicendi readings of verbs like glauben, wünschen, etc. do occur, albeit more markedly. For discussion of the two IP-readings in question, cf. Reinhart (1983a: 175ff.), Reis (1996a: 71ff.), Steinbach (1999b: 9ff.). 9. There exist no systematic studies of the prosodic properties of German parentheticals. For some pertinent observations, cf. Schönfeld (1993). As for the English counterparts to VIPs, Reinhart observes a prosodic split between the two readings in question: while the parenthetical remains unintegrated with respect to property (13va) in the ‘subject oriented’ reading, it is fully integrated in the ‘speaker oriented reading’. Besides, ‘quotative inversion’ is only possible with the first reading (1983a: 178f.). Unfortunately, matters are not as clear-cut in German. 10. As for so-/wie-parentheticals, see Pittner (1995), Brandt (1997), Zimmermann (1997). An explicit comparison of their ‘integrated uses’ is still a desideratum. 11. For a more comprehensive account of iterative IP data, see Reis (1996a: 66f.). Again thanks to J. Pafel for insisting on VIPCs having a peculiar scope behavior and carrying out a pertinent test. 12. Espinal (1991) identified this postsyntactic level as a discourse grammatical level. Given the division of syntactic labor as envisioned in Chomsky (2000), one might speculate on whether PF could do the job. 13. Binding data are notoriously murky, and whether they completely pattern alike in VIPs and so-/wie-IPs (once the relevant factors ‘prosodic integration’, ‘speaker-oriented IP’ are controlled, which is unclear in previous studies), is by no means certain. Note, however, certain differences between bona fide embedding structures and IPCs that are hard to explain under an embedding+slifting analysis of the latter (which in order to also account for the ‘inversion’ pattern alias V1 pattern amounts to postulating that in German the slifted host clause occupies the preverbal position of the IP): (i) Daß eri beliebt ist (/), glaubt jederi/Hansi ‘Everybody/Hans believes that he is popular’ (ii) *Eri ist beliebt, glaubt jederi/Hansi. he is popular believes everybody/Hans For binding data in IPCs vs. bona fide ECs, see below, note 17. 14. This niching restriction was already pointed out by Espinal (1991: 752). Obviously, this is not all to be said on niching restrictions, but this is beyond the scope of this article. (For some pertinent observations w.r.t. German IPs, see Altmann 1981, Pittner 1995, Reis 1995; w.r.t. Dutch IPs Corver 1994: 8ff.) 15. For an overview over a good part, see Espinal (1991); some further references to earlier work (though by no means exhaustive) are given in Reis (1995: 75). A wealth of additional information on English VIPs and related phenomena is provided by Bolinger (1968, 1972).


16. This section is based on the more comprehensive treatment of EV2-constructions in Reis (1995, 1996a). 17. For details, see Reis (1995, 1996a). Here I will just cite some pertinent binding data: (i) shows that EV2Cs are not as parallel to bona fide ECs as the EC-analysis would predict (see Reis 1996a: 54f.). (ii) shows that the binding data (iia) allegedly proving that EV2Cs are ECs can be duplicated with bona fide IPCs (iib), in other words they are compatible with an IPC-analysis. (i) a.

[Heinz glaubt, daß du die meisten Kollegen schätzt, aber] IHNi, meint Heinzi, dass du ti verachtest / ?IHN meint Heinz , verachtest du t . i i i ‘[H. believes that you think highly of most colleagues, but] himi H.i thinks (that) you despise.’ (example by T. N. Höhle) b. *Weni meinst du, daß seinei Mutter abholt? whom think you that his mother up-picks ?Wen meinst du, holt seine Mutter ab? i i whom think you picks his mother up ‘Whoi do you think (that) hisi mother has picked up?’

(ii) a. *Der arme Kerli glaubt eri, hat immer Pech. b. *Der arme Kerli, so glaubt eri, hat immer Pech. the poor guy (so believes he has always bad-luck 18. Note that there are constructions (first noticed in Reis 1995, 1996a) involving genuine extraction from V2-clauses; cf. (i)–(ii). Since they are so severely restricted (only a subpart of speakers accept them, possible extractees are rare, and only preference predicates allow them), they cannot really count as a normal option of German grammar. But note that they are not subject to the alleged restrictions (60i, ii) on V2-extraction; so, if anything, they strengthen the argument against the EC analysis of EV2Cs. (i) a.

b.

(ii) a.

b.

Dorthini ist es besser, du gehst zu Fuß ti. there-to it is better you go on foot ‘To this place you better walk.’ Worüberi wünschte er, er hätte gleich mehr ti erfahren? where-about wished he he had-subj at-once more heard ‘What did he wish he had known about earlier?’ Dorthini meinte er, daß es besser sei, ich ginge ti zu Fuß. there-to thought he that it better be-subj I went-subj on foot ‘To this place he thought that you better walk.’ … wohin er meinte, es sei besser ich ginge t zu Fuß where-to he thought it be-subj better I went-subj on foot ‘…where he thought that I’d better walk’

19. For possible accounts see Holmberg and Platzack (1995: 80ff.). A possible consequence (spelled out for German in Reis 1997 and for Dutch in Zwart 1997: 237) is that ‘embedded’ V2-clauses are structurally dependent but no complement clauses. 20. This section is largely based on the more comprehensive treatment in Reis (1996b, 2000a).

39

40

Marga Reis

21. Early influential propagators of this analysis were Van Riemsdijk (1982) and McDaniel (1989); judging from the contributions to Lutz, Müller and von Stechow (2000), it is still the majority view. For a detailed comparison with Dayal’s (1994 and later) ‘indirect dependency’ approach (which does not involve extraction), plus critical discussion of both alternatives with respect to German, see Lutz (2001) and Felser (2001), which however come to opposing conclusions. 22. Referring to ‘predicate classes’ rather than to just ‘predicates’ allows for the fact that impossible was-IP-predicates like behaupten ‘claim’, vorschlagen ‘suggest’, argwöhnen ‘suspect’ do show up in was…w-constructions; cf. (62vii). These cases, however, can be considered as analogical extensions of the admissible IP predicate classes (see (16): [simple] verbs of saying, thinking, believing), hence conform to (70). 23. This corresponds to the judgments reported in among others Dayal (1994: 151 note 7, 1996), Brandner (1994: 203); but see Höhle (1996, (27)–(29)), Müller and Sternefeld (1996, (54a)), Fanselow and Mahajan (1996, (43a)) for opposite judgments. Numerous informal tests with native speakers have borne out this division of judgments. 24. Cases with intervening daß are rejected in Van Riemsdijk (1982: 12), also in von Stechow and Sternefeld (1988: 358), Dayal (1994: 140f.), Brandner (1994: 204), Bayer (1996: 228), but idiolects accepting it are reported by McDaniel (1989: 575f.), Höhle (1996, Section 2 (5)), Fanselow and Mahajan (1996, (18)), Müller (1997: 257). 25. This proposal has been made in two slightly different versions by J. Pafel (p.c.) and Fanselow and Mahajan (1996); see Reis (2000a) for a more comprehensive discussion. 26. As illustrated in more detail in Reis (2000a: 395ff.), w-copy constructions are more like w…daß-constructions in tolerating bridge predicates involving es or full NPs, or with respect to wh-/Q-scope ambiguities in the matrix clause. Also, the w-copy-construction does not admit multiple wh-phrases in the matrix clause or intervening daß at all. Moreover, w-copyconstructions exist in languages (cf. for example the relative copy construction in French) that have long wh-extraction but no counterparts to was…w-constructions. While this does not invalidate the parallels between w-copy- and was…w-constructions — which has led to standardly viewing these constructions as variants of each other — it seems clear to me that in view of the divided evidence no attempt to reduce the w-copy-construction to one of the two parallel constructions will be entirely satisfactory. 27. For an overview, see Müller and Sternefeld (2001). 28. A case of crosslinguistic equivalence may be provided by extraction from complements involving complementizer drop as in English, which seems to share a number of use features of bona fide parenthetical constructions (apparently mediated by the respective properties of bare complement vs. that-complement constructions; cf. Hand 1993). If so, this subtype of extraction is perhaps as close a counterpart to German prefinite VIPCs as the bona fide English IP-constructions illustrated in (2).

Van as a marker of dissociation Microvariation in Dutch* Jeroen van Craenenbroeck University of Leiden

1.

Introduction

In this article I discuss various constructions in (certain varieties of) Dutch, all of which contain the word van ‘of ’. I argue that in many of these constructions van functions as a marker of dissociation. The article is organized as follows. In the next section I present as yet unobserved data concerning so-called complementizer van in certain Brabant dialects of southern Dutch. In Section 3 I propose an analysis for these data, based on proposals by Kayne (1999), Rooryck (1997), Hoekstra (1993b) and Zwart (2000). This analysis capitalizes on the parallelism between the use of van in the nominal and that in the verbal domain. In Section 4 I discuss the so-called quotative van construction. I argue that this construction too should be analysed as outlined in Section 3. In Sections 5 and 6 I extend the analysis to non-Brabant complementizer van, and French de and Italian di, respectively. I argue that these languages represent further stages in the grammaticalization of van. In Section 7 I summarize the main findings of this article.

2. Complementizer van in Brabant Dutch The dialects of Dutch spoken in the northern part of Belgium exhibit a lot of variation with respect to the complementizer system. This is not only true of complementizers introducing finite clauses (most notably als ‘if ’, of ‘whether’ and dat ‘that’, cf. De Rooij 1965), but also of those introducing infinitival complements. The RND, a series of Dutch dialect atlases, shows that apart from

42

Jeroen van Craenenbroeck

the standard Dutch om, both voor ‘for’ and van ‘of ’ are frequently used as infinitival complementizers in Belgian varieties of Dutch (Blancquaert and Pée 1930–1982). It is the latter one I will focus on in this section. At first sight, the area in which van is used as a complementizer seems to be a fairly homogeneous one. However, when we look more closely, it turns out that in certain Brabant dialects of southern Dutch van has a wider distribution than in the rest of the area (and also wider than that of om in standard Dutch).1 More precisely, in those Brabant dialects van can be used after raising verbs (1a) and after so-called epistemic verbs (2a), a group of Dutch verbs which more or less corresponds to the English believe-verbs (cf. Bennis and Hoekstra 1989: 131). Neither of these two types of verbs can be followed by van in the rest of the southern Dutch area, nor by om in standard Dutch. (1) a. *Hij schijnt van da plezant te vinden. (only – in Brabant Dutch) he seems of that agreeable to find ‘He seems to like that.’ b. Hij schijnt (*om) dat aangenaam te vinden. (Standard Dutch) he seems (*om that agreeable to find (2) a. *Ik meen van daar iets te kunnen aan toevoegen. I believe of there something to can to add (only – in Brabant Dutch) ‘I believe I can add something to that.’ b. Ik meen (*om) daar iets aan te kunnen toevoegen. I believe (*om there something to to can add (Standard Dutch)

These distributional facts seem to indicate that Brabant van is in some way different from general southern Dutch complementizer van. A closer inspection of the data reveals that this is indeed the case. In the non-Brabant part of the southern Dutch area, the use of complementizer van is completely optional, much like that of om in standard Dutch (cf. Blom 1984, however, for a different view on the optionality of om). In the Brabant dialects, however, there is a semantic difference between sentences with and without van. In what follows I will illustrate this using three different examples, each representative of a certain type of matrix verb. At the end of this section I will generalize over these three cases, arguing that van induces a similar reading in each of them. I will then link this reading to the etymology of van, to the use of evidentiality markers in various languages and to the distinction between factive and non-factive complements.

Van as a marker of dissociation

As can be seen in (1a), Brabant van can be used to introduce the complement of a raising verb. When doing so, it induces a certain reading on the sentence. Consider the data in (3).2 (3) a.

Marie schijnt zwemmen plezant te vinden. Mary seems swim (inf) agreeable to find ‘Mary seems to enjoy swimming.’ b. Marie schijnt van zwemmen plezant te vinden. Mary seems of swim (inf) agreeable to find ‘Mary seems to enjoy swimming.’

In (3b) the speaker only has indirect evidence for the statement he is making. Someone has told him that Mary enjoys swimming (hearsay) or he sees her going to the swimming pool three times a week and on the basis of this indirect evidence he infers that Mary enjoys swimming. He could of course be mistaken, because all that he can base himself on is indirect information. In (3a) the speaker has direct evidence for the truth of his utterance. He sees Mary and he sees how she is having a good time in the swimming pool. That is why he concludes that she enjoys swimming.3 Thus we can say that van induces an indirect reading on the sentence when it introduces the complement of a raising verb.4 If van indeed induces an indirect reading on the complement of a raising verb, then we predict it to be incompatible with phrases which explicitly deny such an indirect perspective. This prediction is borne out in (4). (4) a.

In mijn ogen scheen hij zich gisteren te amuseren. in my eyes seemed he himself yesterday to enjoy ‘In my view he seemed to enjoy himself yesterday.’ b. *In mijn ogen scheen hij van zich gisteren te amuseren. in my eyes seemed he of himself yesterday to enjoy

By using the phrase in mijn ogen (‘in my view’, lit. ‘in my eyes’) the speaker in (4) indicates that he has not acquired his information via some external — and thus indirect — channel. On the contrary, he is basing himself solely on his own perception. Clearly such information is in direct contrast with the indirect reading induced by van. That explains why a sentence such as (4b) is considered ungrammatical by speakers of Brabant varieties of southern Dutch. The second type of matrix verbs the complement of which can be introduced by van are the above-mentioned epistemic verbs. Here too, the presence or absence of van has an effect on the semantics of the sentence. Consider the examples in (5).

43

44


(5) a.

Ik meen voor iedereen te kunnen spreken. I believe for everybody to can speak ‘I believe I can speak for everyone.’ b. Ik meen van voor iedereen te kunnen spreken. I believe of for everybody to can speak ‘I believe I can speak for everyone.’

Just like in (3) there is a subtle meaning difference between the a- and b-sentences in (5). In (5b) the speaker is expressing his belief that he can speak for everyone. He is not entirely certain about this, since it is only his belief and he could be mistaken. In the a-sentence the speaker is much more sure of what he is saying. He is convinced that he can speak for everyone; he is convinced of the truth of the proposition expressed in the complement clause. According to the speaker that proposition holds not only on a subjective level (like in (5b)), but also on an intersubjective level.5 In short, the use of van in the complement of an epistemic verb casts doubt on the truth of the proposition expressed in the complement clause. These observations too can easily be tested, since there are contexts which strongly favour the ‘intersubjective’ reading of menen ‘believe’. On the basis of what we have seen so far we expect them to be incompatible with the use of van. Consider the mini-dialogue in (6). (6) A: Ik vond het daar maar niks. I found it there prt nothing ‘I didn’t like it there.’ B: a. Ik meen mij nochtans te herinneren da gij u daar goe I believe me however to remember that you yourself there well geamuseerd hebt. enjoyed have ‘I seem to recall, however, that you enjoyed yourself there.’ b. *Ik meen van mij nochtans te herinneren da gij u daar I believe of me however to remember that you yourself there goe geamuseerd hebt. well enjoyed have

The construction Ik meen mij nochtans te herinneren, especially when in direct contradiction with the preceding statement, is often used to express that one is quite sure about what one is saying. In the dialogue in (6) B is quite sure that A did have a good time at the place in question. Since van is used to express the opposite (the speaker is not sure about the truth of the complement clause), we expect a sentence such as (6b) to be unacceptable, which it is.


The third and final group of verbs I want to discuss here are the so-called control verbs.6 They are exemplified by proberen ‘try’ in (7). (7) a.

Freddy probeert den auto te repareren. Freddy tries the car to fix ‘Freddy tries to fix the car.’ b. Freddy probeert van den auto te repareren. Freddy tries of the car to fix ‘Freddy tries to fix the car.’

Once again there is a meaning difference between the two sentences. In (7b) the speaker doubts whether Freddy will be able to actually fix the car. Through the use of van he is implying that Freddy’s efforts will be in vain, that they will constitute but an attempt. The sentence in (7a) is perfectly neutral with respect to Freddy’s chances of success, but when contrasted with (7b), it acquires a reading where the speaker is fairly confident that Freddy will succeed. It will be more than just an attempt; Freddy will be able to actually fix the car. We can bring out the reading induced by van more clearly by coordinating a sentence like (7b) with another one. Consider the sentences in (8). (8) a.

Freddy gaat nu proberen van den auto te repareren, Freddy will now try of the car to fix maar ik weet zeker da da nie gaat lukken. but I know sure that that not will succeed ‘Freddy will now try to fix the car, but I’m sure he’ll fail.’ b. #Freddy gaat nu proberen van den auto te repareren, en da zal Freddy wil now try of the car to fix and that will zo’n tien minuutjes duren. about ten minutes last ‘Freddy will now try to fix the car and it will take about ten minutes.’

Although both sentences are grammatical, only (8a) yields a successful utterance for speakers of Brabant Dutch. In the first half of the sentence the speaker implies that Freddy’s efforts will be in vain and in the second half he makes this explicit. The sentence in (8b) sounds very strange. It can only mean that according to the speaker Freddy will try to fix the car for about ten minutes, but after those ten minutes the car will still be broken.7 Thus we can conclude that when van is used in the complement of a control verb, it is uncertain whether the state of affairs described in the complement clause will ever take place. As a result of this, the matrix predicate acquires focus: the speaker emphasizes that Freddy’s efforts constitute but an attempt.

45

46


Although there are clearly differences between the three examples discussed above, there are also enough similarities between them to allow for an overall generalisation. The relevant facts can be captured by the statement in (9). (9) The use of van as an infinitival complementizer in Brabant Dutch indicates that the state of affairs expressed in the complement clause is dissociated from the matrix predicate. This dissociation can be situated on an aspectual or an evidential level, depending on the type of matrix verb.

As it stands, the statement in (9) is still in need of clarification. Therefore, let us reconsider some of the examples we looked at above (they are repeated in (10)). (10) a.

Marie schijnt van zwemmen plezant te vinden. Mary seems of swim (inf) agreeable to find ‘Mary seems to enjoy swimming.’ b. Ik meen van voor iedereen te kunnen spreken. I believe of for everybody to can speak ‘I believe I can speak for everyone.’ c. Freddy probeert van den auto te repareren. Freddy tries of the car to fix ‘Freddy tries to fix the car.’

In (10c) the speaker indicates that Freddy’s attempt at fixing the car will be in vain. It will be but an attempt. Put differently, the activity expressed in the complement clause (the fixing of the car) is dissociated from the event described by the matrix predicate (the trying). The stage where Freddy actually fixes the car will not be reached and is thus dissociated from the stage where he tries to fix the car. Since the reading induced by van here crucially involves the completion or non-completion of the activity expressed in the complement clause, I propose to call it dissociation on an aspectual level. Something similar is going on in (10a–b), though on another level. It is clear that schijnen ‘seem’ in (10a) and menen ‘believe’ in (10b) do not refer to events the way proberen ‘try’ in (10c) does. As a result the dissociation induced by van in (10a–b) is situated not on an aspectual, but on an evidential level. The truth of the proposition expressed in the complement clause is called into question. Thus we see how the choice of the matrix predicate influences the precise effect van has on a sentence, but how dissociation is always at the heart of it. Even though the generalisation in (9) covers the data I discussed quite nicely, one might still object that it seems like a fairly unusual or even idiosyncratic observation, with little or no parallels in the rest of the grammar. Before rounding


off this section, however, I will argue that the readings induced by van are not as extraordinary as they might seem at first sight. First of all, when we take the etymology of van into account, it should not come as a surprise at all that dissociation is the core semantic characteristic of a construction containing this preposition. Although the exact provenance of van is still not entirely clear, etymologists and historical linguists agree on the fact that the original meaning of van is ‘away from’ (and thus dissociation).8 This meaning is still very prominent in a sentence such as (11). (11) Jan komt van Brussel. John comes of Brussels ‘John comes from Brussels.’

This example makes clear that in Brabant Dutch van has evolved from a marker of spatial dissociation in the nominal domain to a marker of aspectual or evidential dissociation in the verbal domain. I will come back to this typology and to the evolution of van extensively in the following sections. Moreover, Brabant Dutch is not the only language in the world with an overt dissociation marker. There are many languages that have evidentiality markers not unlike Brabant van indicating that the speaker only has indirect evidence for the statement he is making or that the truth of the proposition expressed is uncertain or called into question (see in this respect for example Matthews 1965 about Hidatsa, a Siouan language, and the articles in Chafe and Nichols 1986). An even more telling — and probably better-known — case in point of a linguistic phenomenon resembling the Brabant complementizer van facts is the distinction between factive and non-factive complements (cf. Kiparsky and Kiparsky 1970). The former are complements the truth of which is presupposed (cf. (12a)), while in the latter one can’t really say anything about the truth or falsity of the proposition expressed in the complement clause (12b). (12) a. John regrets that the door is open. b. John thinks that the door is open.

While in (12a) we know for a fact that the door is open (and that John regrets this), the only thing we know in (12b) is that it is John’s belief that the door is open. Since John could be mistaken in that belief, we don’t know anything for certain about the state the door is in (i.e. about the truth-value of the embedded proposition). Note that what is going on in (12a) is the exact opposite of what we witnessed in the van-sentences in (3)–(8). While factive verbs such as to regret

47

48


force us to interpret their complement clauses as being necessarily true, van induces a reading in which the truth of the embedded proposition is questionable to say the least. We therefore predict these two to be incompatible, a prediction borne out in (13) for the factive expression het erg vinden ‘to regret’ and for the raising verb blijken ‘to turn out, to be shown’, which has factive characteristics in the sense that it presupposes the truth — or the very high degree of probability — of its complement clause.9 Hij vond het erg (??van) haar daar op dat moment aangetroffen he found it bad (??of her there on that moment met te hebben. to have ‘He regretted having met her there at that moment.’ b. Hij blijkt (*van) er niet bij betrokken te zijn. he turns.out (*of there not with involved to be ‘It turns out he is not involved in it.’

(13) a.

Summing up this section, I have shown that in certain (mainly Brabant) dialects of southern Dutch complementizer van not only has a wider distribution than in the rest of the area, but also contributes to the meaning of the sentence. Van induces a reading in which the complement clause is dissociated from the matrix predicate, either on an aspectual or on an evidential level, depending on the type of matrix verb. This dissociation reading turned out to be a parallel in the verbal domain to the spatial dissociation van expresses in the nominal domain. Furthermore, Brabant Dutch complementizer van interacts with the factive/nonfactive distinction in the sense that it is incompatible with factive complements.

3. An analysis of Brabant complementizer van In this section I will propose an analysis for the data discussed in the previous section. As a starting point I will take the idea that the syntactic structure of a van-sentence should reflect its semantics. Furthermore I want to capitalize on the parallelism between the use of van in the nominal and in the verbal domain. Let us first look at the phrase in (14a) and its corresponding tree structure in (14b).


(14) a.

Jan van Brussel John of Brussels ‘John from Brussels’ PP

b. Spec

P′

DP

P

DP

Jan

van

Brussel

It is clear that the semantic relation introduced by van (i.e. spatial dissociation) is also structurally — and hence syntactically — represented in the tree structure in (14b). As the head of the PP, van establishes a semantic relation between the element in its specifier on the one hand and its complement on the other. More precisely, in this configuration, John is said to be spatially dissociated from Brussels. This admittedly fairly trivial structure becomes more challenging when we transfer it to the constructions discussed in the previous section. Recall that I argued there that in a sentence such as (10c) (repeated here as (15a)) the complement clause is dissociated from the matrix predicate. Following the reasoning outlined above, I propose to base-generate (15a) as in (15b). (15) a.

Freddy probeert van den auto te repareren. Freddy tries of the car to fix ‘Freddy tries to fix the car.’ PP

b. Spec

[PRO den auto te repareren]

P′ P van

RootP DP

Root′

Freddy

Root probeer

Here too, van is base-generated (or merged) as a preposition. Its complement is a Root Phrase containing (the root of) the matrix predicate and the matrix subject, while the complement clause is merged in SpecPP.10 In this configuration, just

49

50


like in (14b), van establishes a dissociation relation between its Spec and its complement, though this time on an aspectual level. The complement clause (the fixing of the car) is dissociated from the matrix predicate (Freddy’s trying). In the next step of the derivation this PP is merged with a CP the head of which is marked for focus. CP

(15) c.

C′

Spec [Freddy probeer]i

PP

C[+Foc] vank

Spec [PRO den auto te repareren]

P′ ti

tk

The RootP moves to SpecCP. Van has to move along for reasons of equidistance.11 Movement of RootP to SpecCP takes place for interpretive reasons: the matrix predicate gets (restrictively) focused with respect to the complement clause. The sentence thus stresses that Freddy’s efforts will constitute but an attempt. In the final step of the derivation, the matrix subject and predicate move out of SpecCP to higher structural positions. (15) d.

vP Spec Freddym

v′

v probeerl

CP

Spec [tm t1]i

C′ C vank

PP

Spec [PRO den auto te repareren]

P′ tk

ti

The CP is merged with a vP and the root probeer incorporates into v.12 Freddy moves to SpecvP. Thus Freddy and probeer get interpreted as the matrix subject and the matrix predicate respectively.


The analysis outlined in (15b–d) is strongly reminiscent of Kayne’s (1999) analysis of French complementizer de and Italian complementizer di (cf. also Section 6, where I will come back to these complementizers). Briefly put, the main idea of Kayne’s article is that de/di does not form a single constituent with the infinitival clause. Instead, it is merged later in the derivation and attracts the infinitival clause to its Spec-position in order to license it. Subsequent movements then lead to the surface word order. There are both similarities and differences between Kayne’s proposal and mine. I agree with Kayne in that the prepositional complementizer does not form a single constituent with the complement clause. Instead, that complement clause has to be licensed in the Spec-position of the complementizer. However, I do not take that licensing relation to be a result of Attract+Move, but of Merge. The infinitival clause is base-generated in the Spec-position of van. Furthermore, I have made the licensing relation Kayne mentions more explicit by linking the syntax of van-sentences to their semantics. That same idea also enabled me to find independent motivations for all the instances of Move in my analysis, thus neutralizing a possible objection one might have against Kayne’s approach. However, as the observant reader undoubtedly has noticed, there is still a considerable problem for the analysis in (15). Although it captures the data pertaining to control verbs quite nicely, it is unclear how it would work for the van-complements of raising and epistemic verbs. Recall that when van introduces the complement of these types of verbs, the dissociation is situated on a more abstract (evidential) level. As a result, the matrix predicate in these constructions plays a more peripheral role in the dissociation relation. In the remainder of this section I will show that these observations can nonetheless be reconciled with (a slightly modified version of) the analysis outlined above. A first point to note is that raising and epistemic verbs indeed form a natural class. Rooryck (1997), for example, argues quite extensively for a uniform treatment of raising-to-subject and raising-to-object type of verbs, i.e. of raising and believe-verbs respectively.13 He considers the latter to be the ‘accusative’ counterparts of the former. (16) a. It seemed to all of us that this was wrong. b. We all believed that this was wrong.

Pointing at similarities between the verb to seem and the verbs of comparison in various languages, Rooryck further notes that the semantics of raising verbs (and hence also of believe-verbs) always involve comparison. This is particularly

51

52


clear when the verb to seem selects a finite complement. Rooryck paraphrases a sentence such as (17a) as in (17b). (17) a. It seemed that/like/as if Alfred has eaten his veggies. b. This instance of Alfred having eaten his veggies (= ‘it’) resembles the ‘typical’ instance of Alfred having eaten his veggies.

It is clear that under this approach it in (17a) is not merely an expletive-like pronoun (as was generally assumed), but a full pro-CP serving as the second element of the comparison. A similar idea can be found in Andrea Moro’s (1997) work on predication structures. He argues that it is a predicative element that is base-generated in the bipartite Small Clause-complement of the verb to seem. He analyses a sentence such as (17a) as in (18) (abstracting away from the lexical content of the complement clause). S

(18)

VP

iti V seems

SC S

ti

In analysing van-complements of raising and epistemic verbs, I want to capitalize on two aspects of the proposals just sketched. The first one is the observation that the semantics of raising and epistemic verbs involve comparison. In terms of the analysis outlined in (15) I take this to mean that the CP-layer introduced in (15c) is now no longer one that is marked for focus (a FocP if you like), but rather a CP-layer involving comparison. That such a layer indeed exists in Dutch is independently argued for by Hoekstra (1993b) (cf. also Hoekstra 1993a, Zwart 2000). The only element still missing from my analysis now is the constituent moving into the Spec-position of this CP-layer. Clearly this cannot be the matrix predicate like in (15), as it is not the matrix predicate that is compared to the complement clause. Here I want to make use of another aspect of Rooryck’s and Moro’s proposals, namely the fact that seem-clauses crucially consist of three elements: the matrix verb, the complement clause and a pronominal-like element.14 Following this line of reasoning I propose to analyse a sentence such as (19a) as in (19b).15


(19) a.

Hij schijnt van da graag te doen. he seems of that gladly to do ‘He seems to enjoy doing that.’ VP

b.

CP

V seem Spec pro-IP

C′ C vank

PP Spec [hij da graag te doen]

P′ tk

tpro-IP

Once again van is merged as the head of a PP the subject of which is the complement clause. The complement of van this time is a phonologically null pro-IP. In this base-generated configuration the complement clause is dissociated from the pro-IP. That pro-IP then moves to SpecCP and van again moves along to the C-position. Thus, a comparison relation arises between the pro-IP in SpecCP and the complement clause in SpecPP. I analyse the semantics of this CP-layer as in Zwart (2000): the interpretation of the element in SpecCP is restricted by (the meaning of) the element in the complement of C. The phrase in (20) provides a simple illustration of this idea. (20) groen als gras green like grass ‘as green as grass’

Here too, the element in the complement of the comparative complementizer (i.e. gras) restricts the interpretation of the constituent in SpecCP (groen). It delimits the extent to which (or the way in which) the unnamed subject of this phrase has the property of being green. In terms of the structure in (19b) this means that the complement clause restricts the meaning of the pro-IP in SpecCP. It delimits its interpretation. Taking all of this into account, we can now — in the spirit of Rooryck (1997) — paraphrase the sentence in (19a) and the structure in (19b) as in (21).

53

54


(21) This instance of him enjoying doing that resembles but at the same time differs from the typical instance of him enjoying doing that.

This paraphrase shows quite clearly what van adds to the ‘normal’ meaning of schijnen ‘to seem’. It indicates that the comparison in the complement of schijnen can never reach the point of identity. It casts doubt on the validity of that comparison. Note that this is how the evidential readings of van-complements of raising and epistemic verbs arise. If the comparison in the complement of schijnen is only partially valid, we interpret this as the truth of the embedded proposition being called into question (because the speaker only has indirect evidence for his statement or because he could be mistaken in his beliefs). Thus the structure in (19b) provides a principled structural account of how van induces the readings it does when introducing the complement of a raising or an epistemic verb. Summing up this section, I have proposed an essentially Kaynian analysis for the data presented in the previous section. Van does not form a single constituent with the infinitival clause it introduces, but rather it is merged as a preposition with the complement clause in its Spec-position. Thus I have been able to capitalize on the parallelism between the use of van in the nominal and in the verbal domain. It has now become clear that the precise nature of the dissociation van induces depends on the elements in its Spec and in its complement. When both are DPs the dissociation is spatial, when the first one is an IP and the second a RootP the dissociation is aspectual and when the first one is an IP and the second a pro-IP the dissociation is situated on an evidential level. In the next section, I will introduce yet another type of dissociation induced by van, this time with a CP in its Spec-position and an (optionally) overt pro-CP as its complement. Before doing so, however, I will show that this construction is indeed similar to the Brabant Dutch complementizer van data discussed in Section 2.

4. Quotative van A construction that is becoming exceedingly popular in present-day spoken Dutch (both northern and southern) is the one exemplified in (22). (22) Hij zei van: ik ga naar huis! he said of I go to home ‘He said something like: I’m going home!’ ‘He was like: I’m going home!’16


At first sight this type of construction bears little or no resemblance to the complementizer van sentences discussed above. The clause following van is a quote displaying Verb Second, which in Dutch is a typical characteristic of main clauses. Thus it is not even clear that van functions as a complementizer in this sentence. It seems fair, then, to say that a ‘normal’ quote such as (23a) has much more in common with (22) than the complementizer van sentence in (23b). (23) a.

Hij zei: “Ik ga naar huis!” he said “I go to home ‘He said: “I’m going home!”’ b. Hij schijnt van da plezant te vinden. he seems of that agreeable to find ‘He seems to enjoy that.’

However, there are also some important differences between (22) and (23a). First of all, in the quotative van construction van is often preceded by zo ‘so’, iets ‘something’ or zoiets ‘like that’. In a real quote construction such an element is not allowed. (24) a.

Hij zei (zo) (iets) van: laat me gerust! he said (so (something of leave me in.peace ‘He was like: leave me alone!’ b. Hij zei (*zo) (*iets): “Laat me gerust!” he said *(so *(something “leave me in.peace

Secondly, in a real quote sentence the clause containing the quotative verb can be used as an interjection. This is not possible in a quotative van construction. (25) a.

“Ik ben”, zei hij, “de allerbeste.” “I am said he “the very.best ‘“I am”, so he said, “the very best.”’ b. *Ik ben, zei hij (zoiets) van, de allerbeste. I am said he (something.like.that of the very.best

Thirdly, and probably most importantly, the clause following van in the construction under consideration here is not necessarily a quote. As the examples in (26) show, any type of embedded clause (note the word order) can follow van. (26) a.

Hij zei zo van dat hij daar al eerder was geweest. he said so of that he there already before was been ‘He said something like that he had been there before.’

55

56


b. Hij vroeg zo van hoe da hij daar moest geraken. he asked so of how that he there had.to get ‘He asked something like how he was supposed to get there.’

These facts indicate that there is less similarity between (22) and (23a) than there appeared to be at first. In fact, as the data in (27) show, van behaves more like a true complementizer (such as dat ‘that’ in (27b)) than like a mere element introducing a quote (such as demonstrative dit ‘this’ in (27c)) in that nothing can intervene between van and the ‘embedded’ clause (i.e. the quote). Hij zei ·*vanÒ voor hij wegging ·vanÒ: ik kom niet meer terug. he said ·*of before he left ·of I come not again back ‘Before he left, he said something like: I will not come back.’ b. Hij beweerde ·*datÒ voor hij wegging ·datÒ hij niet meer he claimed ·*that before he left that he not again zou terugkomen. would return ‘Before he left, he said he would not return.’ c. Hij zei ·ditÒ voor hij wegging ·ditÒ: “Ik kom nooit meer terug.” he said ·this before he left ·this “I come never again back ‘Before he left he said this: “I will never return.”’

(27) a.

There is, however, another type of construction in Dutch which does seem to resemble the quotative van sentences exemplified in (22) both from a formal and from a semantic point of view. Consider the data in (28). (28) a.

Jan zei zoiets als: ik kom nooit meer terug. John said something.like.that like I come never again back ‘John said something like: I will never return.’ b. Jan zei zoiets van: ik kom nooit meer terug. John said something.like.that of I come never again back ‘John said something like: I will never return.’

As the (approximative) English translations in (28) show, the meaning of a sentence in which a quote is being introduced by the comparative complementizer als ‘like’ comes very close to that of a quotative van construction. Furthermore, als — just like van — can (and in fact must) be preceded by iets ‘something’ or zoiets ‘something like that’. However, as much alike as the interpretations of the examples in (28) may be, there are still clear semantic differences between them. While the als-sentence in (28a) expresses only comparison and only with respect to the formal properties of the embedded clause, the meaning


of the van-sentence in (28b) is more complex. The paraphrases in (28¢) illustrate this point. (28¢) a.

John uttered words which (formally) resembled the string ik kom nooit meer terug. b. John made a statement the propositional content of which resembled that of the utterance ik kom nooit meer terug, but the form of which was (most likely) different from this string.17

The speaker in (28a) is trying to reproduce the exact words uttered by John. He is using the comparative complementizer als to indicate that he might be slightly off the mark, but still, that is his main intent. The speaker in (28b) on the other hand is trying to represent the content of what John has said (the gist of his statement), while at the same time indicating that he does not want to represent the form of John’s statement. The difference between als and van is not always clear in a context such as (28), where it is hard to discern between form and content of the quoted sentence. There are, however, contexts in which these two can easily be separated. As it turns out, quotative van occurs not only after quotative verbs, but also after verbs expressing thoughts or feelings, and even after the very general hebben ‘to have’. In all these cases there isn’t really a quoted sentence to begin with (assuming our thoughts and feelings to be non-linguistic), so it makes little sense to refer to the formal properties of these clauses. We therefore predict als to be unable to occur in such contexts. This prediction is borne out in (29). Op zo’n moment denk je zoiets van/ ??als: was ik on such moment think you something.like.that of ??like was I maar daar. prt there ‘At times like those you’re like: I wish I were there.’ b. Dan heb ik zoiets van/*als: laat me gerust! then have I something.like.that of like leave me in.peace ‘Then I’m like: leave me alone!’

(29) a.

An even more striking case in point comes from a construction in various dialects of southern Dutch (Vanacker 1985). It has the form het is (iet) van + CP (‘it is something of ’ + CP) where the CP contains a verb in the imperative. Consider the data in (30).

57

58


(30) a.

’t Is van/*als bedriegt den boer. it is of *like cheat (imp) the farmer ‘Someone is cheating.’ b. ’t Is van/*als likt mijn lipken. it is of *like lick (imp) my lip (dim) ‘The food is very nice.’ c. ’t Is van/*als laat de kat maar zeiken. it is of *like let (imp) the cat prt shit ‘Don’t worry.’

The English translations clearly show that the embedded CPs in (30) should in no way be interpreted as literal quotes. Rather, their content contains elements the speaker wants to use to describe the situation. The message he wants to convey resembles in some way the propositional content of the embedded CP. At the same time, however, he is dissociating himself from the formal properties of the embedded ‘quote’. As the speaker is clearly not interested in the literal (and hence formally accurate) reading of the embedded CP, we expect the use of als to be unacceptable in this construction, which it is. I have now provided ample evidence in support of the claim that the van construction under consideration in this section is different from quote sentences and is only partially similar to als-sentences. When using a quotative van construction a speaker wants to convey a message which in content resembles but in form differs from the sentence he appears to be quoting. Note that there are two elements here: comparison and dissociation. This brings the quotative van construction very close to Brabant Dutch complementizer van in the complement of raising and epistemic verbs. A further indication that this comparison is on the right track is the observation that quotative van, just like Brabant Dutch complementizer van, is incompatible with factive verbs. (31) a. *Hij verklapte (zo) (iets) van: ik kom ook naar het feest. he gave.away (so (something of I come too to the party b. *Hij verklapte (zo) (iets) van dat hij ook naar het feest kwam. he gave.away (so (something of that he too to the party came

The natural conclusion then is that quotative van constructions should be analysed in exactly the same way as van complements of raising and epistemic verbs. This means that a sentence such as (32a) receives the structural analysis in (32b).


(32) a.

Jan zei (zo)(iets) van: ik kom nooit meer terug. John said (something.like.that of I come never again back ‘John said something like: I will never return.’ VP

b. V zei

CP Spec (zo)(iets)i

C′ C vank

PP Spec [ik kom nooit meer terug]

P′ tk

ti

Once again, van is merged as a preposition. Its Spec-position is filled with the embedded clause (a CP this time), while the pronominal-like element zoiets — which I take to be a pro-CP — is merged as the complement of van. In this base-generated configuration the embedded clause (the quote) is (formally) dissociated from zoiets. Note that this is a fourth type of dissociation induced by van. When the Spec-position of van is filled with a CP and its complement with a pro-CP, the dissociation induced by van concerns the formal properties of the embedded clause. Next in the derivation, the pro-CP moves to SpecCP (and van to C), where it enters into a comparison relation with the clause in SpecPP. In this configuration, the interpretation of zoiets is restricted by (the meaning of) the clause ik kom nooit meer terug.18 Before rounding off this section it is worth noting that the analysis just outlined for quotative van is also applicable to the (very similar) construction exemplified in (33).19 (33) a.

Ik denk van wel. I think of so ‘I think so.’ b. Hij zei van ja. he said of yes ‘He said so.’

59

60


c.

Het schijnt van thuis. it seems of home ‘Home, it seems.’ d. Wat is dat voor een stof? Ik geloof van een zuur. what is that for a substance I believe of an acid “What kind of substance is that?” “I believe it is an acid (of some sort)”

This construction has the same semantics as the quotative van construction (or at least a highly similar one), it cannot occur in the complement of factive verbs (cf. Barbiers 2000) and it can have both quote-like (33b) and non-quote-like (33a) complements. Therefore I propose to analyse this construction exactly along the lines of (32b), the only difference being that in (33) a constituent of the clause in SpecPP has been fronted after which the rest of the clause has been deleted. In this section I have shown that the so-called quotative van construction bears a great resemblance to the Brabant Dutch complementizer van data discussed in Section 2. Therefore I analysed this construction exactly along the lines of the proposal outlined in Section 3. Thus I came across a fourth type of dissociation. When its Spec-position is filled with a CP and its complement with a pro-CP, the dissociation van induces concerns the formal properties of the embedded clause.

5. Complementizer van in non-Brabant southern Dutch So far, I have only discussed constructions in which van induces a dissociation reading. However, in doing so, I have ignored the fact that a large part of the southern Dutch dialects (the majority in fact) uses van in more or less the same way as om is used in standard Dutch, i.e. as a completely optional (and hence meaningless) complementizer. This raises the question of how van should be analyzed in these dialects, especially since I have been assuming throughout this article that there is a close parallelism between the syntax and the semantics of van constructions. Note that it is not difficult to interpret the southern Dutch state of affairs concerning complementizer van from a more or less functional or pragmatic point of view. Suppose the use of van as a complementizer originated in the Brabant dialects of southern Dutch.20 It is a well-known observation from the linguistic literature that these dialects exert great influence on the regional variety of Dutch spoken in the entire southern Dutch area (the so-called Algemeen Zuidnederlands ‘general southern Dutch’ or verkavelingsvlaams


‘allotment Flemish’, see Goossens 1970 and Debrabandere 1986). It is not surprising, then, that a typically Brabant feature of southern Dutch, such as complementizer van, should find its way into the rest of the southern Dutch area. In the process of language contact accompanying this spreading of van, the more subtle features of the construction (such as the dissociation reading van induces or the fact that it can occur after raising verbs) got lost. That explains how in the non-Brabant part of Dutch-speaking Belgium, van became a dialectal variant of standard Dutch om. However, appealing though such an account may be, it does not provide us with any insights as to how van functions in the (mental) grammars of the non-Brabant speakers of southern Dutch. Therefore it is not a satisfactory explanation of the data. The most striking difference between the Brabant and the non-Brabant use of complementizer van is van’s lack of meaning in the latter. Looking at the situation from a slightly broader perspective, one could say that van has gone through three stages. At first it was a mere preposition, then it started functioning as a complementizer but still retained its semantic features and finally, it lost those features and became a mere grammatical element. This evolution has all the characteristics of a grammaticalization process. Though working in a different context, Klamer (1998) reaches a similar conclusion when discussing (the historical evolution of) Tukang Besi kua and Buru fen.21 These two elements diachronically derive from verbs meaning to say, and in present-day Tukang Besi and Buru they can be used as quote markers or as complementizers. Klamer summarizes the evolution of fen and kua in a schema such as (34) (= Klamer 1998, (51)) which she calls a ‘grammaticalization chain’ (following Heine 1992, 1993). This grammaticalization chain should in fact be seen as one single (complex) linguistic category, which explains why some of the stages in (34) can co-occur in present-day Tukang Besi and Buru. (34) Stage: Meaning:

I verbal say

II verbal/grammatical say, quote marker

III grammatical conjunction

This approach is very suitable for analyzing the complementizer van facts I have been discussing in this section. In (35) I have adopted and adapted Klamer’s structure for van.22,23

61

62


(35) Stage: Function:

I prepositional

Dissociation: Structural position:

II prepositional grammatical yes merged as a P, moved to C

yes merged as a P, not moved

III grammatical no merged as a C, not moved

The grammaticalization chain of van contains three stages all of which occur in (certain varieties of) present-day Dutch. The first stage is exemplified in (36). In this sentence van is used purely as a preposition. Structurally speaking, it is merged under the P-node and does not move from that position. It retains its full lexical meaning. The sentence in (37) is an example of stage II. It is the type of van construction I have discussed extensively in Sections 2 and 3. I have argued that van is merged as a P, but later in the derivation moves to the C-node (thus having both a prepositional and a grammatical function).24 It has retained its lexical meaning, as it induces a dissociation reading on the sentence. The example in (38) (i.e. the same sentence as in (37) but this time in nonBrabant Dutch) is a representative of stage III. It is analyzed in (39). (36) Jan komt van Brussel. John comes of Brussels ‘John comes from Brussels.’ (37) Jan probeert van da te doen. John tries of that to do ‘John tries to do that.’

(Brabant Dutch)

(38) Jan probeert van da te doen. John tries of that to do ‘John tries to do that.’

(non-Brabant Dutch)

CP

(39) Spec [Jan probeer]i

C′ C van

XP

Spec [PRO da te doen]

X′ X

ti


In (39) van is base-generated (or merged) as a C. It no longer derives from an underlying P-position and hence does not induce a dissociation reading. It has reached the end stage of the grammaticalization process and is now a fullyfledged complementizer. An important result of this approach is that the underlying constituent can now no longer be a PP. I will not go into the precise categorial status of this phrase (that is why I have dubbed it XP in (39)), but it seems to be some kind of predication structure.25 The complement clause is said to be an attempt by John. The paraphrase in (40b) of the sentence in (40a) illustrates this point. (40) a. John tries to fix the car. b. ‘To fix the car’ is an attempt by John.

In this section I have given an account of the non-Brabant use of complementizer van. I have argued that it represents the end of a grammaticalization process. Van is now no longer a prepositional complementizer — in the structural sense I have been using this notion — but a fully-fledged complementizer.

6. Complementizer van cross-linguistically So far, I have not made an appeal to language-specific mechanisms or principles in my discussion of various van-constructions. Thus we predict the proposal outlined above to be valid for van-constructions cross-linguistically. Consider in this respect the French and Italian data in (41)–(42) (cf. Kayne 1999). (41) Jean a essayé de chanter. John has tried of sing (inf) ‘John has tried to sing.’ (42) Gianni ha tentato di cantare. John has tried of sing (inf) ‘John has tried to sing.’

The most striking difference between (41)–(42) and the Brabant Dutch complementizer van data discussed earlier concerns once again the semantics of the construction. De/di does not induce a dissociation reading on the complement clause. In light of the discussion in the previous section I propose to analyze French de and Italian di as more grammaticalized variants of Brabant Dutch complementizer van. Thus the sentences in (41)–(42) receive a structural analysis completely parallel to that of non-Brabant Dutch complementizer van.

63

64


CP

(43) Spec [Jean essayé]i [Gianni tentato]

C′ C de/di

XP Spec [chanter] [cantare]

X′ X

ti

7. Conclusion This article was devoted to various constructions containing the Dutch word van. I have approached the data from two distinct though related viewpoints. On the one hand I focused on the use of van as a marker of dissociation. I argued that the precise nature of the dissociation induced by van depends on the elements in its Spec-position and in its complement. The table in (44) gives an overview of the different possibilities I discussed. (44) Element in Specvan DP IP IP CP

Element in Compvan

Type of dissociation

DP RootP pro-IP pro-CP

Spatial Aspectual Evidential Formal

On the other hand I centered on complementizer van-constructions as being part of a grammaticalization process, whereby van moves from the lexical (P-)domain into the functional (C-)domain. This made it possible to account for the non-Brabant use of complementizer van as well as for van-constructions cross-linguistically. The ideas presented in this article should be seen as part of a larger research program. It is a well-known observation that van — as well as many of its counterparts in other languages — is a very versatile word that occurs in many seemingly unrelated constructions. Rather than treating all these different occurrences of van as homonyms, I have shown that it proves more satisfactory — and more challenging — to try and relate them all to each other. As this line


of reasoning leads to a more restrictive and insightful theory, this is clearly an interesting way to proceed.

Notes * This article is a continuation (in English) of Van Craenenbroeck (2000). I want to thank Lisa Cheng, Sjef Barbiers, Crit Cremers, Marjo van Koppen, Ineke van der Meulen, Gertjan Postma, Johan Rooryck, Guido vanden Wyngaerd and the audience of my TIN-dag-talk in Utrecht (5 February 2000) for insightful comments and stimulating discussion. All errors and shortcomings are my own. 1. The precise demarcation of the area in which van has a wider distribution, is not yet clear to me. Most of the dialects that show this broader use are clearly from the province of Flemish Brabant (the area around Brussels), but there seem to be offshoots in Antwerp and Limburg. 2. As the meaning differences I discuss in this section are often very subtle, I will not try to express them in my English translations of the example sentences. These are only meant to represent the overall meaning of the sentence. 3. Note that (3a) in isolation is ambiguous between the two readings. It is only when contrasted with a van-sentence — which selects the ‘indirect evidence’-reading — that it becomes biased towards the other interpretation. 4. The difference between direct and indirect readings of raising verbs was first noticed — though in a different context — by Tasmowski (1989). 5. Note that the a-sentence in isolation is once again ambiguous between the two readings (see also note 3). 6. Strictly speaking, epistemic verbs are also control verbs (cf. Iki meen PROi de beste te zijn. ‘I believe to be the best’). I distinguish between the two here because epistemic verbs behave unlike ‘regular’ control verbs with respect to distribution (cf. supra) and with respect to interpretation (cf. Section 3). 7. The only context I can think of here is one where the speaker is being ironic or even sarcastic: ‘John will now try to fix the car. That will take him about ten minutes and then he will call a mechanic.’ 8. The massive Dutch Woordenboek der Nederlandsche Taal (Dictionary of the Dutch Language) gives a brief survey of the diffrent opinions concerning the origins of van (WNT, XVIII, 381–382). Van Ginneken (1935) argues quite convincingly that van derives from a combination of af ‘from’ and aan ‘near, to’. 9. For reasons not entirely clear to me, judgments by native speakers with factive verbs and complementizer van are not always as strong as I would expect. The sometimes shaky judgments might be due to influence from the other southern Dutch van-dialects (cf. Section 5). 10. I will not elaborate any further on the precise categorial status of the RootP. It represents a (lexical) phrase in which both nominal and verbal roots can be base-generated. The main advantage of adopting such a phrase is the fact that one can now derive both (i) and (ii) from

65

66


the same underlying structure. As these constructions share the same semantics, this is clearly a desirable result. (i) Jan probeert van da te doen. John tries of that to do ‘John tries to do that.’ (ii) Jans poging van da te doen John’s attempt of that to do ‘John’s attempt to do that’ 11. Or because of some more general principle stating that whenever there is A¢-movement to the specifier of an operator head, that head position must be filled. Note that this movement gives the Kaynian (1999) term ‘prepositional complementizer’ a structural interpretation: a prepositional complementizer is an element that is merged as a preposition and subsequently moves to a C-position. 12. I consider a sentence such as (i) evidence in favor of the possibility of merging a CP with v. The verb maken in (i) then is nothing but an overt spell-out of v. (i) Jan maakt dat Marie blij is. John makes that Marie happy is ‘John causes Mary to be happy.’ 13. Recall that the epistemic verbs are the Dutch counterparts of the English believe-verbs (Bennis and Hoekstra 1989: 131). 14. Note that these proposals might also be compatible with Kayne’s (1999: 56) suggestion that il in (i) is merged as a “double” of the infinitival phrase. (i) Il est possible de comprendre cette question. it is possible of understand (inf) that question ‘It is possible to understand that question.’ 15. I only give the analysis for raising verbs. That of epistemic verbs is essentially the same. Note also that I have not represented the raising of hij ‘he’ to the matrix subject position. 16. The English translations here and in the following examples are but approximations. I will come back to the exact meaning of this construction further on in this section. 17. Note that I am arguing against Verkuyl (1977, 1979) here. He claims that this use of van (which he calls ‘performatief van’) indicates that the speaker is only interested in the strictly formal properties of the sentence he quotes. In what follows I will show that this is not the case. 18. Note incidentally that a quotative van sentence such as (i) might also provide an (indirect) argument in favor of an admittedly rather unusual aspect of the analyses in (15), (19) and (32), namely the fact that a PP is merged with a C. If we were to assume a ‘normal’ sentence structure for (i), with multiple CP-layers hosting the various complementizers (van and da) in the spirit of Hoekstra (1993b), it would be extremely exceptional for more than one SpecCP-position in this multiple CP-structure to be overtly filled. As far as I am aware, this type of construction is unattested in (any dialect of) Dutch.


(i) Hij vroeg zo van hoe da hij daar moest geraken. he asked so of how that he there had.to get ‘He asked something like how he was supposed to get there.’ Furthermore, assuming a sentence such as (ii) to be base-generated as a Small Clause-PP, the structure of this clause too would involve a PP dominated by several functional layers, one of which would be CP. (ii) John is in the garden. 19. The sentence in (33a) is considered standard Dutch, (33b) is a typically southern Dutch variant and the examples in (33c–d), which are taken from the previously mentioned Dictionary of the Dutch Language (WNT, XVIII, 406), are substandard. 20. Possibly due to the influence of the nearby French-speaking part of the country. Recall that de, the French counterpart of van, is the default infinitival complementizer in French (cf. also Section 6). 21. Tukang Besi and Buru are Malayo-Polynesian languages. The former is spoken in Southeast Sulawesi and the latter on the island of Buru in Central Maluku (Klamer 1998: 1). 22. Note incidentally that the parallelism between fen/kua and van may be greater than suggested in the main text. First of all, both van and fen/kua can be used both as a quote marker and as a complementizer. Secondly, when used as a quote marker, both van and fen/kua can follow verbs of mental perception as well as regular quote verbs. Thirdly, Klamer argues that when kua is used as a complementizer, it induces an indirect reading on the complement clause (Klamer 1998: 14)! However, since fen and kua are derived from verbs and not from prepositions, I will not pursue this parallelism any further. 23. Postma (1998) uses the Klamer (1998) data in his analysis of the so-called quotative inversion construction in Dutch. His proposal might be compatible with my analysis of quotative van. Also highly interesting is Postma’s further claim that quote readings and focus compete for the same structural position. This idea might corroborate the two-way split I have proposed in my analysis of van constructions (between the focus and the non-focus variant). 24. The idea that grammaticalization should be structurally analysed as movement from the lexical to the functional domain is one that is frequently found among historical linguists. See for example Postma (1999) and Van Kemenade (2000). 25. It is possible that in this configuration, too, the head of the underlying phrase (in this case X) has to move up to C in order to license the movement of the matrix predicate past the constituent in SpecXP.

67

Expletive subjects in subject relative clauses* Line Mikkelsen University of California, Santa Cruz

1.

Introduction

The goal of this paper is to provide an analysis of subject relative clauses in Danish that maximizes coherence with other constructions in the language, in particular impersonal constructions, while making restrictive assumptions about phrase structure and minimal appeal to lexical specification. The data I am concerned with is exemplified in (1), which is borrowed from Vikner (1991: 112). (1) Vi kender mange lingvister som at der vil læse denne bog. we know many linguists who that there will read this book ‘We know many linguists who will read this book.’

The focus is on the three elements som, at and der, which appear between the external nominal head and the finite verb of the relative clause. For convenience I refer to this domain as the left periphery of the relative clause. The left periphery of a subject relative clause may be realized by certain combinations of som, at, der, while other combinations are impossible. Characterizing the possible realizations is a challenge for syntactic theory, and there have been several proposals in the generative literature including Jacobsen and Jensen (1982), Erteschik-Shir (1984), and Vikner (1991). The discussion has centered on the categorial status of each of the three words som, at, der, and this will also play a crucial role in the analysis developed below. Jacobsen and Jensen (1982), working within the framework of Chomsky (1977) and Chomsky and Lasnik (1977), claim that som is a wh-word, while at der is a “two-word complementizer with no internal structure” (Jacobsen and Jensen 1982: 8). Erteschik-Shir (1984), working within the Government and Binding framework of Chomsky

70

Line Mikkelsen

(1982), argues that som and at are complementizers, while der is an expletive functioning as a subject filler. The restrictions on the possible combinations of som, at and der are derived from conditions on operator adjacency, co-superscripting, proper government and the ECP. Finally, Vikner (1991), working within the Relativized Minimality framework of Rizzi (1990), argues that each of the three elements is a lexical complementizer, each projecting a CP. The distribution and internal ordering of som, at and der is argued to follow from idiosyncratic, lexical properties of each of the three heads. In this paper I combine insights from each of these analyses proposing an analysis that maximizes coherence with the analysis of impersonal constructions, while maintaining a restrictive theory of clause structure. The strategy is to locate the simplest possible assumptions about the categorial status of each of the three elements, som, at and der, that are compatible with this restrictive theory of clause structure, and see how much of the data can be accounted for without introducing any additional descriptive technology. The analysis makes minimal appeal to lexical specification, and no appeal to government or adjacency. The proposal is in outline that som is an invariant operator — an overt counterpart to the invariant null operators frequently postulated in analyses of A-bar movement — at a complementizer and der an expletive element. Under this proposal the subject relative clause in (1) receives the analysis in (2). (2) … [DP mange lingvister [CP somi at [IP der I0 [VP ti vil læse denne bog]]]]

where som occupies SpecCP, at is in the head position of CP, and the expletive der is in SpecIP.1 The analysis accounts for the distribution of som, at and der in relative clauses, as well as the realization of the left periphery of subject relative clauses. In addition to these descriptive advantages the analysis opens up the possibility of a unified analysis of relative clauses and impersonal (or expletive) constructions, avoiding the lexical ambiguity of der stipulated in Jacobsen and Jensen (1982), and Vikner (1991). The main obstacle to a unified analysis of expletive and relative der is that expletive constructions and relative clauses differ with respect to definiteness and transitivity restrictions. Without attempting a complete analysis of these differences, I show how they are consistent with a broader set of observations about definiteness effects in relative clauses and Case satisfaction under A-bar movement. The paper is organized as follows. Section 1.1 outlines the basic claims of the analysis and the facts to be accounted for. In Section 2 I show how the analysis accounts for the distribution of som, at, and der in relative clauses. In Section 3

Expletive subjects in subject relative clauses

I turn to a more detailed examination of subject relative clauses showing how the Extended Projection Principle plays a decisive role in the realization of their left periphery. In Section 4 I consider the possibility of a unified analysis of relative clauses and expletive constructions, and discuss some advantages and challenges to a unified analysis. Finally, Section 5 summarizes the paper. 1.1 Basic facts and assumptions The analysis makes the following claims about the categorical status of som, at, and der in relative clauses: (3) a. som is an invariant operator2 b. at is a complementizer c. der is an expletive3

The distributional facts to be accounted for are as follows: (4) a. b. c. d. e.

der occurs only in subject relative clauses som occurs in subject and non-subject relative clauses som does not participate in pied-piping when som, at, and der are all present they must appear in that order either som or der must be present in a subject relative clause

In the following sections I show how these facts can be accounted for under the categorial status assigned to som, at, and der. I assume a restrictive theory of clause structure, where specifiers are unique and where there is no CP recursion in relative clauses (contra Vikner (1991)). Thus the phrase structure of the left periphery of relative clauses is as in (5). (5)

CP C′ C

IP I′

I further assume the Extended Projection Principle (EPP), which in its most general formulation says that SpecIP must be filled. I do not make any specific assumptions about its technical implementation, since I believe these to be irrelevant for the proposal I am making here. I further assume that there is

71

72

Line Mikkelsen

operator movement to SpecCP in relative clauses, again without making any specific assumptions about what drives this movement (such as feature checking or the wh-criterion), though I assume that it is an instance of A-bar movement.

2. The syntactic category and distribution of som, at, and der In this section I show how their respective syntactic categories affect the distribution of som, at, and der in relative clauses. I start with the distribution of der in Section 2.1, then turn to som in Section 2.2, and finally at in Section 2.3. I conclude with a note on linear order in Section 2.4. 2.1 Der occurs only in subject relative clauses As illustrated in (6) below, der occurs in subject relative clauses (6a), but not in object relative clauses (6b), and also not in relative clauses where the relativized constituent is a PP, irrespective of whether the preposition has been stranded as in (6c) or pied-piped as in (6d). (6) Vi kender de lingvister … we know the linguists a. der vil læse denne bog. there will read this book b. *der han vil besøge. there he will visit c. *der han spurgte efter. there he asked after d. *efter der han spurgte. after there he asked

[subject relative] [object relative] [PP relative, no pied-piping] [PP relative, pied-piping]

The fact that der only appears in subject relative clauses is directly related to the expletive status assigned to der.4 The expletive der occurs only in the position targeted by the EPP, which I, following convention, take to be SpecIP. In nonsubject relative clauses that position is occupied by the thematic subject argument, i.e., the pronoun han (‘he’) in (6b–d). In a subject relative clause, on the other hand, the thematic subject argument is the operator, which occurs in SpecCP at surface structure. Following much recent work, I assume that all arguments originate within the lexical layer of clausal structure. More specifically, I assume that the external argument of a transitive verb originates in the


highest position within the lexical layer — SpecVP in older conceptions (Sportiche 1988), the specifier of v or µ in more recent conceptions (Chomsky 1995; Hale and Keyser 1993; Kratzer 1994; Kratzer 1996). I further assume that the operator in a subject relative clause may move directly from its thematic position to SpecCP.5 Under exactly those circumstances, the expletive der is inserted in SpecIP to satisfy the EPP. This is the analysis of the subject relative clause in (6b), given in (7). (7) … de lingvisteri [CP OPi [IP der [VP ti vil læse denne bog]]] the linguists there will read this book

It is this part of the proposal that opens up the possibility of unifying the analysis of relative clauses and impersonal constructions. I return to this in more detail in Section 4. 2.2 Som is an invariant operator There are two reasons to believe that som is an invariant operator. First, its distribution is strikingly similar to that of the invariant null operator OP, and second, som cannot cooccur with other operators, in particular som cannot co-occur with wh-operators like hvad ‘what’. I develop these two points in turn below. 2.2.1 The distribution of som and OP The distribution of som parallels that of the invariant null operator OP, while differing from the distribution of wh-operators in certain respects.The parallel distribution of som and OP supports the claim that som is an invariant operator, while the distributional differences between som and wh-operators support the distinction drawn between the two above (cf. footnote 2). There are at least five points of similarity between som and OP. These are listed in (8). (8) som and OP both a. occur in subject and non-subject relative clauses b. occur with any external head irrespective of its animacy, gender, number, etc. c. allow preposition standing d. disallow pied-piping e. lack a genitive form

I do not attempt to explain why invariant operators have exactly this distribution. For present purposes the goal is to establish that som is an invariant

73

74

Line Mikkelsen

operator, and then make the methodological assumption that whatever ultimately accounts for the distribution of null invariant operators will account for the distribution of som as well. Invariant operators in subject and non-subject relative clauses. As illustrated in (9) below, som occurs in subject as well as non-subject relative clauses: (9) Vi kender de lingvister … we know the linguists a. som vil læse denne bog. who will read this book b. som han vil besøge. who he will visit c. som han spurgte efter. who he asked after

[subject relative] [object relative] [PP relative, no pied-piping]

This is also true of the invariant null operator OP, with the important additional requirement that when OP occurs in a subject relative clause der must be present. I return to this fact section in 3, where I propose that it can be explained in terms of the Extended Projection Principle. For now, concentrate on the fact that OP, like som, can occur in subject and non-subject relative clauses, as show in (10). (10) Vi kender de lingvisteri … we know the linguists a. OPi der ti vil læse denne bog. there will read this book b. OPi han vil besøge ti. he will visit c. OPi han spurgte efter ti. he asked after

[subject relative] [object relative] [PP relative, no pied-piping]

As operators, som and OP originate in VP-internal position and move to SpecCP. The VP-internal position of the operator depends on the thematic properties of the operator; SpecVP for subject relatives and sister of V and P for object and PP relatives, respectively. Under these assumptions relative clauses with som and OP receive similar structural analyses, as illustrated for object relative clauses in (11). (11) … de lingvisteri the linguists a. [CP somi [IP han [VP vil besøge ti]]] he will visit


b. [CP OPi [IP han [VP vil besøge ti]]] he will visit

Lack of restrictions on external head. Unlike wh-operators, invariant operators are not specified for semantic and syntactic properties like animacy, gender, number, and case, and they thus occur with external nominal heads irrespective of the semantic and syntactic properties of that nominal. (The following abbreviations are used in the glosses: pl = plural, sg = singular, neu = neuter gender, com = common gender.) (12) … de folki /den pigei /det husi /den byi the.pl people/the.sg girl /the.neu house/the.com house a. somi han vil besøge ti. he will visit b. OPi han vil besøge ti. he will visit

Preposition stranding and pied-piping. Danish generally allows preposition stranding as well as pied-piping under A-bar movement: (13) Hvemi gav du bogen til ti? who gave you book.def to ‘Who did you give the book to?’ (14) [Til hvem]i gav du bogen ti? [to who gave you book.def ‘To who(m) did you give the book?’

However, som and OP differ from wh-operators with respect to preposition stranding and pied-piping in relative clauses. The invariant operators som and OP are perfect under preposition stranding, while wh-operators are at best marginal (Vikner 1991: 112). (15) … de lingvister the linguists til ti. a. somi han gav bogen he gave book.def to b. OPi han gav bogen til ti. he gav book.def to c. ??hvemi han gav bogen til ti. he gave book.def to

75

76

Line Mikkelsen

Conversely, invariant operators do not participate in pied-piping, while wh-operators do: (16) … de lingvisteri the linguists tj. a. *[til somi]j han gav bogen [to he gave book.def b. *[til OPi]j han gav bogen t j. [to he gave book.def c. [til hvemi]j han gav bogen tj. [to who he gave book.def

Note that in English the invariant null operator also does not participate in pied-piping, while overt wh-operators do: (17) This is the booki a. *[in OPi]j I read it tj. b. [in whichi]j I read it tj.

Whatever the explanation for the inability of invariant operators to participate in pied-piping, the identical behavior of som and OP with respect to pied-piping and preposition stranding lends strong support to the claim that som is an invariant operator, one which differs from OP only in being overt. No invariant operators in genitive relative clauses. Invariant operators do not occur in genitive relative clauses, while wh-operators do: (18) Jeg kender en forfatteri I know an author a. *[somi bog]j du har læst tj. book you have read b. *[OPi bog]j du har læst tj. book you have read c. [hvisi bog]j du har læst tj. [whose book you have read

Again, a similar contrast is found in English between the invariant null operator and overt wh-operators: (19) I know an authori a. *[OPi book]j you have read tj. b. [whosei book]j you have read tj.


2.2.2 Co-occurence with wh-operators The second argument for som being an invariant operator comes from the fact that it cannot co-occur with wh-operators like hvad ‘what’, hvem ‘how’, and hvis ‘whose’. Interestingly, der differs from som in being able to co-occur with these wh-operators. This is entirely consistent with the analysis propsed here, where der is an expletive element occuring in SpecIP. As shown in (20)–(22) there are at least three instances where der co-occurs with a wh-operator. In none of these can som occur. Embedded subject relatives (example from Vikner (1991: 119)): (20) Jeg ved ikke … I know not a. hvem du tror der har gjort det. who you think there has done it b. *hvem du tror som har gjort det. who you think has done it

Genitive subject relative clauses (example from Vikner (1991: 111), see also Erteschik-Shir (1984)):6 (21) Jeg kender en pige … I know a girl a. ?[hvis hund] der spiser æbler. [whose dog there eats apples b. *[hvis hund] som spiser æbler. [whose dog there eats apples

Appositional subject relatives with clausal antecedents (example (22a) from the news paper Weekendavisen 03/17/00). (22) … vores eget fly var gået definitivt i stykker our own plane was gone definitively in pieces ‘… our own plane had definitively broken down’ a. hvad der udløste voldsomme protester … what there caused strong protests ‘which caused strong protests … b. *hvad som udløste voldsomme protester … what caused strong protests

Under the analysis developed here, the contrast between der and som in (20)–(22) relates directly to the categorial status of der and som. In relative

77

78

Line Mikkelsen

clauses operators undergo A-bar movement to SpecCP. There is only one SpecCP, since specifiers are unique and there is no CP-recursion in relative clauses. Under these assumptions it follows that som cannot co-occur with a wh-operator, since both must occur in SpecCP, and there is only one SpecCP, which cannot host two operators. The fact that der can co-occur with wh-operators is consistent with the present analysis: as an expletive, der occurs in SpecIP which is an A-position. Thus its presence is not expected to interfere with A-bar movement to SpecCP: (23) … [CP hvemi tror du [CP ti [IP der [VP ti har gjort det?]]]] who think you there has done it (24) … en pigei [CP [hvisi hund]j [IP der [VP tj har spist æblet]]] a girl [whose dog there has eaten apple-def (25) … [CP hvadi [IP der [VP ti udløste voldsomme protester …]]] what there caused strong protester

2.3 The complementizer at Let us finally examine the categorial status of at. The assumption that at is a complementizer is probably the least controversial part of the analysis. It is shared by all previous analyses of Danish relative clauses that I know of, including Jacobsen and Jensen (1982), Erteschik-Shir (1984), Taraldsen (1986, 1991), and Vikner (1991).7 Like the English complementizer that, at occurs in complement clauses and subject clauses: (26) Mikkel sagde [CP at kartoflerne var færdige.] Mikkel said that potatoes.def were done (27) [CP At Skotland kommer med til VM] er helt that Scotland comes with to World.Cup is completely udelukket. out.of.the.question ‘That Scotland makes the World Cup is completely impossible.’

Moreover, subject extraction from embedded clauses containing at gives rise to an impairment of acceptability similar to that attributed to the that-trace effect in English (see also Taraldsen (1986: 161. fn. 9)): (27) Hvemi tror du at hun har inviteret ti til festen? who think you that she has invited t to party.def ‘Who do you think that she has invited to the party?’


(28) ??Hvemi tror du at ti har inviteret hende til festen? who think you that t has invited her to party.def

I will return to the possible relevance of the (th)at-trace effect in the next section, but let us first consider the linear order of the three elements som, at, and der. 2.4 Linear order So far I have argued that som is an invariant operator, at a complementizer, and der an expletive. As an operator, som occurs in SpecCP, the complementizer at occupies the head position of CP, and the expletive der occupies SpecIP: CP

(28) DP som

C′ C at

IP DP

I′

der

The structure in (28) allows us to make a prediction about linear order: whensom, at, and der are all present in a relative clause they must appear in that order. The prediction is borne out, as shown in (29). (29) Vi kender de lingvister … we know the linguists a. *som der at vil læse denne bog. will read this book b. *at der som … c. *at som der … d. *der som at … e. *der at som …

This unique linear order is harder to account for on analyses where som and/or der are also complementizers. In particular, Vikner (1991: 120) has to postulate very specific requirements on co-indexation and spec-head relations to account for unique linear order. On the present analysis, linear order follows directly from the categorial status assigned to each of these elements and standard X-bar theoretic assumptions about the structure of CP.

79

80

Line Mikkelsen

3. EPP effects The left periphery of a subject relative clause may be realized by different combinations of the three elements som, at and, der with varying degrees of grammaticality. Based on the detailed judgements given in Jacobsen and Jensen (1982: 10) I have organized the data into three categories: fully grammatical, grammatical with some degree of degradation (?–??), and ungrammatical (*), as illustrated in (30).8 (30) Vi kender de lingvister we know the linguists a. som … c. ?som der … b. der … d. ?at der … e. ?som at der f. ??som at …

g. *at … h. *…

I will consider the difference between grammatical and ungrammatical realizations, i.e. (30a–f) vs. (30g–h), and then turn to the variation within the set of grammatical realizations. The data in (30) reveal an important descriptive generalization, stated in (31). (31) Descriptive Generalization: either som or der must appear in a subject relative clause.

This requirement is fulfilled in (30a–f) which are all grammatical — though to varying degrees — but not in the ungrammatical (30g–h). I propose that the generalization in (31) is a direct reflection of the Extended Projection Principle (EPP). In a subject relative clause the EPP can be satisfied in two ways. Either the operator moves directly from its thematic position to SpecCP and the expletive der occurs in SpecIP to satisfy the EPP, or the operator moves to SpecCP via SpecIP, and no expletive is inserted. If the operator is overt, i.e. som, it satisfies the EPP in the intermediate position. I assume that the null operator OP cannot satisfy the EPP, consistent with the fact that Danish does not allow pro-drop. To illustrate the proposal in more detail, let us go through some of the individual cases. Consider first the analysis of (30a): (32) [CP somi [IP ti [VP ti …]]]

Here the invariant operator som moves from its thematic position, SpecVP, to SpecIP, where it satisfies the EPP. From SpecIP it moves to SpecCP. This is a familiar instance of A-movement followed by A-bar movement of the (head of


the) resulting A-chain. Since som satisfies the EPP in its intermediate position no expletive is inserted. Note also that the head position of the CP is empty. This is consistent with the descriptive generalization that at is generally optional in embedded clauses, similar to the situation with English that. Here I remain agnostic as to whether this is due to optional deletion of at or alternation with a null complementizer. Next consider the analysis of (30b) given in (33). (33) [CP OPi [IP der [VP ti …]]]

Here the null operator has moved directly from its thematic position to SpecCP, and the expletive der occurs in SpecIP, satisfying the EPP. Again at is absent, indicating either deletion or a null C head. Consider next the ungrammatical (30g). Given what we established about direct and indirect operator movement above, there are two possible derivations for (30g). These are given in (34). (34) a. *[CP OPi at [IP [VP ti …]]] b. *[CP OPi at [IP ti [VP ti …]]]

In (34a) the empty operator has moved directly from SpecVP to SpecCP leaving SpecIP empty, resulting in an EPP violation. In (34b) the empty operator has moved to SpecCP via SpecIP, leaving a trace in the intermediate position. The question is, why the derivation in (34b) is not grammatical. The answer I would like to pursue here is that like (34a), (34b) violates the EPP. The crucial assumption behind this answer is that the null operator cannot satisfy the EPP. Support for this assumption comes from a fact about English null operator constructions first noted by Arlene Berman, and later discussed in detail by Stowell (1984) and Browning (1987). Berman’s observation is that the trace of an infinitival null operator may never originate in SpecIP: (35) *This candidate is hard for [CP OPi [IP people to believe [VP ti to be competent]]]

In (35) the null operator starts out in the specifier of the lower IP. From here it undergoes A-bar movement to the closest SpecCP. There is no obvious reason why the sentence should be ungrammatical, but it is. If we assume that, at least in non-pro-drop languages like English and Danish, null operators cannot satisfy the EPP, the ungrammaticality can be attributed to an EPP violation in the lowest clause. Let us summarize what we have seen so far: either som or der can satisfy the EPP in a subject relative clause. In the first case the operator moves to SpecCP

81

82

Line Mikkelsen

via SpecIP satisfying the EPP in the intermediate position. In the second case the operator moves directly to SpecCP and the expletive is inserted in SpecIP where it satisfies the EPP. If neither som nor der is present the derivation is ruled out by the EPP, since the null operator by assumption cannot satisfy the EPP, and there is no expletive to do the job. Turning now to the analysis of the four cases of intermediate grammaticality (30c–f), let us first note that all of these satisfy the EPP: (36) ?[CP somi[IP der [VP ti …

(= (30c))

(37) ?[CP OPi at [IP der [VP ti …

(= (30d))

(38) ?[CP somi at [IP der [VP ti …

(= (30e))

(39) ??[CP somi at [CP ti

(= (30f))

[CP ti …

In the first three cases, (36)–(38), the EPP is satisfied by the expletive in SpecIP, the operator has moved directly to SpecCP. In (39) the EPP is satisfied by the overt operator som in its intermediate position, and no expletive is inserted. If we first compare (36)–(39) to (32) and (33) above, another descriptive generalization emerges: realizations with more than one overt element in the left periphery are degraded compared to realizations with exactly one overt element (i.e. som or der). I do not have a concrete suggestion as to why this should be, but one could imagine several possible explanations, including economy effects, processing effects, and influence from prescriptive grammars (cf. Vikner 1991: 132, fn. 15). Since I have nothing substantial to say I do not pursue the issue further. I will, however, make a suggestion as to why (39) is worse than (36)–(38). Relying on the observation made in Section 2.3 about (th)at-trace effects in Danish, I want to suggest that relative illformedness of (39) in comparison with (36)–(38) can be understood as reflecting the contribution of the (th)at-trace effect. In (39) at immediately precedes the operator trace in SpecIP, giving rise to a typical (th)at-trace configuration. Such a configuration is not found in any of (36)–(38). In (36) at is not present, and in (37) and (38), where at is present there is no trace in SpecCP, which is instead occupied by the expletive. Non-subject relative clauses. Above I argued that the obligatory presence of either som or der in subject relative clauses is due to the EPP. If this is correct, we expect there to be no such requirement in relative clauses where the EPP is satisfied independently, i.e. in non-subject relative clauses. This is indeed the case, as illustrated by the data in (40).


(40) Vi kender de lingvister … a. [CP OPi [IP han [VP vil besøge ti.]]] he will visit b. [CP OPi [IP han [VP spurgte efter ti.]]] he asked for

[object relative clause] [PP relative clause]

In both cases the thematic subject han satisfies the EPP, and there is no need for an overt operator or expletive.

4. A unified analysis of expletive and relative der Danish has a range of impersonal constructions of the schematic form in (41). See Vikner (1995a: 168ff) for comprehensive analysis and discussion of these constructions. (40) der expletive

v finite verb

(V) (non-finite verb)

DP (XP) DP (adverbial)

The expletive der occurs in the initial position, and the thematic subject DP appears after the rightmost verb. The analysis of der as an expletive in subject relative clauses opens up the possibility of a unified analysis of relative clauses and these impersonal constructions. Compare the subject relative clause in (42) to the embedded impersonal construction in (43). (42) Vi kender de lingvister [CP OP [IP der vil læse denne bog.]] we know the linguists there will read this book ‘We know the linguists that will read this book.’ (43) Vi ved [CP at [IP der vil blive læst mange bøger.]] we know that there will be read many books ‘We know that many books will be read.’

What I propose here is that in both constructions der is an expletive, occuring in SpecIP to satisfy the EPP. There are several advantages to a unified treatment of relative clause constructions and impersonal constructions, but also some challenges. The first advantage is methodological: there is no need to stipulate that there are two different lexical items pronounced der, one of which is a complementizer and the other a nominal expletive element. This stipulation is made in Vikner (1991:110) and Jacobsen and Jensen (1982:6). The second advantage is that this unification brings together two classes of constructions that have previously been largely studied separately, and with separate theoretical issues in mind.

83

84

Line Mikkelsen

Impersonal constructions have primarily been investigated from the point of view of A-movement, and the related issues of Case and theta-role assignment (e.g. Vikner 1995a), while relative clauses have been studied primarily from the point of view of A-bar movement and operator-variable binding. A unified analysis of relative clauses and impersonal constructions allows us to investigate directly the interaction of phenomena associated with A- and A-bar movement. The main challenge to the kind of unified analysis of relative and impersonal constructions sketched above is to account for cases where they appear to behave differently. As pointed out by Vikner (1991: 115), there are at least two such cases, involving definiteness restrictions and transitivity restrictions. Definiteness restrictions. Danish impersonal constructions display a definiteness restriction on the postverbal argument of the sort documented for English by Milsark (1979). This is illustrated by the contrast in (44). Subject relative clauses containing the expletive der do not appear to display any such restriction, as illustrated by the lack of contrast in (45). (44) Vi ved at … we know that a. der vil komme mange lingvister. there will come many linguists b. *der vil komme de lingvister. there will come the linguists (45) Vi kender … we know a. mange lingvister der vil komme t. many linguists there will come b. de lingvister der vil komme t. the linguists there will come

Transitivity restrictions. Impersonal constructions also display a transitivity restriction in Danish. The verb in an impersonal construction cannot be transitive. Only intransitives and passive forms are allowed; cf. Vikner (1995a: 168). In subject relative clauses on the other hand, there is no transitivity restriction, and any verb can appear:9 (46) Vi ved at … we know that a. der bliver læst mange bøger. there are read many books


b. *der mange lingvister læser denne bog. there many linguists read this book c. *der læser mange lingvister denne bog. there read many linguists this book (47) Vi kender … we know a. mange bøger der bliver læst. many books there are read b. mange lingvister der læser denne bog. many linguists there read this book

In the following two sections I try to show that these differences are in fact consistent with the analysis of relative clauses proposed above. The strategy in both cases is to show that the difference in question derives from differences in the syntax of impersonal and relative clauses. The discussion is based on observations made independently for other languages about the definiteness value of relative operators and Case requirements of operators that undergo A-bar movement. 4.1 The lack of a definiteness restriction in relative clauses In this section I argue that the lack of definiteness effects in a subject relative clause like (48) is consistent with der being an expletive. If this argument can be made successfully it removes one of the objections to treating der as an expletive in relative clauses. (48) Vi kender de lingvister [CPOPi [IP der [VP vil komme ti.]]] we know the linguists there will come

I begin with an observation about English relative clauses made in Browning (1987). Based on data like that in (49), Browning observes (p.130) that “[g]aps in relative clauses appear to be indefinite even when the relative heads are definite”. (49) a. *There were the men in the garden. (= Browning’s (164a)) b. The men that there were in the garden were all diplomats. (= Browning’s (165a))

(49a) is an existential construction and it violates Milsark’s definiteness restriction on existential constructions, since it has a definite noun phrase in the postcopular position. In (49b) the definite noun phrase has moved out of the existential clause, forming a relative clause with a gap in the post-copular position.

85

86

Line Mikkelsen

Crucially, (49b) is grammatical, which is what leads to Browning’s observation that the gap “appear[s] to be indefinite” even though the relative head (= the man) is definite. The suggestion I want to make here is that Browning’s observation about English carries over to Danish, even though the two languages seem to differ in how relative clauses like (49b) are interpreted (see the discussion in Section 4.1.1 below). Thus the suggestion is that the Danish data in (50), are, in the relevant respects, parallel to the English data in (49). (50) a. *Der kommer de lingvister til festen. there come those linguists to party.def b. De lingvister der kommer til festen er alle fra Sverige. the linguists there come to party.def are all from Sweden

Browning (1987) considers two possible explanations for the lack of definiteness effects in relative clauses like (49b).10 One possibility she considers is that the trace of the relative operator is inherently indefinite. This means that a structure like (51) does not violate the definiteness restriction (as long as the definiteness restriction is understood as a restriction against definite elements in the post-copular position of an existential clause). (51) The men [CP OPi that there were ti in the garden] were all diplomats

Another possibility, which she ends up proposing as the correct one, is that the lack of definiteness effects in relative clauses stems from the fact “that the antecedent of the relative operator is NP …, a category which is unspecified for definitness”, (p.131). Browning is proposing a structure for relative clauses as in (52). DP

(52) D′ D

CP NP

More recently Bianchi (1999) has reached similar conclusions about the lack of definiteness effects in relative clauses, working within the antisymmetry framework of Kayne (1994). Bianchi (1999) argues for a raising analysis of relative clauses, and considers in this connection the possibility of reconstruction of the external determiner to the gap position inside the relative clause, which could give rise to definiteness effects in relative clauses. Based on detailed empirical investigations she concludes (p.46) that “… the determiner introducing the relative “head” is generated outside the relative clause, and hence it is not


reconstructed in the trace position.” Bianchi (1999: 80–96) also argues, following Reinhart (1987), that relative wh-words like English who and which are not definite, which is related to Browning’s first suggestion that traces of relative clause operators are inherenly indefinite. Working within rather different frameworks the two authors reach the same conclusion as regards definiteness effects in relative clauses, namely that these are not to be expected. I take this as evidence that the analysis of der as an expletive in subject relative clauses is consistent with the lack of a definiteness restriction in these constructions. 4.1.1 A semantic aside Carlson (1977) claims that relative clauses like (49b), which involve relativization of the postverbal argument of an existential there-construction form a separate class called amount relatives, which differ from ordinary restrictive relative clauses semantically in involving quantification over degrees rather than individuals. Heim (1987) follows Carlson and proposes (53) as a semantic characterization of the definiteness restriction. (53) *There be x, where x is an individual variable.

(53) rules out (49b) on a reading where the variable corresponding to the postcopular trace ranges over individuals, but allows it on a reading where the trace translates as a variable over degrees. No such restriction is evident in Danish subject relatives, which are structurally analogous to (49b) if der is analyzed an expletive. Note though that discussions of ‘amount relatives’ have been exclusively concerned with existential constructions, not with impersonal constructions more generally. (Heim’s generalization in (53) is careful to make this explicit.) If Carlson’s restrictions derive from the semantics of existentials (as argued by e.g. Arnold 2000), we do not expect to see evidence of them in cases such as (50b). 4.2 The lack of a transitivity restriction in relative clauses The lack of a transitivity restriction in subject relative clauses is probably the most serious empirical and theoretical challenge to a unified analysis of relative clauses and impersonal constructions. I cannot offer an explanation of this difference, but in this section I try to place the Danish data in a wider set of observations about the interaction of A- and A-bar movement.

87

88

Line Mikkelsen

Let us start with the fact that Danish does not allow transitive expletive constructions (TEC): (54) *Vi ved at [der mange lingvister har læst denne bog.] we know that [there many linguists have read this book

Vikner (1995a: 152–153, 188–190) argues that this is due to the lack of V0 to I0 movement in the language, which prevents the subject DP mange lingvister from being Case licensed in a position lower than SpecIP (a similar argument is made in Bobaljik and Jonas (1996) with respect to the licensing of SpecTP as a subject position). On the other hand, subject relative clauses may be transitive: (55) Vi kender mange lingvister [der har læg denne bog.] we know many linguists [there have read this book

If we pursue a unified analysis of impersonal constructions and relative clauses, where der is uniformly an expletive in SpecIP this leads to a puzzle: (56) Why are transitive subject relative clauses possible, when transitive expletive constructions are not possible?

Following Vikner’s account of the lack of TECs, I suggest that the relevant difference has to do with Case. More specifically, I suggest that the Case requirements of the relative operator differ from those of the thematic subject in a transitive impersonal construction. There is cross-linguistic evidence that Case requirements are weakened in constructions involving A-bar movement. This was observed for French and English by Kayne (1980: 79–80), and discussed by Rizzi (1990: 60). For more recent discussion see also Alexiadou and Anagnostoupoulou (2001a) and Richards (1999). (57) a. Whoi can you assure me ti to be competent? b. *I assure you them to be competent. (58) a.

Jean, que Marie croit être intelligent, … John who Mary believes to.be intelligent b. *Marie croit John être intelligent.

(= Kayne’s (35)) (= Kayne’s (36))

What the data in (57) and (58) seem to show is that an A-bar chain need not originate in a Case position (though see Postal (1993) for a different interpretation). This observation is relevant to the puzzle set out above, since relative clauses involve A-bar movement of the relative operator, whereas impersonal constructions (typically) do not:


(59) Vi kender mange lingvister we know many linguists [CPOPi [IP der [VP ti læser denne bog.]]] there read this book (60) *Vi ved [CP at [IP der [VP mange lingvister læser denne bog.]]] we know that there many linguists read this book

What I want to suggest is that the difference between (59) and (60) is parallel to the difference between (57a) and (57b) above. I maintain, with Vikner (1995a), that (60) is ungrammatical because the thematic subject DP mange lingvister is not Case licensed, just like the DP them is not Case lisensed in (57b). (59) is grammatical for whatever reason (57a) is. If the suggestions about Case and A-bar movement made above are on the right track we expect TEC to be allowed when the subject undergoes A-bar movement. However, this expectation is generally not borne out: (61) *[CP [Mange lingvister]i læstej [IP der tj [VP ti tj denne bog.]]] many linguistsi readj there tj ti tj this book [topicalization] (62) *[CP Hvemi læstej [IP der tj [VP ti tj denne bog?]]] whoi readj there tj ti tj this book

[interrogative]

Taraldsen (1991:105–106) argues that the ungrammaticality of (61) and (62) has to do with the fact that the finite verb has moved to C0, which is again related to the verb second nature of Danish matrix clauses (Vikner 1995a). If the ungrammaticality of (61) and (62) can be explained independently, it might be possible to maintain the approach to the lack of transitivity restrictions in relative clauses sketched above, but the matter clearly needs further investigation. Before leaving the issue of transitive expletive constructions, I would like to point out a fact about older stages of Danish that lends indirect support to the unified analysis of expletive constructions and relative clauses suggested here. As illustrated by the data in (63)–(65), earlier stages of Danish seem to have allowed transitive expletive constructions. (63) … der skal dog Ingen narre mig i Sligt. there shall though no.one fool me in such.matters ‘… though no one will fool me in such matters.’ (C. Hauch (1873–4) Samlede Romaner og Fortællinger)

89

90

Line Mikkelsen

(64) Saadan en Elendighed har der aldrig Nogen hørt. such a misery has there never anyone heard ‘No one has ever heard of such misery.’ (Blicher (1893) Samlede Noveller og Skizzer.) (65) Har der nogen Mennesker paa Jorden hørt Mage. has there any people on earth heard like ‘Has anyone on earth heard the like.’ (Skjoldborg (1911) Ideale Magter)

Within the same period the expletive der also occurred in non-subject relative clauses, as the example in (66) shows: (66) Alle de Sløjder, der jeg kan dem lærer jeg eder saa gerne. all the wood.skills there I can them teach I you so willingly ‘I am very willing to teach you all the wood skills that I master.’ (Danske Folkeviser i Udvalg [Collected Danish Folk Songs] Olrik and Falbe-Hansen (eds.) 1899–1909).

Whatever the explanation for the ability of older Danish to license subject DPs in positions below SpecIP, the correlation between TEC and non-relative clauses with expletive der supports a unified account of expletive constructions and relative clauses.

5. Conclusion In this paper I have proposed an analysis of subject relative clauses in Danish where som is an invariant operator, at a complementizer, and der an expletive. The analysis accounts for the distribution of these elements, their internal ordering, co-occurence with wh-operators, and the restrictions on the realization of the left periphery of subject relative clauses. In addition to these descriptive advantages the analysis opens up the possibility of a unified analysis of relative clauses and impersonal constructions. A unified analysis raises several questions, the most serious one being the lack of a transitivity restriction in subject relative clauses. Without offering an explanation of this difference, I have tried to show how it is consistent with a broader set of observations about the interaction of Case and A-bar movement, which are still in need of a satisfactory account.


Notes * This paper has benefited greatly from discussions with Jim McCloskey and Geoff Pullum. Thanks to Matthias Kromann for help with locating and dating the examples from older Danish, and to Chris Potts for his expert advice on style and grammar. I would also like to thank the audience at the 15th Comparative Germanic Syntax Workshop in Groningen, May 26–27, 2000 for useful comments. All remaining errors are mine. 1. Taraldsen (1991: 99–107) also argues, contra Taraldsen (1986), that der is an expletive occuring in SpecIP in relative clauses. 2. I distinguish between invariant operators (som and op) and wh-operators (hvad ‘what’, hvem ‘who’, hvilken ‘which’). The distinction is based on how much lexical information is encoded in the operator. Wh-operators are inherently marked for semantic and syntactic distinctions like animacy, number, specificity, and case, while invariant operators provide none of this kind of information. When the type of operator (invariant or wh-) is not relevant I use ‘operator’ as a cover term, on the understanding that we are considering only the class of operators involved in constructions involving A-bar movement, including relative clauses, clefts, and interrogatives. 3. Danish also has a place adverbial der which differs prosodically from the expletive der in being stressed. Since adverbial der does not play a role in the analysis developed here, I use der throughout to refer to the unstressed expletive der only. 4. A similar argument is made in Taraldsen (1991: 101). 5. This raises some issues about Case, in particular about the Case requirements of operators. I postpone this discussion to Section 4.2. 6. Vikner (1991: 111) gives (21a) a question mark, and I agree that the example is slightly degraded. Like Vikner (1991) I cannot offer an explanation for the relative degradedness of (21a), except to note that the degree of degradation is similar to that found in subject relative clauses where der co-occurs with som (cf. Section 3). Under the present analysis, these cases are parallel to (21a) in that der co-occurs with an overt operator in SpecIP. Without venturing a concrete proposal I believe that this is in fact the source of the slight deviance of these examples. The question remains though why (20a) and (22a) are not similarly degraded. 7. Erteschik-Shir (1984: 137) actually says that relative clauses with at are ungrammatical, but she nonetheless considers at a complementizer. 8. As pointed out by Vikner (1991) many prescriptive grammars consider the middle category ungrammatical, which may influence speaker judgments. There nonetheless seems to be a clear enough difference between degraded (?–??) and ungrammatical (*) to warrant this distinction. 9. There are several potential positions for a thematic subject in a transitive expletive constructions, but none of these is possible in Danish, as shown by Vikner (1995a: 189ff). Here I concentrate on the immediately pre-verbal position, but I believe that the observations made below carry over to the other potential subject positions discussed by Vikner (1995a). 10. This and related issues are discussed in detail in Bianchi (1999) within the framework of Kayne (1994), see especially pp.41–49 and 80–86.

91

Syntactic versus semantic control* Susi Wurmbrand McGill University

1.

Introduction

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.1 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 CP-material, control infinitives were generally considered to be clauses (whether they involve CP-material or not). The so-called PRO-theorem provided another theoryinternal 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 & Lasnik 1993, Martin 1996) one of the major theoretical motivations for a CP-structure for control infinitives disappears. Many contemporary syntactic approaches assume

94

Susi Wurmbrand

that complementizerless and non-wh-infinitives are smaller categories. To take just one approach, Boškovic´ (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 control principle. The main aim of this article is to investigate the interaction of the syntax and semantics of infinitival constructions. While we will provide further support for Chierchia’s claim that (certain) infinitives are subjectless predicates syntactically and semantically, we will also conclude 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. Obligatory vs. non-obligatory control 2.1 The classification of obligatory vs. non-obligatory control Since Williams (1980), it has been widely accepted that the control phenomenon subsumes two sub-cases: obligatory and non-obligatory control. While most authors agree that these two forms of control have to be distinguished to account adequately for the properties of control constructions, the actual classification of infinitival constructions as obligatory vs. non-obligatory control infinitives is subject to some debate. To avoid confusion, we will first outline how we will use the obligatory vs. non-obligatory control distinction in this article. The examples in (1) and (2) illustrate a number of control phenomena that have been distinguished in the literature. (1) a. John tried to leave b. It is dangerous for babies to smoke around them c. It was difficult to leave

Syntactic versus semantic control

d. John persuaded Mary to leave together e. We thought that the chair preferred to gather at 6

(1a) represents an instance of exhaustive control (cf. Landau 1999, 2000) — i.e., the infinitival subject refers exhaustively to the matrix subject John. Examples such as the one in (1b) from Kawasaki (1993) involve arbitrary control — i.e., the reference of the infinitival subject is left unspecified. (1c) is an instance of 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). (1d) involves split control — i.e., the infinitival subject is controlled jointly by the matrix subject and object. And (1e), 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).2 Finally, the German examples in (2) represent the phenomenon of control shift or variable control which is found in many languages.3 (2) a.

Ichi habe ihmj angeboten [PROi/j mich zu erschießen] Ii have himj-dat offered [PROi/j me to shoot] ‘I offered him to shoot myself.’ ‘I offered him that he could shoot me.’ b. Ichi habe ihnj gebeten [PROi/?j erschossen zu werden] Ii have himj-acc asked [PROi/?j shot to become] ‘I asked him to allow/arrange that I get shot (by somebody).’ ?‘I asked him to allow/arrange that he get shot (by somebody).’

As is illustrated in (2a), 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 Section 2.2), as well as certain thematic and pragmatic properties of the construction (see for instance Farkas 1988, Petter 1998, Landau 1999, 2000 for extensive discussions). As can be seen in (2b) for instance, constructions involving the verb bitten ‘ask’ allow control shift best when the infinitival predicate is passivized or involves a modal predicate — i.e., when the (surface) infinitival subject bears a non-agentive theta role (a similar effect has been noted in English for promise-constructions). As far as we are aware, the classification of exhaustive control as in (1a) as obligatory control is uncontroversial. Similarly, arbitrary control as in (1b) is generally considered as the prototypical case of non-obligatory control. Concerning the examples in (1c–e) and (2), on the other hand, the classification is less straightforward. Like obligatory and unlike non-obligatory control infinitives,

95

96

Susi Wurmbrand

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.4 We thus assume that the obligatory/non-obligatory control distinction is a basic semantic distinction. Note that by grouping partial, split, and variable control with non-obligatory control, we do by no means claim that non-obligatory control is a homogeneous phenomenon or that all forms of non-obligatory 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 non-obligatory 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/non-obligatory 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 properties. 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 (see also Wurmbrand 1998b, 2001). 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 non-obligatory 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. non-obligatory 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 Two classes of infinitival constructions While variable control (control shift) is quite readily available in German, there are nevertheless constructions that clearly block this form of control. As we have seen above, infinitives combining with verbs like offer, ask (also propose, persuade, promise etc.) allow either subject or object control (modulo certain thematic restrictions in the infinitive). Infinitives combining with verbs like forbid, allow, or recommend, on the other hand, do not have this option. As is illustrated in (3), these verbs prohibit any kind of flexible interpretation and only allow object control.

97

98

Susi Wurmbrand

(3) Ichi habe ihmj verboten … Ii have himj-dat forbidden a. … [subj*i/j mich zu erschießen] [subj*i/j me to shoot] ‘I forbade him to shoot me.’ *‘I forbade him that I should/could shoot myself.’ b. … [subj*i/?j erschossen zu werden] to become] [subj*i/?j shot ?‘I forbade him to allow/arrange that he get shot (by somebody).’ *‘I forbade him to allow/arrange that I get shot (by somebody).’

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 (4a) vs. (4b)). (4) 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 (5)), whereas others block this form of interpretation (cf. the examples in (6)). As is illustrated in (5a), 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). (5) a.

Der Direktori hat ihmj vorgeschlagen the principali has himj proposed [PRO sich im Schloß zu versammeln] [PRO self in-the castle to gather ‘The principal proposed to him to gather in the castle.’ [PROi, PROj, PROi+k (e.g., the teachers), PROj+k (e.g., the students), PROi+j+k (e.g., the whole school)] b. Eri beschloß/erwog/plante hei decided/contemplated/planned [PROi+k sich im Schloß zu versammeln] [PROi+k self in.the castle to gather ‘He decided/contemplated/planned to gather in the castle.’


c.

Hansi beschloß/erwog/plante Johni decided/contemplated/planned [PROi+k sich gemeinsam zu bewerben] [PROi+k self together to apply ‘John decided/contemplated/planned to apply together.’

(6) 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 [sich gemeinsam zu bewerben] [self together to apply ‘The clerk disallowed/allowed/recommended (to) him to apply together.’

Thus, variable and partial control split infinitival constructions into two groups which we labeled obligatory vs. non-obligatory control. The interesting question that arises from this distinction is why there is such a distinction and what it tells us about the grammar of infinitival constructions.

3. Syntactic vs. semantic 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 1983, Lebeaux 1984, 1985, Hornstein 1999, Landau 1999, 2000).5 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 needs

99

100 Susi Wurmbrand

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 infinitival subject and the antecedent. Possibly the most common way to do so is by assuming that obligatory control expresses an anaphoric relation (cf. Manzini 1983, Bouchard 1984, Koster 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 & Roussou 2000). In contrast, non-obligatory control is subject to looser conditions, which is generally attributed to the pronominal (Manzini 1983, Bouchard 1984, 1995, Koster 1984, Hornstein 1999) or logophoric (Grinder 1970, Kuno 1975, Lebeaux 1984, 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 section, 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 difference 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 (7) (where j is a context dependent modal operator; (7b) is a rough paraphrase of (7a)). (7) a. try’ (P) (x) Æ j P (x) Chierchia (1984a: 34) b. whenever x tries to bring about P, then in all the contextually relevant situations (namely those where what x tries actually succeeds), x does P

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. In the following sections, we will concentrate on German infinitival constructions which will provide support both for and against a direct translation of a Chierchia style account into syntactic terms. In particular, we will show that a certain class of infinitival constructions, 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 also see that the correlation is only a one-way correlation, since obligatory control 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.

101

102 Susi Wurmbrand

The proposal that we would like to put forward here 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, non-obligatory 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 non-obligatory control infinitives) that include a syntactic subject are translated into semantic properties. This line is taken for instance by Chierchia (1989a) who suggests that PRO functions essentially as a 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

Syntactic versus semantic control 103

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.6 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 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 will first show that properties targeting the syntactic structure (movement and binding) support a syntactic distinction between subjectless infinitives and infinitives with a subject, which correlates in the way presented above with the semantic properties of the infinitival constructions. Secondly, we will discuss a property targeting the semantic structure (the interpretation of elided infinitives) and show that this property supports the semantic distinction between obligatory vs. non-obligatory control as suggested here. 3.2 Restructuring The first piece of evidence for the lack of a syntactic (PRO) subject comes from infinitival constructions that display clause union (cf. Aissen & Perlmutter 1976, 1983) or restructuring (cf. Rizzi 1978, 1982) effects. The property of restructuring constructions we will focus on here is a movement operation often labeled long passive. As is illustrated in (8), passivized (cf. (8a)) and unaccusative (cf. (8b)) restructuring verbs affect the argument structure of the embedded predicate in the following way: the embedded object is assigned nominative case and agrees with the matrix auxiliary.7 Note that the infinitive does not show passive morphology.

104 Susi Wurmbrand

(8) a.

Der Lastwagen und der Traktor wurden/*wurde zu the-nom truck and the-nom tractor were /*was to reparieren versucht repair tried ‘They tried to repair the truck and the tractor.’ b. Der Brief und der Bericht sind/*ist mir auf Anhieb the-nom letter and the-nom report are /*is me-dat straightaway zu entziffern gelungen to decipher managed ‘I managed straightaway to decipher the letter and the report.’

What is crucial about object movement of this sort is that in restructuring infinitives, the case properties in the embedded predicate 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 for the embedded object. Note also that object movement of this sort is restricted to restructuring infinitives. As the ungrammaticality of (9) shows, long passive is impossible in infinitival constructions involving a non-restructuring verb such as plan, decide, announce.8 (9) a. *… dass der Traktor zu reparieren geplant wurde that the-nom tractor to repair planned was ‘…that they planned to repair the tractor.’ b. *… dass die Traktoren zu reparieren geplant wurden that the tractors to repair planned were ‘…that they planned to repair the tractors.’

The only way to express passive with a non-restructuring verb is an impersonal passive construction as in (10) — i.e., a construction without long object movement.9 (10) a.

… dass [den Traktor zu reparieren] geplant wurde that [the-acc tractor to repair planned was ‘…that they planned to repair the tractor.’ b. … dass [die Traktoren zu reparieren] geplant wurde/*wurden that [the tractors to repair planned was /*were ‘…that they planned to repair the tractors.’

“Long” object movement of this sort raises a number of questions for approaches that consider restructuring infinitives as full clauses — i.e., as structures that involve an independent embedded argument structure and in particular an


embedded subject. The most urgent questions are why the embedded object loses accusative case when the matrix verb is passivized, why the embedded subject does not block this A-movement operation, and why object movement is impossible in non-restructuring infinitives. The analysis that we will pursue is built on the idea that restructuring infinitives do not involve clausal projections but are simple VPs, as has been suggested in a number of works on restructuring.10 Taking VP-approaches a step further, we claim that a restructuring infinitive does not only lack C and T projections, but also does not include an embedded structural case position or assigner or an embedded subject. Let us first consider a sentence with an active restructuring verb such as the one in (11a). Under the assumption that restructuring infinitives lack an embedded structural case position/assigner, the obvious question arising is how the embedded object is assigned accusative case. Our analysis is illustrated in (11b). The only structural object case position in a sentence with an active restructuring verb is provided by the matrix vP. We thus assume that it is indeed the matrix case position that is responsible for the structural case of the embedded object. Assuming a movement approach to case checking (see also Wurmbrand 2000), the embedded object raises (overtly or covertly) to the “matrix clause” in order to check its accusative case. (11) a.

… weil Hans den Traktor zu reparieren versuchte since John the-acc tractor to repair tried ‘…since John tried to repair the tractor.’ TP

b.

T′

NOM John

T0

vP ACC

v′ SUBJ tJohn

v′

V0 tried

VP OBJ the tractor

v0

VP

V0 to repair

106 Susi Wurmbrand

TP

c. NOM

T′ VP

VP OBJ the trucks

V0 tried managed

T0 were is

V0 to repair

Returning now to the passive and unaccusative examples in (8), the case and agreement facts are no longer surprising. Since restructuring infinitives do not include 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 or unaccusative, structural accusative is unavailable, leaving the (matrix) T as the only case assigner for the embedded object (cf. (11c)).11 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. Non-restructuring constructions such as (10), on the other hand, project an embedded subject and have their own functional structure (vP, TP, or CP-infinitives depending on the “grade” of non-restructuring; cf. note 8). Assuming standard locality conditions, the embedded object enters into a case/agreement relation with the closest functional head — i.e., the embedded v0 rather than a head in the matrix clause — resulting in accusative rather than nominative case on the object and the lack of agreement with the matrix auxiliary. Furthermore, A-movement of the embedded object across the embedded subject is blocked. Before turning to the consequences of long object movement for control, we will conclude the discussion of restructuring with another argument for the analysis as proposed here. So far, we have seen that 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. What we will show now is that this case and agreement relation is not only possible but in fact obligatory. To do so, consider first the examples in (12) which at first sight appear to challenge the analysis presented here. Given the claim that restructuring infinitives do not include a structural object case/ agreement position, we would expect that accusative should never be available for the embedded object in passive or unaccusative contexts. However, as is


shown in (12), accusative objects are perfectly grammatical in infinitival complements combining with passivized or unaccusative restructuring verbs. (12) a.

… dass versucht wurde/*wurden that tried was/*were den Traktor und den Lastwagen zu reparieren the-acc tractor and the-acc truck to repair ‘…that they tried to repair the tractor and the truck.’ b. … dass es ihm gelungen ist/*sind that it him managed is/*are den Traktor und den Lastwagen zu reparieren the-acc tractor and the-acc truck to repair ‘…that he managed to repair the tractor and the truck.’

Assuming our analysis is correct, the obvious answer to the question of why accusative is possible and agreement with the matrix T is not necessary in the examples in (12) is that these examples are indeed instances of non-restructuring infinitives (i.e., infinitives that include an embedded subject and project a structural object case/agreement position). In other words, verbs like try, manage have to be considered as ambiguous — they either are restructuring verbs that combine with a bare VP-complement, or they are non-restructuring verbs that combine with a vP (or bigger) complement. 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. One such property is 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 (13a), the embedded object can be scrambled from an extraposed restructuring infinitive; if, on the other hand, the matrix verb is a non-restructuring infinitive as in (13b), scrambling is impossible (cf. (13c)). (13) a.

… dass Hans den Traktor versucht hat zu reparieren that John the-acc tractor tried has to repair ‘…that John (has) tried to repair the tractor.’ b. … dass Hans bedauert hat den Traktor reparieren zu müssen that John regretted has the-acc tractor repair to must ‘…that John (has) regretted having to repair the tractor.’ c. *… dass Hans den Traktor bedauert hat reparieren zu müssen that John the-acc tractor regretted has repair to must ‘…that John (has) regretted having to repair the tractor.’

108 Susi Wurmbrand

Returning to the examples in (12), the impossibility of scrambling can be straightforwardly illustrated. In contrast to (13a), the examples in (14) which correspond to the examples in (12) (i.e., examples lacking a case and agreement relation between the embedded object and the matrix T) clearly block scrambling of the embedded object from the extraposed infinitive. These examples thus provide strong support for the claim that 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. (14) a. *… dass den Traktor versucht wurde zu reparieren that the-acc tractor tried was to repair ‘…that they tried to repair the tractor.’ b. *… dass (es) ihm den Traktor gelungen ist zu reparieren that (it him the-acc tractor managed is to repair ‘…that he managed to repair the tractor and the truck.’

To sum up, although CP-approaches have various ways of building in assumptions that will derive long object movement, the major advantage of a VP-approach is that it not only straightforwardly accounts for long object movement but also that it predicts and explains the existence of this movement operation. 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. Before we will answer this question, let us first look at the predicates that allow long object movement in German summarized in Table 1. Table 1.Long A-movement Verb

Verb

A-movement

beginnen vergessen wagen gelingen [unacc]

begin forget dare manage

beabsichtigen versuchen versäumen mißlingen [unacc]

intend (want) + try + miss, neglect + fail +

empfehlen verbieten gestatten

recommend forbid permit

erlauben

allow

untersagen

prohibit

+ + %


Looking at the constructions that allow long A-movement in German, what is striking is that all predicates in Table 1 trigger obligatory control — i.e., variable, partial, or split control are prohibited in constructions involving these predicates. Under the analysis proposed above, this correlation comes as no surprise.12 We have suggested that the omission of an infinitival subject in syntax (which is a necessary condition for long-object movement) is only possible when the interpretation of the subject is specified elsewhere — i.e., as part of the meaning of the higher predicate. Since the meaning of non-obligatory control predicates does not include a pre-specified antecedent for the embedded subject, a (non-obligatory control) structure lacking a syntactic subject would not be interpretable since the meaning of the embedded subject could not be determined (i.e., neither lexically nor syntactically). We thus expect that restructuring (i.e., long-object movement) is only possible with obligatory control predicates. Going back to Chierchia’s approach, recall that (obligatory control) infinitives are properties (i.e., subjectless predicates) semantically which become associated with a subject post-semantically (via lexical entailments built into the meaning of the selecting verbs). Assuming a one-to-one syntax/ semantics correspondence (but see Section 3.4), the properties of German restructuring then provide additional evidence for Chierchia’s view: long object movement indicates that these infinitives lack a syntactic subject, and the obligatory (i.e., exhaustive) control relation of these constructions indicates that these constructions are properties semantically. 3.3 Binding The second and crucial piece of evidence we will provide for the lack of an infinitival subject in restructuring infinitives and the presence of a subject in non-restructuring infinitives comes from the binding properties of infinitival constructions in German. Let us start with an example such as (15a) 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 involve an embedded infinitival PRO subject), the anaphor is bound by the embedded subject as illustrated in (15b). Under our analysis, on the other hand, there is no embedded subject in restructuring infinitives and the embedded anaphor is bound directly by the matrix subject (cf. (15c)).

110

Susi Wurmbrand

(15) a.

… weil der Hans sich zu rasieren versuchte since the John self to shave tried ‘…since John tried to shave himself.’ b. since John [PROi selfi to shave] tried c. since Johni [selfi to shave] tried

[+PRO]-approach [−PRO]-approach

Thus, both the [+PRO] approach in (15b) and the [−PRO] approach in (15c) provide a way of accounting for the coreference relation between the embedded anaphor and the matrix subject in examples such as (15a). 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 with the matrix subject thus being established only indirectly. While (empirically) indistinguishable in examples such as (15a), 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, since anaphors are indeed illicit in restructuring infinitives (in contrast to non-restructuring infinitives) when there is no appropriate antecedent in the matrix predicate. 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 3.3.2.13 3.3.1 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 (16), embedded anaphors can occur in impersonal passive constructions (cf. (16a)) and impersonal adjectival constructions (cf. (16b)). Furthermore, embedded anaphors are possible in the impersonal non-restructuring versions of constructions involving a potential restructuring verb (cf. (16c)); recall that infinitives involving a structural accusative object in passive or unaccusative matrix contexts are nonrestructuring infinitives).


(16) a.

Es wurde beschlossen it was decided [PROi sichi den Fisch mit Streifen vorzustellen] [PROi selfi the-acc fish with stripes to-imagine ‘They decided to imagine what the fish would look like with stripes.’ b. Es war notwendig [PROi sichi einen Wagen zu kaufen] to buy] it was necessary [PROi selfi a-acc car ‘It was necessary to buy oneself a car’ c. Es wurde versucht [PROi sichi den Fisch mit Streifen it was tried [PROi selfi the-acc fish with stripes vorzustellen] to-imagine ‘People tried to imagine what the fish would look like with stripes.’

As indicated in the examples, we assume that the anaphors in (16) 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 (16). As is shown in (17b), implicit passive arguments cannot bind anaphors in German.14 (17) a.

Frederiki hat sichi ein Haus gekauft Frederici has selfi a house bought ‘Frederic bought himself a house.’ b. Ein Haus wurde (*sich) gekauft a house was (*self) bought ‘A house was bought (*oneself).’

The examples in (16) and in particular the contrast between these examples and the ungrammatical (17b) thus provides evidence for the existence of an embedded subject in non-restructuring infinitives (a further argument for this claim will be provided in the next section). Since the examples in (16) are nonrestructuring infinitives (as witnessed 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. (18a)). Furthermore, as is shown in (18b), benefactive anaphors associated with the implicit subject are impossible (however, the sentence is grammatical without the anaphors).

111

112

Susi Wurmbrand

(18) a. *… weil {sich} der Fisch {sich} vorzustellen versucht wurde since {self} the-nom fish {self} to-imagine tried was ‘…since somebody tried to recall the image of the fish.’ b. … weil {*sich} der Turm {*sich} zu bauen versucht wurde since {*self} the-nom tower {*self} to build tried was ‘…since somebody tried to build (*himself) the tower.’

Assuming that restructuring infinitives lack an embedded (syntactic) subject and implicit arguments cannot bind anaphors in German, the contrasts in (16) and (18) can straightforwardly be accounted for. Since restructuring infinitives do not involve 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. 3.3.2 Ditransitive constructions The final 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 as is illustrated in (19). Whatever the nature of this restriction is, it will allow us to draw certain conclusion about the presence or absence of infinitival subjects. (19) … weil der Hansh der Mariam sichh/*m auf dem Photo since the-nom Johnh the-dat Marym selfh/*m in the picture zeigte showed ‘…since John showed Mary himself/*herself in the picture.’

Consider first the example in (20a) with the (potential) restructuring verb manage. (20) a.

Es ist ihm gelungen [PROi sichi einen Turm zu bauen] it is him managed [PROi selfi a-acc tower to build ‘He (has) managed to build himself a tower.’ b. Sie hat dem Hans erlaubt she has the-dat John allowed [PROi sichi den Fisch mit Streifen vorzustellen] [PROi selfi the-acc fish with stripes to-imagine ‘She allowed John to imagine what the fish would look like with stripes.’


c.

Sie hat dem Hans verboten she has the-dat John forbidden [PROi sichi den Turm anzuschaun] [PROi selfi the-acc tower to-look-at ‘She forbade John to look at/investigate (refl.) the tower.’

Since the infinitive involves an accusative object which cannot have been casemarked 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 nonrestructuring infinitives project an embedded syntactic subject which can function as the antecedent for the embedded anaphor. Similarly, the ditransitive constructions in (20b,c) involving potential restructuring verbs allow embedded anaphors which we assume are bound by the embedded PRO subject.15 Importantly, the grammaticality of the examples in (20) (i.e., examples involving a dative “controller” for the embedded subject) provides a further argument for the presence of an embedded subject in non-restructuring contexts. If one were to assume that obligatory control infinitives are necessarily subjectless predicates — i.e., if examples such as the ones in (20) would lack an embedded syntactic subject, it would not be clear how the anaphors are bound in these examples. Since, under this view, the dative arguments would be the only potential binders for the embedded anaphors, however, dative arguments cannot bind anaphors in German (cf. (19)), the embedded anaphors would end up without an antecedent, and hence the sentences should be ungrammatical. We thus conclude that the presence vs. absence of a syntactic subject is tied to the restructuring/non-restructuring distinction rather than the obligatory/nonobligatory control distinction. Turning finally 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 (20) are changed to (clear) instances of restructuring infinitives, anaphors become impossible. This is illustrated in (21) by the impossibility of reflexive predicates in infinitives involving “long” passive (i.e., restructuring infinitives). Note in particular (21c) which shows that long passive (and hence restructuring) is possible, however, only when the embedded predicate is not reflexive.16

113

114

Susi Wurmbrand

(21) a. *… weil der Fisch dem Hansi sichi mit Streifen vorzustellen since the-nom fish the-dat Johni selfi with stripes to-imagine gelungen ist managed is ‘…since John managed to imagine what the fish would look like with stripes.’ b. *… weil der Fisch dem Hansi sichi mit Streifen vorzustellen since the-nom fish the-dat Johni selfi with stripes to-imagine erlaubt wurde allowed was ‘…since somebody allowed John to imagine what the fish would look like with stripes.’ c. … weil ein Turm dem Hansi (*sichi) anzuschauen verboten since a-nom tower the-dat Johni (*selfi) to-look-at forbidden wurde was ‘…since John was not allowed 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 position/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 — a property that is independently motivated by the interpretation of these constructions (but see next section) — 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 this contrast without further assumptions. 3.4 Syntax π semantics In the previous sections, 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 (16c) and (18a) (repeated here as (22a,b): (22) a.

Es wurde versucht [PROi sichi den Fisch mit Streifen it was tried [PROi selfi the-acc fish with stripes vorzustellen] to-imagine ‘People tried to imagine what the fish would look like with stripes.’ b. *… weil {sich} der Fisch {sich} vorzustellen versucht wurde since {self} the-nom fish {self} to-imagine tried was ‘…since somebody tried to recall the image of the fish.’ c. … dass versucht wurde [PRO den Traktor und den that tried was [PRO the-acc tractor and the-acc Lastwagen zu reparieren] truck to repair ‘…that they tried to repair the tractor and the truck.’

While embedded anaphors are licensed in non-restructuring infinitives such as (22a), they are impossible in restructuring constructions lacking an antecedent in the matrix predicate (cf. (22b)). We have argued that this contrast is due to the presence (22a) vs. absence (22b) of a syntactic antecedent for the anaphor — an embedded syntactic PRO subject as indicated in the examples. Most importantly, however, both examples in (22) are obligatory control constructions. Similarly, the non-restructuring versions of examples involving a potential restructuring verbs (i.e., examples such as the ones in (12), repeated as (22c) can only receive an obligatory control interpretation. Assuming our analysis is correct, the examples in (22) 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).17 Thus, the presence of a syntactic 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,

115

116

Susi Wurmbrand

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 (cf. also note 12), they nevertheless seem to support a lexical/ semantic approach to obligatory control. Let us see how the obligatory control effect in (22a,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. 3.5 “It” anaphors In this last section of the article, we will focus on the interpretation of non-overt infinitives that are represented by it-anaphors, and we will see again that the syntactic structure cannot be fully reduced to the interpretation of infinitives. Chierchia (1984a, 1984b) has noted that it-anaphors replacing an infinitive or a gerund in examples such as (23) display an ambiguity between a strict and a sloppy interpretation.18 (23) Ezio likes fooling around, but I would bet that his wife Mimi doesn’t like it at all a. it: Mimi fooling around sloppy reading b. it: Ezio fooling around strict reading

In (23), the antecedent of it can refer to a situation in which the understood subject of the non-expressed 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).


The situation is different when the matrix predicate is changed to a verb like begin. As can be seen in (24) (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. (24) Ezio began playing the violin for fun and Nando began it out of lust a. it: Nando’s playing sloppy reading b. it: *Ezio’s playing *strict reading

The analysis Chierchia (1984a, 1984b) provides to account for the (im)possibility of the strict interpretation of it-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 it-anaphors are interpreted essentially like pronouns, i.e., as variables that are either bound by a structural antecedent or receive their value from context. Let us first see how the two interpretations are accounted for in Chierchia’s theory. For the sloppy interpretation in (23) (repeated in detail in (25)), Chierchia assumes that it represents a property in syntax/semantics which is bound by the property fooling around in the preceding conjunct (cf. (25a)). 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. (25b)). (25) Ezio likes fooling around … Mimi doesn’t like it at all a. Ezioe likes [fooling around]k … Mimim doesn’t like itk at all it-association Ezioe likes [fooling around]k … Mimim doesn’t like [fooling around]k b. Ezioe likes [subje fooling around]k … Mimim doesn’t like [subjm fooling around]k meaning postulate

For the strict reading in (23) (repeated in (26)), 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. That is, as is illustrated in (26a,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 it-anaphor (cf. (26c)).

117

118

Susi Wurmbrand

(26) Ezio likes fooling around … Mimi doesn’t like it at all a. Ezioe likes [subje fooling around]k … Mimi doesn’t like itPROPOSITION b. Ezioe likes [subje fooling around]k … Mimi doesn’t like itPROPOSITION meaning postulate c. Ezioe likes [subje fooling around]k … Mimi doesn’t like itk it-association Ezioe likes [subje fooling around]k … Mimi doesn’t like [subje fooling around]k

Thus, the antecedent of it in (26) is not a structural entity (i.e., it is neither present in syntax nor in semantics) but rather an object that is created postsemantically. 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 (26). However, as we will see below, the mechanism of associating it with a non-structural 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 it-anaphors in begin-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. (27a,b)); verbs like begin, on the other hand, are only compatible with a VP-complement (cf. (27c,d)).19 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 it-anaphor, it can only stand for a property and hence will always have to refer back to a property, yielding only a sloppy interpretation. (27) a. b. c. d.

… like … like … begin *… begin

[PROPERTY [PROPOSITION [PROPERTY [PROPOSITION

INF] *John likes Mary’s fooling around INF] *John likes fooling around INF] *John began fooling around INF] *John began Mary’s fooling around

Before we discuss Chierchia’s analysis, let us first give an overview of the situation in German, where it-anaphors can replace infinitives. The interpretation of it-anaphors in German shows two crucial generalizations. First, as is summarized in Table 2, the strict vs. sloppy interpretation of it-anaphors depends solely on the type of predicate combining with it; the nature of the antecedent


of it is irrelevant. Second, we will see that strict interpretations are only possible when the it-anaphor combines with a non-obligatory control predicate (according to our classification in Section 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 2.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

+ + − −

+ + + +

(28a) (28b) (28c) (28d)

Examples illustrating the four possibilities in Table 2 are given in (28). (28) a.

Hans beschloß zu heiraten John decided to get-married [nachdem Peter es angekündigt hatte] [after Peter it announced had (i) ‘John decided to get married after Peter had announced that he, Peter, would get married.’ (ii) ‘John decided to get married after Peter had announced that John would get married’ b. Hans wagte zu heiraten John dared to get-married [nachdem Peter es angekündigt hatte] [after Peter it announced had (i) ‘John dared to get married after Peter had announced that he, Peter, would get married’ (ii) ‘John dared to get married after Peter had announced that John would get married’ c. Hans beschloß zu heiraten John decided to get-married [nachdem Peter es gewagt hatte] [after Peter it dared had (i) ‘John decided to get married after Peter had dared to get married.’ (ii)*‘John decided to get married after Peter had dared that John would get married.’

119

120 Susi Wurmbrand

d. Hans wagte zu heiraten John dared to get-married [nachdem Peter es versucht hatte] [after Peter it tried had (i) ‘John dared to get married after Peter had tried to get married.’ (ii)*‘John dared to get married after Peter had tried that John would get married.’

In (28a,b), it combines with non-obligatory control predicates and a strict interpretation is possible (the first paraphrase of the examples in (28) represents the sloppy interpretation, the second paraphrase represents the strict interpretation). As an aside, the two interpretations in (28a,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 (28c,d), on the other hand, it combines with obligatory control predicates and a strict interpretation is impossible. The same situation holds for ditransitive constructions: (29) a.

Peter hat sich geweigert Peter has self refused [für 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.’ b. Peter hat angeboten Peter has offered [für die Freunde seiner Mutter Holz zu hacken] [for the friends of-his mother wood to chop nachdem sie es ihnen befohlen hatte after she it them ordered had (i) ‘Peter offered to chop wood for the friends of his mother, after she had ordered them to chop wood.’ (ii)*‘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 [sich im Schloß zu versammeln] Peter has promised [self in-the castle to gather nachdem der Direktor es beschlossen/*versucht hatte after the principal it decided/*tried had (i) ‘Peter promised to gather in the castle after the principal had decided that Peter and his associates would gather in the castle.’ (ii) ‘Peter promised to gather in the castle after the principal had decided that he [the principal] and his associates would gather in the castle.’ (iii) *‘Peter promised to gather in the castle after the principal had tried that he [the principal] (and his associates) would gather in the castle.’

As can be seen in (29a), non-obligatory control predicates such as offer allow a strict and a sloppy interpretation for the it-anaphor.20 It-anaphors combining with an obligatory control predicate such as order (cf. (29b)), 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, (29c) shows that partial control patterns with nonobligatory control — a partial control interpretation is possible for both the strict and the sloppy interpretation of the it-anaphor. Finally as noted in (29c), obligatory control verbs such as try cannot combine with an it-anaphor that refers back to a collective predicate when the subject of the obligatory control verb is singular.21 Since obligatory control predicates prohibit a partial control interpretation and only allow a sloppy interpretation in ellipses contexts, the understood subject of the infinitive replacing the it-anaphor in (29c) can only be 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. The examples in (28) through (29) 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 it-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 3.4), the answer will be negative, and

121

122 Susi Wurmbrand

it-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 it-anaphors that combine with an obligatory control predicate but that at the same time give rise to the assumption that the it-anaphor 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 it-anaphor contexts — what is crucial is the semantic property of obligatory control. To show that an it-anaphor represents a clausal structure syntactically is not straightforward and can only be shown indirectly. In examples such as (30a), the it-anaphor combines with the obligatory control predicate manage. (30) a.

Peter hat beschlossen [sich den Fisch mit Streifen vorzustellen] Peter has decided [self the-acc fish with stripes to-imagine nachdem es der Maria gelungen ist after it the-dat Mary 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 selfM the fish-acc to-imagine] *Mary-datM managed [subjP selfP the fish-acc to-imagine]

Note first that the only interpretation it can receive in (30a) 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. (30b)). Secondly and most importantly, (30a) 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 (30c) — i.e., a structure in which it has been associated with an antecedent. (30) c.

Mary-datM managed [SUBJM SELFM the Wsh-acc to-imagine]

d. Mary-datM managed [SUBJM SELFM the Wsh-acc to-imagine] X

As we have seen in the previous section, dative arguments cannot bind anaphors in German. Assuming that the same binding conditions hold in (30c) (i.e., for the associate of it), the grammaticality of (30) then provides evidence for the presence of an infinitival subject as an antecedent for the embedded anaphor in (30c). If the infinitive does not include a PRO subject at the level


where binding applies (as in 30d), the anaphor would end up without an antecedent and the structure would be predicted to be ungrammatical (like (22b)). We thus conclude that the correct structure for it in (30a) is the clausal structure in (30c) rather than the VP-structure in (30d).22 To summarize, an analysis that ties the possibility of a strict interpretation of an it-anaphor to the syntactic structure (in particular the presence of a syntactic subject) seems to make the wrong prediction for the interpretation of it-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 witnessed by the movement and binding properties). However, as a result of the inherent lexical/semantic property of obligatory control predicates, which identifies the embedded subject with a prespecified 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 it-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 it-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 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 it-anaphor will thus generally block a strict interpretation of their complements.

4. Conclusion Table 3 provides an overview of the properties found in infinitival constructions in German that we have discussed in this article. Restructuring refers to infinitives without an embedded (PRO) subject; non-restructuring refers to infinitives with an embedded (PRO) subject. The main claim we have made in this article is that different properties target different aspects of the syntactic and semantic structure. Movement and binding (principle A) target the syntactic restructuring vs. non-restructuring distinction, whereas the interpretation of it-anaphors targets the semantic obligatory vs. non-obligatory control distinction. As can be seen in Table 3,

123

124 Susi Wurmbrand

Table 3.Syntactic and semantic properties of infinitival constructions Control relation

Syntactic structure

Strict Long reading passive

Obligatory Obligatory Non-obligatory Non-obligatory

Restructuring infinitive − Non-restructuring infinitive − Non-restructuring infinitive + Restructuring infinitive N/A

+ − − N/A

Embedded anaphors (& no matrix antecedent) − + + N/A

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 non-restructuring contexts. Whether the infinitive is an obligatory or non-obligatory 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). The opposite situation holds for the interpretation of it-anaphors. A strict reading for it-anaphors requires a non-obligatory control context. Whether the infinitive is a restructuring or non-restructuring infinitive is irrelevant for this property.

Notes * Earlier versions of this article were presented at the University of Potsdam, at the Österreichische Linguistentagung in Vienna, and at the 15th Comparative Germanic Syntax Workshop in Groningen. I would like to thank the audiences of these talks, as well as Idan Landau and Jonathan Bobaljik for many helpful comments and discussions. Partial funding for this research was provided by a grant from the Social Sciences and Humanities Research Council of Canada. All errors of fact and interpretation are my responsibility. 1. Throughout this article, 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). 2. For many speakers, partial control is a marked interpretation, and there are also speakers that reject this from 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 article are from speakers that readily allow partial control. 3. As has been noted by Comrie (1984), the possibility of control shift is subject to variation (both across and within languages).


4. This classification requires some clarification concerning implicit control. As we will see below, implicit control in impersonal try-constructions (i.e., constructions involving an exhaustive control predicate) and in examples such as (1c) (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 of 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. 5. Some 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 2.1, 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. 6. 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. 7. The verb ‘manage’ is an unaccusative verb in German which requires the auxiliary be in the perfective. 8. We will consider the possibility of “long” object movement as a major criterion to distinguish between restructuring and non-restructuring. A discussion of other restructuring criteria and a finer grained classification of infinitival constructions has to be put aside here. The determination of restructuring is subject to some debate which is largely due to the fact that there are different grades of (non-)restructuring; furthermore, the labels “restructuring” vs. “non-restructuring” are not used uniformly when referring to these groups. Since different “restructuring” criteria target different syntactic and semantic properties, the criteria often pick out different classes of infinitival constructions, and classifications that do not distinguish between these grades of (non-)restructuring will therefore give conflicting results. The reader is referred to Wurmbrand (2001) for an extensive discussion of the restructuring criteria and a systematization of the different grades of (non-)restructuring. 9. The examples in (10) require an intonational break before and after the infinitival complement. In general, German speakers prefer extraposition of non-restructuring infinitives, however, the intraposed position of the infinitive as in (10) is also grammatical. 10. Proponents of VP-approaches include Strozer (1976, 1981), Zagona (1982), Picallo (1985, 1990), Haider (1986a, 1986b, 1986c, 1991a, 1991b, 1993), DiSciullo & Williams

125

126 Susi Wurmbrand

(1987), Rochette (1988, 1990, 1999), Rosen (1989, 1990), Moore (1990, 1994), Cinque (1997, 2001), Wurmbrand (1998a, 1998b, 1998c, 1999, 2000, 2001). 11. For reasons of simplicity we present passive and unaccusative constructions here as lacking a vP altogether. However, we do not commit ourselves to this structure. The only important fact for the analysis here is that there is no structural case position/assigner in passive constructions (i.e., no SpecvP, see also Zwart 2001b). Whether a passive construction involves or does not involve a passive v0 is orthogonal to the discussion of restructuring. 12. Note that the correlation is only a one way correlation — i.e., obligatory control does not entail restructuring (however, restructuring entails obligatory control). As we discuss in detail in Wurmbrand (2001), restructuring is subject to a number of syntactic and semantic conditions that all have to be satisfied. Thus, while obligatory control is a necessary condition for restructuring it is not a sufficient one. 13. Ditransitive constructions in which the verb combines with an infinitive and an accusative argument will be ignored since these constructions are generally non-restructuring infinitives (cf. Sabel 1996). 14. There are some well-known counterexamples to this claim. In particular, anaphors bound by an implicit argument seem to improve in generic contexts (at least for some speakers). Since all the examples we use are non-generic, the judgments are consistent and this interfering factor does not affect the argument made in the text. 15. Note that the presence of accusative case in (20b,c) (vs. 20a) does not a priori entail that these infinitives are non-restructuring infinitives, since the matrix predicates are potential structural case assigners and hence accusative could be assigned by the matrix predicate. However, in addition to the binding properties which we will discuss in the text, there is independent support for our claim that these examples are indeed non-restructuring infinitives. Although we cannot reproduce our analysis and the arguments for it here, we want to point out that in Wurmbrand (2000), we provide a way to distinguish whether case is assigned by the matrix predicate or the embedded predicate. In short, we show that in restructuring constructions, case checking has to be met by movement, and that movement for case reasons does not allow reconstruction in German. Since the embedded objects in (20b,c) occur inside the infinitives and take scope under the matrix predicates, it can be concluded (given the analysis in Wurmbrand 2000) that structural case is not assigned in the matrix predicate. 16. Some speakers might find long passive as in (21c) slightly marked. As mentioned in note 15, case movement in restructuring constructions does not allow reconstruction; the only way constructions with “long” passive can be interpreted is with the object taking scope over the matrix predicate. This resulting interpretation is often pragmatically marked (as in (21c) with the object taking scope over forbid). The markedness of (certain) long passive constructions is thus expected and does not affect the analysis here or our account of restructuring. 17. By clausal complement, we refer to a structure that includes a subject (i.e., at least vP). Whether these infinitives involve other functional projections such as TP, CP, is left aside here. 18. In English, it-anaphors can only replace gerunds. As we will see below, this is 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. 19. (27d) is acceptable under the (irrelevant) interpretation where begin is interpreted as cause. 20. 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. 21. We thank Idan Landau for pointing us to this prediction. 22. Note that if one were to assume that binding takes place after the meaning postulate applies (i.e., (30d) is the syntactic structure for (30) and binding applies to the structure in (30c) 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 (22b), which would (incorrectly) be predicted to be grammatical. The same problem would arise for the other examples provided in Section 3.3 as evidence against a syntactic PRO subject.

Movement and Morphology

Parametric variation and scrambling in English Roland Hinterhölzl Humboldt University

1.

Introduction and outline

In the Principles and Parameters approach, word order differences between languages have been explained by assuming parametric variation in the headcomplement order. To explain the differences in word order between German and Dutch on the one hand and English on the other hand, it was assumed that in the Westgermanic SOV-languages the functional and lexical projections in the IP domain are head-final. In a framework subscribing to the universal base hypothesis (Kayne 1994, Chomsky 1994), word order differences have to be related to other properties. In the different installments of the Minimalist Program (Chomsky 1993, 1994), the desideratum is established that parametric differences between languages be reducible to morphological differences of the lexical items of these languages. Along these lines, Vikner (1997) tried to relate the extent of overt V-movement in languages to the “strength” of the morphological paradigm of regular verbs in those languages. Zwart (1993) proposed that OV-orders are the result of overt movement of the object into AgrOP, while VO-orders result from covert movement of the object to a licensing agreement position. In a similar vein, the difference in the position of the direct object vis-a-vis the verb between English and German can be related to the “strength” of the case morphology of the nouns in these languages. However, if one looks at the situation in Dutch — OV-order but no Case-morphology — one realizes that the strength of Case-features in Dutch comes out as a rather arbitrary property.

132

Roland Hinterhölzl

Moreover, if we look at the different positions that verb-particles, small clause predicates and argument-PPs occupy in, say, English and German, then it becomes doubtful whether any morphological property of these elements can be found that could be held responsible for these differences. A property that correlates with the position of the object with respect to the verb and that has received little attention in this connection so far, is the position of so-called event related adverbs, that is, Time, Place and Manner adverbs. These adverbs occur preverbally in the order T > P > M in OV-languages but postverbally in the exact mirror image in VO-languages, as is illustrated in (1). (1) a. C b. C

TPM—V V—MPT

OV-languages VO-languages

It is clear that this difference cannot be simply explained by differences in the scope of V-movement. Haider (1993, 2000), subscribing to the universal base hypothesis, proposed that SOV is the underlying base order from which SVO orders are derived by head movement of the verb, necessary to license its arguments in the canonical licensing direction of the language (left in German, but right in English). The distribution of event related adverbs in these languages — if one order is to be derived from the other — must involve movement of more complex categories, since the unmarked order of these adverbs is TPM in German but MPT in English. Given Cinque’s seminal work on adverbs, these adverbials are peculiar in several respects (cf. Cinque 1997): a. they appear to differ from adverbphrases (AdvPs) proper in not being rigidly ordered; b. in contrast to AdvPs proper, they can be interchangeably in each other’s scope, as is illustrated in (2); (2) a. They met students everyday of the week in a different university b. They met students in each university on a different day

c. they differ from AdvPs proper in being typically realized in the form of PPs or bare NPs. I assume that this property explains the former two properties: being PPs or NPs, they have the potential for scrambling which may move them out of their base position to other positions according to their quantificational or referential properties. And

Parametric variation and scrambling in English

d. scope may go from right to left (cf. (3a)), but binding only from left to right, as is illustrated by the contrast in (3b,c). (3) a. John met Mary in a (different) park every Sunday b. *Sue met Mary in his house on everybody’s birthday c. Sue met Mary on everybody’s birthday in his house

The standard account of postverbal adverbs in VO-languages was given in terms of layered adjunction to the VP on the right. This account not only violates the Linear Correspondence Axiom (LCA) of Kayne (1994), but also fails to account for binding relations as illustrated in (4). (4) She met everyi student on hisi birthday

Pesetsky (1995) proposed to resolve this paradoxical situation by assigning a dual structure to each clause: a layered one and a cascading one. The layered structure accounts for the basic constituent structure pertaining to postverbal complements and adjuncts. The cascading structure provides the relevant c-command relations for binding. In this article, I argue that there is a better solution to Pesetsky’s paradox that does not involve a dual structure and that this solution implies that English has scrambling. Furthermore, I argue that scrambling of DPs in English is ‘silent’, that is to say, that it involves Spell-out of the lower copy. The article is organized in the following way. In Section 2, I discuss Pesetsky’s paradox in more detail. I present two arguments that the cascading structure is untenable and that the layered structure in English is to be derived (via successive cyclic intrapositon of verbal projections) from the German order (cf. (1)) obeying the LCA. Finally, I propose that the binding facts should be accounted for in terms of scrambling of the arguments of the verb. In Section 3, I outline a minimalist theory of parametric variation that accounts for the differences in verb movement and scrambling between English and Continental West Germanic. In Section 4, I give an overview of scrambling operations that occur in Dutch and German and discuss their properties and restrictions. From this discussion, it will follow that there is no reason to assume that English could not have the types of scrambling operations observed in Dutch. Finally, in Section 5, I provide some more arguments for silent scrambling in English. First, I discuss certain scope facts and argue that — in the absence of covert movement — they can be given the simplest explanation if it is assumed that scrambling is silent in English. Then, I discuss the phenomenon of antecedent-contained deletion (ACD) and show

133

134 Roland Hinterhölzl

that scrambling, contrary to approaches based on Quantifier Raising (QR) and Case movement, can resolve all cases of ACD.

2. Pesetsky’s paradox or What is the correct analysis of event related adverbs? The standard approach to the syntax of adverbs in English assumes that adverbs are right-adjoined to VP or I, as indicated in (5). (5) [IP SU [VP [VP V DO] Adjunct]]

Right-adjunction structures, either base generated or derived by movement, are incompatible with the universal base hypothesis. Independently of the universal base hypothesis, Larson (1988), Stroik (1990) and Pesetsky (1995) have argued that the standard approach to the syntax of adverbs is mistaken, since it fails to account for basic c-command relations between them and the complements of the verb. Typical c-command diagnostics, such as negative polarity item (NPI) licensing (6a) and quantifier-bound pronouns (6b), indicate that postverbal adjuncts are in the c-command domain of postverbal complements. (6) a. John saw no student in any classroom b. John visited everyone on his birthday

Since in the representation in (5) the direct object fails to c-command the postverbal adjunct, Larson (1988) proposed that event related adverb(ial)s are part of a (multi-) layered VP-shell in which these elements are deeper embedded than the complements of the verb as is indicated in (7). (7) [VP SU V [VP DO tV Adjunct]]]

2.1 Against a Larsonian approach In the Larsonian approach, event related adverbs are analyzed as a sort of (optional) complements in the VP. While the Larsonian approach neatly accounts for the c-command effects illustrated in (6), it fails to account for standard constituency tests such as VP-fronting. The latter process indicates, contrary to the state of affairs in (7), that the verb and the direct object form a constituent which excludes adjuncts (8c). On the other hand, constituents of Larsonian shells motivated by binding do not permit fronting (8d). These observations led Pesetsky into proposing an additional — layered — structure.


(8) John promised that he would visit them in Vienna on Friday, and… a. visit them in Vienna on Friday, he did b. visit them in Vienna, he did on Friday c. visit them, he did in Vienna on Friday d. *them in Vienna on Friday, he visited

To explain (8c) without an additional structure, it has to be assumed that the adjuncts are moved out of the VP, that is, are evacuated, before VP-fronting. This account raises several questions. First, it is not clear what the motivation for this evacuating movement should be. Secondly, it is not clear why this movement is restricted to the right-most constituent in each case. (9) shows that evacuating an intermediary adjunct leads to reduced acceptability. (9) ??visit them on Friday, he did in Vienna

In (9), the adjunct in Vienna has been extracted from a specifier position before VP-fronting takes place. To explain the contrast between (8b) and (9), one may propose that only complements may be evacuated. However, this account cannot be generalized, since the subject and the object, in the presence of an adjunct, occupy a specifier position in the Larsonian shell, but can be freely extracted from the VP. Another problem with the Larsonian approach concerns the anti-c-command requirement in the licensing of parasitic gaps: the trace of the licensing operator may not c-command the parasitic gap. This derives the descriptive generalization that objects can license parasitic gaps in adjunct clauses, while subjects fail to do so (cf. (10a) versus (10b)). Since in the Larsonian approach, both subjects and objects c-command adjuncts, it is unclear how to account for this generalisation. (10) a. Which article did you criticize before reading? b. *Who saw you before you recognized?

The above noted problems with the Larsonian approach I consider minor or technical problems that may be overcome in a relatively simple fashion. In the following I present two arguments which I consider more detrimental to the Larsonian approach. The first argument involves the semantic interpretation of event related adverbs and the second argument pertains to the comparative dimension of accounting for the different distribution of these adverbs in OVand VO-languages. First, the Larsonian approach to the syntax of event related adjuncts raises questions about the proper interpretation of these elements. In a Larsonian

135


shell structure, temporal adverbs are deeper embedded than manner adverbs, as is shown in (11). (11) a. John wrote the letter carefully today b. [VP John wrote [VP the letter tV [VP carefully tV today]]]

Following Ernst (1998), Haider (2000) and others, I assume that the attachment of adverbs is determined by their scopal properties. The scopal requirements of an adverb include selection for a clausal argument of a particular type. Ernst (1998) specifies a schema of abstract clausal entities relevant for the interpretation of adverb(ial)s. (12) Speech Act > Fact > Proposition > Event > Specified Event

From (12) it follows, for instance, that evaluative adverbs like unfortunately selecting for a fact cannot attach lower to the clausal skeleton than modal adverbs like probably selecting for a proposition, though they can otherwise occupy various positions in the clause, as is illustrated in (13). (13) a. (Unfortunately) Eddie (unfortunately) has (?unfortunately) left b. *Probably Eddie unfortunately has left

From a semantic point of view, manner adverbs specify an aspect of only part of the event, namely the process component of the event, while temporal adverbs situate the entire event with respect to the utterance time. Thus, standard assumptions about the interaction of syntactic structure and semantic interpretation predict that temporal adverbs should attach to the clause higher than manner adverbs, not lower, as in the Larsonian approach. Secondly, here is what I call the comparative argument. If the English order is basic, then it is not clear how the German order is to be derived. A roll-up structure that moves a constituent containing the temporal adverb in front of the manner adverb and subsequently moves that larger constituent in front of the final position of the verb fails to account for the scopal properties of these adverbs in the middle field. In the German middle field an adverb always has scope over the adverb to its right. A derivation in terms of movement of the adverbs by themselves raises several questions. The straightforward derivation violates cyclicity as illustrated in (14).


(14) [TP

[VP SU [V [DO V1 [Manner [V2 Temp]]]]]] 2 1

Though there is a derivation that obeys cyclicity — the one that moves the manner adverb and the direct object into the middle field before the temporal adverb is moved —, both derivations suffer from the circumstance that it is not clear what the motivation behind these obligatory movements should be. In other words, why is it that event related adverbs are apparently licensed in situ in English but have to undergo licensing movement in German? To the extent that we cannot find a satisfactory answer to this question, the Larsonian approach is rendered unattractive. 2.2 For a comparative approach Given the above considerations, in particular, taking serious the semantic argument, it seems that the order of event related adverbs observed in German, namely T > P > M, is closer to the base than the English order. Thus, I propose that the English order is derived from the German order via successive cyclic intraposition of verbal projections.1 In the following, I assume that manner adverbs are base generated in the VP while Time and Place adverbs are base generated above VP as is indicated in (15). (15) [Temp … [Loc … [v [Manner [V DO]

Under these assumptions, the English sentence in (16a) is derived from the base structure in (16b) via successive intraposition as indicated in (16c). (16) a. John visited them in Vienna on Friday b. [IP Johni [on Friday [in Vienna [VP ti visited them]]]] c. [IP John [[[visited them] in Vienna] on Friday]]

The representation in (16c) accounts for the VP-fronting facts illustrated in (8) and (9), but fails to account for the binding facts. To explain the data in (6), I assume that English has “silent” scrambling. Silent scrambling means that there is an overt operation that moves the direct object of the verb into middle field, just like German scrambling moves the direct object out of the VP to a position preceeding adverbs in the middle field, but, unlike German, spells out the lower copy, as indicated by the underlining in (17).

137

138

Roland Hinterhölzl

(17) a. [IP Oi (Adv) [VP V Oi]] scrambling in German b. [IP Oi [VP V Oi] ‘silent’… the lower copy is spelled out

First note that scrambling, being an operation of A-movement, creates a new binding position to license bound pronouns and NPIs. As is shown in (18a,b), scrambling of the direct object across the temporal adverb can license the bound pronoun contained in the adverb. (19) shows that scrambling can license an NPI. (18) a. *weil die Maria an seinemi Geburtstag jedeni Freund besuchte since the Maria on his birthday every friend visited b. weil die Maria jedeni Freund an seinemi Geburtstag besuchte since the Maria every friend on his birthday visited (19) a. *weil Hans jemals keinen Studenten traf since Hans ever no student met b. weil Hans keinen Studenten jemals traf since Hans no student ever met

Also note that if we accept the semantic argument sketched above and assume that the English order is to be derived from the German order, in other words, if we dispense with the Larsonian approach to the syntax of event related adverbs, then scrambling is the only option to derive the binding facts in (6): the scrambling analysis is superior to an analysis in terms of a covert operation at LF, since quantifier raising would lead to a Weak Crossover (WCO) effect in the case of (6b). To account for the contrast in (3bc), I assume that the restriction on pronoun binding displayed in these examples can be subsumed under a leftness condition on variable binding (cf. Hornstein 1994). But this restriction is independent of the scrambling approach and a problem for any analysis of the binding facts that obtain between postverbal adjuncts and arguments in English. 2.3 Against the (revised) standard approach Given the option of silent scrambling, the question arises whether the standard approach to the syntax of adverbs in English in terms of right-adjunction would be simpler. In this account manner adverbs would indeed be attached lower than temporal adverbs. This alternative is viable but it raises a couple of nontrivial questions. One has to assume not only that there is a parameter that specifies the attachment site of event related adverbs but also that this parameter is linked to

Parametric variation and scrambling in English 139

the head-complement parameter. It is not at all clear why exactly event related adverbs but not the other adverbs of the clause (quantificational, aspectual, discourse related or focusing adverbs as well as predicational adverbs that take propositions and facts as arguments) are parametrized this way. Secondly, there is the typological fact that scrambling, viewed as movement out of the VP into different positions — interspersed with adjuncts — in the middlefield, is found in OV-languages but absent in VO-languages. In the comparative approach in which it is assumed that adverbs are left-attached to the clause, the absence of audible scrambling in VO-languages can be explained in the following way. Right-peripheral adverbs require the assumption of VP-intraposition which process would “mask” prior scrambling of an argument across these adverbs for the language learner. In the standard approach, in which it is assumed that adverbs can be either left- or right-adjoined to the clause there is no connection between scrambling and the licensing of adverbs. In this approach, it is not clear why arguments cannot move into the middle field in VO-languages. Another property of adverbs in VO languages that is in need of an explanation is the fact that adjuncts that can occur between the subject and the VP in VO-languages are subject to restrictions absent in OV-languages: (20) a. John (more) often (*than Peter) read the book b. Hans hat öfter (als der Peter) das Buch gelesen Hans has more-often (than the Peter) the book read

Descriptively speaking, the head of the adjunct must not have material to its right. This is only possible if the adjunct appears in sentence final position, that is to say, if the adjunct is right-adjoined. An option, on the other hand, that is not available in OV-languages as the contrast illustrated in (21) shows. (21) a. John read the book more often than Peter b. *Hans hat das Buch gelesen öfter (als Peter) Hans has the book read more-often (than Peter)

The above facts can be captured by the stipulation in (22) if we assume that the canonical position for adjunction is to the right in VO-languages and to the left in OV-languages. (22) Only in the canonical adjunct position may the head of the adjunct take a complement


(22) is problematic in at least two respects. First of all, it is not clear why there should be such a particular restriction like (22) in languages. Secondly, there are numerous counterexamples to this generalization in both English and German. In German, PPs and clauses, when extraposed, can freely occur in rightadjoined position though their respective heads P and C take a complement in these cases. In English, topicalized constituents which are analyzed as being leftadjoined to IP, that is, which are in a non-canonical adjunction position, can, of course, contain complements, as is illustrated in (23). (23) a. When Mary came in, Peter went out b. In the morning, Peter went out c. More often than necessary, Peter went out

Thus the stipulation in (22) is far too general and it is not clear at all how to restrict it in the appropriate fashion. In the comparative approach adverbs can only be left-attached. There is intraposition of verbal projections to license the verb in VO-languages as we will argue in Section 3. Assuming that the restriction exemplified in (20) has something to do with heavy material occuring in sentence-medial position, I propose that — since intraposition of verbal projections is available in VO-languages in general — it can be used for an operation that possibly applies at PF and moves light predicates across heavy adverbs such that the latter end up at the right periphery. Since the grammar of German does not have at its disposal an operation of intraposition of verbal projections, (21b) is ungrammatical. In the following section, I will outline a theory of parametrization within which the differences between German and English can be explained without recourse to the head complement parameter or the notion of feature strength.

3. Parametric variation and Which factors determine the Spell-out of copies? I assume that languages are essentially alike differing only in how they satisfy universal constraints. That is to say that I assume that all movement operations are overt. In particular, I assume that languages may differ in only two macroparameters: A. Each feature may be checked via XP- or X0-movement B. In checking a feature either the higher or the lower copy may be spelled out


With respect to Parameter A, I assume that there is a constraint ‘Attract Closest’ which implies that XP-movement is the unmarked option in feature checking. X0-movement is dependent on special requirements of the attracting head, for instance, on the attracting head being an affix. Parameter A is relevant for the positioning of adverbs. When the English verb undergoes licensing movement, it does so by XP-movement (the unmarked option) which will result in inverted orders. When the German verb undergoes licensing movement, it does so by X0-movement, followed by remnant XP-preposing (cf. Hinterhölzl 2000, Haegeman 1999), which leaves preexisting orders unchanged. With respect to Parameter B, I assume that in checking an uninterpretable feature, the higher copy must be spelled out.2 In checking an interpretable feature either copy may be spelled out unless the attracting head has a positional feature. A positional feature requires that the attractee is spelled out in the checking domain of the attractor (this is different from the notion of strong features). Typical examples of positional features are [wh], [neg] and [focus]. To illustrate the notion of a positional feature, let us look at the distribution of definite DPs in German. Definite DPs always have to move across negation, as is shown in (24ab). We may assume that this movement occurs to check a specificity feature of the NP. However, if the definite DP is focused, checking of the specificity feature may seemingly be dispensed with (24c). (24) a.

weil der Hans das Buch nicht gelesen hat since the Hans the book not read-part has b. *weil der Hans nicht das Buch gelesen hat since the Hans not the book read-part has c. weil der Hans nicht das BUCH gelesen hat since the Hans not the book-focus read-part has

This is a problem, since a feature that needs to be checked, should be checked alike in all contexts. A neat analysis of the facts in (24) is obtained, if we assume that the DP has in fact moved into the domain where specifics are licensed (that is above sentence adverbs including negation) but due to the positional feature of focus, is spelled out in the specifier position of a functional head which occurs below negation. Which head in a given language has a positional feature is subject to crosslinguistic variation. I assume that Parameter B is relevant for the instantiation of scrambling. In particular, I assume that Spell Out is guided by economy considerations in the

141


following manner. If a language has the basic word order XY and there is interpretative evidence for movement of Y, but no evidence for movement of X then the simplest analysis is that Y is spelled out in the lower copy. These Spell Out choices are fixed by an optimizing learning algorithm in the process of language acquisition. A child acquiring English will, on encountering interpretive evidence for scrambling but no evidence from the PF-side, assume that scrambling spells out the lower copy. In the next section, I will provide a typology of scrambling operations in West-Germanic and show how scrambling in English fits into the picture.

4. Silent scrambling: A typology of scrambling operations A careful investigation of scrambling operations in German and Dutch reveals that there are (at least) three types of scrambling operations whose properties are summarized in the table given below. Name

Motivation

Permutation Reconstruc- moved item creates new Depends on of tion bears stress binding Case arguments position

Permutation Familiarity/Scope yes Extension Familiarity/Scope no F-scrambling Contrast yes

no no yes

no no yes

yes yes ?

yes no no

First, there is the type of scrambling known from German which allows movement of the object across the subject, as is illustrated in (25a,b). It is this operation that can be held responsible for voiding WCO-effects (25b,c) and Superiority effects (25d) between subjects and objects in German. (25) a.

weil den Hans die Maria liebt since the-acc Hans the-nom Maria loves [since Mary loves John] b. weil jedeni seinei Mutter liebt since everyone-acc his-nom mother loves [since his mother loves everyone] c. Weni liebt seinei Mutter nicht? whom loves his mother not [who does his mother not love]


d. Wen hat wer geliebt? whom has who loved [who does who love] e. weil jeden eine Frau liebt since everyone-acc a-nom woman loves [since some woman loves everyone]

I call this operation permutation since it allows for the arguments of the verb to be generated in any order. It is an operation of A-movement since it creates a new binding position (cf. (25b)) that cannot be reconstructed, as can be seen from the fact that sentences like (25e) are unambiguous. Permutation is unavailable in Dutch, as is illustrated in (26a,b). (26) a.

dat Jan de boeken niet koopt that Jan the books not buys b. *dat de boeken Jan niet koopt that the books Jan not buys

But Dutch allows for scrambling of the arguments of the verb across adverbs that occur in the middle field as long as scrambling preserves the order of the arguments. I call this operation extension. Extension is a type of A-movement, as we have seen in (18) above, that like permution does not allow for reconstruction, since the resultant structures are unambiguous. With permutation and extension, the moved item is unstressed. That there is a third type of scrambling is most clearly shown by Dutch. As was shown in (26), the direct object can generally not move across the subject. However, there is a type of scrambling which Neeleman (1994) called focus-scrambling, but really involves movement of a contrastive topic, which exactly does that, as is shown in (27a). This operation comes with a special intonational contour (the so-called hat contour, with a rise on the moved item and a fall on the focused element), obligatorily requires a (contrastively) focused element in the remainder of the clause and is to be conceived as an A-bar movement operation, since it may not only affect arguments, as permutation and extension exclusively do, but also predicates, as is illustrated in (27b). (27) a.

dat zulke boeken zelfs JAN niet koopt that such books even Jan-focus not buys b. dat rood zelfs JAN het hekje niet verft that red even Jan-focus the fence not paints


One may wonder why two types of A-scrambling have to be distinguished and what their respective motivations are. As has been noted before (cf. Diesing 1992), scrambling often comes with a specificity effect. (I use the term familiarity to encompass DPs that are familiar from the linguistic as well as the nonlinguistic context). That A-scrambling cannot be reduced to the sole trigger of specificity/ familiarity is shown in (28). (28) weil jeder zwei Bücher einer Frau schenkte since everyone-nom two books-acc a-dat woman gave (intermediate scope possible)

In (28), the direct object has scrambled across the indirect object but remains in the scope of the subject. If A-scrambling necessarily came with a specificity effect, then the direct object should have only widest scope in (28). Since an intermediate scope-reading is possible A-scrambling must be able to have an additional trigger which I identify with scope-taking. I assume that permutation and extension can target the same positions. In German, additional positions are available which allow for the permutation of arguments. I identify these positions as the specifiers of functional heads hosting clitics and assume that these positions are accessible for DPs in a language, if Case in that language distinctively marks grammatical functions. This is the case in German but not in Dutch or Icelandic. Older stages of Dutch and English had scrambling of the German type. Dutch probably lost permutation when it lost its Case distinctions in the nominal system (which are partially preserved in the pronominal system). Given these observations, it seems reasonable to assume that English preserved scrambling of the Dutch type, that is, extension and F-scrambling. In the final section, I present an additional argument for the conjecture that English has scrambling, as well as some observations that speak for the silent character of scrambling in English. 5. Evidence for silent scrambling Let me first expand on why I think scrambling is silent in English. If English has scrambling of the Dutch type, then we expect that the direct object can be separated from the verb by adverbs in the middle field. This is generally not the case. In particular, the direct object cannot be separated from the verb by a postverbal manner adverb. Note that due to VP-intraposition this state of


affairs is actually expected, as is illustrated in (29). In (29c), the VP has to move higher than the scrambled direct object (DO) in accordance with the extension condition on derivations. (29) a. [C … [Manner [VP V DO]]]] base structure scrambling b. [C … DOi [Manner [VP V DOi]]]] c. [C … [VP V DOi] DOi [Manner tVP]]] VP-intraposition

(29c) does not tell us anything about whether scrambling is silent or ‘audible’ in English: Spell Out of either copy of the direct object derives the correct word order. However, if we look at more complex structures, then we realize that scrambling in English must be silent. (30) shows that the direct object may have scope with respect to two adverbs, while the expected word order that the scope taking scrambling operation (across both adverbs) yields, is impossible. (30) a. John met every woman twice on a Sunday b. *John met twice every woman on a Sunday

In (30), the direct object may have scope over Sunday and twice, which is actually the prevalent reading (W > S > t). If scrambling in English were audible then the wide scope reading in (30) would require additional movement of the verb and the direct object to derive the correct word order. The simplest option of deriving this reading with the word order given in (30a) is to assume that scrambling is silent in English. Note, by the way, that the impossible English order in (30b) is licit in another VO-language, Italian. Diana Pili (2000) reports that the order V – Manner – DO – higher Adv is possible in Italian. Finally, I would like to discuss a phenomenon that supports my conjecture that English has scrambling. The assumption of (silent) scrambling has the advantage that it can account for all cases of antecedent-contained deletion (ACD). There are two types of explanations that posit covert movement for ACD-resolution: QR and covert Case driven movement to AgrOP. Within these approaches it is agreed that movement of the direct object out of the VP is necessary to create the proper antecedent, the V-trace-configuration, that can be copied into the ellipsis site, as is illustrated in (31). (31) a. Mary invited everyone that I did b. Mary [everyone that I [e]] invited t

[e] = [invited t]

May (1985) has argued that the phenomenon of ACD provides good evidence for the existence of QR, since only quantified expressions but not referential

145


expressions can resolve cases of ACD, as the contrast between (31a) and (32a) shows. (32) a. *Mary invited Peter who I did b. Mary invited Peter, who Susan also did

However, Cases of ACD involving names are not impossible in general. (32a) improves considerably if the relative clause expressed some contrast to the previous clause, that is, if it contains some focused element, as is illustrated in (32b). Since names are not the type of elements that can reasonably be assumed to be in need of QR for the sake of their interpretation, QR-based approaches have problems with ACD involving referential expressions like names (Vanden Wyngaerd and Zwart 1991). This observation led Hornstein (1994) into proposing that cases of ACD are resolved by Case-driven movement into AgrOP in English, the argument being that both referential and quantificational DPs are in need of Case. Note, however, that Case-movement-based explanations have problems with NP-contained ACD, as has been pointed out by Kennedy (1997) and Pesetsky (2000). The interpretation of (33a), given in (33b), involves movement of the PP out of the containing NP a report, which cannot possibly be motivated with the need for checking Case. (33) a. Mary read a report on every murder the police did b. every murder Mary read a report on, the police read a report on c. weil Maria über jeden Mordfall einen Bericht las since Maria about each murder a report read

Note, however, that scrambling can be argued to be able to resolve all the cases of ACD illustrated in (31)–(33). Names and QPs scramble alike in German, though due to different motivations as we have seen in the previous section. And scrambling may also move a PP out of an NP, as is illustrated in (33c), as long as the NP is indefinite, which is exactly the restriction that applies to cases of NP-contained ACD. The assumption of silent scrambling may even provide us with an explanation for the restriction on ACD involving names that is illustrated in (32ab). Let us make the reasonable assumption that each movement operation must be either PF- or LF-interpretable. Silent scrambling never has a PF-effect and will thus be only licit if it has an LF-effect. However, scrambling of names is not interpretable in terms of scope, since a name, independently of its position, is always interpreted in the same way, namely, with widest scope. It thus follows that silent scrambling of names cannot be extension-scrambling but must be


F-scrambling which, as we showed in the previous section, necessarily involves a focused constituent which sets up the required contrast. Since I propose that cases of ACD are better not treated as being resolved by the operation of QR, the question arises whether QR plays any role in English and in language in general. In the spirit of Kayne (1998), it would be advantageous to have a grammar without postcyclic rules, that is to say, a grammar in which all operations occur overtly in a cyclic fashion. Two phenomena immediately come to one’s mind which were traditionally taken as evidence for the existence of QR: cases of inverse scope and cases of inverse linking, illustrated in (34a,b), respectively. (34) a. Someone loves everyone b. Some inhabitant of every city hates its traffic

Let us talk about cases of inverse scope (34a) first. The ambiguity of (34a) should not be explained in terms of reconstruction of the subject to its base position below the object in SpecAgrOP, since then we would not expect a Weak Crossover effect in cases like (35a) below. Thus the ambiguity should be explained in terms of movement of the object across the subject. Such an operation is QR or scrambling as is shown in the German example (35b). (35) a. *Hisi mother loves everyonei b. weil jedeni jemand ti liebt since everyone someone loves ‘for everyone there is someone that loves him’

Let us assume that QR does not exist. If scrambling is to do the job (involving a Spell-out of the lower copy marked by the trace in (35b)), then the inverse scope in (34a) can only be achieved by applying f-scrambling, since extension, as we have seen in (26) for Dutch, cannot move objects across subjects. F-scrambling being an operation of A-bar movement will induce a WCO-effect in (35a). Thus we derive the desired effect, that no WCO-effect is induced with respect to binding of pronouns contained in event related adjuncts (since the A-movement operation of extension is available there), while a WCO-effect obtains when a bound pronoun is contained within the subject. Furthermore, we predict that the inverse scope reading in (34a) is not freely available but comes with a special intonation. Most native speakers I have asked agree with this judgment but I have not been able to find out whether the special intonation is the same intonational contour that is typically related with f-scrambling in Dutch and German.


The standard account of the case of inverse linking in (34b) involves QR of the universal quantifier that adjoins it to IP, from which position it can bind the pronoun in the object. Note that this case is more recalcitrant, since the parallel operation of overt scrambling of the possessive DP out of the containing DP is not available in all cases in German. Extraction of a PP out of a transitive subject is generally ungrammatical, as is shown for the parallel case to (34b) in (36a). In addition, binding of a pronoun outside of the containing DP is impossible in this case in German, as is shown in (36b). (36) a. *weil von jeder Stadt [ein Einwohner t] den Verkehr haßt since of each city [an inhabitant it hates b. *weil ein Einwohner von [jeder Stadt]i ihreni Verkehr haßt since an inhabitant of [every city its traffic hates

Furthermore, note that, abstracting away from binding the pronoun, (36b) is not ambiguous in German. It can only have the absurd reading that someone who lives in every city hates (some) traffic.3 If QR is the solution to cases like (34b), then the question arises why QR is available in English, but not in parallel cases in German. Instead, I would like to argue that the difference between (34b) and (36b) be related to another difference between German and English, which points to a difference in the structure of DP between the two languages. In German, a prenominal possessor, contrary to the state of affairs in English, cannot bind a pronoun outside of the containing DP, as is shown in (37b). Again, if (37a) is to be explained in terms of QR, then the question arises, why QR is possible in English but not in the parallel German case. (37) a. Every mani’s mother loves himi b. *Jedermannsi Mutter liebt ihni

Kayne (1994: 26) proposes that the English DP contains two specifier positions, a lower one hosting genitive subjects and a higher one which he identifies as an operator position and which in his system allows for c-command out of the DP. We may assume that it is movement to this position which is responsible for the inverse reading in (34b) and that this position, for some reason, is unavailable in German. If this is correct, then QR is not needed to explain cases of inverse linking in English: movement of the relevant DP to the high operator position is sufficient to license inverse scope as well as the bound pronoun and makes extraction out of DP superfluous. In German, inverse linking and a DP external bound pronoun are only possible if extraction is possible. This is illustrated in (38) where the PP is extracted from a direct object DP.


(38) weil Maria über jeden Schauspieler [eine Broschüre t] an seinen since Maria about each actor [a brochure to his Agenten schickte agent sent

6. Conclusion To summarize, I have outlined Pesetsky’s paradox which involves the assignment of a dual structure to sentences containing event related adjuncts. I presented two arguments showing that the Larsonian approach to event related adverbs, which derives Pesetsky’s cascading structure to account for the c-command properties between postverbal complements and postverbal adjuncts is mistaken. Then, I argued that Pesetsky’s layered structure is to be derived from a corresponding right-branching structure, directly observable in German and Dutch, via successive cyclic VP-intraposition. I proposed that the interpretational effects yielded by Pesetsky’s cascading structure are derived by the operation of scrambling in English. I argued that scrambling in English is silent, that is to say, must involve Spell-out of the lower copy, and provided further evidence for silent scrambling by showing that silent scrambling, in distinction to QR or Case-based movement can account for all cases of ACD in English. Furthermore, I outlined a minimalist theory of parametric variation within which the differences in object and adverb placement in English on the one hand, and Dutch and German on the other hand, can be accounted for without recourse to an underlying head-complement parameter or the notion of feature strength. Finally, I addressed the question of the (residual) role of QR in the syntax of English. I discussed cases of inverse scope and inverse linking and argued that they are better accounted for without making use of the operation of QR.

Notes 1. There is another base structure given in (i) that respects the semantic condition outlined in the text: (i) [[[VP [e Manner]] e Loc] e Temp] The structure in (i), however, raises several questions. First, it is not clear how the empty heads of verbal type which introduce event related adverbs as their complement and take the VP hosting the arguments of the lexical verb as their specifier, are identified. Secondly, we need to allow for projection to proceed via the specifier, since we want that functional heads


above VP belong to the extended projection of the lexical verb. Thirdly, it suffers from the same defect as the Larsonian structure: to derive the German (Dutch) order, individual obligatory movement of the adverbs into the middle field has to be assumed, the motivation of which remains obscure. 2. Following Nuñez (1995), I propose that in a checking operation only the pertinent features of the merged copy are checked. The uninterpretable features of the lower copy can then only be gotten rid of by phonological deletion, which encompasses formal features in Nuñez’ system. 3. The inverse scope reading becomes more readily available, if the DP involves a clear contrast in terms of focused and nonfocused constituents (cf. ia). Under these conditions also extraction, nota bene out of a passive subject in (ib), is quite natural. (i) a. b.

weil nur der BÜRgermeister von jeder Stadt eingeladen war since only the mayor of every city invited was weil von jeder Stadt nur der BÜRgermeister eingeladen war since of every city only the mayor invited was

No such information-structural condition has been reported for cases of inverse linking in English. The relevance of this extra requirement in German can then be related to the lack of an (unspecialized) operator position in the German DP as is argued below.

V2 and Holmberg’s Generalization* Øystein Nilsen Utrecht Institute of Linguistics OTS

1.

Introduction

The standard (symmetric) analysis of verb second (V2) (Holmberg and Platzack 1995) is that the finite verb (Vf) head-moves to the highest functional projection of the clause. Then some other constituent, for instance the subject or an adverbial, has to move to the specifier of that same projection. These two movement steps, in addition to a general ban on adjunction to CP, ensure that Vf will always end up in the second position of the clause. (1) [CP XPi [CP Vfj+C [IP …tj…ti…]]]

It is the purpose of the present article to argue against a head movement analysis of V2. The main argument for an XP-movement analysis will come from the fact that certain (apparent) V2-violations in Mainland Scandinavian seem to pose severe problems for a head movement analysis. The problematic data involve focus particles that can intervene between the finite verb and the first constituent. It will be argued that these cannot be treated as a clitic on the verb and that the V2 violations are real. The interaction between V2 violations with focus particles and argument shift of weak pronouns will be used to show that the verb does not move to second position as a head. From this conclusion it follows that weak pronouns, in fact, do not shift. When they appear to have moved, it is a larger constituent containing the VP that has moved. Thus, pronoun shift and V2 are treated as surface reflexes of one and the same operation. This gives a very simple explanation for Holmberg’s Generalization to the effect that object shift cannot cross phonetically realized material from the VP: it cannot do so because it is the VP itself, or rather an XP containing it, that moves.

152

Øystein Nilsen

The traditional view of V2 is also challenged by the fact that there are topicalization-like processes and wh-movement effects that seem to require a sentence internal landing site. Furthermore, some facts concerning subject-verb inversion are problematic for the standard treatment of V2. Inversion is usually analyzed as V-to-C movement, with the subject in SpecIP or equivalent position. This leads to the expectation that inversion can only occur when the verb really is in C. However, there are cases in which the verb is arguably much lower than this, and the subject still has to follow it. The proposed account builds on recent work by Kayne (1998, 1999), Cinque (1999), Rizzi (1997), Koopman and Szabolcsi (2000) and others. In particular, no covert movement is used, all movements are to the left and the analysis relies heavily on the use of ‘remnant’ movement, i.e. movement of a constituent containing a trace. More specifically, the proposal is that Rizzi’s (1997) functional projections FocP, TopP and FinP are merged below sentential adverbs. V2 consists in successive raising of TopP around sentential adverbs, carrying the verb-initial FinP along. One of the key features of the analysis is thus that it renders V2 sensitive to the properties of individual classes of adverbs. The article is organized as follows. In Section 2, the basic data concerning the V2 violating focus particles and their interaction with pronoun shift is presented. Section 3 presents a ‘first approximation’ to an analysis that can handle the facts. Section 4 suggests that the problems encoutered with the first analysis is that fronting operations and S-V inversion should be able to access quite low positions. In Section 5 the main proposal is developed. Section 6 sums up and concludes.

2. Some data 2.1 V2-violations with focus particles As has been discussed by Egerland (1998), there exist certain apparent exceptions to the V2-generalization in Mainland Scandinavian involving so-called ‘focus-sensitive adverbs’ or ‘focus particles’ (henceforth fpt). The phenomenon is illustrated in (2) below with data from Norwegian.1 (2) a.

Jens bare gikk. J. just left b. Jens nesten gråt. J. almost cried

V2 and Holmberg’s Generalization

It is also possible to have the fpt after Vf, as in (3). Neither of the two orders appears to be marked or degraded in any way. (3) a.

Jens gikk bare. J. left just b. Jens gråt nesten. J. cried almost

Other expressions that exhibit the same behavior include til og med ‘even’ (lit. ‘to and with’), minst ‘at least’, utelukkende ‘exclusively’, ikke mer enn såvidt ‘not more than barely’, simpelthen ‘simply’. Thus we are not dealing with a quirk of a couple of words. Below are examples. (4) a.

Han til og med leste den. he even read it b. Han utelukkende sover hele dagen he exclusively sleeps whole the-day c. Han ikke mer enn såvidt berørte den. he not more than barely touched it d. Han simpelthen tok den. he simply took it

Egerland (1998) notes that with nesten ‘almost’ there is a truth-conditional difference corresponding to its different positions. Consider the following examples. (5) a.

Jens nesten brølte hurra. J. almost roared hooray b. Jens brølte nesten hurra. J. roared almost hooray

(5a) can only mean that Jens pronounced the word “hurra” in a manner that almost qualifies as roaring it. Let us refer to this as the ‘manner’ reading. The most salient reading of (5b) is that he didn’t cry “hurra”, although he was about to, i.e. a ‘modal’ reading. It can also get the manner meaning if pronounced with heavy stress on the verb. See Rapp and Von Stechow (1999) for discussion of these and other readings of the German adverb fast (‘almost’). This pattern can also be taken to indicate that the fpt is not adjoined to C’, since the manner reading presumably results from attaching the adverb lower, not higher, than the site responsible for the modal reading. Egerland (1998) analyzes this phenomenon in terms of Cinque’s (1999) Universal Hierarchy. The cases involving nesten and those involving bare are

153

154

Øystein Nilsen

given different analyses. Egerland maintains a standard analysis of V2 in terms of head movement to the highest FP in the clause (ForceP in Rizzi’s (1997) ‘split CP’ framework). He analyses the adverb nesten (‘almost’) as a specifier of a modal projection in the IP-layer. For (5a), he claims that Vf can remain in or below that modal head when nesten is in its specifier. The adverb bare (‘only’, ‘just’), he treats as a syntactic clitic on Vf. The analysis of bare as a clitic is supported by two facts. The first is that the adverb can be phonetically reduced into the monosyllabic ba’ in Swedish (not possible in Norwegian). (6) Per ba’ gick. P. just left

[Swe]

The second argument is that, according to Egerland, bare cannot appear in front of auxiliaries. (7) *Per bara/ba’ har gått. P. just has left

[Swe]

The same applies to Norwegian as long as the example is read with neutral intonation, but if the auxiliary is stressed (8), the result is much better; also if a less semantically impoverished auxiliary is used (9).2 (8) Jens bare HAR gått. J. just has left ‘It just IS the case that Jens has left.’ (9) Jens bare måtte gå. J. just must-past leave ‘Jens just HAD TO leave.’

There are other problems with assuming that bare is a clitic. First, it can be modified. (10) Jens simpelthen bare gikk. J. simply just left

After Kayne (1975), one of the defining characteristics of syntactic clitics has been taken to be that they cannot be modified. Secondly, other adverbs that cannot plausibly be taken to modify bare can also precede Vf when bare does; in fact, only when bare does. (11) a.

Jens vanligvis bare svarer ikke. J. usually just answers not b. *Jens vanligvis svarer ikke. J. usually answers not


This points to the conclusion that bare should be treated on a par with nesten, so that, when these adverbs are present, Vf can remain in a low position in the IP-field. If the position of bare is lower than that of vanligvis, we can also explain why the latter adverb actually has to precede Vf when bare does. Compare (12) to (11a): (12) *Jens bare svarer vanligvis ikke. J. just answers usually not

Since the negation must follow Vf in the relevant construction, it cannot be regarded as V-in-situ, either. (13) a.

Jens bare liker ikke fiskekaker. J. just likes not fishcakes b. *Jens bare ikke liker fiskekaker. J. just not likes fishcakes

So far, we can conclude that V2-violating bare is not a clitic on Vf; that bare occupies a position lower than vanligvis; and that Vf can remain below that position when bare is present, although it cannot remain in situ. 2.2 Weak Pronoun Shift In the next few paragraphs, we will see that there are reasons to think that V2 does not involve head movement of Vf, but rather movement of a phrasal category. A corollary of our observations will be that weak pronouns, in fact, do not shift. Consider the pattern below. Derfor vanligvis bare svarte ’n ’a ikke. therefore usually just answered he her not b. Derfor svarte ’n ’a vanligvis bare ikke. therefore answered he her usually just not c. *Derfor ’n ’a vanligvis bare svarte ikke. therefore he her usually just answered not d. *Derfor svarte vanligvis bare ’n ’a ikke. therefore answered usually just he her not

(14) a.

In (14) we see that when the subject and the object are both realized as weak pronouns, they must remain immediately right-adjacent to the verb. In (14a), the verb+arguments complex remains below the position of bare, whereas in (14b), the entire complex is moved around bare and vanligvis to the V2-position.

155

156 Øystein Nilsen

(14c,d) are added to show that the complex cannot be split up. This seems to indicate that it is moving as a constituent. The alternative would be to say that the verb and the arguments move separately, but that the verb somehow blocks further movement of the pronouns (cf. Chomsky’s 1995 Shortest Move/ Minimal Link Condition). One would need an extra landing site for the pronouns, higher than the negation, but lower than the other adverbs. The pronouns would move as high as they could without crossing the verb in the overt syntax, and then proceed to the higher position(s) covertly. The features attracting the pronouns to the higher positions would have to be ‘optionally strong’. It is obviously simpler to say that the Vf and the pronouns are moving as a constituent. That analysis obviates the need for extra landing sites and optionally strong features. Taking V2 to be derived by XP-movement, we arrive at the following result: (15) Weak pronouns do not shift.

When a weak pronoun appears to have moved across adverbs, it is something else containing the pronouns that has moved. Consider now the following pattern where the subject and the object appear to be moving as a constituent without Vf. Derfor svarte Jens ’a vanligvis ikke. therefore answered J. her usually not b. Derfor svarte vanligvis ikke Jens ’a. therefore answered usually not J. her c. *Derfor svarte Jens vanligvis ikke ’a therefore answered J. usually not her d. *Derfor svarte ’a vanligvis ikke Jens. therefore answered her usually not J.

(16) a.

The subject and the weak, pronominal object can follow the adverbs as long as they remain adjacent. In Swedish, (16c,d) are also possible. The two arguments need not remain adjacent in that language. In Danish, only (16a) is grammatical. For Norwegian, then, the subject and the object seem to make up a constituent in these examples.

3. First approximation A simple way to derive these constituents, which will be shown to be inadequate shortly, would be to assume the following. There is an XP dominating VP into


which Vf always moves. This XP, in turn, moves to SpecFin prior to fronting of some constituent to SpecTop. The derivations would go as follows, ignoring the base position of the adverbial derfor (‘therefore’). Phonetic material is in boldface and cyclicity is ignored in this derivation. (17) [TopP Top [FinP Fin [IP not [XP X [VP he answered her]]]]] Vf moves to X fi (18) [TopP Top [FinP Fin [IP not [XP answeredi+X [VP he ti her]]]]]

XP moves to SpecFin and some constituent topicalizes fi (19) [TopP therefore Top [FinP [XP answeredi+X [VP he ti her]] Fin [IP not tXP]]]

This gives us the verb+arguments constituent we demonstrated in (14) above. If the subject bears focus, the remnant VP is extracted into the IP-field prior to movement of XP to SpecFin. (20) [TopP Top [FinP Fin [IP not [XP answeredi+X [VP Jens ti her]]]]] VP scrambles into IP fi (21) [TopP Top [FinP Fin [IP not [VP Jens ti her] [XP answeredi+X tVP]]]] XP moves to SpecFin and some constituent topicalizes fi (22) [TopP therefore Top [FinP[XP answeredi+X tVP] Fin [IP not [VP Jens ti her] tXP]]]

Thus the constituent made up by the subject and the object in (15) is the remnant VP. If bare is merged into the IP, it attracts XP (cf. Kayne 1998).3 In this case, either XP or IP moves further up to SpecFin. 3.1 Problems One of the attractive features of an analysis along these lines is that, prima facie, it seems to explain Holmberg’s Generalization (Holmberg 1986, 1999). (23) Holmberg’s Generalization (HG) Argument shift cannot cross any phonetically realized material from the VP, i.e. the verb, a verbal particle, a dative preposition, or other arguments of the verb, although it can cross traces of these, as well as sentential adverbs.

Employing a nonmovement analysis of argument shift, HG seems to follow trivially. Weak pronouns cannot cross any phonetically realized material from the VP, simply because it is the VP itself (or something bigger) that moves.

157

158

Øystein Nilsen

Unfortunately, the explanation offered of HG from this account breaks down when one looks at the system more closely. This is because of what would happen in non-V2 contexts. Vf would presumably move to X in these contexts as well, and then nothing prevents the remnant VP from scrambling into IP, yielding ungrammatical orders such as the following: (24) a. *… at Jens ’a ikke svarer that J. her not answers b. *… at Jens ikke fiskekaker liker. that J. not fishcakes likes

In order to prevent such orders, we would have to make extraction of VPs or objects from XP somehow contingent upon subsequent raising of XP to SpecFin. That is, we would have to reintroduce a notion of HG which is what we set out to derive. Similarly, one would need to account for the unavailability of subject-verb inversion in non-V2 contexts with weak pronominal subjects. Suppose that the finite subordinator at ‘that’ is generated in Fin and somehow blocks XP-toSpecFin movement as well as topicalization. As it stands, our account would lead to the incorrect expectation that the following order should be grammatical with the structure in (25b). (25) a. *… at ikke svarte ’n ’a that not answered he her b. [TopP Top [FinP that Fin [IP not [XP answeredi+X [VP he ti her]]]]]

Another problem is that it is not clear that the account explains root-V2. The extent to which it succeeds in doing so depends on whether TopP is the only projection dominating FinP. In e.g. Rizzi (1997), which is where the names TopP, FinP originate from, two other FPs are postulated, both of which dominate FinP, i.e. FocP and ForceP. Thus, either we would have to show that these projections do not exist, or that, for independent reasons, they cannot be filled in the relevant cases. These considerations are taken to show that a more radical departure from standard assumptions is needed.

4. More data: How initial is the initial position? Before we turn to our second approximation to the proper analysis of V2, some data will be reviewed that purport to show that the initial position, i.e. SpecCP in traditional analyses, need not be construed as a base-generated initial position.

V2 and Holmberg’s Generalization 159

The evidence we will review suggests that operations like wh-movement and topicalization are done in (at least) two separate movement steps, one targeting an IP-internal position, and a second one whose nature we try to elucidate in the remainder of the article.4 Consider the following contrasts:5 (26) a. Al very probably won. b. *How probably did Al win? c. How probable is it that Al won? (27) a. Al quite possibly won. b. *How possibly did Al win? c. How possible is it that Al won? (28) a. *How [probably/possibly/fortunately/necessarily/evidently/maybe/ frankly/usually] did Al win?

It seems that wh-movement of higher adverbs (cf. Cinque 1999) is systematically impossible. This contrasts with the behavior of lower adverbs, which do allow this kind of movement: (29) How [quickly/effortlessly/often/soon/frequently] did Al win?

Given that degree modification of the higher adverbs is possible, and that the adjectival counterparts of, e.g. (28) (i.e. ‘how fortunate/usual/… is it that…’) are grammatical, it seems that we are dealing with a syntactic phenomenon, rather than a (purely) semantic one. This suggests that wh-movement is a composite operation, consisting of two parts: One movement step targeting a relatively low, “IP-internal” position, call it P1 and another step targeting the first position of the clause, (P2). If P1 is generated lower than adverbs like probably, the ungrammaticality of examples like (28) would follow from the Extension Condition (Chomsky 1995). In other words, the only way for them to obey the Extension Condition would be to merge with P1, the wh-position directly, and then raise to their base-position. There are some indications that the same reasoning applies to topicalization. That is, there are topicalization-like processes that seem to target an IP-internal position in embedded clauses only. One case in point is so-called ‘stylistic fronting’ in Icelandic (cf. Holmberg 2000, among others). Another embedded, “IP-internal” topicalization process is illustrated by the following Norwegian examples. These may seem slightly contrived, but the contrast between (30b) and the other two is quite sharp. In (30a), we see an adverbial modifier (bak låven), modifying the most deeply embedded predicate, but appearing displaced from it, in the mittelfeld of the next clause up. In (30b)

160 Øystein Nilsen

we see that this kind of displacement is unavailable if the target is the root clause. In this case, the displaced adverbial will appear in the very first position, as illustrated in (30c). Det jeg sa var at jeg bak låven aldri har skjønt what I said was that I behind the-barn never have understood hvorfor han plantet tulipaner. why he planted tulips b. *Jeg har bak låven aldri skjønt hvorfor han plantet I have behind the-barn never understood why he planted tulipaner tulips c. Bak låven har jeg aldri skjønt hvorfor han plantet behind the-barn have I never understood why he planted tulipaner. tulips

(30) a.

It is tempting to say that bak låven is, in some sense in the ‘same position’ in (30a) and (30c). Yet another indication that something is wrong with the standard view of the left periphery is what appears to be subject-verb inversion in a quite low position. Consider the following Norwegian examples where the object has been topicalized around a V2-violating fpt. (31) a. b. c. d. e.

Meg vanligvis bare svarte ikke Jens. me usually just answered not J. Meg vanligvis bare svarte Jens ikke. me usually just answered J. not *Meg vanligvis bare Jens svarte ikke. me usually just J. answered not *Meg vanligvis Jens bare svarte ikke. me usually J. just answered not *Meg Jens vanligvis bare svarte ikke. me J. usually just answered not

Given the discussion in Section 2, it appears that Vf cannot be in C. One could try to say that, in (31a) the subject is inside VP. This would not work for (31b), since here, the subject precedes the negation as well. Thus, the standard ‘V-to-C’ analysis of subject-verb inversion cannot handle this phenomenon. The same point could be made with ‘distributive’ conjunctions like (32) (cf. Zamparelli 2000).


Meg vanligvis både slo de og sparket. me usually both beat they and kicked ‘They usually both beat and kicked me.’ b. *Meg vanligvis både de slo og sparket. me usually both they beat and kicked c. *Meg vanligvis de både slo og sparket. me usually they both beat and kicked d. *Meg de vanligvis både slo og sparket. me they usually both beat and kicked

(32) a.

For reasons of space, we will not enter into a discussion of distributive conjunction here. I refer the reader to Zamparelli’s work for an analysis that is congenial to the analysis to be presented here.

5. Second approximation Taking our conclusions so far quite literally, we arrive at the following picture of ‘clausal architecture’: (33) [adv* [WP [FocP [TopP [FinP VP]]]]

Here, the heads Foc, Top, Fin are the ones argued for by Rizzi (1997). W is the head introduced by Kayne (1998) to deal with scope phenomena involving the fpt only. We do not invoke Rizzi’s (1997) head ‘Force’, mainly because it is not necessary for our purposes. Adverbs attract TopP to their specifiers. We assume that, in case nothing is focus, or the entire sentence is, TopP is attracted to SpecFoc, and subsequently to SpecW. Thus in these cases, the sequence W, Foc will simply be omitted. Let us see how the system works by going through some derivations. 5.1 Root clauses We begin by deriving simple root clauses like the following: (34) Jens svarte meg vanligvis. J. answered me usually (35) [VP John answered me] merge Fin and move V fi [FinP answered [VP John me]] merge Top and move John fi

161

162 Øystein Nilsen

[TopP John [FinP answered [VP me]]] merge usually and move TopP fi [AdvP [TopP John [FinP answered [VP me]]] usually]

In this derivation, the entire sequence John answered me climbs around the adverb. The only movement step which is always necessary is movement of Vf to Fin. The object could also have moved to SpecTop. In that case, the subject would either remain in-situ, or move to SpecFoc. Consider now a sentence with an indefinite object noun phrase. In these, the indefinite will move to SpecFoc, prior to movement of TopP to SpecW. (36) Jens leste vanligvis en bok. J. read usually a book (37) [VP John read a book] merge Fin and move V fi [FinP read [VP John a book]] merge Top and move John fi [TopP John [FinP read [VP a book]]] merge Foc and move a book fi [FocP a book [TopP John [FinP read [VP]]]] merge W and move TopP fi [WP [TopP John [FinP read]] [FocP a book]] merge usually and move TopP fi [AdvP [TopP John [FinP read]] usually [WP[FocP a book]]]

It is easy to see that addition of more adverbs would only lead to iteration of this last step, so V2 is derived for these cases. If the subject bears focus, we now expect the ungrammatical (38a). (38) a. *Derfor gjenkjente meg vanligvis en student. therefore recognized me usually a student b. Derfor gjenkjente vanligvis en student meg. therefore recognized usually a student me

The derivation of (38a) would go as follows: (39) [VP a student recognized me] merge Fin and move V fi [FinP recognized [VP a student me]] merge Top and therefore fi


[TopP therefore [FinP recognized [VP a student me]]] merge Foc and move a student fi [FocP a student [TopP therefore [FinP recognized [VP me]]]] merge W and move TopP fi [WP [TopP therefore [FinP recognized [VP me]]] [FocP a student]] merge usually and move TopP fi [AdvP [TopP therefore [FinP recognized [VP me]]] usually [WP[FocP a student]]]

Recall from the discussion in Section 2 that such orders are actually attested in Swedish. Thus we suggest that this is how they are derived. In order to rule them out in Norwegian we will resort to the strategy suggested in Section 3, i.e. that the subject pied-pipes the VP to SpecFoc. This makes the subject and the object behave as a constituent and prevents the subject from inverting with the object: (40) [VP a student recognized me] merge Fin and move V fi [FinP recognized [VP a student me]] merge Top and therefore fi [TopP therefore [FinP recognized [VP a student me]]] merge Foc and move VP fi [FocP [VP a student me] [TopP therefore [FinP recognized]]] merge W and move TopP fi [WP [TopP therefore [FinP recognized]] [FocP [VP a student me]]] merge usually and move TopP fi [AdvP [TopP therefore [FinP recognized]] usually [WP [FocP [VP a student me]]]]

In Danish subjects must precede all adverbs. Objects follow them unless they are weak pronouns, in which case they must also precede them. We would like to suggest that in Danish, SpecFoc has been ‘grammaticalized’ as a case position for objects. Weak pronouns, being marked for case, do not have to move there. This means that there are but two positions the subject can choose from: it can move to SpecTop, or it can remain in situ. In the first case, it will end up in the V2-initial position, as in the derivation (35) above. In the latter, it will remain immediately right-adjacent to the verb, much as the weak object pronoun in (35). Thus we have three levels of freedom with regard to subjects and SpecFoc. In Swedish the subject can move there alone, in Norwegian it must pied-pipe the VP, and in Danish it cannot move there at all.

164 Øystein Nilsen

5.1.1 bare Let us now turn to the problematic cases with V2-violating focus particles. In general, we expect these to appear in the Adv* area of (33). We suggest that they differ from other adverbs in that, instead of attracting TopP, they attract their focus associate, and come with a W-projection into which bare itself and TopP moves, i.e. we essentially adopt the treatment in Kayne (1998). Suppose that we add bare to the derivation in (40). Then, if en student is the associate, we would have the following derivation, corresponding to the grammatical sentence in (42). (41) [VP one student recognized me] merge Fin and move V fi [FinP recognized [VP one student me]] merge Top and therefore fi [TopP therefore [FinP recognized [VP one student me]]] merge Foc and move VP fi [FocP [VP one student me] [TopP therefore [FinP recognized]]] merge W and move TopP fi [WP [TopP therefore [FinP recognized]] [FocP [VP one student me]]] merge just and move FocP fi [bareP [FocP [VP one student me]] just [WP [TopP therefore [FinP recognized]]]] merge W and move just and TopP fi [WP [TopP therefore [FinP recognized]] just [bareP [FocP [VP one student me]] [WP]]] (42) Derfor gjenkjente bare én student meg. therefore recognized just one student me

Adding usually, we get the following continuation, corresponding to the grammatical sentence (44). (43) merge usually and move TopP fi [AdvP [TopP therefore [FinP recognized]] usually [WP only [bareP [FocP [VP one student me]]]]] (44) Derfor gjenkjente vanligvis bare én student meg. therefore recognized usually only one student me

Consider now the derivation of (11a) above, repeated here, in which the verb is the associate. In this case, FinP moves to Specbare.


(11) a.

Jens vanligvis bare svarer ikke. J. usually just answers not

(45) [VP John answers] merge Fin and move V fi [FinP answers [VP John]] merge Top and move John fi [TopP John [FinP answers [VP]]] merge not and move TopP fi [AdvP [TopP John [FinP answers]] not] merge just and move FinP fi [bareP [FinP answers] just [AdvP [TopP John] not]] merge W and move just and TopP fi [WP [TopP John] just [bareP [FinP answers] [AdvP not]]] merge usually and move TopP fi [AdvP [TopP John] usually [WP just [bareP [FinP answers] [AdvP not]]]]

The essential difference between these examples and ‘ordinary’ V2 sentences is thus that the verb has been pulled out of TopP. Therefore, when TopP moves around higher adverbs, the verb is left behind. Of course, the last step of (45) can be skipped, yielding the equally grammatical sentence Jens bare svarer ikke (‘J. just answers not’). Since weak pronouns do not move, they will be left inside the VP. This is what causes them to stay adjacent to the verb. The partial derivation in (46) for (14a), illustrates this. (14) a.

Derfor vanligvis bare svarte ’n ’a ikke. therefore usually just answered he her not

(46) [AdvP [TopP therefore [FinP answered [VP he her]]] not] merge just and move FinP fi [bareP [FinP answered [VP he her]] just [AdvP [TopP therefore] not]] merge W and move just and TopP fi [WP [TopP therefore] just [bareP [FinP answered [VP he her]] [AdvP not]]] merge usually and move TopP fi [AdvP [TopP therefore] usually [WP just [bareP [FinP answered [VP he her]] [AdvP not]]]

Nothing special needs to be said about the low subject-verb inversions noted in Section 4. There are two cases to consider. If the subject is a weak pronoun, it

166 Øystein Nilsen

remains in the VP, and moves with the verb, as in the previous derivation. If it is a full noun phrase, as in (31a) it can move to SpecFoc, giving rise to the derivation in (47). (31) a.

Meg vanligvis bare svarte ikke Jens. me usually just answered not J.

(47) [VP John answered me] merge Fin and move V fi [FinP answered [VP John me]] merge Top and move me fi [TopP me [FinP answered [VP John]]] merge Foc and move VP fi [FocP [VP John] [TopP me [FinP answered]]] merge W and move TopP fi [WP [TopP me [FinP answered]] [FocP John]] merge not and move TopP fi [AdvP [TopP me [FinP answered]] not [WP [FocP John]]] merge just and move FinP fi [bareP [FinP answered] just [AdvP [TopP me] not [WP [FocP John]]]] merge W and move just and TopP fi [WP [TopP me] just [bareP [FinP answered] [AdvP not [WP [FocP John]]]]] merge usually and move TopP fi [AdvP me usually [WP just [bareP [FinP answered] [AdvP not [WP [FocP John]]]]]]

As the reader may have noticed, nothing has been said, so far, about examples like (14b), where Vf does appear in the second position even though there is an fpt modifying it. (14) b. Derfor svarte ’n ’a vanligvis bare ikke. therefore answered he her usually just not

In order to handle these cases, we propose that FinP, when it moves to Specbare, can pied-pipe TopP. This has the effect of making bare behave as other adverbs with respect to V2. This is illustrated in the partial derivation below. (48) [AdvP [TopP therefore [FinP answered [VP he her]]] not] merge just and move TopP fi [bareP [TopP therefore [FinP answered [VP he her]]] just [AdvP not]] merge W and move just and TopP fi


[WP [TopP therefore [FinP answered [VP he her]]] just [bareP [AdvP not]]] merge usually and move TopP fi [AdvP [TopP therefore [FinP answered [VP he her]]] usually [WP just [bareP [AdvP not]]]]

5.1.2 Auxiliaries and the derivation of HG There is a large literature on the syntactic treatment of auxiliaries (cf. Cinque 2000, Koopman and Szabolcsi 2000, Julien 2000 for recent discussion). One controversy is whether or not to treat e.g. participial constructions as biclausal (Kayne 1993). We remain agnostic about this question here, simply analyzing auxiliaries by stacking them onto the VP. We do treat them as ‘raising verbs’ in the sense that they attract the subject from the inner VP. This is in order to exclude sentences like (49) where the subject follows an axiliary. (49) *Derfor kan ikke ha ’n sett Jens. therefore can not have he seen J.

We assume that participial VPs move to SpecFoc unless the participial verb is a topic. Apart from this, auxiliaries do not pose any special problems for our account. We need to discuss them, however, in order to show that the account really derives HG, which we repeat here for convenience. (23) Holmberg’s Generalization (HG) Argument shift cannot cross any phonetically realized material from the VP, i.e. the verb, a verbal particle, a dative preposition, or other arguments of the verb, although it can cross traces of these, as well as sentential adverbs.

We have said that weak pronouns never move. What is crucial is that they never shift. Otherwise, they are free to move, as it were. In particular, they can move to SpecTop, which ultimately places them in the initial position of the clause. They cannot move to SpecFoc for the simple reason that they are weak, and, in Danish, because they bear morphological case. Consider now the contrast between (50) and (51), a typical example of HG. (50) a.

Jeg så ’n ikke. I saw him not b. *I så ikke ’n. I saw not him

168 Øystein Nilsen

(51) a. *Jeg har ’n ikke sett. I have him not seen b. Jeg har ikke sett ’n. I have not seen him

We have already seen how it comes about that the weak pronoun must precede the negation in (50a). We need to demonstrate that it cannot do so if there is an auxiliary. The first steps in the derivation of (51b) are as follows: (52) [VP I have [PtcP seen him]] merge Fin and move V fi [FinP have [VP I [PtcP seen him]]]

At this point, one of two things can happen: either the subject moves to SpecTop, or the object does. As we shall see below, the participial verb can also do this. If the subject moves, we get the following continuation, deriving (51b). (53) merge Top and move John fi [TopP I [FinP have [VP [PtcP seen him]]]] merge Foc and move PtcP fi [FocP [PtcP seen him] [TopP I [FinP have [VP]]]] merge W and move TopP fi [WP [TopP I [FinP have]] [FocP [PtcP seen him]]] merge not and move TopP fi [AdvP [TopP I [FinP have]] not [WP [FocP [PtcP seen him]]]]

In the alternative case, the object moves to SpecTop, yielding the following continuation of (52), corresponding to the grammatical sentence in (55).6 (54) merge Top and move him fi [TopP him [FinP have [VP I [PtcP seen]]]] merge Foc and move PtcP fi [FocP [PtcP seen] [TopP him [FinP have [VP I]]]] merge W and move TopP fi [WP[TopP him [FinPhave [VP I]]] [FocP [PtcP seen]]] merge not and move TopP fi [AdvP [TopP him [FinP have [VP I]]] not [WP [FocP [PtcP seen]]]] (55) Han har jeg ikke sett. him have I not seen


If the participial verb is a topic, it moves to SpecTop on its own. This implies that heads can move to specifier positions, contra standard assumptions. It is possible that this problem would disappear if we adopt a biclausal structure for participial constructions, (Kayne 1993). Another alternative would be to say that the participle adjoins to Top, and that this serves the same purpose as moving it to SpecTop. We choose to live with this problem for the purposes of this article. Consider the contrast below. (56) a.

Sett har jeg ’n ikke. seen have I him not b. *Sett har jeg ikke ’n. seen have I not him

When the participle is fronted, the pronoun must precede the negation. This fact is problematic for the traditional account which treats weak pronoun distribution as pronoun shift and V2 as V-to-C with subsequent topicalization. In order to rule out (51a), such an account must block pronoun shift over phonetically realized material from the VP. Suppose that the proper formulation of HG is in entirely phonological terms. The fact that pronoun shift is allowed around traces might follow rather elegantly from such a formulation, since traces do not have phonological content. But the fact that they can cross adverbs seems rather mysterious; the more so because they must cross adverbs if they can. It seems unlikely that there should be any phonological property distinguishing adverbs from all other material. To be on the safe side, we will show that there isn’t. The Norwegian word fortsatt ambiguously represents the adverb ‘still’ and the participial verb ‘continued’. In the adverb interpretation it forces pronoun shift and in the participial interpretation it blocks it. (57) a.

Han har fortsatt det. he has continued it b. Han har det fortsatt. he has it still

Rejecting a phonological formulation, there are two scenarios to consider according to whether or not bare heads are allowed to topicalize. If they are, one would need to assume that there is some domain extension mechanism (cf. Chomsky’s 1995 notion of ‘equidistance’) which would be triggered by such topicalization.7 If head movement to SpecCP is not endorsed, one would need to assume that pronoun shift cannot cross the participle for syntactic reasons, but that this can

170 Øystein Nilsen

be violated if the (remnant) VP subsequently topicalizes. On our account, nothing special needs to be said. We assume that heads can move to SpecTop with the caveat noted above. The derivation of (56a) runs as follows. (58) [FinP have [VP I [PtcP seen him]]] merge Top and move Ptc fi [TopP seen [FinP have [VP I [PtcP him]]]] merge not and move TopP fi [AdvP [TopP seen [FinP have [VP I [PtcP him]]]] not]

This concludes our discussion of root clauses. In the next sub-section, we will discuss how (finite) embedded clauses might be derived. 5.2 Finite embedded clauses This subsection is quite tentative. The proper treatment of embedded finite clauses requires a theory of bridge verbs and why they allow embedded V2 etc. This is a complicated issue which will not be dealt with here. We will limit ourselves to adding some observations and a simple account of non-V2 embedded clauses In embedded V2 clauses (EV2), the complementizer is obligatory. That is, it is obligatory as long as the EV2 is not fronted. In that case, the complementizer is impossible. (59) a.

Hun sa *(at) hun har ikke sett meg. she said *(that she has not seen me. b. *(At) hun har ikke sett meg, sa hun. (that she has not seen me said she

In non-V2 embedded clauses (NV2), the pattern is almost the opposite. (60) a.

Hun sa (at) hun ikke har sett meg. she said (that she not has seen me b. *(At) hun ikke har sett meg, sa hun. (that she not has seen me said she

If the matrix predicate is not a ‘bridge’ verb, EV2 becomes impossible, and the complementizer is obligatory in both places. (61) a.

Det er trist *(at) hun ikke har sett meg. it is sad *(that she not has seen me


b. *Det er trist (at) hun har ikke sett meg. it is sad (that she has not seen me

Suppose that the complementizer in NV2 clauses adds an extra Foc, Top, Fin layer on top of the original one. It attracts the (nominative) subject, and moves to Top, preventing anything else from moving there. Then either FinP or the lower WP moves to SpecFoc. If the subject is in the lower SpecTop, we get derivations like the following:8 (62) [WP1 [TopP John likes] [FocP it]] merge that and move John fi [FinP John that [WP1 [TopP likes] [FocP it]]] merge Top and move that fi [TopP that [FinP John [WP1 [TopP likes] [FocP it]]]] merge Foc and move FinP fi [FocP [FinP John [WP[TopP likes] [FocP it]]] [TopP that]] merge W2 and move TopP fi [WP2 [TopP that] [FocP [FinP John [WP1 [TopP likes] [FocP it]]]]] merge not and move TopP fi [AdvP [TopP that] not [WP2 [FocP [FinP John [WP1 [TopP likes] [FocP it]]]]]] (63) … at ikke Jens liker det that not J. likes it

If we had moved WP1 to SpecFoc instead of FinP, John would have ended up adjacent to the complementizer, i.e. the following order. (64) … at Jens ikke liker det that J. not likes it (65) [TopP that [FinP John [WP1 [TopP likes] [FocP it]]]] merge Foc and move WP1 fi [FocP [WP1 likes it] [TopP that [FinP John]]] merge W2 and move TopP fi [WP2 [TopP that [FinP John]] [FocP [WP1 likes it]]] merge not and move TopP fi [AdvP [WP2 [TopP that [FinP John]] not [FocP [WP1 likes it]]]]

As for the EV2 orders, one might suppose that an account closer to Kayne’s (1998) in which the complementizer is merged to the matrix, rather than the embedded predicate would work. Then it would fail to intervene with movement

171

172 Øystein Nilsen

in the embedded clause. This would indicate that EV2 clauses are quite different from NV2 clauses. One rather subtle phenomenon that would fall out quite naturally with such a dichotomy is that EV2 clauses appear to be assertive, while NV2 clauses are not. For example, I cannot felicitously utter (66) if I really think Jens is the murderer. No such presupposition seems to be associated with (67). (66) Jeg har hørt at Jens har ikke drept ham. I have heard that J. has not killed him (67) Jeg har hørt at Jens ikke har drept ham. I have heard that J. not has killed him

Less subtle is the fact that EV2 clauses are strong islands for extraction, while NV2 clauses are not. (68) a.

Hvilken mann har du hørt at Jens muligens har drept? which man have you heard that J. possibly has killed b. *Hvilken mann har du hørt at Jens har muligens drept? which man have you heard that J. has possibly killed

These are all facts that the theory should ultimately be able to handle. We will leave them for future investigations.

6. Summary and conclusions We have seen that the traditional analysis of Scandinavian V2 runs into problems on a number of counts. Taken separately these problems need not be devastating. However, when taken together, they suggest quite strongly that the account is wrong in important respects. The proposed account attempts to deal with the problems by replacing two widely acclaimed assumptions about how V2 is derived with the following two: CP type projections are better treated by merging them below sentential adverbs, and the verb moves to second position as a phrase, rather than as a head. We have tried to maintain an analysis according to which the same movements apply in all sentences in the different languages; the potential difference being the ‘size’ of the moved constituents.


Notes * I am very grateful to Guglielmo Cinque, Anders Holmberg, Hilda Koopman, Eric Reuland, and Peter Svenonius for helpful comments and discussion about this work. All errors are my own. 1. Unless otherwise indicated, all examples are from Norwegian. 2. According to Christer Platzack (p.c.), these sentences are grammatical in Swedish as well. 3. That is, of course, bare which focuses on the verb. If it focuses on some other constituent, it would attract that. 4. Reinhart (1981) argues for two separate COMP positions on the basis of Hebrew data. 5. Since the class of adverbs involved is identical in English and Norwegian, I use English examples here. As far as I have been able to determine, they are the same in Dutch, German, Japanese, Russian and Serbo-Croatian as well. 6. As pointed out in Cardinaletti and Starke (1995), weak pronouns in initial position cannot be spelled out in their most reduced form. Thus (55) contrasts with (i). (i) *’n har jeg ikke sett. We take this to be a morphophonological effect. 7. Note that equidistance, as formulated by Chomsky (1995), would not do. 8. I assume that it is impossible to move from SpecFoc to a case position. Thus, if the subject is not in SpecTop, the derivation crashes.

173

The distribution of declarative verb second in Germanic Olaf Koeneman University of Groningen

1.

Introduction*

Recent developments in linguistic theory fail to shed a new light on verb movement parameters. Although the language variation in this area has been well documented (cf. Vikner 1995a and Koeneman 2000 for an overview), theoretical approaches tend to be descriptive rather than explanatory in nature. The standard way of capturing verb movement nowadays is to assume that the verb moves from its base position and adjoins to a functional head that c-commands the projection of the verb, as in (1). F

(1) F V

VP F

t

Given this format, the question becomes what causes cross-linguistic differences in verb placement. In checking theory (Chomsky 1995), it is assumed that features residing in the functional head F can be either strong or weak (cf. Zwart 1993, 1997 for such an approach applied to verb movement in Germanic). In the former case, the verb is forced to move overtly. In the latter case, this operation can be postponed until after Spell Out. Although this captures parametrization in verb placement, the account is descriptive rather than explanatory since definitions of ‘strong’ and ‘weak’ are crucially lacking. This entails that questions like those in (2), which I put central in this article, only receive ad hoc answers.

176 Olaf Koeneman

(2) a.

Why do some verb second (henceforth: V2) languages have V2 in main clauses only and some also in embedded contexts? b. Why is English the only Germanic language without declarative V2?

It seems therefore fair to conclude that the formulation of alternative proposals on verb movement parametrization would be no luxury. In Koeneman (2000) it is argued that as an explanation for verb movement parametrization (1) is the wrong format (cf. also Ackema et al. 1993). Instead of assuming that the verb moves in order to “check” some feature F, it is proposed that the verb moves in order to project it, thereby satisfying some constraint. The lexicon makes available complex heads, consisting of the verbal stem and functional features, but leaves unspecified which of the two should become the syntactic head after insertion: output conditions determine that. The representation of a verb taking two functional features, for instance, would look as in (3). (3) V

F1

F2

After insertion into syntax it is V that must project first in order to discharge its θ-grid, given standard assumptions: VP

(4) V

DP

Suppose now that some output condition demands that some feature, F, heads a separate projection. In that case, the verb moves and merges with its own projection. No ambiguity results as a consequence of this self-attachment (cf. Chomsky 1995, Chapter 4) if it is the functional feature rather than V that projects, as illustrated in (5). F

(5) F V

VP F

t

Although this operation leads to a representation that is similar to the one obtained in the standard approach (compare (5) to (1)), the crucial difference

The distribution of declarative verb second in Germanic 177

is that there is no longer a prefabricated head position that the verb moves to at some point in the derivation. This makes it possible to formulate triggers that are ‘positional’ in nature. The verb moves because one of its features must occupy a particular structural position with respect to other elements in the clause. There is no abstract head doing that already. This idea will be taken up in the next section, where concrete triggers for verb movement are developed.

2. Triggers There is robust distributional evidence for two verb movement operations, generally coined V-to-I and V-to-C. This entails that there must minimally be two features other than V that the verb can project after movement. Koeneman (2000) claims that the relevant features are the two stated in (6). (6) a.

V-to-I movement is an operation that the verb undertakes in order to project Agreement (= Agr) features. b. V-to-C movement is an operation that the verb undertakes in order to project Tense (= T) features.

As they stand, the statements in (6) are merely descriptive. We still have to motivate output conditions, i.e. triggers, that make projection of these features necessary. Given space limitations as well as the fact that the claim in (6a) has been explicitly worked out elsewhere (Koeneman 1997), I will adopt it here as an assumption. It is well-known that in some languages, the finite verb precedes VP-adverbs, negation and floating quantifiers, whereas this is not the case in others (cf. Emonds 1976; Pollock 1989). The order Vfinite ADV/NEG/FQ has been taken to reflect verb movement, as indicated by the trace in (7a): (7) a. Subject Vfinite ADV/NEG/FQ tV b. Subject ADV/NEG/FQ Vfinite

Whether a language has the operation in (7a) or not, leading to the contrast between (7a) and (7b), appears to be far from arbitrary.1 There is an impressive line of research indicating a correlation between overt verb movement and inflection. Languages that have a rich subject agreement paradigm have the order in (7a), whereas languages without rich agreement tend to leave the verb in situ (Kosmeijer 1986; Pollock 1989; Platzack and Holmberg 1989, 1995; Roberts 1993; Rohrbacher 1994, among others. See Alexiadou and Fanselow,

178 Olaf Koeneman

this volume, and Bobaljik 2000 for recent critiques). The existence of a correlation between rich agreement and verb movement is corroborated by synchronic as well as diachronic evidence, which can be found in the references mentioned (cf. also Section 2.2 for some data). The statement in (6a), then, is intended to capture this generalization: languages with rich agreement must project an independent Agr-phrase.2 The empirically most precise definitions of the rich/ poor dichotomy are those in Rohrbacher (1994) and Koeneman (2000). Under both proposals, the languages under discussion are distributed as indicated in the table below: (8) Agreement type

Rich agreement

Poor agreement

Languages

Icelandic, Yiddish, German

English, Mainland Scandinavian, Dutch

The data showing the correlation with verb movement will be presented as we proceed. Let us now turn to the claim in (6b) and formulate an output condition that will, among other things, capture V-to-C movement in Germanic. I propose that, at least in languages that morphologically encode T, the condition in (9) holds. (9) The Tense condition The Tense features of the predicate (e.g. [±past]) must be visible on a head that COMMANDs both the subject and the predicate.

The idea that Tense is realized VP-externally is a traditional one in generative syntax (cf. Chomsky 1957, Stowell 1981). The more precise intuition behind (9) is that T is a characteristic of an event or proposition, interpreted distinctly from the verb itself. It is not a feature of the denotation of V and as such does not form a semantic unit with it. This assumption is, I believe, relatively uncontroversial. The specific claim made here is that this property is syntactically encoded (cf. Evers 1982). This leaves two options. Either T is present on a head that takes the predicate as its complement and the subject as its specifier. In that case, T can distinctly command the subject and predicate from one position by m-command (cf. (10a)). Alternatively, T can be present on a head that takes as its complement a category that dominates both the subject and the predicate. In (10b), T c-commands both of these elements from one position.3


FP

(10) a. SU

F′

[+T] F

PRED F′

b. [+T] F

XP SU

PRED

Let us assume that tense features are introduced by that element of which they are inherently part, namely the tense affix located on the verbal head. If T must COMMAND the subject and the predicate, this entails that, for interpretation to proceed correctly, the tense features must minimally be made visible beyond the position in which they are inserted. One mechanism ensuring this is movement. V-to-C, then, is analyzed as a way to satisfy (9). Under the assumption that the Tense condition becomes relevant after the subject has been merged with the predicate, it follows naturally that the verb consequently moves up to a position higher than the subject and the predicate in order to project T and satisfy (9) under c-command. T′

(11) T T V

VP Agr

SU

VP

T

Since T puts no restrictions on what can appear as its specifier, any XP can move there. The result is a V2 effect. (I will postpone discussion about what might trigger XP-fronting until Section 6.) Under the assumption that (9) applies in all Germanic languages, we are able to capture main clause V2 effects in Icelandic, Yiddish, Mainland Scandinavian and Dutch: Bókina ·keyptiÒ Jón ·*keyptiÒ ekki books ·bought John bought not b. Dos bukh ·shikÒ ikh ·*shikÒ avek the book ·send I ·*send away

(12) a.

(Icelandic) (Yiddish)

180 Olaf Koeneman

Boken ·köpteÒ Ulf inte ·*köpteÒ books ·bought Ulf not ·*bought d. Dat boek leest Jan op vakantie that book reads Jan on holiday

c.

(Swedish) (Dutch)

If tense features of the predicate must take scope over the subject and the predicate, we expect the verb to move in embedded clauses as well. In the same vein, then, we are able to capture the fact that V2 effects show up in embedded domains in Icelandic and Yiddish. (13) a.

að í herberginu hefur kyrin staðið that in the.room has the.cow stood b. az morgn vet dos yingl zen a kats that tomorrow will the boy see a cat

(Icelandic) (Yiddish)

What does not follow are those cases for which the proposal now overgenerates. For the purposes of this article, two cases must be dealt with, which are basically the questions in (2). First of all, although embedded V2 effects are rather unrestrictedly attested in Yiddish and Icelandic, this is not the case for Mainland Scandinavian, Dutch and German. In Swedish, V2 under a complementizer only occurs under so called bridge verbs and these clauses are usually analyzed as ‘embedded main clauses’ in some sense. In Standard Dutch, V2 under a complementizer is ruled out across the board. Ungrammatical examples are given in (14): (14) a. *Jan beklagar att den här boken hade jag läst Jan regrets that this here book had I read b. *Jan betreurde het dat dit boek had ik gelezen Jan regrets it that this book had I read

(Swedish) (Dutch)

Now, if the Tense condition is universal, the question becomes how the tense features of the predicate take scope over the subject and the predicate if no verb movement takes place in embedded domains. Likewise, no V2 is observed in English declarative clauses, again raising the question of how the Tense condition is satisfied. (15) *Books will John never reads

The proposal is as follows. Language-specific properties that can be independently motivated block movement if they make it possible for the syntax to satisfy the relevant constraint in a cheaper way. The specific claims that I will work out in subsequent sections are those in (16):

The distribution of declarative verb second in Germanic

(16) A V-to-C type verb movement is blocked… a. in embedded clauses in Dutch and Mainland Scandinavian since the Tense condition can be met through the complementizer (Section 3). b. in German embedded clauses since its OV character allows generation of inflection in a separate VP-external position to the right of VP (Section 4). c. in English due to the presence of an empty Tense-marker selecting VP (Section 5).

3. The root/non-root asymmetry Since Den Besten’s (1983) proposal, it is a commonly held opinion that the presence of the complementizer blocks verb movement in embedded clauses, leading to root/non-root asymmetries in Dutch and German. Of course, this idea faces the opposite question (namely how symmetric V2 languages can exist) but I nevertheless believe that Den Besten’s insight is basically correct. From the present perspective, it means that the complementizer is apparently involved in the satisfaction of the Tense condition in some but not all languages. What I propose is that the distinction between symmetric and asymmetric V2 is the consequence of a locality condition on head-head dependencies. Let me explain. If triggers for verb movement are positional in nature, the operation no longer takes place in order to establish a dependency relation between two heads, one functional and one lexical. This makes it less obvious that verb movement is necessary at all to establish such dependency relations. In fact, checking theory as defined in Chomsky (2000) no longer assumes that a functional head can enter into a checking relation with the verb only if it has been attracted to it. Hence, two heads can “see” each other without a movement operation taking place. Against this background, I will assume that a head can indeed enter into a dependency relation with another head without moving to it but that some locality condition, formulated as in (17), is at play:4 (17) Accessibility Accessible to a head α are (i) β, β being the closest (segment of a) head in α’s c-command domain (ii) every complete head γ, where γ is a sister of (a segment of) β.

181

182 Olaf Koeneman

The intuition behind (17) is as follows. A head α can only enter into a dependency relation with the first complete head in its c-command domain (cf. (i)), not with segments of a head. If constructing the first complete X0-category entails that α automatically comes across other complete heads, these heads will be equally accessible to α (cf. (ii)): α cannot help but see them. Let me illustrate this with a few examples: aP

(18) a. a

bP b

gP g

...

aP

b. a

bP b

... g

b

aP

c. a

bP b b

g g

...

d

The structure in (18a) is rather straightforward. The first head in α’s c-command domain is β. Since α cannot look any further, γ is not accessible to α, only β is. In (18b), the head closest to α is again β, since the closest head node in α’s c-command domain is a segment of β. Since α can only enter into a dependency relation with complete heads, it must construct the complete category β. Therefore, the daughters of the top segment β become relevant pieces of structure. In this process, α will hence automatically encounter γ, which is a complete category and a sister of β, hence by (17ii) accessible to α, like β. In


(18c), the first complete head that α constructs is β. In this process, no other complete heads are encountered, only one segment of γ, which is a sister to the lowest segment of β. Therefore, γ and δ are inaccessible to α, only β is. Let us return to the Tense condition. Suppose that we have the following structure, consisting of a subject-predicate combination selected by a complementizer, as in (19a): CP

(19) a. C

VP Su

VP V

Ob

V

Agr T [±past]

V

CP

b. [±past] C

VP Su

VP V

V V

Ob Agr

T [±past]

The structure in (19a) depicts an embedded clause in for instance Swedish, where no verb movement has occurred. The question is how the requirement that T take scope over the subject and the predicate is met. Let us assume that a complementizer selecting a finite clause is marked for T. However, it will at most express [+tense]: C is for instance not inherently marked for [±past] in the languages under discussion and therefore cannot anchor the event denoted by the embedded proposition in time. It is these features that the Tense condition refers to. The proposal is therefore that the two tense-marked elements, the complementizer and the tense affix, enter into a head-head dependency relation.

184 Olaf Koeneman

The definition of accessibility (cf. (17ii)) allows it (T being a sister of a segment of V). The consequence is that C receives a specification for [±past], as indicated in (19b). As the tense features of the predicate now take scope over the subject and the predicate, as required, verb movement is blocked. If the complementizer can assist in the satisfaction of the Tense condition, thereby making verb movement redundant, why does embedded V2 still take place in Icelandic and Yiddish? My answer is that these languages have, besides V-to-C, another verb movement, namely V-to-I. Given that these languages have rich agreement inflection, they must project an independent AgrP. It is this independent property that causes Icelandic and Yiddish to display symmetric V2 effects. Distributional evidence for V-to-I is hard to obtain. It is usually assumed that V-to-I movement is visible because the finite verb ends up in a position to the left of adverbs that supposedly mark the left edge of VP. Hence, French can be assumed to show this movement in contrast to English (Emonds 1976; Pollock 1989): Jean ·*souventÒ embrace ·souventÒ Marie Jean ·*often kisses ·often Marie b. John ·oftenÒ kisses ·*oftenÒ Mary

(20) a.

Now if the verb undergoes V2, as in Yiddish and Icelandic, it will always end up to the left of adverbs and there is no direct way of establishing whether the verb has stopped in intermediate positions. Nevertheless, there are several clues suggesting that V-to-I does indeed take place in these languages. First of all, Icelandic and Yiddish have a rich agreement paradigm, showing five different agreement distinctions in the present tense. The correlation between V-to-I and rich agreement, which has been extensively argued to play a role in verb positioning in Germanic, can be demonstrated clearly with languages that lack V2. Romance languages, for instance, are generally richly inflected and evidence for verb movement is robust (cf. Emonds 1976 and Pollock 1989 for French and Belletti 1992 for Italian). English agreement is clearly poor and it is therefore expected that the verb remains to the right of adverbs (cf. (20b)). The same is true for Mainland Scandinavian once V2 is controlled for (i.e. in embedded clauses). Second, Vikner (1995a) argues that expletives can only co-occur with transitive predicates in languages that have both V-to-I and V-to-C. Both Mainland Scandinavian, which only has V-to-C, and French, which only has V-to-I, lack these constructions in contrast to Icelandic and Yiddish:


(21) a.

Það hafa margir jólasveinar borðað búðing there have many Santa Clauses eaten pudding b. Es hot imitser gegesn an epl there has someone eaten an apple c. *Det har någon ätit ett äpple there has someone eaten an apple d. *Il a un homme mangé une pomme there has a man eaten an apple

(Icelandic) (Yiddish) (Swedish) (French)

Although these data do not directly show that Icelandic and Yiddish have V-to-I, the idea that transitive expletive constructions require a more extensive functional domain seems very plausible (cf. Bobaljik and Thráinsson 1998 for such a claim). Especially the present theory of verb movement allows for an elegant description of the data: only if the verb moves twice will there be two specifiers available for hosting both the subject and the expletive. I refer to Koeneman and Neeleman (2001) for a worked-out account along these lines. Thirdly, Icelandic and Yiddish disallow expletives in third position, in contrast to Mainland Scandinavian: (22) a.

Í gær hefur (*það) komið strákur yesterday has (*there come a.boy b. Nekhtn iz (*es) gekumen a yingl yesterday is (*there come a boy c. Igår er der kommet en dreng yesterday is there come a boy d. Idag har det kommit många lingvister hit today have there come many linguists here

(Icelandic) (Yiddish) (Danish) (Swedish)

The contrast can again be related to the presence versus absence of an AgrP. Suppose that in Icelandic and Yiddish the verb moves in order to project Agr. Then one could argue that rich Agr licenses a pro-expletive in its specifier, which is preferred to generation of an overt expletive (cf. Cardinaletti 1991). In Mainland Scandinavian, there is no rich Agr licensing a pro-expletive in its specifier and we always see an overt expletive surfacing in either first or third position.5 Given these observations, let us assume that Icelandic and Yiddish have two verb movements, despite the lack of distributional evidence for one of them.6 One movement takes place to project Agr, the other to project T. The question being addressed here is why both have to take place in embedded domains, unlike in Mainland Scandinavian. The answer can now be related to this independent factor distinguishing Icelandic and Yiddish from Mainland Scandinavian.

185

186 Olaf Koeneman

Suppose that the verb moves once, projects Agr and that the subject subsequently fills SpecAgrP. After that, the result is embedded under the complementizer, as in (23): CP

(23) C

AgrP DP

Agr′ Agr

V V

VP Agr

t

OB

T

It will be clear that the tense affix does not have COMMAND over the subject and the predicate, since it is dominated by the category Agr. At the same time the structure is such that a dependency relation between C and T is blocked. As in Icelandic and Yiddish Agr projects after the first verb movement, Agr is the first head that C encounters in its c-command domain. Once the complete category Agr has been constructed, C has in addition only come across one segment of V. Crucially then, T is not accessible to C, given the locality on head dependencies. Consequently, [±past] will not become visible on C. Since the Tense condition fails to be met, a second verb movement is required. This movement will project TP, just like in main clauses. For the account to work, the notion accessibility must for the moment be stipulated. In defence of it, two remarks are in order. First, the advantage of the proposal, compared with the standard view which stipulates that complementizers block verb movement in only a subset of the Germanic languages (Den Besten 1983), is that V2 now receives a uniform trigger: projection of T in the absence of a complementizer. It is an independent property of Icelandic and Yiddish that leads to embedded V2 effects. Second, as I will show in Section 5, accessibility plays an important role in the verbal syntax of English.

4. The role of the OV base order in German Although the analysis presented in the previous section accounts for the difference between symmetric and asymmetric V2 languages by referring to an


independently motivated property, German now poses us with an interesting paradox. On the one hand, it is expected to pattern with Icelandic and Yiddish. It has the same richness of agreement (that is, five distinctions in the present tense paradigm), it has transitive expletive constructions in main clauses (cf. (24a)) and no overt expletives show up in third position (cf. (24b)): (24) a.

Es hat jemand einen Apfel gegessen there has someone an apple eaten b. Gestern ist (*es) ein Junge gekommen gestern ist (*there a boy come

(German)

These facts suggest that Agr must head an independent projection. If, however, the verb moves in order to project Agr, we consequently expect that a second movement becomes necessary for the projection of tense features, as in Icelandic and Yiddish. Nevertheless, German is an asymmetric V2 language. In embedded clauses, the verb appears at the end of the clause. (25) … dass Sabine ihn liebt that Sabine him loves

(German)

To explain the difference between Icelandic and Yiddish on the one hand and German on the other, we would again like to find an independently motivated property that distinguishes these languages. What is relevant, I believe, is that Icelandic and Yiddish are VO languages, whereas German is an OV language. More precisely, the fact that objects precede the verb in embedded clauses makes it possible to generate verbal inflection separately from the verb in German, as indicated in (26): C

(26) C

AgrP DP

Agr′ VP

OB

V

Agr T

Agr

Note that in this structure Agr heads an independent projection, as required. At the same time, T c-commands both the subject and the predicate: it is not dominated by Agr, only by one segment of it, and it can therefore c-command

188 Olaf Koeneman

out of the head that it is part of (cf. the definitions in note 3). Hence both the conditions on Agr and T are satisfied by merging inflection to the right of V, from where it projects a head-final projection. Why would it be possible in German to generate inflection in a distinct position? And why must it be? There are two properties of inflection that are relevant here. First of all, inflection is affixal, meaning that T and Agreement must be spelled out on a verbal stem (Lasnik’s 1981 stray affix filter), a condition that most naturally holds at PF. Second, inflection is suffixal in that T and Agreement appear after the verbal stem rather than in front of it. Both conditions are straightforwardly met by the structure in (26). Since the stem of the finite verb appears at the right edge of VP, it is adjacent to [Agr [T] Agr] at PF, so that inflection can be spelled out appropriately. Second, these affixes are situated to the right of the verbal stem if one flattens the structure. Hence, they can be spelled out as suffixes, as required. Although it follows from the properties of inflection that the structure in (26) is allowed, this in itself does not explain why this option must be realized: Why is moving the verb leftward to project Agr, as in Icelandic and Yiddish, blocked altogether? There is a straightforward answer to this question, namely economy. Note that both conditions are met by the strategy taken. What does not take place is verb movement. This operation has become completely redundant since it fulfils no purpose. In short, OV languages with rich agreement generate inflection to the right of VP, since it is the most economical way of satisfying the relevant conditions: merge over move (Chomsky 1995). In VO languages, on the other hand, the same strategy is blocked for obvious reasons. If the inflectional material is generated on the right, like in German, internal arguments would generally intervene, so that the adjacency requirement is violated. If inflectional affixes are generated distinct from the verb but to the left, they cannot be spelled out as suffixes since they linearly precede the verbal stem when one flattens the structure.7 In terms of economy, then, nothing is gained by generating Agr in a position distinct from the verbal stem.

5. The lack of declarative V2 in English As already noted, English is the exception within the Germanic language group in not having generalized V2 in declarative clauses. Since in English verbs are marked for tense but can nevertheless stay in situ, the question is how the Tense condition is satisfied without verb movement. English is a VO language, so that


generating the tense affix separately from the verb, a strategy similar to the one employed by German, is not an option. Although one could assume that the complementizer is involved in the satisfaction of the Tense condition, as in Mainland Scandinavian, this still does not explain the lack of verb movement in main clauses. Like before, we would like to find an independently motivated factor that explains why English behaves differently. The lack of V2 is not the only property that makes English stand out within Germanic. It is also the only language where a particular class of elements, consisting of modals, auxiliaries have and be, finite forms of do and an infinitival particle to, show some common behavior. Most strikingly, they can all precede negation (cf. (27)) and license VP-ellipsis (cf. (28)). Furthermore, the elements in this class are mutually exclusive (cf. (29)): (27) a. John will not go to work today b. John does not go to work today c. John decided to not go to work (28) a. Mary would never wear anything that stupid but John certainly will b. Mary never wore that stupid party hat but John did c. Mary told John to lose that stupid hat but John did not want to (29) a. *Mary decided to will work today b. *Mary may will not go to work

Generalizing, one can say that there is a particular head position above negation that is realized by the above-mentioned class of lexical heads. The crucial observation is that in declarative clauses the main verb appears after VP-adverbs, indicating that it has not moved (cf. (30a)). When negation is generated, however, it triggers the presence of a finite head above negation which is distinct from the lexical main verb. Without the presence of such a head, the sentence is ungrammatical, as can be observed in (30b): (30) a. John never goes to work b. *John not goes to work

The standard account for the paradigm in (27)–(30) is that negation blocks some (morpho)syntactic process, so that generation of another finite head becomes necessary. This idea is already present in Chomsky’s original affix hopping analysis (Chomsky 1957): inflectional features above negation cannot be associated with the verb in its base position, since negation intervenes. As these features have to be picked up (or spelled out, or checked), a last resort operation must take place. Apparently, movement of the main verb is not an

190 Olaf Koeneman

option in English. The sentence can only be rescued by a head distinct from the lexical verb. Under the assumption that indeed some blocking effect takes place in (30b) but not in (30a), there must be some VP-external element in the structure with which the finite verb can be related in (30a) but not in (30c). In other words, the contrast reveals the presence of an element that is not phonologically realized in (30a), an empty head. The properties of this head position must at least be compatible with those elements that are in a complementary distribution with it, namely auxiliaries, modals and the infinitive marker to. Since these lexical heads divide across the [±tense] dimension, a natural hypothesis to make is that there is a head which abstractly expresses T. Given that all these elements compete for the same position, only one element from this class can precede negation, so that (29) is accounted for. Hence, the structure in (31) is empirically motivated: TP

(31) SU

T′ T

VP V

V V

OB Agr

T [–past]

The fact that English lacks declarative V2 can now be derived from this structure, since the Tense condition is met by the assumptions we have made so far. Although empty T is marked for [±tense], it is not further specified for this property. Like a complementizer, it lacks a specification for [±past]. If we were to assume that it could, we would be forced to postulate several empty elements T, each carrying a different tense specification (i.e. one expressing present tense, one past tense, etc.) which together form a tense paradigm. Under the reasonable assumption that a paradigm cannot be made up of several distinct null morphemes only, T by necessity lacks a specification for [±past]. Therefore, T cannot anchor the event in time. Recall now from the discussion of Mainland Scandinavian that C could enter into a dependency relation with the tense affix since this affix is accessible to C: it is encountered by C in the construction of


the complete head V. Under exactly the same reasoning, empty T can get a specification for [±past] from the tense affix on V. A dependency relation between T and the tense features on V ensures that the [±past] specification becomes visible on the empty head. The representation of John kissed Mary therefore looks as in (32): TP

(32) DP

T′ T [–past]

VP V

John

DP

V

T [–past]

kiss

-ed

Mary

Once the tense features of the predicate become visible on T, the Tense condition is satisfied: [−past] on T has COMMAND over both the subject and the predicate under m-command. It follows, therefore, that no additional projection has to be created through verb movement in English declarative clauses and that subject verb inversion remains absent. Note that the difference between English and a regular V2 language is that in the former the Tense condition is met under m-command whereas verb movement satisfies the constraint under c-command in the latter. In a V2 language, the Tense condition becomes relevant once the subject has been merged into the structure: satisfaction of it follows subject insertion. In English, on the other hand, empty T is merged into the structure before the subject is inserted: Satisfaction of the Tense condition coincides with insertion of the subject. Why, then, does merger of empty T not follow insertion of the subject in English? I think that the answer lies in the nature of the empty element involved. Notice that English is different from the other Germanic languages in having a number of modal heads which together form a class. This makes it likely that the empty head postulated is actually licensed by it. In other words, it is the existence of this modal paradigm that generates a semantically and phonologically empty modal. Two facts then follow. First of all, as other Germanic languages do not have this modal paradigm, such a semantically

191

192 Olaf Koeneman

vacuous modal cannot become available as a consequence of paradigmatic licensing. At least within the Germanic language group, the element is unique to English. Hence, we capture the fact that English alone lacks generalized V-to-C movement (i.e. V-to-C movement in both interrogatives and topicalizations).8 Second, as the empty tense marker belongs to the same paradigm as the modals, it is in complementary distribution with these heads. Under the assumption that the modals select a VP-predicate rather than a proposition, empty T has the same distribution as will, can, and must and is inserted before the subject is. The consequence of this is that the Tense condition is satisfied earlier in the derivation in English than in a V2 language, namely once the subject has been merged in the structure.9 Let us now return to the do-support paradigm. Under the standard assumption that, at least in English, negation heads its own projection, it will be the closest head that empty T sees when it is generated. The empty T lacks a feature value for [αpast] and will therefore try to enter into a dependency relation with a head that has a feature value for [αpast], namely the tense affix on the verb. However, empty T cannot enter into a dependency relation with this feature on the finite verb since negation intervenes. The resulting structure is as in (33). TP

(33) SU

T′ T

notP not

VP V

V

OB T [–past]

As will be clear, [−past] does not have COMMAND over the subject or the predicate. For this reason, a head specified for at least [±past] must be generated above negation: the empty T-marker is not an option. Hence, either a modal or, alternatively, a form of do must be used in negated contexts.10 Since these heads are all specified for [±past], they are able to satisfy the Tense condition: They occupy the right structural position to have COMMAND over the subject and the (negated) predicate.11


6. XP-fronting In this section I offer some speculations about what triggers XP-fronting in V2 languages. The striking fact is that the verb appears in second position and is preceded by some XP. Although an account has been provided for the verb movement in V2 constructions, I have so far ignored what might trigger XP-fronting. In principle I could adopt a condition requiring that the highest specifier position simply must be filled given some criterion (Rizzi 1996) or checking operation (Zwart 1993, 1997). Although such an analysis is, I believe, correct in its intuition, it is certainly not straightforward. First of all, it is not obvious what the relation is between the head, T in the present analysis, and the specifier, since basically any XP can occur in first position. Second, it would ignore the fact that declarative verb first orders are attested in V2 languages as well, which would be in violation of the hypothetical condition. Verb first orders typically arise in narrative contexts. An example is given in (34): (34) Komt een man de kamer binnen… comes a man the room in

(Dutch)

In order to overcome a violation of the hypothetical condition on the highest specifier, an empty operator could be postulated. There are two other verb first environments for which the presence of such an operator has been assumed. Yes/no-questions are often taken to have a covert counterpart to a wh-operator (cf. (35a)). In the literature on conditionals (Heim 1982, Kratzer 1986) a covert adverb of quantification, a generic or necessity operator, has been empirically motivated (cf. (35b)). (35) a.

OP heeft Jan dit boek gelezen? has Jan this book read b. OP mocht Harry nog komen, dan kan hij doodvallen. should Harry still come then can he dead-drop

(Dutch)

Under the assumption that this operator is realized as an XP in SpecTP, the examples in (35) are not really verb first structures. It is far less straightforward that an operator should be present in examples like (34). It is unclear how that would account for their interpretation and their restriction to particular contexts. Moreover, note that in English, a language without V2, verb movement can be observed in examples parallel to (35), whereas a structure similar to (34) is lacking:12

194 Olaf Koeneman

(36) a. b. c.

Has John read this book? Should John come, tell him to drop dead. #Does a man come in the room

In general, the postulation of empty operators should be very restricted. Otherwise it becomes hard to see why there is no empty operator that can be used precisely in declarative main clauses, as Roberts and Roussou (to appear) remark (cf. also Weerman 1989 for this point). What I would like to suggest instead is the following. The verb movement projecting T implies the possibility of adding a specifier. This has as a consequence that the overall structure counts as unfinished. The implication of a specifier corresponds to an unsaturated proposition at LF. More specifically, I propose that an implied specifier is interpreted as a semantic variable. It is this variable that has to be assigned a value. XP-fronting, then, is required as a last resort operation.13 However, given the characterization of verb first clauses as unfinished propositions with an unsaturated variable, it is unclear how narrative inversion (cf. (34)) fits in. In these cases, verb first structure can be felicitously used at the beginning of a story or joke, so that it is unclear what would saturate the variable introduced by verb movement. Closer scrutiny, however, reveals that this is not the only environment in which verb first clauses show up. They typically appear embedded in conversations (Den Besten 1983, Sturm 1986, Iris Mulders p.c.). Examples are given below:14 (37) a.

Weet je nog dat ik gisteren naar Amsterdam zou gaan? recall you still that I yesterday to Amsterdam would go Piet had eindelijk tijd voor me. Piet had at.last time for me Kom ik daar aan. Wat denk je? Piet is er niet. come I there prt what think you Piet is there not b. Je kent die vreemde man die bij de supermarkt werkt, niet? you know that strange man that at the supermarket works not Ik sta daar laatst te wachten. Begint hij opeens tegen me I stood there recently to wait begins he suddenly to me te praten. to talk

The verb first structures here have the effect of making the link to previous discourse more tight (cf. Zwart 1997). The absence of a syntactically marked link to previous discourse in the highest specifier in the last sentence seems to


express that the link is completely obvious and not so much that the information conveyed by the clause is new, as suggested by Roberts and Roussou (to appear) for narrative inversion.15 These verb first structures crucially appear after the discourse setting has been introduced. Similar observations have been made for verb first declaratives in Scandinavian. Both Platzack (1985) and Sigurðsson (1990) remark that these structures are prompted by discourse cohesion. As an indication of this, Sigurðsson notes for Icelandic that these sentences typically have pronominalized subjects. The fact that these verb first structures are so tightly connected to previous discourse reveals how they fit into the analysis developed so far. For pronouns, it is a generally accepted viewpoint that they are semantic variables which either obtain a value by being (semantically) bound or by having a contextually specified value (cf. Chierchia 1995 for discussion). If this state of affairs applies to pronouns, it would not be surprising to find that the same strategies are available for assigning a value to the variable introduced by verb movement. I therefore propose that in verb first structures, it is the situation expressed in the previous discourse (by approximation the time and place set up in it) that assigns a value to this variable. The variable in examples like (37), then, has a value defined by the linguistic context. For this to be felicitous, the link to previous discourse must be prominent. This is comparable to the fact that, given a particular discourse, pronouns can be felicitously used if and only if an antecedent is accessible, i.e. can be reconstructed from the previous discourse (cf. Ariel 1990). Let us now return to the question of how narrative inversion fits in. Since these verb first clauses typically appear at the beginning of a story or joke, it seems impossible to assign the variable a discourse value. A way of understanding it is to again draw a parallel with the distribution of pronouns in discourse. Although these elements usually refer back to an accessible antecedent in previous discourse, it is not uncommon to find them at the beginning of a novel, for instance.16 (38) a.

Hij staat elke dag om half zeven op. he gets every day at half seven up-prt (from Eerst grijs dan wit dan blauw, Margriet de Moor) b. De boerenmeid (of vrouw) had tenslotte niet geprotesteerd toen hij the farmer’s girl (or wife had after all not protested when he zijn kin op haar schouder liet rusten. his chin on her shoulder let rest (from De tranen der acacia’s, W. F. Hermans)

196 Olaf Koeneman

Although the pronouns in (38) do not have antecedents, we do not judge these sentences as infelicitous or, worse, ungrammatical. Apparently, they are appropriate ways to begin a story. What the writer achieves by the use of a pronoun in the first sentence is the suggestion of a shared discourse, giving the reader the feeling that (s)he is put right in the middle of things. We can then say that the pronoun is interpreted through accommodation (cf. Heim 1982) in this abstract discourse, which minimally contains the presupposition that there is a male entity. In this light, narrative inversion can be seen as a stylistic device with similar properties. A verb first structure can be used as a syntactic way of suggesting a shared discourse, just like the use of a pronoun in (38). The fact that these structures occur at the beginning of a story or joke is accounted for. One may wonder why the use of verb first constructions is relatively restricted. It is not a priori clear why relying on previous discourse for assigning a value to the variable is dispreferred to XP-fronting. After all, linking pronouns to antecedents in previous discourse is very common. A way of understanding the marked status of verb first structures is as follows. Superficially they look like yes/no-questions in that they display similar word order. If both verb first and verb second clauses are in principle interpretable as declarative clauses, a pragmatic principle then favours the one that is most clearly distinct from a non-declarative clause, i.e. a yes/no-question. For the present purposes, reference to Gricean maxims (‘avoid obscurity and ambiguity’) suffices to make the point. Hence, the V2 order is the unmarked declarative and the occurrence of verb first declaratives is relatively restricted.

7. Discussion In this article, it was argued that differences in verb positioning across languages are the result of (universal) output conditions interacting with languagespecific properties. Put central in this article is the distribution of V2. V-to-C is analyzed as a movement operation triggered by a condition on the tense properties of the predicate. The verb, then, moves with the purpose of projecting tense features. After some XP has been put into the specifier of the moved verb, a V2 effect ensues in Germanic main clauses. Languages differ as to whether V2 effects are found in embedded clauses. It was argued that in principle the Tense condition can be satisfied by a complementizer entering into a dependency relation with the tense affix on the verb. In languages with rich inflection, however, the verb moves independently to project Agr. This


operation makes the complementizer unable to assist in satisfaction of the condition on T and a second movement becomes necessary. The fact that no embedded V2 takes place in German, despite the need for an independent Agrprojection, is due to it being an OV language: inflection is generated in a distinct position so that the constraints on Agr and T are satisfied without verb movement. In English, the lack of V2 was linked to the presence of an empty head, an element licensed by the English modal paradigm. It should be noted that the view on functional projections and the specific triggers for verb movement argued for here are logically independent. Irrespective of one’s formulation of output conditions, however, the flexible view on functional structure has a number of conceptual advantages. First, since the presence of prefabricated head positions is no longer required for providing a trigger for verb movement, it becomes possible to reduce feature redundancy. In standard approaches, a functional head has a feature make-up that corresponds to similar features on the verb. This entails that these features are represented twice. Under the alternative view, they are only represented once, namely on the verb. This view becomes possible once we allow these features to project after movement. Note that an immediate consequence is the possibility of postulating positional triggers: Features on the verb must be brought in a particular VP-external position. In an approach that postulates abstract T and Agr nodes, these features would either be in that position already or would become visible in that position as a consequence of a dependency relation (AGREE) with the inflected verb. In both events, verb movement would be left unexplained. Moreover, by allowing a restricted (that is, paradigmatically licensed) occurrence of empty heads, an explanation for the lack of V2 in English becomes possible. This explanation would be hard, or at least unnatural, to state in an approach that allows empty heads more unrestrictedly. Second, violations of Chomsky’s (1995) Extension Condition are reduced. This condition requires that movement always extends the tree structure. Movement of a subject to SpecIP, for instance, creates a specifier and thereby extends the root of the clause. When heads are taken from the lexicon, they are merged with the top node of the representation already built, never somewhere in between. Verb movement blatantly violates this condition if analyzed as an operation that adjoins the verb to an empty head, as standardly assumed. In the alternative conception, however, verb movement is just as creating as XP-movement is: The operation extends the root of the clause just like insertion of a head does. Hence, verb movement no longer violates the extension condition.

198 Olaf Koeneman

Third, the proposal allows a unification of functional projections. In standard analyses, the I-position generally corresponds with overt morphology on the verb (T and Agr features), whereas C does not. In asymmetric V2 languages, for instance, the verb looks the same in main and embedded clauses. For this reason it is often assumed that C corresponds to abstract morphology. It is a priori not clear why the categories I and C should differ in this respect (cf. Bobaljik 1995: 299 on this point). The dichotomy between C and I is eliminated in the present analysis since C is replaced by T. Hence, V-to-C and V-to-I movement are related to morphology to the same extent. Both operations take place to project some feature of the verb and in both cases this feature correlates with existing morphology. Future research will have to indicate if and how the present proposal relates to more standard approaches to functional projection and to what extent unification is possible. The issue is far from trivial. If the number of verb movements correlates with the number of functional projections present, the conclusion must be that languages differ in the amount of functional structure that they generate, at least in overt syntax. Nevertheless, it is a widespread belief, perhaps best known as the universal base hypothesis, that all languages are fundamentally similar at some level of representation. This hypothesis implies that word order differences are merely a surface phenomenon and that apparent cross-linguistic differences are ultimately reduced or eliminated, for instance by covert checking operations (Chomsky 1995). If so, clausal structure must be much more uniform than the alternative proposal sketched here suggests. It is hard to directly evaluate the flexible approach in the light of the universal base hypothesis and checking theory. The latter are conceptual ideas and as such not immediately falsifiable. Showing that a piece of structure is not used in a particular language is not the same as showing that it does not exist. Likewise, although the focus in this article has been on V-to-C and V-to-I, movements that find strong distributional support, additional operations can never be categorically excluded. The goal of this article is not to argue against either the universal base hypothesis or checking theory in any direct way. They are not truly incompatible with the current proposal anyway, because the only claim made here is that, even if the universal base hypothesis and checking theory are entirely correct at some fundamental level, they have contributed little to our understanding of word order parametrization. Proponents of these views are therefore invited to look upon the theory developed here as a proposal on the workings of overt syntax which, in one form or another, is needed anyway. The fact that the


flexible approach to verb movement is able to make testable predictions about (differences in) verb placement suggests to me that it is worth exploring.

Notes * I would like to thank the audience at the 15th Comparative Germanic Syntax workshop for their comments and Jan-Wouter Zwart for proofreading the manuscript. All errors are mine, obviously. 1. I abstract away from V2 here. In contexts where V2 does not take place, some languages still move the verb whereas in other languages the verb remains in situ. 2. The statement in (6a) does not provide a rationale for the movement as such. In Koeneman (1997) it is argued that rich agreement affixes count as subjects for the grammatical system. Under the assumption that they must enter into a predication relation with VP (Williams 1980 and later work), verb movement becomes necessary. In its base position, Agr is dominated by the category with which it must establish a predication relation. 3. Throughout this article, I adopt the notions of c-command and m-command from Chomsky (1986): (i) a. b.

A category α c-commands a category β if and only if (i) α does not dominate β, and (ii) every category that dominates α dominates β. A category α m-commands a category β if and only if (i) α does not dominate β, and (ii) every maximal category that dominates α dominates β.

4. Like Chomsky (2000) I assume that heads can enter into dependency relations with other heads even if no movement takes place, but the notion accessibility used here cannot be equated with agree. The main difference is that in the present proposal the head-head dependency established does not involve feature identity but rather feature sharing. Since it is beyond the scope of this article to attempt a detailed comparison, let alone a unification, of checking theory and the present proposal, I will follow my own course here and note the crucial difference. 5. In the present theory, where verb movement creates functional projections, Mainland Scandinavian lacks an AgrP altogether and third position expletives are simply adjoined to VP. See Koeneman and Neeleman (2001) for a detailed discussion and alternative analysis of the contrast in (22). 6. Vikner (1995a) observes for Icelandic and Yiddish that in embedded questions subject verb inversion leads to ungrammaticality: (i) a. *Ég veit ekki af hverju í herberginu hefur kýrin staðið I know not why in the.room has the.cow stood a¢. Ég veit ekki af hverju kýrin hefur oft staðið í herberginu I know not why the.cow has often stood in the.room b. *Ikh veys nit ven in tsimer iz di ku geshtanen I know not when in the.room has the cow stood b¢. Ikh veys nit ven di ku iz oyfn geshtanen in tsimer I know not when the cow has often stood in the.room

(Icelandic)

(Yiddish)

200 Olaf Koeneman

Since (ia¢,ib¢) are well formed, Vikner concludes that here verb second has not applied. As the verb precedes the adverbs, the examples show distributional evidence for V-to-I. However, if declarative V-to-C is triggered by some general output condition, which I think would be a desirable result, it should apply equally in all finite clauses, including (ia¢,ib¢). In that event, the ungrammaticality of the examples in (ia,ib) must have a different source. One possibility can be found in Vance (1989) and De Bakker (1997). They try to account for the rare occurrence of verb second effect in embedded wh-questions in Old French and suggest that it is due to a pragmatic clash. Clause-initial XPs are generally prominent constituents, indicating what the link is to previous discourse. Wh-constituents, however, are prominent in much the same way. Suppose that as a consequence of this shared function there is a strong tendency to let the embedded wh-operator be followed by a non-prominent XP. In that case, it is not surprising to find that subject-initial clauses give better results, since subjects are default topics. 7. The analysis proposed for German was inspired by Bobaljik’s (1995) account of verb movement. In the present account, however, the possibility of generating affixes in a position distinct from the verb is restricted to OV languages with rich inflection. Bobaljik assumes that PF-adjustment rules can revert an Agr-V order into a well-formed morphological object with the affix appearing after the verbal stem. Obviously, adopting such rules here would destroy the explanation offered for the difference between symmetric and asymmetric verb second. 8. The analysis now seems to predict that all languages not paradigmatically licensing this null element should have V-to-C movement, which is clearly wrong for the Romance languages. In Koeneman (2000) it is shown how the lack of such a movement can be related to an independent property of these languages, namely the pro-drop phenomenon. 9. It should be noted that the empty modal is different from the overt modals in not blocking Tense and Agr features from occurring on the main verb: (i) a. John Ø never kisses Mary b. *John did/will kisses Mary It is argued in Koeneman (2000) that this contrast follows from the way the modal paradigm in English is set up. More specifically, modals are negatively marked for Agr and have a value for T, whereas the empty modal is underspecified for both properties. Under the assumption that morphological features must be realized if no clash ensues, the correct result obtains. 10. Since finite forms of do, as well as of be and have, can all be used in the same position, it seems more appropriate to call the relevant paradigm a class of T-markers rather than a modal paradigm. What makes it possible for finite forms of do, have and be to be in the same class as the modals may be what Pollock (1989) suggests as an answer to the question of why in his analysis these elements can raise in contrast to main verbs: their lack of thematic properties. This still leaves unsolved some notorious issues (e.g. constructions like John hasn’t a car, where has precedes negation and nevertheless appears to be thematic), but these are not directly relevant to the issue at hand, the lack of verb second in English. 11. I will assume that negation is part of the predicate in (33): After all, a negated predicate is also a predicate. Although the presence of negation blocks a dependency relation between T and the tense affix, it does not block c-command of the predicate by [±past], since it is itself part of this category.


12. Of course, one could think of sentences like Comes a man into the room, where SpecTP is filled by a hypothetical null operator. The point, however, is that these structures are only possible with unaccusative predicates, which makes them distinct from the operator constructions in (36a,b). If a null operator were generally available for narrative constructions in English as well, it is unclear why the unaccusative restriction should hold for narrative inversion but not for questions. 13. It is beyond the scope of this article to discuss the syntax of verb first languages like Irish and Welsh in any detail. Given the present analysis, they must have an independent property that makes it possible for verb first structures to surface as unmarked declaratives. Both Irish and Welsh have an intricate system of sentence-initial particles. If these elements provide the value for the variable introduced by verb movement, the contrast with regular V2 languages can be derived (see Koeneman 2000 for an analysis along these lines). 14. As can be observed in (37), definite subjects are perfectly felicitous in these contexts. This makes it less likely that verb first constructions arise as a consequence of expletive er having been dropped from first position, which would have been a potential analysis for (34). 15. Sturm (1986: 356, footnote 19) remarks about these sentences that, though not unnatural or unusual, they are hard to define. 16. Thanks to Jan Frans van Dijkhuizen and Bertram Mourits for providing these examples.

A verb’s gotta do what a verb’s gotta do! On Scandinavian infinitivals and the AGR parameter* Øystein Alexander Vangsnes University of Tromsø

1.

Introduction

The purpose of this article is to discuss how V-to-I-movement in Icelandic control infinitivals can be seen as an effect of the same syntactic requirement that triggers V-to-I-movement in (embedded) finite clauses. To attribute these two instances of V-to-I-movement to the same source is not an entirely straightforward matter. The reason for that is that whereas finite V-to-I-movement can be related to the rich inflection on Icelandic finite verbs, showing number and person agreement with the sentential subject, there is no such variation on the moving verb in control infinitivals, i.e. on an infinitive. This is then our puzzle: why is there V-to-I-movement in Icelandic control infinitivals when there is no visible agreement on the moving verb? The solution that I will advocate is conceptually simple, but it requires some rethinking of the syntactic mechanisms involved, in particular concerning the nature of functional categories. The essence of my account is the following. I will argue that V-to-I-movement represents one possible way of meeting a formal requirement related to a particular functional category (to be identified with AgrS), a strategy preferred by UG when enhanced by overt morphology. Unlike earlier accounts I take it that the particular functional category (AgrS) is present in control infinitivals, and this assumption will be semantically motivated. Finally, I argue that although overt morphology does not enhance verb movement as the chosen strategy in control infinitivals, it is chosen for conceptual/ economical reasons: the grammar does not have to specify a different strategy

204 Øystein Alexander Vangsnes

for control infinitivals than for finite clauses. In other words, the (leftmost) verb moves in Icelandic control infinitivals because it does so elsewhere.

2. Verb movement and Scandinavian infinitivals The fact that the (leftmost)1 verb in an Icelandic control infinitival occurs to the left of sentence adverbs like aldrei ‘never’ constitutes the evidence for saying that the verb has moved to a VP-external position since such adverbs are either taken to mark the VP-border or to occur to the left of VP. The examples in (1) illustrate that the verb cannot occur to the right of such an adverb. (1) a.

Jólasveinarnir lofuðu að borða aldrei búðing. Santa-Clauses-def promised to eat never pudding b. *Jólasveinarnir lofuðu að aldrei borða búðing. Santa-Clauses-def promised to never eat pudding ‘The Santa Clauses promised to never eat pudding’

(Icelandic)

In this respect Icelandic contrasts with Mainland Scandinavian where the infinitive of a control infinitival may not raise across a sentential adverb. This is shown by the Norwegian examples in (2) — Norwegian is representative of Mainland Scandinavian in general. (2) a. *Julenissen lovte å eta aldri graut. Santa-Claus-def promised to eat never porridge b. Julenissen lovte å aldri eta graut. Santa-Claus-def promised to never eat porridge ‘Santa Claus promised to never eat porridge.’

(Norwegian)

In turn the situation found in control infinitivals parallels the situation found in embedded finite clauses where there is finite verb movement in Icelandic but not in Mainland Scandinavian. This contrast is illustrated in (3) and (4). (3) a.

… að jólasveinarnir borðuðu aldrei búðing. that Santa-Clauses-def ate never pudding b. *… að jólasveinarnir aldrei borðuðu búðing. that Santa-Clauses-def never ate pudding ‘… that the Santa Clauses never ate pudding.’

(4) a. *… at julenissen åt aldri graut. that Santa-Claus-def ate never porridge

(Icelandic)

(Norwegian)

On Scandinavian infinitivals and the AGR parameter 205

b. … at julenissen aldri åt graut. that Santa-Claus-def never ate porridge ‘… that Santa Claus never ate porridge.’

Since Holmberg and Platzack (1988) it is widely accepted that the difference between the languages on this point in some way or another is related to a very clear difference in the verbal morphology of Icelandic on the one hand and Mainland Scandinavian on the other: Icelandic has a rich inflectional system with both person and number subject-verb agreement on finite verbs, whereas Mainland Scandinavian finite verbs only carry tense morphology. Holmberg and Platzack (1995) captures this correlation by postulating the so-called “AGR-parameter”, and the idea is basically that a positive setting for the parameter triggers V-to-I-movement whereas a negative setting does not. Holmberg and Platzack also relate a number of other syntactic properties to this parameter, but these will not concern us here. The full paradigms of tensed forms of the verbs corresponding to ‘eat’ in Icelandic and Norwegian are given in Table 1. Table 1.Finite inflection in Icelandic and Norwegian Icelandic

1. sg. 2. sg. 3. sg. 1. pl. 2. pl. 3. pl.

Norwegian

present

past

present

past

borða borðar borðar borðum borðið borða

borðaði borðaðir borðaði borðuðum borðuðuð borðuðu

et et et et et et

åt åt åt åt åt åt

Translated into a checking based framework (Chomsky 1995), Holmberg and Platzack’s theory would in essence amount to saying that INFL (in effect AgrS) carries a strong agreement feature which must be checked and deleted before Spell-Out, the overt syntactic reflex of which is that the verb occurs to the left of sentence adverbials and the overt morphological reflex of which is that the moving verb is richly inflected. Correspondingly, in Mainland Scandinavian the agreement feature would be weak and verb movement would take place covertly, i.e. after Spell-Out. The overt position of the verb is therefore its base-generated position.


Thráinsson (1993) discusses Scandinavian infinitivals within such a checking based theory, and he points out that it is unclear how the verb movement seen in Icelandic control infinitivals can be related to richness of agreement since there is neither person nor number agreement on infinitives in Icelandic. Thráinsson’s point can be illustrated by the examples in (5) which show that the infinitive has the same form regardless of the person and number of the constituent controlling PRO (i.e. the matrix subject). (5) a.

Ég lofaði að borða aldrei búðing. I promised to eat never pudding ‘I promised to never eat pudding.’ b. Þiú lofaðir að borða aldrei búðing. c. Hann lofaði að borða aldrei búðing. d. Við lofuðum að borða aldrei búðing. e. Þiið lofuðuð að borða aldrei búðing. f. Þieir lofuðu að borða aldrei búðing.

‘You promised…’ ‘He promised…’ ‘We promised…’ ‘You-pl promised…’ ‘They promised…’

Instead of relating V-to-I-movement in Icelandic control infinitivals to agreement (and consequently AgrS), Thráinsson develops an alternative account based on a comparison between control infinitivals and other complement types introduced by the complementizer að, more specifically modal infinitivals and finite embedded clauses. As Thráinsson points out, most Icelandic modal verbs take an infinitival complement introduced by the complementizer að. Some examples are given in (6). (6) a.

Jón ætlar *(að) lesa bókina. John intends *(to read book-def ‘John will read the book.’ b. Jón verður *(að) lesa bókina. John must *(to read book-def ‘John must read the book.’ c. Jón á *(að) lesa bókina. John ought *(to read book-def ‘John should read the book.’ d. Jón þlarf *(að) lesa bókina. John needs *(to read book-def ‘John needs to read the book.’

(Icelandic)

However, unlike control infinitivals and finite complements, there is no evidence for verb movement in the modal infinitivals. Thráinsson gives the examples in (7a–d) to illustrate this fact — the example in (7c) shows that the


infinitive will raise past the adverb oft ‘frequently’ in a control infinitive, whereas (7d) shows that this is not possible in a modal infinitive — the infinitive will have to follow the adverb as in (7e). (7) a. b. c. d. e.

Risarnir segja að þleir hafi stundum étið rikisstjórnir. (finite) giants-def say that they have sometimes eaten governments (finite) Risarnir segja að þleir étii stundum ti rikisstjórnir. giants-def say that they eat sometimes governments Risarnir lofa að étai oft ti rikisstjórnir. (control) giants-def promise to eat frequently governments *Risarnir eiga að étai oft ti rikisstjórnir (modal) giants-def ought to eat frequently governments Risarnir eiga að oft éta rikisstjórnir. (modal) giants-def ought to frequently eat governments

Thráinsson relates this and other differences between the complement types first of all to a difference in the number of functional projections each complement type has, and secondly to the placement of the complementizer. In brief, his analysis is that finite complements have CP, AgrSP, and TP; control infinitivals have AgrSP and TP but not CP; and modal infinitivals have TP but not CP and AgrSP. In turn the element að heads CP in embedded clauses, AgrSP in control infinitivals, and TP in modal complements. Sentence adverbs like aldrei, stundum, and oft Thráinsson takes to be adjoined to VP. Furthermore, he assumes that að heads the highest/leftmost functional projection in the various complement types, and that T0 carries strong verbal features in Icelandic. The effect of the latter assumption is that “finite as well as non-finite verbs have to move [to T0] to check the relevant V-features, presumably tense.” (Thráinsson 1993: 198). This gets the facts right. In a finite complement the complementizer heads C and the verb moves to T (and on to AgrS). In a control infinitival the complementizer heads AgrS and the infinitive moves to T. Finally, in a modal complement the complementizer heads T and this position can therefore not be targeted by the verb, and hence there is no verb movement. The analysis of the sentences in (7) can thus be rendered as in (8). (8) a. b. c. d.

… segja [CP að … segja [CP að … lofa [AgrSP að … eiga [TP að oft

[AgrSP þeir hafii [TP stundum [VP ti étið rikisstjórnir [AgrSP þeir étii [TP stundum [VP ti rikisstjórnir [TP éta oft [VP ti rikisstjórnir [VP éta rikisstjórnir


Although Thráinsson states that it is unclear how ‘richness of agreement’ can account for verb movement in Icelandic control infinitivals, this notion is, in my opinion, implied by Thráinsson’s own analysis. For unless this is what supports the assumption that T carries strong verbal features in Icelandic, it is unclear why Mainland Scandinavian shouldn’t also have strong T and, following from that, verb movement to T in control infinitivals. Although Mainland Scandinavian does not have person and number agreement, Mainland Scandinavian verbs carry tense morphology to the same extent as in Icelandic.

3. A compositional approach to the functional domain In the following I will develop an alternative analysis of Scandinavian infinitivals which to some extent is in the spirit of Thráinsson. More specifically, I will follow his idea that the differences between the various complement types should be related to differences in the number of functional projections found. However, I will claim that the combination of functional categories is different, and that it is semantically driven. The core idea is that there is an explicit correlation between functional categories and semantic properties, and that the actual composition of the functional domain of a clause, and phrase structural objects in general, reflects its semantic properties. Two claims are crucial for the analysis in this respect: (i) T correlates with tense and ‘temporal anchoring’, and (ii) AgrS correlates with an ‘event’ interpretation, and accordingly the presence of AgrS in a clause entails that the clause denotes an event. The crucial point here, then, is that an event interpretation is not dependent on temporal anchoring: these two denotational aspects are factored out, and given distinct syntactic correlates. As for the status of C, I will return to that in a later section. Furthermore, I propose that whereas embedded clauses contain both AgrS and T in addition to C, control infinitivals have AgrS but not T in addition to C. Modal infinitivals have neither AgrS nor T, only C. According to the claims about the correlation between functional categories and semantic properties I then suggest that neither control infinitivals nor modal infinitivals receive a temporal anchoring, and moreover that control infinitivals, but not modal infinitivals, receive an event interpretation. Given these assumptions the set-up in (8) can now be recast as in (9).


(9) a.

… segja [CP að

[AgrSP þeir hafii [TP stundum [VP ti étið rikisstjórnir (finite) b. … segja [CP að [AgrSP þeir étii [TP stundum [VP ti rikisstjórnir (finite) [AgrSP éta oft [VP ti rikisstjórnir (control) c. … lofa [CP að d. … eiga [CP að oft [VP éta rikisstjórnir (modal)

The claim that control infinitivals are not temporally anchored may seem to go against the insight in Stowell (1982), recalling Bresnan (1972), that control infinitivals involve an “unrealized” tense. In that respect they contrast with gerunds, a phenomenon that Stowell illustrates by the examples in (10) and (11). (10) a. Jenny remembered [PRO to bring the wine]. b. Jenny remembered [PRO bringing the wine]. (11) a. Jim tried [PRO to lock the door]. b. Jim tried [PRO locking the door].

About these sentences Stowell (1982: 563) says the following: In each case, the tense of the infinitival complement is understood as being unrealized with respect to the tense of the matrix; thus in [10a], Jenny has not yet brought the wine at the point at which she remembers to do so, while in [11a], Jim does not succeed in locking the door when he tries to do so. In contrast, the understood tense of the gerund is completely malleable to the semantics of the governing verb. Since one normally remembers things about the past, the tense of the gerund is understood as past with respect to the matrix in [10b]; in [11b] the tense is ambiguous between present and unrealized with respect to the tense of the matrix.

The difference between control infinitivals and gerunds can be further illustrated when we consider cases where the semantic properties of the matrix verb require that the state-of-affairs in the complement cannot happen simultaneously with that of the matrix. In such cases gerunds are not possible. Consider the examples in (12). (12) a. Jim promised to lock the door. b. *Jim promised locking the door.

These observations are only seemingly incompatible with the present approach. If control infinitivals denote an event which is distinct from that of the matrix, one could argue that the “unrealized tense” of control infinitivals is but an effect of this: the event of a control infinitival is, by hypothesis, temporally


unanchored whereas the event of the matrix clause is anchored, and since the denotation of the complement necessarily must relate to the denotation of the matrix clause, the result is an interpretation where the complement denotation is “unrealized” with respect to the matrix. Pursuing this line of reasoning I would say that gerunds do not denote events that can be interpreted independently of the matrix clause, and consequently, within the present apporach, that they do not contain AgrS. Semantically speaking that seems correct: the temporal interpretation of the gerund in the examples in (13) varies with that of the matrix tense. (13) a. Jim tried locking the door. b. Jim tries locking the door. c. Jim will try locking the door.

It is also worth noticing that in these examples the locking of the door necessarily takes place — their interpretation is not compatible with a situation where Jim does not succeed in locking the door. Moreover, on a textual note, the implication of the sentence is that John’s attempt is one of achieving something by locking the door. In parallel examples with control infinitivals things are different. The interpretation of the sentences in (14) is one where the attempt is one of getting the door locked, and this interpretation is indeed compatible with a situation where Jim does not manage to lock the door. (14) a. Jim tried to lock the door. b. Jim tries to lock the door. c. Jim will try to lock the door.

In that sense the temporal interpretation remains constant: the act of locking the door is hypothetical in all cases since it is an open question whether or not the door actually got locked. Another way of phrasing the difference between control infinitivals and gerunds is to say that the matrix/gerund structure involves a singular (albeit complex) event whereas the matrix/control structure involves two events. Returning now to control versus modal infinitivals it appears straightforward to say that a clause structure with a modal matrix verb and an infinitival complement involves just a single event and furthermore that there is no separate temporal anchoring for the infinitival complement. (15) a. Jim will open the door. b. Jim ought to open the door.

On Scandinavian infinitivals and the AGR parameter

In the present system these (lacking) semantic properties are reflected by the absence of AgrS and T from the structure of the infinitival complement: in terms of event and tense interpretation the infinitive and the (finite) modal both relate to the matrix AgrS and T. Summarizing so far, finite complements and control infinitivals both have the functional category AgrS whereas modal infinitivals do not. This constitutes a first step towards a unified account of verb movement in finite and non-finite complements: we can identify the target of the moving verb as AgrS. The next step is to find a unified explanation for why the verb moves.

4. Identification Theory In Vangsnes (1999) I argue that functional categories are abstract heads listed in the lexicon. The abstract functional categories are either extensional or categorial, and extensional functional heads determine semantic properties of the phrase whereas categorial ones determine the categorial status of a phrase structural object (PSO) and provide a boundary towards other PSOs. Put differently a categorial functional head closes off the extended projection of a lexical head, whereas extensional functional heads combine with one or more lexical heads to yield a referring expression. In clauses the functional category heading C, κ, is categorial, whereas the categories heading AgrS and T, σ and τ, respectively, are extensional. With reference to the preceding discussion in Section 3 the presence/absence of AgrSP and TP is a question of whether or not σ and τ are present in or absent from the numeration: σ anchors the state-of-affairs denoted by the verb to an event, whereas τ provides a temporal anchoring for the same state-of-affairs. Importantly, the presence of the abstract functional heads must be licensed through a syntactic mechanism termed identification. This is stated as The Identification Principle in (17), and the definition of identification is given in (18). (17) The Identification Principledef A functional category must be identified in overt syntax. (18) Identificationdef A functional category F in an extended projection P is identified iff a constituent of P that contains at least one feature relevant for F is merged in either the head or the specifier position of F.

211


The Identification Principle is therefore a possible trigger for movement: a constituent already merged in a part of the extended projection c-commanded by the functional category may contain a feature relevant for the identification of the functional category, and it may therefore move to the projection of that functional category. Furthermore, I make a distinction between lexical features and agreement features. This distinction resembles the [±interpretable] distinction in Chomsky (1995: 277ff) and subsequent work: a lexical feature then corresponds to a [+interpretable] feature and an agreement feature to a [−interpretable] feature. The distinction becomes relevant when there is a choice between more than one possible identifier for a functional category. The basic intuition behind the distinction is that a lexical feature is a feature that a constituent possesses independently of other constituents whereas an agreement feature is a feature that a constituent receives from or shares with another constituent, or, in effect, with a lexical feature of that other constituent. Certain features may have a dual status: they may be a lexical feature on one constituent, and an agreement feature on another constituent within the same PSO. A case in point is [num(ber)] which I take to be a lexical feature on nouns and pronouns, but an agreement feature on verbs.2 The features that come into play in the present discussion are the following: (19) Lexical features: Agreement features:

{[deixis], [Case], [tense], [num]lex, [V(erbal)] …} {[person], [num]agr …}

Importantly, by [Case] I understand an abstract, semantic property of noun phrases along the lines discussed by De Hoop (1992), and not morphological case.3 That justifies its classification as a lexical feature. Morphological case one could argue is an agreement feature. Moreover, I take [deixis] to be a characteristic morphological feature of definite noun phrases, including pronouns. As for the extensional categories σ and τ the features relevant for their identification are the following. (20) σ must be identified by an element containing at least one of the following features: [deixis], [Case], [person], [V]. (21) τ must be identified by an element containing at least one of the following features: [tense], [number]agr/lex, [V].

In situations where there is a choice of more than one identifier for a functional category, the following algorithm applies for choosing the identifier.


(22) Preferred identifierdef When there are several candidate constituents for identifying a functional category F, the preferred identifier for F will be (i) the constituent containing the largest number of agreement features relevant for F; else, if there is no such constituent, (ii) the constituent containing the largest number of lexical features relevant for F; else, if there is no such constituent, (iii) the constituent containing the smallest number of irrelevant lexical features; else, if there is no such constituent, (iv) an X0.

In effect, this algorithm assigns a more prominent status to agreement features than to lexical features, and, moreover, heads are preferred over phrases, other things being equal. The more prominent status of agreement features captures the intuition that agreement features unlike lexical features are redundant, so if a language has agreement features in some domain, they should be there for some reason: the reason is that they render constituents suitable for identification of functional categories. On this background let us now consider the identification of the clausal extensional categories in Scandinavian, in particular the identification of σ.

5. The identification of σ in Scandinavian and the notion of generalized strategy All argument noun phrases carry the lexical feature [Case]. Countable noun phrases carry the lexical feature [num], whereas mass denoting noun phrases do not (cf. note 2). Definite noun phrases carry the lexical feature [deixis], and this distinguishes them from indefinite noun phrases which do not carry this feature. As for finite verbs they carry the lexical features [tense] and [V]. On the basis of paradigms like the ones in Table 1 above I will posit that Icelandic finite verbs carry the agreement features [person] and [number], and that Mainland Scandinavian finite verbs do not. One may then envisage the feature compositions in Table 2 for various constituents in the two varieties of Scandinavian. I will return to infinitive verbs shortly. Given this feature composition the finite verb will be the identifier of τ in both Icelandic and Mainland Scandinavian. Since identification must apply in overt syntax this means that I assume V-to-T-movement in all varieties of Scandinavian, and moreover that the position of sentence adverbs, including

213


Table 2.The feature composition of constituent types in Scandinavian

Vfin DPdef-count DPindef-count DPdef-mass DPindef-mass

Icelandic

Mainland Scandinavian

[tense], [person], [num]agr, [V] [Case], [deixis], [num]lex [Case], [num]lex [Case], [deixis] [Case]

[tense], [V] [Case], [deixis], [num]lex [Case], [num]lex [Case], [deixis] [Case]

negation, is higher than (i.e. to the left of) T/τ. Notice that although an Icelandic finite verb carries one more feature relevant for the identification of τ than Mainland Scandinavian, namely [num], this does not affect the choice of the Mainland Scandinavian finite verb as the preferred identifier for this functional category: none of the various DP types is a head with more relevant features (and no other non-V(erbal) X0 with both a tense and a number feature is likely to be found within the grammar either, cf. (21) and (22)). Things are different when we consider the identification of σ, however. The Icelandic finite verb is an X0 with an agreement feature ([person]) relevant for the identification of this functional category, and since the feature is the only agreement feature relevant for identification of σ, this makes the Icelandic finite verb a hard candidate to beat when it comes to choice of identifier. Again, since identification is a principle affecting overt syntax, there is overt T-toAgrS-movement in Icelandic. In Mainland Scandinavian on the other hand, the lexical feature [V] is the only feature relevant for the identification of σ carried by the finite verb. This means that most of the DP types listed in Table 2 will be preferred over the finite verb as identifiers of σ in this variety of Scandinavian. Accordingly, there is no T-to-AgrS-movement in Mainland Scandinavian. Instead, σ is identified by the subject DP in this language. What then if the subject is a mass denoting indefinite subject? That would be a case where the finite verb would be a better suited identifier even in Mainland Scandinavian, so one could expect to find T-to-AgrS-movement in such cases. But one does not. In fact, one does not find truly indefinite mass denoting subjects at all. To the extent that the sentence in (23a) is well formed it can only be a generic statement about beer, and if such a reading is to be avoided one would leave the DP in a VP-internal position and merge an expletive in AgrSP as in (23b).


(23) a. ??… at øl aldri har vore i kjøleskåpet. that beer never has been in fridge-def ‘… that beer has never been in the fridge.’ b. … at det aldri har vore øl i kjøleskåpet that expl never has been beer in fridge-def ‘…that there has never been beer in the fridge.’

I thus claim that σ is never identified by the finite verb in Mainland Scandinavian. The reason for this is that to the extent possible the language learner will avoid construction specific strategies for identifying functional categories. Since σ is identified by the subject DP in the majority of cases a similar strategy is also used in cases where the (potential) subject is less suited for identification than some other constituent: the expletive is, by hypothesis, an element carrying both the feature [Case] and the feature [deixis], and merger of this XP is resorted to in cases involving an indefinite mass-denoting argument noun phrase. In order words, identification of σ by an XP is the generalized strategy in Mainland Scandinavian. The generic reading that results in cases like (23a) can be accounted for by positing that these noun phrases in fact are equipped with a [deixis] feature with the value [+generic]. Discussion of related issues can be found in Vangsnes (2001). Consider then control infinitivals. τ is absent, but σ is present, and there are in principle two candidates for the identification of σ: the infinitive verb and PRO. PRO carries, qua pronoun, the feature [deixis] (cf. Section 4), and given the view that PRO is in fact Case-marked (cf. Sigurðsson 1991), one may argue that it also carries the feature [Case]. Both of these features are relevant for the identification of σ. What about the infinitive? Surely it will carry the lexical feature [V], but does it carry other features? I will argue that it does so in Icelandic but not in Mainland Scandinavian. More specifically, Icelandic infinitives agree with PRO and there is subject verb agreement between PRO and the infinitive just like there is between a subject and a finite verb. In motivating this I again resort to issues concerning language acquisition, and the basic line of reasoning is the following. The infinitive belongs to the same inflectional paradigm as finite forms. The language learner does not distinguish infinitives from finite forms with respect to syntactic properties — both types of verb forms cooccur with subjects, and even though the infinitives do not have an overt correlate for the subject, they are specified (in the lexicon) as carrying the same agreement features as finite forms. The only difference is in fact that the infinitive does not carry any tense specification.

215


So the overall situation in the language determines the syntactically relevant feature specifications for infinitives. In Icelandic the language learner finds evidence for specifying verbs, finite and infinitive, as carrying the agreement features [person] and [num] since in the majority of cases there is an overt morphological correlate. In Mainland Scandinavian there is not, and accordingly finite verbs and infinitives are not specified as carrying these features. In turn this gives us a straightforward answer as to why the infinitive of a control infinitival moves in Icelandic: it is the preferred identifier for σ. In other words the verb’s gotta do what it does elsewhere: identification of σ by the verb is the generalized strategy in Icelandic. In Mainland Scandinavian the infinitive does not move. Instead σ is identified by PRO, the preferred identifier in this variety of Scandinavian.4

6. Conclusion I have now accomplished the main objective of this article: to develop a uniform account of why there is verb movement in Icelandic control infinitivals and embedded finite clauses and of why there is no such movement in the corresponding Mainland Scandinavian constructions. The logic of the account is conceptually simple. First of all I compared the phrase structure of control infinitivals and (embedded) clauses and followed a line of reasoning according to which their phrase structure comprises the functional category AgrS: I argued that AgrS correlates with an event reading and that both control infinitivals and embedded clauses denote events that are separate from that of the matrix clause. Second, I identified AgrS as the target for verb movement in Icelandic sentential complements, and Section 4 and 5 presented a theory which provides answers to what the purpose of such movement is and to why a grammar will choose such a strategy over another in order to fulfill the purpose. The theory-specific side of the account may be inadequate, but I do believe that the fundamental general insight is correct: the verb moves in Icelandic control infinitivals because it does so elsewhere. That is an easy “rule” for the language learner to capture and follow.


Notes * I am grateful to Anders Holmberg and Peter Svenonius for discussion and comments. Yet I alone am responsible for any inconsistencies, inadequacies, or shortages of exposition in the article. 1. A control infinitival may contain an (aspectual) auxiliary verb in Scandinavian. In that case it is the auxiliary verb that moves in the Icelandic control infinitival. This is exemplified in (i). (i) a.

Pétur lofaði að hafa ekki borðað allan búðinginn áður en ég kæmi. Peter promised to have not eaten all pudding-def before than I came ‘Peter promised to not have eaten all the pudding before I came.’ b. *Pétur lofaði að ekki hafa borðað allan búðinginn áður en ég kæmi.

Notice that the auxiliary carries the infinitive form whereas the main verb carries the perfect form as required by the auxiliary hafa ‘have’. 2. The [num] feature of noun phrases viewed as a closed off PSO counts as a lexical feature, although various constituents within the noun phrases may carry an agreement [num] feature, “received/shared” with the lexical [num] feature of the noun. In effect, only countable noun phrases will be assumed to carry such a [num] feature, cf. Vangsnes (1999: 46ff). 3. Notice that [Case] then would be a [+interpretable] feature in the Chomskyan sense, contrary to the position Chomsky takes. 4. PRO resides within AgrSP in Icelandic also, but that is for matters pertaining to definite subjects that I will not go into here. See Vangsnes (2001) for discussion.

On the correlation between morphology and syntax The case of V-to-I* Artemis Alexiadou and Gisbert Fanselow Universtity of Potsdam

In this article we revisit V-to-I-movement in Germanic and beyond. We examine and evaluate the hypothesis that there is a correlation between richness of verbal inflectional morphology and the obligatory movement of the finite verb to Infl, which has been adopted in much recent literature. We show that this hypothesis is empirically inadequate, and that in fact V-to-I movement across languages is independent of morphology. We formulate an explanation for the observable tendencies/links between syntactic movement and morphological richness by considering how inflectional morphology arises, how verb second (V2) languages transform into non-V2 SVO-languages, and how V-to-I-movement can get lost in the history of a language.

1.

V-to-I movement and word order variation

It has been proposed that SVO languages fall into two groups when one considers the relative order of the finite verb and a ‘low’ adverb, which is taken to mark the left edge of the VP. In languages such as Icelandic, the finite verb in a subordinate clause obligatorily precedes such adverbs (2), while in languages such as English, the finite verb obligatorily follows these adverbs (3) (data from Vikner 1995b, Haegeman 1997). (1) a. S V Adv O Æ Icelandic, French, Italian b. S Adv V O Æ English, Danish, Faroese, Norwegian, Swedish

220 Artemis Alexiadou and Gisbert Fanselow

(2) a. b. c. d. e. f.

En. That John [VP often eats tomatoes] surprises most people Da. At Johan [VP ofte spiser tomater] overrasker de fleste Fa. At Jón [VP ofta etur tomatir] kemur óvart á tey flestu Ic. *Að Jonas [VP oft borðar tómata] kemu flestum á óvart Fr. *Que Jean [VP souvent mange des tomates] surprend tout le monde It. *Gianni [VP spesso mangia pomodori]

(3) a. b. c. d. e. f.

En.*That John eats [VP often tomatoes] surprises most people Da.*At Johan spiser [VP ofte tomater] overrasker de fleste Fa.*At Jón etur[VP ofta tomatir] kemur óvart á tey flestu1 Ic. Að Jonas borðar [VP oft tómata] kemu flestum á óvart Fr. Que Jean mange [VP souvent des tomates] surprend tout le monde It. Gianni mangia [VP spesso pomodori]

Under the assumption that in both language groups the adverb is attached at the same site (see e.g. Alexiadou 1997, Cinque 1999 for recent discussion), this pattern suggests that in the first group the verb has undergone head movement to (at least) Infl, while in the second group the verb remains in VP (Emonds 1976, Travis 1984, and Pollock 1989, among others). Another difference between these two language groups concerns their verbal inflectional paradigms. In French, Italian, and Icelandic, on the one hand, verbs are marked for a range of distinctions for person and number, while in English and Danish this is not the case. (4) Danish høre ‘hear’ jeg du han vi I de

hører hører hører hører hører hører

(5) Italian sentire ‘hear’ io tu lui noi voi loro

sent-o sent-i sent-e sent-iamo sent-ite sent-ono

English walk (past) I you he we you they

walk-ed walk-ed walk-ed walk-ed walk-ed walk-ed

French marcher ‘walk’ (past)

Icelandic kasta ‘throw’

je march-ais [-7] tu march-ais [-7] il march-ait [-7] nous march-ions vous march-iez ils march-aient [-7]

eg þú hann við þið þeir

kasta kasta-r kasta-r köst-um kast-ið kasta

On the correlation between morphology and syntax 221

The above observations have led to the proposal that the presence of rich inflectional morphology is responsible for the presence of overt V-to-I movement in a language. Kosmeijer (1986) presents the first attempt in the generative literature to establish such a link between V-to-I movement and the relevant richness of verbal inflection. According to this view, V-to-I movement is observed in exactly the languages with rich verbal inflection. This hypothesis has been expressed in several ways in the literature, depending on whether or not one adopts a minimalist perspective, and is repeated here in (6)–(7), see e.g. Pollock (1989), Ouhalla (1988), Platzack (1988), Platzack and Holmberg (1989), Chomsky (1993). (6) Overt V-movement is triggered by rich verbal morphology (7) Infl has a strong V-feature iff Infl has rich morphology

More generally, on the basis of contrasts such as the ones illustrated above, (8) has been proposed to be a principle of Universal Grammar (UG). (8) Syntactic movement is triggered by morphology (Haegeman 1997: 8)

In this article we argue that (8) cannot be a UG principle. To the extent that it holds at all, we show that the correlation between overt movement and the presence of rich morphology is a consequence of the mechanisms by which rich morphology comes into being in the diachronic development of languages, and of the restricted nature of the mechanisms by which syntactic processes such as verb movement can be lost. We present potential triggers for the loss of V-to-I movement that make no direct reference to (the loss of) morphology. The article is organized as follows. First, we consider the issue of whether the correlation between morphology and overt movement proposed for V-to-I movement extends to other domains of grammar, and we will arrive at a negative result (Section 2). We then show that the problem of exactly characterizing “rich” morphology has not been solved satisfactorily. What can at most be defended is that all SVO languages with “rich” suffixal inflection have overt verb movement — no other correlations hold (Sections 3 and 4). We then present the diachronic explanation for the connection between overt verb movement and presence of rich inflection. This is complemented by a demonstration that the loss of overt verb movement is only indirectly influenced by a loss of inflection (Section 5).


2. Morphology and movement from a broader perspective 2.1 Problem A: Is (8) a correlation only for V-to-I movement? In this section we show that the type of correlation proposed for V-to-Imovement between rich morphology and overt movement is not observed in any other domain of grammar. Thus, V-to-I movement would be the only process for which such a link between overt displacement and rich morphology holds. But even if restricted to that domain, the correlation does not hold universally: there are languages (e.g. Irish) that show V-movement to I (see McCloskey 1996) in the complete absence of inflectional morphology. Consider first the phenomenon of verb second movement (V2), the movement of the finite verb to the second position of the clause (yielding subject-verb inversion), a process that can, e.g., be observed in main clause questions in all Germanic languages, and also in the other main clauses in all Germanic languages except English (9). It is standardly assumed that the finite verb occurs in the Comp position (C0) in these constructions (Thiersch 1978; see Vikner 1995a for a recent discussion and references). (9) a. *Maybe has Peter read this book b. Måske har Peter læst denne bog maybe has Peter read this book c. Vielleicht hat Peter dieses Buch gelesen maybe has Peter this book read

(Danish) (German)

Note that German has a rich system of agreement, whereas there is no agreement at all in Danish. Nevertheless, the finite verb moves to Comp both in Danish (9b) and in German (9c). In other words, the presence or absence of V2 does not predict whether inflectional morphology is rich. Likewise, neither rich (German vs. Italian) nor weak (Danish vs. English) morphology implies that verb second movement applies in all main clauses. The V2 property is unconnected to any morphological property. Turning to a further instance of head movement, consider N-movement within the DP. Since the establishment of the ‘DP hypothesis’ (see e.g. Abney 1987, Horrocks and Stavrou 1987, Szabolcsi 1987), research on the noun phrase opened up questions similar to those asked about the number, the types, and the role of functional projections in CP/IP in the light of Ouhalla’s (1988) and Pollock’s (1989) ‘split INFL’ hypothesis. In particular, there has been extensive discussion of N-movement to D or an intermediate functional category (if there


is one). It has been argued that N-movement is parametrized across languages much like V-movement is. The conceptual motivation for it lies in the general spirit of the parallelism between the nominal and the clausal domain: N-movement parallels V-movement. A powerful hypothesis arguing for the existence of N-movement is based on the placement of the noun with respect to a number of adjectives modifying it. This is tied with specific assumptions about the status of adjectives. These are: a) adjectives are XPs, and b) adjectives appear in strict order DP-internally, which follows from an analysis in which these are placed in the unique specifiers of the various functional projections established independently within DP (see Cinque 1993). On the basis of this hypothesis we again observe that languages fall into two groups if one considers the relative order of the noun and an adjective modifying it (10a vs. 10b). Assuming that adjectives are situated at the same position across languages (Cinque 1993), the order in (10b) has been attributed to the application of head movement of the noun to a functional head inside the DP (10c) (Bernstein 1993, Cinque 1993, among others). Thus, the contrasts in (11) and (12) suggest that Spanish has N-movement, while Greek lacks it. (10) a. Art Adj N b. Art N Adj c. [DP D [FP N [Adj t]]] (11) a.

la muchacha americana the girl American b. *la americana muchacha

(Spanish)

(12) a.

(Greek)

i amerikanida gineka the American woman b. *i gineka amerikanida

However, this state of affairs does not correlate with the richness of noun morphology in the two languages. Greek nouns bear ‘rich’ inflection, as they inflect for number, gender and case (13), while their Romance counterparts inflect only for number and gender (14) (assuming that o and a function as gender markers; but cf. Harris 1991): (13) a.

i amerikanid-a ginek-a the-fem:sg:nom American-fem:sg:nom woman-fem:sg:nom b. ton amerikanid-on ginek-on the-fem:pl:gen American-fem:pl:gen woman-fem:pl:gen


muchach-o muchach-o-s boy boys b. muchach-a muchach-a-s girl girls

(14) a.

Thus, while Spanish nouns are well behaved for (8), this does not hold for Greek nouns. Of the three well-studied cases of head movement, only one, viz. V-to-I-movement, does not falsify (8) in a straightforward way. DP-movement does not seem to fall under (8) either. Movement to SpecIP is independent of morphology. (15)–(16) illustrate raising in English and Icelandic. In both languages the subject DP moves from the lower clause to SpecIP of the matrix clause for reasons of the Extended Projection Principle (EPP) and/or Case checking. However, nouns inflect for Case and number in Icelandic only, while English nouns lack all signs of Case inflection. Likewise, the attracting head Infl has rich morphology in Icelandic only, but not in English. (15) Steve seems to love Madrid (16) Ólafuri virtist [ti lesa margar bækur] Olaf-nom seemed read many books

(Icelandic) (Sigurðsson 1989)

Furthermore, Irish VSO and Chinese SVO clauses (see Aoun and Li 1993, among others) exemplify the absence of NP-movement in the absence of inflection, while Greek VSO orders present a case where no NP-movement to SpecIP takes place in the presence of morphology: although Greek nouns and verbs bear inflection (see above), movement to the subject position is not necessary. Evidence for this comes from a combination of adverbial and participial placement facts in periphrastic constructions. As shown in (17), the order of constituents in these environments in Greek is: auxiliary – aspectual adverb – participle – light manner adverb – subject. In Alexiadou (1997), the relative order between the light manner adverb, which marks the left edge of the VP, and the participle is taken as evidence showing that the participle has moved outside the VP domain. There it is argued that the participle reaches a functional head Asp (for ‘aspect’).2 The subject in (17) follows both the light manner adverb and the participle. From this we can conclude that the subject stays in its VP internal position (see Alexiadou and Anagnostopoulou 1998 for further discussion). (17) an ehi idhi diavasij [VP kala [VP o Petros tj to mathima]]. if has already read well Peter-nom the lesson ‘If Peter has already read the lesson well.’


A¢-movement certainly does not lend support to (8) either. If there is a tendency for wh-movement, it certainly goes in the opposite direction of (8). The overt manifestation of a [+wh] feature of Comp (as we find it in Japanese) normally implies a wh-in-situ strategy (see e.g. Cheng 1991), but the reverse does not hold (no overt wh-movement in the absence of ‘scope marking’ in Chinese or Hindi). There are no overt features of the wh-phrase which one could correlate with the presence or absence of movement. Focus movement suggests a similar pattern. As (18) shows, DPs can be fronted across languages to a Focus/Topic position in the absence of any relevant morphology: (18) a.

Den Ball hat David Beckam nicht getroffen the ball has D. B. not kicked b. AZ ÙJSÁGOT dobtam el the newspaper-acc threw-I away ‘It is the newspaper that I threw away’

(German) (Hungarian)

On the other hand, (19) illustrates that a language that marks focus morphologically may lack focus movement. (19) Raaman ippoozee varunnulluu Rama now-foc come-pres

(Malayalam)

If scrambling is the result of an overt movement process at all (see Boškovic´ and Takahashi 1998, Fanselow 2001, for opposing views), it is certainly not linked to a morphological property of the attracting head. Scrambled DPs often show “rich” Case morphology, but there are languages with a rich Case system without scrambling (Icelandic), and languages with scrambling but lacking rich Case distinctions (Bulgarian). On the basis of the above, one has to conclude that V-to-I movement seems to be the only movement operation taking place due to richness of inflection. Note that such a correlation would more or less be unexpected in a minimalist concept of syntax: movement is triggered by the need to check strong attracting (categorial) features of functional categories, and there is no a priori reason for there being an implication between the strength of an attracting feature and the morphology of the attractor or the attractee. From a minimalist perspective, the two domains should rather be independent of each other. The absence of a clear correlation such (8) is thus expected from a theoretical point of view — but this already renders it quite unlikely that (8) should suddenly become valid when confined to V-to-I movement.


2.2 Problem B: How to characterize rich inflection? Even if there is no general correlation between movement and morphology, a generalization such as (6) (= (8) restricted to V-to-I movement) connecting positional options of the verb and its inflection calls for an explanation if it can be defended. In fact the evidence presented in Haegeman (1997) seems to clearly suggest that whenever the verbal inflection in a language is poor or even absent, no verb movement takes place. Two issues arise here, however. First, even if one confines one’s attention to the Germanic languages and their various dialects, it turns out that defining rich/strong morphology is not straightforward. In principle one would want to identify a necessary and sufficient condition for V-to-I movement. In the literature there have been several such attempts (see Platzack 1988, Platzack and Holmberg 1989, Roberts 1993, Falk 1993, Vikner 1995, Rohrbacher 1999, Koeneman 2000). It has been suggested that the presence of V-movement is linked to the presence of person distinctions in a verbal paradigm (Platzack and Holmberg 1989), while for other researchers number marking seems to play a crucial role (Roberts 1993). Rohrbacher (1994) proposes that a language has V-to-I iff first and second person are differentiated in the verbal paradigm. Whether such proposals are empirically adequate need not concern us for the moment. What is important is that these approaches have in common that they take number/ person distinctions in a verbal paradigm to be the trigger for V-movement to Infl — V movement is not linked to a property of the individual morphemes (see Bobaljik 2000 for a discussion and criticism of such approaches). More generally, the question to what extent syntactic variation is or should be expressed as language specific knowledge about morphology, even where this knowledge is not encoded with individual lexical entries, is not really dealt with satisfactorily (see Snyder 1995, Rohrbacher 1999). As Snyder points out, it is not clear whether there is a principled reason to expect the particular feature combinations distinguished by a given morphological paradigm to have direct consequences for language specific properties of syntax. One could imagine that the implications are completely reverse. Furthermore, as has been pointed out by Vikner (1995a), all these approaches face a problem on the ground that there are languages like Late Old Swedish (Falk 1993: 172f.) or Faroese (Vikner 1995a) in which V-to-I movement is optional. For instance, Late Old Swedish, and the Norwegian dialect of Hallingdalen make number distinctions, but may lack V-movement as illustrated in the following examples.


(20) Late Old Swedish (15th century) sg. kräver ‘demand’ pl. kräva (21) at cristet folk mz truldoms listom forvinna alla pinor that christian people with witchcraft overcome all pains (Falk 1993: 173) (22) Hallingdalen sg. kasta ‘throw’ pl. kastæ (23) at me ikkje kjøpæ bokje that we not buy book-the

Moreover, approaches based on paradigms cannot explain the difference between main verbs and auxiliaries in English. As is well known, auxiliaries move to Infl in English, while main verbs do not. (24) a. *John kisses often Mary b. John has often kissed Mary

3. V-to-I in correlation with the inventory of functional projections Bobaljik and Jonas (1996) explored the possibility that the main trigger for V-movement should not be formulated in terms of counting fine distinctions in the paradigms involved, but rather by concentrating on the inventory of functional elements or projections that a language might make use of. Bobaljik (1995), Thráinsson (1996), Bobaljik and Thráinsson (1998) propose that languages may vary as to whether they have a pre-pollockian unsplit IP or an IP containing an Agreement phrases distinct from Tense (25) (the so called Split Infl Parameter). (25) a.

IP I′ I

VP


b.

AgrP Agr′ Agr

TP T′ T

AgrP Agr′ Agr

VP

In particular, Bobaljik and Thráinsson (1998) argue that there are a series of straightforward consequences to assuming such a parameter, both for the syntax and for the morphology, namely there are more specifier positions in (25b) than in (25a), there are non local relations among Infl-type heads in (25b), and there are more terminal nodes in (25b) than in (25a). On this view, on might expect two VP external subject positions (the specifiers of AgrP and TP) and perhaps a VP external DP object position in languages that have structure (25b), but not in those that have (25a). The authors claim that this is the correct interpretation of multiple subject position and object shift phenomena in Icelandic — as opposed to Mainland Scandinavian and English. Hence Icelandic licenses SpecTP as an intermediate subject position, allows for object shift, and exhibits transitive expletive constructions. On the other hand, languages such as English lack object shift, transitive expletive constructions, and do not license SpecTP as a further subject position. (26) a.

Það klaruðu margar mys [VP alveg ostinn] there finished many mice completely the.cheese b. Það klaruðu margar mys ostinn [VP alveg] there finished many mice the.cheese completely ‘Many mice completely finished the cheese’

TEC OS

(27) *There read a man a book

Furthermore, the authors develop an account of verb movement, according to which feature checking can take place both with and without movement (between a head and its complement). In (25a) feature checking is thus allowed without movement, since Infl and V stand in a head complement relationship.


Economy conditions will then imply that feature checking must not take place via movement. Therefore, there is no V-to-Infl movement in languages with an unsplit Infl. In (25b), however, only the lowest functional head is in a complement relation with the verb, hence verb movement seems required for feature checking with higher heads. Per se, this forces overt or covert movement of the verb, so no predictions concerning a visible displacement of V are made as such. However, Bobaljik and Thráinsson assume that verb movement to Infl is always overt. In recognizing that it is the terminal nodes of syntax that correspond to abstract morphemes which may be the locus of lexical insertion, the Split Infl Parameter predicts that languages with only one inflectional head will be limited to one inflectional affix after the verb stem, while languages with a split Infl may (but need not) allow more. In a language like Icelandic, the verb shows clear marking for both tense and agreement identified by the presence of two separate morphemes for these categories (28). This however is not the case in English, where tense and agreement morphemes are in complementary distribution: (28) 1sg 2sg 3sg 1pl 2pl 3pl

Icelandic kasta ‘throw’ PRESENT PAST kasta kasta-ði kasta-r kasta-ði-r kasta-r kasta-ði köst-um köstu-ðu-m kast-ið köstu-ðu-ð kasta köstu-ðu

English tremble PRESENT PAST tremble trembl-ed tremble trembl-ed tremble-s trembl-ed/*-d-s tremble trembl-ed tremble trembl-ed tremble trembl-ed

Note, however, that it is only Icelandic that shows the full set of properties linked to a split Infl (transitive expletives, object shift, and verb movement). French, Italian (and Catalan) are languages which have distinct Agr and Tense morphology (Belletti 1990: 28): (29) je romp-er-ai I break-fut-1sg

The above reasoning correctly predicts that V-movement takes place in these languages. However, as is well known, these languages lack the properties associated with Icelandic, i.e. transitive expletives and object shift (30). Thus the correlation here is weakened. (30) *il achetait un enfant un livre there has.bought a child a book


If only languages with a split Infl allow more than one inflectional morpheme attached to V, then the prediction is made that V-movement should take place in such language. The distinction should not in principle be taken to make reference only to Tense and agreement morphemes, but also to other inflectional categories. For instance, in Greek there is no distinct agreement and tense morphology, but V-movement takes place (Rivero 1994; (32)). Note, however, that the language marks Aspect morphologically. In (31), e- is a phonetic vowel, and -s is the perfective Aspect marker. On the other hand, we do not find any evidence for object shift of the type found in Icelandic (see Alexiadou 1999), and there are no transitive expletive constructions in Greek. (31) Greek rihno ‘throw’ PRESENT PAST IMPERFECTIVE rihn-o e-rihn-a rihn-is e-rihn-es rihn-i e-rihn-e rihn-ume rihn-ame rihn-ete rihn-ate rihn-un rihn-ane

PAST PERFECTIVE e-rik-s-a e-rik-s-es e-risk-s-e rik-s-ame rik-s-ate rik-s-ane

(32) riksame amesos ti bala threw-perf:1pl immediately the ball-acc ‘We threw the ball immediately’

The empirical force of the Split-Infl-approach is thus weakened, since its predictive success is confined to V-movement (see also Alexiadou and Anagnostopoulou 2001b); note though Bobaljik (2000) is certainly correct in pointing out that, of the properties considered, only V-movement is forced by a split Infl. It should be added, however, that the evidence for V-to-I movement in subordinate clauses is not uncontroversial for German or Dutch (see Haider 1993, Koopman 1995, Zwart 1997: 52 for counterarguments). It may thus very well be the case that there ARE languages with distinct Tense and Agreement morphology, transitive expletive constructions, and object shift that disallow verb movement. Moreover, according to Vikner (1995a), the proposal that distinct Tense/Agreement morphology implies overt verb movement is incompatible with Faroese. In Faroese past tense, the verb clearly inflects for tense and agreement (33), but no V-movement takes place in one of the two dialects of the language, while in the other dialect, verb movement is only optional:3

On the correlation between morphology and syntax

(33) Faroese kasta ‘throw’ (Bobaljik 2000) PAST kasta-ð-i kasta-ð-i kasta-ð-i kasta-ð-u kasta-ð-u kasta-ð-u

It is also far from clear that the Split-Infl idea can be successfully extended to other instances of head movement. Thus, Infl moves to Comp in (34) in spite of the fact that Infl is the head of the complement of Comp, so that the features of Comp selecting Infl would have to be checked without movement in an approach that extends feature checking to the head-complement relation. The strict correlation between the distance of two heads H and K and the application of overt head movement would therefore have to be confined to the interaction of V and Infl.4 (34) what [Comp will] she [Infl _ —] say?

For obvious reasons, the Split-Infl idea finally fails to be compatible with the assumption that negative expressions and adverbs that appear higher than VP but lower that IP are heads or specifiers of headed projections (see e.g. Alexiadou 1994, 1997, Cinque 1999). In (35), VP is not the complement of Infl, so that the establishment of a checking relation between Infl and V seems to presuppose movement even if one grants that heads may check features of their complements. (35) [IP… Infl … [XP NEG/Adv [VP V …

To the extent that negation and adverbs are inherently related to head positions in the sense just specified, the Split-Infl idea faces a serious problem.

4. A defensible implicational generalization Some of the evidence presented above already suggests that the connection between “rich morphology” and overt movement is not a bi-conditional one. In fact, Vikner (1994) has come to the conclusion that only an implicational relation — the one indicated in (36) — is defensible.5

231


(36) The verb may only stay in V if there is no inflectional morphology.

Such a formulation linking rich inflection to V movement but not vice versa comes very close to what one finds empirically (see below). Within the Germanic language group, richly inflected Icelandic shows V-to-I-movement — while the generalization cannot be strengthened to a bi-conditional because of the notorious Kronoby dialect of Swedish, which displays V-to-I-movement in the absence of verbal agreement inflection: (37) he va bra et an tsöfft int bootsen it was good that he bought not book-the

Second, V-to-I-movement takes place in Irish, although verbs do not bear inflectional morphemes (see McCloskey 1996: 269): (38) a.

Chuala Róise go minic roimhe an t-amhrán sin heard Róise often before-it that song ‘Róise had often heard that song before’ b. Gheall sé go bhfillfeadh sé ar an bhaile promised he that return he on home ‘He promised that he would return home’

Likewise, there are some SVO creole languages (see Baptista 1999) that show verb movement in the absence of agreement or any inflectional morphology. Therefore, in line with Vikner’s (1994) cautious formulation, one must conclude that languages without rich verbal morphology may or may not show verb movement to Infl. The absence of rich inflection does not predict anything with respect to verb movement. Likewise, infinitival verbs can move to Infl, too, although they lack inflection. As shown in (39), infinitivals do move in Italian, but not in French, although in both languages they lack inflectional features: (39) a. *ne lire pas le livre neg read not the book b. ne pas lire le livre c. non leggere più il libro neg read more the book d. *non più leggere il libro

Vikner’s generalization still implies, however, that richly inflected verbs must not stay in situ. For the Germanic and Romance SVO languages, this implication indeed seems to hold (at least, there is no known counterexample), but —


as we have already pointed out — it is not at all obvious that richly inflecting SOV language like German fall in line with (36). Presumably, there are also SVO-languages that do not support (36). Noonan (1992: 122) describes the situation in Lango, where V-to-I movement appears to fail to take place in the presence of rich agreement which is pre-fixal: adverbial adjuncts are placed either at the beginning or end of a sentence (cf. (40)), and “the negative particles (…) are normally placed between the subject and the verb”. (40) lóc`6 òg`7ò gwôk àwó’ró man 3sg-kick-perf dog yesterday ‘The man kicked the dog yesterday.’

Therefore, (41) seems to be the only generalization that can be defended (as before, to the extent that German has V-to-I movement in embedded clauses). (41) Suffixal rich inflection implies V-to-I-movement.

5. A historical perspective One now has to identify the factor that is responsible for the validity of (41). In the light of the present discussion it should be obvious that (41) is an unlikely candidate for an independent principle of Universal Grammar. Note that it certainly need not be the case that all properties characterizing natural languages in general must be due to such independent principles of UG. What is a possible grammar is also influenced by factors such as processability or learnability: natural languages must be produced and understood by their speakers, and they must be acquired by children. Furthermore, natural language grammars are not invented and designed by language engineers, and put to use only later — rather, they are the products of diachronic processes, the results of a series of steps of language change. It is at least conceivable that the mechanisms of language change are such that certain combinations of properties simply cannot arise as the product of a natural diachronic process. Such a scenario is rather likely if — as seems to be generally accepted nowadays — syntactic change is the result of shifts in the analysis of constructions of the adult language by language acquiring children, in such a way that the surface outputs of the two grammars should differ minimally only at first. This restricts the nature of a possible syntactic change considerably, so that certain grammatical systems, although compatible with UG, can never be observed among the


world’s languages, because they cannot arise naturally. It is our claim that the V-to-I-movement generalization exemplifies a case of such an array of properties that cannot arise in natural language change processes. Our argument involves three steps. First, it needs to be established how strong systems of suffixal agreement morphology come into being; second, the emergence of SVO order must be investigated. Finally, reasons for the loss of V-to-I movement must be identified, with a discussion of possible relations to the impoverishment of verbal inflection. 5.1 The direct path to an SVO system with rich inflection It seems to have been established that suffixal inflectional systems arise via the cliticization of a subject pronoun clitic (see Givón 1976, Corbett 1995). The sequence V+clitic is reinterpreted as V+AGR. For there to be a chance for the suffixal re-interpretation of such a clitic, the verb must appear in front of the subject pronoun, either in Infl or in C. (42) a. [Infl verb] [vP subject …] b. [Comp verb] [IP subject …]

Thus, strong suffixal inflection can arise only if there is verb movement. If V does not move to I or C, the subject pronoun simply is not on the correct side of the verb to be reanalyzed as a suffixal inflection by language-acquiring children. At the point of the emergence of rich inflection, at least, it is clear why a correlation such as (41) holds — not because of a principle of UG, but because rich suffixal agreement could not arise in a language without verb movement. No similar necessity for the verb being placed outside VP exists for a process of building up prefixal agreement, however. Therefore, languages like Lango which rich inflection but lacking verb movement pose no problem at all. Do we have synchronic evidence for the view that rich inflection arises from a reinterpretation of subject clitics? We will present data from two languages that exemplify the transition from a clitic pronoun to a verbal inflection. The first case to be considered here is Irish. The language exhibits complementarity of what might appear to be person-number agreement on the verb, and the appearance of a phonologically expressed subject (pronominal or DP). (43) a.

D’eirigh Ciaran rose-3sg Ciaran b. D’eirigh na girseachai rose-3sg the girls

(Irish)


c.

D’eirigh me rose-3sg I d. d’eiriodar rose-3pl e. *chuirfine mé put(cond).1sg I

McCloskey and Hale (1984) show that (43d) can be the result of the incorporation of a subject pronoun into the adjacent verb — it does not involve agreement morphology in a strict sense. The status of the verbal ending as an incorporated subject is still visible, as it is, e.g., possible to conjoin the inflectional subject with a lexically specified NP (McCloskey and Hale 1984): (44) dá mbeinn-se agus tusa ann if be:cond.1sg-contr and you there ‘if you and I were there’

The complementarity of the clitic subject pronoun and full lexical NPs, combined with further evidence like the one just presented, makes the non-agreement status of the phonological material attached to the verb still quite evident for the language acquiring child. A somewhat different case in point comes from various varieties of Arabic, where left dislocated subjects involving clitic doubling develop into A-subjects (data from Aoun, Benmamoun and Sportiche 1994). In Standard Arabic verbs give the impression of exhibiting full (person, number and gender) agreement with the subject in SVO orders at first glance, but they show partial/deficient ‘agreement’ (i.e. agreement only in gender and number) in VSO orders: (45) a.

Naama l-‘awlaad-u slept-3S the.children-nom ‘The children slept’ b. ‘al-‘awlad-u naamuu the.children-nom slept-pl

Ouhalla (1994) proposes that (45b) involves a left dislocated subject with clitic doubling involving a resumptive pronoun rather than simply agreement. For example, the fact that one cannot find indefinite NPs in preverbal position (see (46)) strongly supports this view.


(46) a. *‘ahad-un raqas-a maca-ha one-nom danced-3sg with.her ‘Someone danced with her’ b. raqas-a ‘ahad-un maca-ha danced someone-nom with.her

The canonical Standard Arabic clause pattern and the agreement system in a strict sense is thus exemplified in (45a) only. Notice, however, that cliticized subject pronouns and full lexical NPs are not complementarily distributed in Standard Arabic, so that there is a certain likelihood that the input pattern (47a) will be reinterpreted as an agreement pattern (47b) by language-acquiring children. (47) a. [FP S [IP V+clitic …]] b. [IP S V+Agr …]

This indeed seems to have happened in Moroccan Arabic, where both SVO and VSO orders show full agreement between the verb and the subject. (48) a. *N‘as le-wlaad slept.3sg the.children b. na‘su le-wlaad slept.3pl the.children c. le-wlaad na‘su the.children slept.3pl

(Moroccan Arabic)

Moroccan Arabic also lost the difference in distribution of definite and indefinite NPs in preverbal position — which shows no restrictions concerning specificity or definiteness. In other words, the preverbal position has acquired the status of a standard subject position. The former clitic pronoun has been reinterpreted as an agreement morpheme. (49) hetta wahed maza even one neg-came.3msg ‘Anyone didn’t come.’

We may thus take Irish, Standard Arabic, and Moroccan Arabic to represent three stages in the transition from a grammar with subject clitics to a grammar with agreement. First, the clitic status of the morpheme M is still obvious because of the complementarity of M with a full lexical subject NP. “Later”, clitic doubling is introduced, which blurs the evidence — simple distributional facts no long rule out an agreement interpretation, but — as in Standard Arabic — there may be other evidence E like restrictions for indefinites that make the


clitic status of M still learnable. When E is lost, there is no reason for the children to not go for the (simpler) analysis in which M is an agreement morpheme. 5.2 The path via V2 systems We have just seen how suffixal agreement morphology arises, and why this process is intimately linked to the presence of verb movement to a pre-subject position. In its initial stage, the suffixal agreement system can only occur in a language in which the verb moves in front of the subject. For SVO systems that emerge because of a reinterpretation of left dislocated subjects of VSO clauses, the link between rich morphology and V-to-I movement thus seems established. However, the verb can also be placed into a pre-subject position by moving to Comp (the verb-second option), and while V-to-C movement seems to always proceed via Infl, we need to show why the diachronic shift from a V2-grammar to an SVO-grammar always leads to languages with V-to-I movement. Note that the transition from V2 to SVO may take place at a time when verbal inflection is rich. If we want to capture the implication of verb movement by rich verbal morphology, we must show that no language without V-to-I movement but with rich inflection can arise in this context. This point is particularly relevant for us, since Germanic and Romance languages (which constitute the major bulk of evidence for the correlation under discussion) originated from an Indoeuropean (IE) origin, which probably was a richly inflected verb-second language (with a verb-final base order), as Krisch (1997) deduces from the fact that Sanskrit, Latin, Hittite, and other early IE languages show some V2 effects (apparently more reminiscent of what one presently finds in Kashmiri — see Bhatt 1999 — than German). Romance and some Germanic languages then made the transition from V2 to SVO. The transition from V2 systems to SVO-systems primarily seems to involve the reanalysis of a string involving a clause-initial subject followed by a verb. During the V2 stage, (50a) is the correct analysis of such a string; when the language shifts to the SVO type, (50b) becomes relevant. (50) a. [CP subject [Comp verb] …. {adverb, negation} …] b. [IP subject [Infl verb] …. {adverb, negation} …]

Note that the reinterpretation of (50a) as (50b) leaves the position of adverbs/ negation relative to the finite verb constant: when a phonetic sequence subjectverb-X-adverb-Y is no longer interpreted as a CP by children, the string verb …


adverb/negation constitutes perfect evidence for an SVO language with V-to-I movement (and against a grammar without V-to-I movement). There are many reasons for why such a reanalysis may take place (see e.g. Kroch 2000). An increase of the number of left-dislocation structures at the cost of the simple topicalization of objects may, e.g., blur the evidence for the V2 analysis of a subject-verb string to such an extent that language acquiring children start preferring (50b) (this may have happened during the history of French). Whatever the precise individual reason may be in each case that such a transition is made, the simple reanalysis in (50) implies the greatest possible extent of harmony in terms of the surface strings between the old grammar and the new one (the new one produces a subset of the strings allowed by the old one, and nothing more). Therefore, there can be no strong positive evidence in the linguistic input that might encourage those children who made the transition to (50b) to revise their grammars later on the basis of some adult sentence. The development in (50) is thus not an unlikely one. In contrast, a transition from a V2 grammar to an SVO grammar without movement to Infl would imply that the surface strings cannot remain constant. The output (51) of the new grammar would not be acceptable to the adults’ grammar, and the adults’ output subject-verb-adverb … would be ruled ungrammatical by the grammar of the children. It is highly unlikely that a child hearing verb-adverb order only in his/her input would stick to a grammar licensing (51) in the light of the evidence he/she is constantly confronted with. (51) [IP subject adverb [Infl verb] …]

In other words, a natural process of language change in the process of language acquisition by children can transform a V2 system into an SVO-system with V-to-Infl movement only. No richly inflected verb can escape from movement to Infl, in this kind of language change at least. 5.3 The Loss of V-to-I and the Loss of Inflection The correlation presently under consideration that rich inflection implies V-to-I movement has been derived for the initial stages of SVO systems that arise from VSO or V/2 grammars in a diachronic process.6 A simple reflection reveals that a later loss of V-to-Infl movement is not a very likely incident. Verb movement is not lost “spontaneously”, i.e., a child confronted with verb adverb-order will not postulate a grammar lacking verb movement without reasons. A loss of verb movement will occur only if the grammar allows an


additional movement process that changes the order of the two elements frequently enough, so that the evidence for V-to-I movement becomes less and less transparent. If standard movement is leftward bounded, the pertinent movement operation can be adverb preposing (targeting a position immediately in front of the verb) only. Stylistic Fronting (SF) is such an operation that is grammatical in Icelandic, and was so in Mainland Scandinavian, too, by which adverbs may end up in front of the verb. As pointed out, e.g., by Holmberg (2000), SF targets the closest element to the right of the finite verb (which could be a verbal particle, a negation, an adverb, or a PP), and places it into the first position of the clause. SF applies only when the subject position is unfilled. On the one hand, this is the case in impersonal constructions (as in (52) from Old Swedish), but there are also constructions in which the verb appears in third position (as in (53)). The latter type of data always involve pronominal subjects. As Platzack (1988) argues, they are due to the fact that the weak pronoun has cliticized to the Comp position, which leaves the subject slot empty, so that SF can and must apply. (52) att icke haffver feladh på manheet och modh iblandh vårt folk that not has failed on manhood and courage among our people (Falk 1993: 227) (53) œn han eigh bannaþe ihesus namne if he not cursed Jesus’ name

V3 with pronominal subject (Falk 1993:188)

Data such as (53) are, we believe, the key to an understanding of the loss of V-to-I movement. In the target grammar, the proper analysis of (53) has the finite verb bannaþe in Infl, but it is preceded by negative operator eigh due to SF, which itself was triggered by the cliticization of the subject pronoun onto Comp. But the string also allows another and indeed simpler analysis: the pronoun is in the subject position, and the adverb precedes the verb because the latter has not left VP. This is nothing but the loss of V-to-I movement. As Falk (1993: 184) observes, the latter process (which reinterprets the results of SF) coincides temporarily with the loss of SF in Modern Swedish. (54) a. [CP [Comp œn han] [IP eigh [I¢ bannaþe [VP …]]]] b. [CP œn [IP han [VP eigh bannaþe …]]]

(old) (new)

By the reinterpretation of structures with V in I preceded by material that has undergone SF, as simple clauses with V in situ, verb movement can get lost in the history of a language — and apparently has gone lost in exactly that way in Mainland Scandinavian. There are other conceivable ways by which an adverb


may get directly in front of Infl (e.g., adjunction to IP in a structure with an inverted subject), but the reflection on SF already suggests that the loss of V-to-I is linked to the fulfillment of a number of non-trivial presuppositions so that the process is probably a rare one. Having identified a likely reason for the loss of verb movement, we now have to deal with the question of why this diachronic process apparently affects languages without a rich inflectional system only. We have no definitive answer to this question, but the following reflection may be relevant. Note that (53) suggests that the condition triggering SF is a prosodic one: exactly one prosodic word/phrase must precede the finite verb. When the subject is a clitic, this condition is not met, and SF applies. From a purely syntactic point of view, however, SpecIP might be left empty, as (52) suggests: SpecIP is not filled by a DP with a phonetic matrix. Furthermore, SF can meet the requirements of a prosodically defined finite-second constraint only, it cannot normally satisfy the needs of a syntactic principle (requiring there to be a DP in the specifier position of IP), since the category moved in front of Infl is usually not a DP. If a DP is needed, an overt expletive would have to be inserted. We therefore suspect that a language with SF respects a prosodically defined Infl-second constraint, but it crucially allows the proper subject position to be phonetically unfilled, i.e., the language is of the pro-drop-type. But pro-drop languages normally have a rich inflectional system. Thus, when inflection gets weak, the pro-drop property gets lost, too, so that SF is no longer a proper way of dealing with unfilled SpecIP positions in front of the verb. In such a situation, SF ceases to be a grammatical process, so that adv verb order in the input can no longer be attributed to SF by children. It then has to be the base generated order. 5.4 Loose ends The scenario described in the preceding subsection is a possible explanation for the apparent impossibility of V-movement getting lost when inflection is still strong. There may be other possible causes, and there is no evidence to decide this issue at present. Likewise, V movement might get lost in different ways from the ones considered above, and for each of them, it would have to be shown why it fails to be viable when inflection is strong. Another topic that will have to be dealt with is the array of facts in English. In contrast to main verbs, English auxiliaries can appear in Inf (55), which suggest that the blocking of V-to-I movement is more intimately connected to properties of lexical items than is predicted by the present approaches of verb movement.


Likewise, adverbs may intervene between the verb and an object provided the latter is not a DP, as illustrated in (56).7 (55) John has not bought the book (56) John laughs often at Mary

6. Conclusion In this article, we have argued that there is no synchronic UG principle that establishes a link between inflection and verb movement. To the extent that such a link can be defended empirically (rich inflection is incompatible with there being no verb movement), we have shown that it reduces to laws of the diachronic development of languages: systems of rich inflection cannot arise without there being verb movement in the language, and, arguably, verb movement cannot get lost as long as rich inflection exists. We can derive these laws from simple assumptions concerning syntactic change. Grammars violating them cannot arise in a natural process of language change.

Notes * The present article is a slightly revised and enlarged version of a talk we presented at the Germanic Syntax Workshop in Groningen, May 2000. For comments and helpful suggestions, we thank Karin Donhauser, Caroline Féry, Susann Fischer, Tonjes Veenstra, and Sten Vikner. The research reported here was partially supported by DFG grants INK 12/B1 to the Innovationskolleg “Formale Modelle Kognitiver Komplexität”, and to the Forschergruppe “Konfligierende Regeln”, both at the University of Potsdam. 1. Vikner (1995a: 149) cites the following Faroese example from Barnes (1987): (i) ??at studentarnir skuldu skrivliga svara spurninginum that students-the should in.writing answer question-the According to his informants (i) might be acceptable in a rather formal style. See below for a discussion. 2. In Cinque (1999), it is argued that participles raise in all Romance languages as well. 3. Bobaljik and Thráinsson (1998) actually claim that, although there is a vowel alternation in the past tense, this alternation is an instance of vowel change, and not a sequence of two affixes. Hence the language allows an analysis with unsplit Infl, and can therefore lack V-movement. If we disregard the optionality problem discussed above, this approach is certainly able to explain away this problem. Note here that Jonas (2002) makes a distinction between two dialects of Faroese. Both have V-movement, but they differ with respect to the


height of negation attachment. In the dialect where negation is low V-to-I movement takes place. In fact the dialect that has less morphology seems to have more movement. 4. One might try to counter this line of reasoning by placing an additional functional head between Comp and Infl in (34), but — in addition to the issue of its independent motivation — one question is why the presence of this head does not force Infl-to-Comp movement in subordinate clauses as well. 5. In later work, Vikner (1995b) proposes the following generalization: (i) V-to I iff person morphology is found in all tenses. This faces a problem with the data to be discussed below. 6. It is also conceivable that an S-Adv-O-V language develops into an S-Adv-V-O system by the frequent extraposition of all objects and PP adjuncts, so that the input could be reinterpreted as an SVO system without V-to-I movement by children. In principle, the transition could be made at any stage of the inflectional paradigm, so that a grammar violating the correlation we discuss might arise in this way. The non-existence of such grammatical systems must thus be due to the fact that this kind of transition from SOV to SVO is excluded. A reason could be that the headedness of VP is also (and perhaps primarily) encoded in the order of verbs and auxiliaries — for which it is hard to see how it might be changed by extraposition. 7. Several options could be suggested. For instance, in (56) the adverb could be argued to adjoin to VP and not to vP. Moreover, in (55) it could be argued that the auxiliary is a light element that raises obligatorily (see Roberts 1997).

Language Variation and Change

Observations on the loss of Verb Second in the history of English* Eric Haeberli University of Reading

1.

Introduction

As often observed in the literature (cf. e.g. van Kemenade 1987, Stockwell 1984), Old English (OE) has word order patterns which are reminiscent of the Verb Second (V2) phenomenon found in the modern Germanic languages. In particular, fronting of some element to clause-initial position often leads to subject-verb inversion and, hence, to the occurrence of the finite verb in second position. This word order property is illustrated in (1).1 [ðæs halgan weres stefne] gehyrde Theoprobus þa [the holy man’s voice heard Theoprobus then ‘Theoprobus then heard the holy man’s voice.’ (Gregory H, 140.17.140.30) b. [þinre meder] geheolp þin halga geleafa [your mother helped your holy faith ‘Your holy faith helped your mother.’ (Ælfric’s Lives of Saints, I, 212.28) c. And [egeslice] spæc Gregorius be ðam… and [sternly spoke Gregorius about that ‘And Gregorius spoke sternly about that …’ (Wulfstan, 202.46) d. [On þæm dagum] wæs Alexander geboren on Crecum… [in those days was Alexander born in Greece ‘At that time, Alexander was born in Greece …’ (Orosius, 104.21)

(1) a.

In (1a) an accusative object is fronted to initial position, in (1b) a dative object, in (1c) a manner adverb and in (1d) a temporal PP adjunct. In all these cases,

246 Eric Haeberli

the fronting of a consituent goes together with subject-verb inversion as in the modern Germanic V2 languages. In Modern English, the corresponding word orders would be ungrammatical. V2 patterns therefore seem to have been lost in the history of English and this loss is an issue which has received considerable attention in the literature (cf. e.g. Fuss 1998, van Kemenade 1987, Kroch et al. 2000, Lightfoot 1995, 1997, Platzack 1995, Roberts 1993, Stockwell 1984). However, the discussions in the literature raise two main problems. First of all, detailed data describing the change are rare. And secondly, no satisfactory explanation has been found so far as to why this change occurred. The main goal of this article is to make a contribution to the first point (for the second issue, cf. Haeberli 2002). More particularly, I will discuss the status of subject-verb inversion in various prose texts from the Old and Middle English periods in order to provide a general picture of how V2 was lost in the history of English. The article is organized as follows. Section 2 presents some general aspects of the syntax of OE which will allow us to determine exactly what was lost in the history of English with respect to the syntax of V2. Section 3 then deals with subject-verb inversion in OE in more detail. In Section 4, the status of V2 in Middle English (ME) is discussed. Finally, Section 5 concludes the article.

2. What was lost? Before we start our discussion of the loss of V2 in English, some remarks concerning the syntax of the earliest attested period of English, i.e. Old English, are necessary so that we can determine exactly what was lost in the course of the history of English. In the literature on the syntax of OE, two V2 contexts have generally been distinguished (cf. e.g. van Kemenade 1987, Pintzuk 1991): (i) V2 in the context of operator fronting (i.e. with wh-elements, negation but also some adverbs which are not typically operators such as þa (‘then’), þonne (‘then’) and nu (‘now’)); (ii) V2 in the context of non-operator fronting. This distinction is based on the different behavior of pronominal subjects in the two contexts and in the recent literature context (i) has generally been analyzed as involving V-movement to C whereas context (ii) has been analyzed as involving V-movement to an inflectional head below C (cf. e.g. Cardinaletti and Roberts 1991, Haeberli 2000, Hulk and van Kemenade 1997, Kroch and Taylor 1997, Pintzuk 1991, 1993). What is important for our purposes here is that in context (i) we still can

Observations on the loss of Verb Second in the history of English 247

find what has been referred to as “residual V2” in Modern English (cf. e.g. When will John leave?). Fronting of interrogative or negative constituents leads to fronting of a verbal element to the left of the subject. The main difference between OE and Modern English is that the fronted verbal element cannot be a main verb any more in Modern English, but this restriction is the consequence of a more general development affecting the movement properties of main verbs (cf. e.g. Kroch 1989) rather than a substantial change concerning the syntax of V2. The crucial context for changes in the V2 syntax of English is therefore context (ii) in which a non-operator is fronted, as illustrated in the examples in (1) above. The Modern English equivalents of these examples would be ungrammatical even if an auxiliary followed the fronted non-operator. Thus, what was lost in the history of English is the frequent occurrence of V2 patterns when a non-operator is fronted.2 However, not all cases in which a non-operator is fronted are relevant for our purposes. As often discussed in the literature (cf. e.g. van Kemenade 1987, Pintzuk 1991), subject-verb inversion generally does not occur when the subject is pronominal. This is illustrated in (2). [þæt] þu meaht swiðe sweotole ongitan (Boethius, 88.14) [that you can very easily understand b. and [mid gelæredre handa] he swang þone top mid swa micelre and [with skilful hand he swang the top with such great swiftnesse (Apollonius, 20.13.22) swiftness

(2) a.

In (2a) an object is fronted and in (2b) an adjunct PP occupies the clause-initial position, but in both cases no subject-verb inversion occurs. The word orders in (2) therefore correspond to surface word orders found in Modern English. Given the systematic lack of subject-verb inversion with pronominal subjects in OE already, clauses with subject pronouns do not undergo any substantial changes in the history of English (but cf. Section 4 below for some additional observations on this point). Hence, the diachronic developments which are of interest to us only concern clauses with non-pronominal, i.e. full DP, subjects. Finally, a general point concerning the notion of V2 should be made here. In the examples we have considered so far (cf. (1)), subject-verb inversion leads to V2 orders because only one constituent has been moved to the beginning of the clause. However, it is not the case that subject-verb inversion always leads to V2 in OE. Instead, two (or more) constituents can sometimes precede the finite verb. Two illustrations are given in (3).

248 Eric Haeberli

[On þæm dagum] [on Tracia þæm londe] wæron twegen cyningas [in those days [in Thrace the land were two kings ymb þæt rice winnende (Orosius, 114.15) about that kingdom fighting ‘In those days, in Thrace, two kings were quarrelling about that kingdom.’ b. [Ðysne yrming] [æfter his forðsiðe] wurðodon þa hæðenan eac [this poor-wretch [after his decease worshiped the heathens also for healicne god (Wulfstan, 223.58) instead-of high God ‘After his decease, the heathens also worshiped this poor wretch instead of God.’

(3) a.

Although such examples are not very frequent and V2 is the standard pattern in subject-verb inversion contexts, they nevertheless suggest that the V2 syntax is not very rigid in OE.3 This observation will be confirmed in the following section. The term “loss of V2”, as generally used in the literature, therefore might be slightly misleading in the sense that a V2 syntax as we know it from the modern Germanic languages never existed in the attested periods of English (cf. also the patterns in example 2 and Section 3 below). It therefore seems more adequate to describe the developments in the history of English as the loss of certain subject-verb inversion patterns and the concomitant loss of V2 orders. In the remainder of this article, I will therefore use the more general term of ‘subject-verb inversion’ rather than V2, thereby implying that, although most of these constructions are at the same time V2 structures, they may also occasionally involve the presence of two non-subjects to the left of the finite verb. In conclusion, the main issue that arises with respect to the loss of V2 in the history of English is the question how subject-verb inversion was lost in clauses containing a fronted non-operator and a non-pronominal subject. In the remaining sections I will therefore focus on the status of such constructions throughout the history of English and more particularly during the OE and ME periods.

3. Old English As observed in the previous section, subject-verb inversion in contexts of nonoperator fronting in OE generally occurs only if the subject is a full DP. However, even if the subject meets this condition, subject-verb inversion is by no means categorical. This is shown in (4).


[ðone] Denisca leoda lufiað swyðost (Wulfstan, 223.54) [that Danish people love most ‘The Danish people love that one most’ b. [Eallum frioum monnum] ðas dagas sien forgifene (Laws 2, 78.43) [all free persons these days be given ‘These days should be given to every free person’ c. ge [eac] [hwilum] þa yflan bioð ungerade betwuh him selfum and [also [sometimes the evil are discordant between them selves ‘And sometimes the evil people are also discordant among themselves’ (Boethius, 134.26) d. [æfter þan] þæt lond wearð nemned Natan leaga [after that that land was named Natan lea ‘After him, that land was called Netley.’ (Chronicle A, 14.508.1)

(4) a.

Although the occurrence of patterns like (4) has sometimes been observed in the literature (cf. e.g. Allen 1990: 150, Bean 1983: 62/81, Eythórsson 1996: 114f, Haeberli and Haegeman 1995: 85, Kiparsky 1995: 145, Kroch and Taylor 1997: 304), no attention has generally been paid to them in the theoretical analyses of OE word order (cf. e.g. van Kemenade 1987). However, as already shown by Koopman (1998), the word order patterns shown in (4) do occur quite frequently. Table 1 below provides some quantitative data concerning subject-verb inversion in clauses containing a fronted non-operator and a nonpronominal subject in ten text samples taken from Pintzuk et al. (2000). The figures in Table 1 show that the frequencies of non-inversion are by no means negligible. Non-inversion occurs in 15.2% (Apollonius) to 59.5% (Gregory C) of all main clauses in which a non-operator is fronted and the subject is a full DP. Some additional observations should be made with respect to the data in Table 1. First of all, it could be argued that clauses which lack subject-verb inversion are actually V-final main clauses. Although V-final orders are most frequent in subordinate clauses, they do occur in main clauses (cf. e.g. Koopman 1995) and it could be assumed then that such clauses remain V-final (and hence lack inversion) even if a non-operator is fronted.4 However, the V-final option does not provide a likely explanation for the frequent occurrence of non-inversion in the data in Table 1. As has often been pointed out, V-final main clauses are particularly frequent in coordinate clauses (cf. e.g. Mitchell 1985: 710ff.), and the figures in Table 1 include such clauses (cf. example 4c). We therefore would expect that, since coordinate clauses favor V-final order, statistical data based on non-coordinated main clauses only should show

250 Eric Haeberli

Table 1.Main clauses with non-operators preceding non-pronominal subjects in samples of ten OE texts5 Text (date)

inversion (XP-V-SU)

no inversion (XP-SU-V)

% uninverted

Bede(950)

43 26 28 23 67

9 13 5 15 20

17.3% 33.3% 15.2% 39.5% 23.0%

Total

466

188

28.7%

Total (before 950)/ Without Chronicle A

279/ 127

126/ 94

31.1%/ 42.5%6

Total (after 950)

187

62

24.9%

significantly lower frequencies of non-inversion if V-final order indeed was a crucial factor favoring the lack of subject-verb inversion. However, this expectation is not borne out. Once we exclude all main clauses introduced by the conjunction ‘and’ from the data in Table 1, the frequencies of non-inversion for the individual texts change only slightly and the change can be either a decrease or an increase.7 The figures from all OE texts taken together give a frequency of non-inversion of 30.6% in non-coordinated clauses (vs. 28.7% in Table 1). In the texts before 950, we obtain a frequency of 34.3% or 40.3% without the Chronicle (vs. 31.1% and 42.5% respectively in Table 1). Finally in the later texts, the frequency of non-inversion is 26.1% (vs. 24.9% in Table 1).8 Thus, coordination and, hence, V-final orders do not seem to be crucial for the occurrence of non-inversion in OE. This conclusion is confirmed by some quantitative data provided by Koopman (1995: 139) and Pintzuk (1993: 22, fn. 22). These authors estimate that the frequency of V-final orders in non-coordinated main clauses is not more than 6% of all main clauses. As just shown, the frequency of non-inversion in non-coordinated main clauses is around 30% and thus considerably higher than 6%. This clear contrast would be unexpected if non-inversion was

Observations on the loss of Verb Second in the history of English

closely related to V-final orders. In conclusion, both the irrelevance of coordination and the contrast in frequency between V-final orders and non-inversion in non-coordinated clauses suggest that the frequent lack of inversion in nonoperator fronting contexts is not simply a consequence of the availability of V-final orders in OE. Instead, there seems to be genuine optionality as to whether or not subject-verb inversion takes place when a non-operator is fronted, with inversion being the more frequently used option. Let us now turn to a different issue that Table 1 raises. Given that subjectverb inversion in the contexts considered here is a word order option which was lost in the history of English, we may wonder whether this loss was already under way in the OE period, i.e. whether there was a decrease of inversion. As a matter of fact, the data in Table 1 suggest that this was not the case. If anything happened during the OE period, it rather seems to be a strengthening than a weakening of subject-verb inversion constructions. Thus, the average frequency of non-inversion in the earlier texts (before 950) is around 10% higher than in the later texts (after 950) (cf. Table 1 and footnotes 5 and 8). This looks like a development towards a more rigid V2 grammar during the OE period. However, such a conclusion will have to be confirmed by further research based on larger text samples and a larger number of texts. Finally, Table 1 raises an additional issue. As observed above, the data in Table 1 suggest that there is optionality as to whether fronting of a nonoperator leads to subject-verb inversion or not. However, inversion is still the clear majority pattern in OE. The question that arises then is whether any factors can be identified which determine the occurrence of non-inversion. At first sight, it is not clear that the answer to this question is positive (cf. also Koopman 1998: 145ff.). For example when we consider the type of fronted element in non-inversion constructions, we can observe that non-inversion occurs most frequently with fronted adjuncts (adverbs, PPs). However, fronted arguments also regularly do not give rise to subject-verb inversion. In the text samples studied here, 130 clauses contain a fronted object and in 22 cases (16.9%) non-inversion occurs (cf. also Koopman 1998: 136 for additional data).9 As for the type of subject involved in non-inversion constructions, we can observe that it tends to be relatively “light”. Thus, in 55 out of the 188 noninversion clauses the subject consists of a single word (generally a name) and in 68 clauses the subject consists of two words (i.e. 64.9% of the subjects are either one or two word subjects.). Yet, the lack of this property again does not mean that non-inversion is impossible. Heavier subjects also can occur in such constructions, as the examples in (5) show.

251

252 Eric Haeberli

[þa] [æfter þære mæssan] seo modor and seo dohtor astrehton [then [after the mass the mother and the daughter prostrated hi on gebedum… (Ælfric’s Lives of Saints, I, 210.20) themselves in prayers b. [Eac] [on þam ylcan timan] sum preost Aquinensis þære cyricean [also [at the same time some priest Aquinensis of-the church wearð gedreht mid deofolseocnysse. (Gregory H, 134.16.134.22) became tormented by demoniacal possession

(5) a.

Thus, simple distinctions like argument vs. adjunct fronting or heaviness of subject do not provide any simple answers to the question as to when subjectverb inversion does not apply in OE. However, an analysis based on more finegrained distinctions and the use of more sophisticated statistical tools may identify certain factors which at least favor the occurrence of non-inversion in a significant way. I will leave this issue for future work. In summary, we have seen that already in OE there is a substantial number of clauses with a fronted non-operator and a full DP subject which do not exhibit subject-verb inversion. It is therefore not entirely adequate to talk about the “loss of V2” in English since there is no attested period in the history of this language during which it had the properties of a typical V2 language (cf. also Section 2). However, subject-verb inversion is the clear majority pattern in clauses with a non-pronominal subject and a fronted non-operator in OE.

4. Middle English Let us now consider the development of subject-verb inversion after the OE period. The situation in Early Middle English (EME) is still comparable to that found in OE. Kroch and Taylor (1997: 311) discuss the frequencies of subjectverb inversion in seven texts from the early 13th century. These frequencies show that, as in OE, inversion is still predominant with full DP subjects when a non-operator is fronted. If all of Kroch and Taylor’s figures for the different texts are taken together, we obtain a frequency of non-inversion of 28.6%. This figure is very close to the figure given in Table 1 for the total numbers obtained from the different OE texts (28.7%). Hence, the status of subject-verb inversion does not yet seem to have changed substantially at the beginning of the Middle English (ME) period. However, during the ME period the frequency of inversion in contexts of non-operator fronting decreases rapidly. Van Kemenade (1987: 183ff.) therefore


suggests that V2 starts being lost by around 1400. This observation is confirmed to a large extent by the quantitative data in Table 2 obtained from the PennHelsinki Parsed Corpus of Middle English.10 Table 2 gives the numbers and frequencies for the (non-)occurrence of subject-verb inversion in clauses with a fronted non-operator and a full DP subject in samples from 33 ME prose texts from the 14th and 15th centuries.11,12 The ranking is based on the frequency of non-inversion, the texts at the top being those with the highest frequencies of non-inversion (i.e. those in which the change has advanced most). In 23 out of the 33 texts listed in Table 2, the frequency of non-inverision is above 50%. By the 15th century, inversion in contexts of non-operator fronting has thus become a minority pattern, although there is still considerable variation among the different texts which does not seem to be of a clear dialectal or chronological nature. With respect to the percentages in Table 2, it is important to note that the frequency of non-inversion is not expected to reach 100% during the history of English. Sentences in which a non-operator is fronted and the subject follows the finite verb can still be found in Modern English. Some typical contexts are shown in (6) (examples from Bresnan 1994: 78, Schmidt 1981: 6/8/9, Stockwell 1984: 581)). (6) a. b. c. d. e. f.

[Plainly detectible] were the scars from his old football injury. [Stolen] were all of the newlyweds’ gifts. [Across the river] lived seven dwarfs. [In this rainforest] can be found the reclusive lyrebird. [Thus] ended his story. [In the year 1748] died one of the most powerful of the new masters of India.

In (6a/b), a predicate is fronted and finite be precedes the subject. Examples (6c/d) are cases of locative inversion. And (6e/f) illustrates inversion with certain adjuncts such as thus or point time adverbials. The contexts shown in (6) are also frequent contexts for inversion in 14th and 15th century ME already, as the examples in (7) illustrate. [bettur] is schort payne þen longe. ‘A short pain is better than a long one.’ b. [blessed] be God!

(7) a.

(Siege, 86.514) (Brut, 221.409)

254 Eric Haeberli

Table 2.Main clauses with non-operators preceding non-pronominal subjects in samples of texts from the late 14th and the 15th century — type I Text (date)

no inversion inversion

% uninverted

Purvey, Prologue to the Bible (c1388) Old Testament (a1425) Life of St. Edmund (c1450 (1438)) New Testament (c1388) Documents (1380–1420) Mirk’s Festial (a1500 (a 1415)) Trevisa, Polychronicon (a1387) Malory, Morte Darthur (a 1470) Book of Margery Kempe (c1438) In Die Innocencium (1497) Phlebotomy (c1400–1425) Gregory’s Chronicle (c1475) ME Sermons, ms. Royal (c1450 (c1415)) Siege of Jerusalem (c 1500) Caxton, Prologues/Epilogues (1477–1484) Hilton, On Perfection (a1450 (a1396)) Private letters (1448–1480) Julian of Norwich (c1450 (c1400)) Wycliffite Sermons (c1400) Rolle, Prose Treatises (c1440 (a1349)) Capgrave’s Chronicle (a1464) Brut (c1400) Cyrurgie of Guy de Chauliac (?c1425) Fitzjames, Sermo Die Lune (?1495) Cloud of Unknowing (a1425 (?a1400)) Chaucer (c 1380–1390) Vices and Virtues (c1450 (c1400)) Earliest English Prose Psalter (c1350) Mandeville’s Travels (?a1425 (c1400)) Robert Reynes (1470–1500) Caxton, Reynard the Fox (1481) Mirror of St. Edmund, ms. Thornton (c1440 (?1350)) Mirror of St. Edmund, ms. Vernon (c1390)

29 107 56 96 64 28 72 82 35 26 18 59 33 47 19 23 48 25 122 20 54 34 18 19 30 64 15 24 15 14 8 17 13

96.7% 94.7% 91.8% 89.7% 84.2% 82.4% 77.4% 75.9% 72.9% 72.2% 72.0% 71.1% 68.8% 68.1% 67.9% 67.6% 67.6% 64.1% 63.5% 60.6% 60.0% 56.7% 56.3% 46.3% 44.4% 43.0% 38.5% 29.3% 28.3% 26.9% 23.5% 21.3% 13.4%

1 6 5 11 12 6 21 26 13 10 7 24 15 22 9 11 23 14 70 13 36 26 14 22 33 85 24 58 38 38 26 63 84

NB a = before, c = around

And [before the Emperoures table] stonden grete lordes & riche barouns & othere. (Mandeville, 143.317) ‘And in front of the emperor’s table stood great lords and rich barons and others.’ d. for [in þe serkil] was writin hir name. (Capgrave, 210.19) ‘For on the ring was written her name.’ c.


e. f.

[thus] endeth the book named Proloconycon… (Caxton, Prologues, 41.582) [Þat Šere] deide þat worthy man Beda þe preost. (Polychronicon, VI, 219.77) ‘In that year died that worthy man Bede, the priest.’

Given that the constructions in (6) still occur in Modern English, it is not surprising that very similar constructions can also be found at the time when subject-verb inversion is generally on its decline. Hence, what may be more revealing for determining the status of subject-verb inversion in ME is to count only those cases of inversion which have disappeared in the history of English, i.e. to exclude the constructions shown in (7). I will call these cases Type II inversion (vs. Type I which includes the constructions in (7)). The quantitative data for Type II inversion are given in Table 3.13 Once the Modern English inversion contexts in (6) are not included, we obtain frequencies of non-inversion which are above 50% in 28 out of the 33 text samples. In 19 text samples, the frequency of non-inversion is even above 75%. The ME data in Tables 2 and 3 show that by the 15th century subject-verb inversion has become a clear minority pattern and the OE/EME system in which subject-verb inversion was predominant in contexts of non-operator fronting is being lost. Two main questions now arise: (i) Are there any specific contexts in which the remaining instances of subject-verb inversion occur in the late ME texts? (ii) Are there any explanations for the frequency differences between the various texts (cf. e.g. Edmund, ms. Vernon (c1390) 21.3% noninversion vs. Purvey/New Testament (c1388) 96%)? With respect to the first question, the following main observations can be made. First of all, if we consider the remaining instances of subject-verb inversion in the two texts which have the highest frequency of non-inversion in Table 3, we can observe that they both involve a passive construction. The two examples are given in (8). [Forsothe] [to Adam] was not foundun an helpere lijk hym. (OTest, II, 20G.97) ‘Surely, a helper like him was not found for Adam.’ b. And [in this maner] was bothe hys shurte and hys breche imade (Life of St. Edmund, 166.99) ‘And his undergarments were made in this way.’

(8) a.

In (8a) and (8b), a non-operator is in clause-initial position while the subject either follows both the finite auxiliary and the participle (8a) or it occurs

256 Eric Haeberli

Table 3.Main clauses with non-operators preceding non-pronominal subjects in samples of texts from the late 14th and the 15th century — type II Text (date)

no inversion inversion

Old Testament (a1425) Life of St. Edmund (c1450 (1438)) Purvey, Prologue to the Bible (c1388) New Testament (c1388) Mirk’s Festial (a1500 (a 1415)) Documents (1380–1420) Trevisa, Polychronicon (a1387) Malory, Morte Darthur (a 1470) Book of Margery Kempe (c1438) Caxton, Prologues/Epilogues (1477–1484) Brut (c1400) Gregory’s Chronicle (c1475) Capgrave’s Chronicle (a1464) Rolle, Prose Treatises (c1440 (a1349)) Siege of Jerusalem (c 1500) In Die Innocencium (1497) Private letters (1448–1480) ME Sermons, ms. Royal (c1450 (c1415)) Phlebotomy (c1400–1425) Hilton, On Perfection (a1450 (a1396)) Julian of Norwich (c1450 (c1400)) Cyrurgie of Guy de Chauliac (?c1425) Wycliffite Sermons (c1400) Mandeville’s Travels (?a1425 (c1400)) Cloud of Unknowing (a1425 (?a1400)) Robert Reynes (1470–1500) Fitzjames, Sermo Die Lune (?1495) Chaucer (c 1380–1390) Earliest English Prose Psalter (c1350) Vices and Virtues (c1450 (c1400)) Mirror of St. Edmund, ms. Thornton (c1440 (?1350)) Caxton, Reynard the Fox (1481) Mirror of St. Edmund, ms. Vernon (c1390)

107 56 29 96 28 64 72 82 35 19 34 59 54 20 47 26 48 33 18 23 25 18 122 15 30 14 19 64 24 15 17 8 13

1 1 1 4 2 7 9 14 6 4 8 14 13 5 12 7 13 9 5 8 11 8 62 9 19 13 18 64 28 22 31 15 48

% uninverted

99.1% 98.2% 96.7% 96.0% 93.3% 90.1% 88.9% 85.4% 83.3% 82.6% 81.0% 80.8% 80.6% 80.0% 79.7% 78.8% 78.7% 78.6% 78.3% 74.2% 69.4% 69.2% 66.3% 62.5% 61.2% 51.9% 51.4% 50.0% 46.2% 40.5% 35.4% 34.8% 21.3%

between the two verbal elements (8b). Such constructions can also be found with a fairly high percentage among Type II inversions in other text samples such as the New Testament (1 passive construction out of 4 Type II inversions), Documents (3/7), Polychronicon (4/9), Malory (3/14), Gregory (11/14), Capgrave (5/13), In Die Innocencium (4/7), Private Letters (5/13), ME Sermons (4/9), Brut (4/8), Hilton (3/8), Julian of Norwich (4/11), Cyrurgie (4/8), Mandeville (3/9) or


Cloud (5/19). Thus, it seems that passive constructions favor the occurrence of the subject in a low structural position and hence in a position which follows the finite verb. In some other texts, some other preferences with respect to the verbal context can be observed in inversion contexts. For example, the presence of copula be frequently gives rise to inversion in Siege (8/12) and Mandeville (4/9), whereas clauses containing a finite modal often exhibit inversion in Mirk (2/2), Kempe (5/6), Wycliffite Sermons (13/62), Cloud (7/19), Vices (6/22). However, as the examples in (9) to (11) below will show, inversion can be found in any kind of context, in particular also with transitive verbs. With respect to the fronted element, inversion occurs in various contexts. In (9), different types of adverbs are fronted. And [þerfore] saide Maister Arnalde þat he … (Cyrurgie, 577.217) ‘And therefore Master Arnald said that…’ b. [Wonderfuly] is a mans affeccion varied in goostly felyng of þis nouŠt… (Cloud, 122.588) ‘A man’s emotion is wonderfully varied in the spiritual feeling of this nothing…’ c. and [sone þerafter] were messangers i-sent to Avyon to þe pope (Polych, 352.410) ‘And soon after that, messengers were sent to Avignon to the pope.’ d. and [oftentyme] deyn men (Reynes, 160.104) ‘And often, people die.’

(9) a.

Furthermore, subject-verb inversion also can be found with various types of PP adjuncts. And [accordyng to the same] saith Salamon that the nombre of foles is infenyte. (Caxton, Prologues, 11.3) ‘And according to the same, Solomon says that the number of fools is infinite.’ b. [So] [with that] departed the damesell (Malory, 47.92) ‘So with that, the damsel left.’ c. [In þis wyze] bene all good levers called þe frendes of God (ME Sermons, 16.74) ‘In this way, all those who live righteously are called the friends of God.’ d. [In this] wille oure lorde that … (Julian, 62.330) ‘With this, our lord wishes that …’

(10) a.

258 Eric Haeberli

[Fro þat place] was þe king led to London to þe Tour. (Capgrave, 213.71) ‘From that place, the king was led to London to the Tower.’ f. But [at þe deth of Cryst] was Tyberyis Emparowr of Rome (Siege, 73.90) ‘But at the time of Christ’s death, Tiberius was the Roman emperor.’ g. [In þis Šere, in þe seuene day of May], cam þe Emperor Sigemund to London (Capgrave, 247.376) ‘In this year, on the seventh day of May, the emperor Sigmund cam to London.’ e.

And finally, fronting of an argument can also trigger subject-verb inversion in many of the ME text samples listed in Tables 2 and 3. [This] seith Austyn there. (Purvey, I, 56.108) ‘Austin says this there.’ b. [Thyse wordes] sayd our sauyour Ihu Cryst of the temple of his holy body. (Fitzja, A5V.82) ‘Our Savior Jesus Christ said these words about the temple of his holy body.’ c. and [muche sorow] had sir Gawayne to avoyde his horse (Malory, 201.420) ‘And Sir Gawain had much difficulty to dismount from his horse.’ d. [Of þese men] spekiþ Seynt Petir þus: … (Hilton, 14.99) ‘Saint Peter says the following about these men: …’

(11) a.

But many of the fronted non-operators which give rise to inversion in the examples in (9) to (11) can also be found (sometimes in the same text) in clauses in which no subject-verb inversion has taken place. This is shown in (12). (12) a.

And [þerfore] the lore and þe manere of knowynge of symple þinges is Šeuen of Galien in þe firste bokes of Symple Medecynes … (Cyrurgie, 576.193) ‘And therefore the instructions for simple things are given by Galen in the first books of Simple Medicines…’ b. and [sone þerafter] þe schap of þe cros was i-seie forsake þe baner … (Polychronicon, VIII, 89.204) ‘And soon after that, the shape of the cross was seen how it left the banner…’ c. [Wyth that] sir Raynolde gan up sterte … (Malory, 200.370) ‘With that, Sir Reynold sprang up to his feet…’


d. [at þe tyme of his passion] Pylat send hyme to Herrode

e.

f.

(Siege, 76.152) ‘At the time of the Passion, Pilate sent him to Herod…’ Eke [in þis Šere] Thomas, duke of Clarens, cam hom fro Gian (Capgrave, 238.174) ‘Also, in this year, Thomas, duke of Clarens, came home from Gian.’ [This þingis] God send to hyme for þis cavssys, (Siege, 73.83) ‘God sent him these things for these reasons.’

In (12a), the adverb ‘therefore’ has been fronted but does not trigger subjectverb inversion whereas in (9a) inversion takes place in the same text. Similar variation can be found in (12b) to (12e) (identical or similar fronted nonoperators as in (9c), (10b), (10f), (10g) in the same texts) and in (12f) (similar fronted object as in 11b but in a different text). The data in (9) to (12) thus suggest that, with respect to the type of fronted non-operator, there is no clearly identifiable factor which determines the presence or absence of subject-verb inversion in ME. However, as pointed out already in our discussion of OE (Section 3), it may be that by using more detailed statistical evidence and tools some factors can be identified which at least favor the occurrence of inversion. I will return to this issue in future work. Let us finally consider the status of the subject in subject-verb inversion constructions in ME. Again, the general observation based on data like those in (9) to (11) is that subject-verb inversion is not simply restricted to some specific type(s) of subject. Although most subjects in the examples above are definite, indefinite subjects also occur in inversion constructions (cf. e.g. 9c, d). Similarly, most of the subjects in (9) to (11) are fairly light, but, not unexpectedly, heavier ones also frequently follow the finite verb (cf. e.g. 11b). Finally, it is interesting to note that the class of subjects which occur in subject-verb inversion constructions in ME even includes subject pronouns (cf. also van Kemenade 1987: 198). This observation is fairly surprising from a diachronic point of view. As mentioned in Section 2 and as often discussed in the literature, fronting of a non-operator generally does not lead to subject-verb inversion with pronominal subjects in OE. This observation is confirmed by the following quantitative data obtained from the text samples studied in Table 1 above (cf. also Kroch and Taylor 1997: 311 for some data for EME). In seven out of the ten text samples in Table 4 and in all the five late texts, subject-verb inversion with a pronominal subject never occurs at all. In Bede and the Chronicle, there is one exception to the restriction on inversion with subject pronouns (but cf. note 14). The only text in which such inversions

260 Eric Haeberli

Table 4.Main clauses with non-operators preceding pronominal subjects in samples of ten OE texts Text (date)

no inversion14 % uninverted % uninverted with inversion full DP subjects (Table 1)

Bede (950)

37 91 25 41 27 30 20 45 38 31

1 0 1 0 4 0 0 0 0 0

97.4% 100.0% 96.2% 100.0% 87.1% 100.0% 100.0% 100.0% 100.0% 100.0%

50.0% 27.5% 17.4% 59.5% 39.5% 17.3% 33.3% 15.2% 39.5% 23.0%

occur with some frequency is Orosius (12.9% inversion). However, the general picture that arises is that subject pronouns generally do not invert with the finite verb when a non-operator is fronted in OE. In ME, the situation is considerably different, as Table 5 shows. Among the 27 text samples which exhibit Type II inversion with non-pronominal subjects relatively frequently (non-inversion below 90%), only 3 completely lack inversion with subject pronouns. In the other 24 texts, inversion with a pronominal subject can be found at least once and in general several times. Although the frequency of Type II non-inversion with full DP subjects is lower than non-inversion with subject pronouns in most texts,15 the data in Table 5 nevertheless suggest that subject-verb inversion with pronouns is an option which was generally available in ME and that therefore the fairly clear-cut contrast between pronominal and non-pronominal subjects found in OE/EME has disappeared in later ME. A few illustrations of subject-verb inversion with pronominal subjects in ME are given in (13). [On þe same maner] schalt þou do wiþ þis lityl worde GOD. (Cloud, 78.323) ‘You should do the same with this little word GOD.’ b. and [þe cherch of Lincoln] gaue he to Herry Beuforth… (Capgrave, 210.11) ‘And he gave the church of Lincoln to Herry Beuforth…’ c. And [herof] am J sure (Caxton Prologues, 89.186) ‘And I am sure about this.’

(13) a.


Table 5.Main clauses with non-operators preceding pronominal subjects in samples of texts from the late 14th and the 15th century text

no inversion

inversion16 % uninverted % uninverted with full DP subjects (type II, Table 3)

Old Testament (a1425) Edmund (c1450 (1438)) Purvey (c1388) New Testament (c1388) Mirk (a1500 (a 1415)) Documents (1380–1420) Polychronicon (a1387) Malory (a 1470) Kempe (c1438) Caxton (1477–1484) Brut (c1400) Gregory’s Chronicle (c1475) Capgrave (a1464) Rolle (c1440 (a1349)) Siege of Jerusalem (c 1500) In Die Innocencium (1497) Private letters (1448–1480) Sermons (c1450 (c1415)) Phlebotomy (c1400–1425) Hilton (a1450 (a1396)) Julian (c1450 (c1400)) Cyrurgie (?c1425) Wycliffite Sermons (c1400) Mandev (?a1425 (c1400)) Cloud (a1425 (?a1400)) Reynes (1470–1500) Fitzjames (?1495) Chaucer (c 1380–1390) Earliest Psalter (c1350) Vices (c1450 (c1400)) Edmund, Thornton (c1440 (?1350)) Caxton, Reynard (1481) Edmund, Vernon (c1390)

46 72 25 103 27 81 48 203 110 38 79 59 29 33 87 32 213 57 23 37 52 45 73 31 169 31 31 95 47 49 95 48 126

1 0 0 0 1 5 0 30 16 4 6 0 31 6 4 2 18 4 6 8 14 2 13 1 42 0 12 95 16 19 105 28 23

97.9% 100.0% 100.0% 100.0% 96.4% 94.2% 100.0% 87.1% 87.3% 90.5% 92.9% 100.0% 48.3% 84.6% 95.6% 94.1% 92.2% 93.4% 79.3% 82.2% 78.8% 95.7% 84.9% 96.9% 80.1% 100.0% 72.1% 50.0% 74.6% 72.1% 47.5% 63.2% 84.6%

99.1% 98.2% 96.7% 96.0% 93.3% 90.1% 88.9% 85.4% 83.3% 82.6% 81.0% 80.8% 80.6% 80.0% 79.7% 78.8% 78.7% 78.6% 78.3% 74.2% 69.4% 69.2% 66.3% 62.5% 61.2% 51.9% 51.4% 50.0% 46.2% 40.5% 35.4% 34.8% 21.3%

262 Eric Haeberli

d. And [on a tyme] was he taken bi pirates in the see. (Fitzjames, B3R.154) ‘And once he was captured by pirates on the sea.’ e. & [many tymes] haue I feryd þe wyth gret tempestys of wyndys (Kempe, I, 51.110) ‘And I have frightened you many times with great tempests.’ f. And [many mervayles] shall he do (Malory, 47.79) ‘And he shall do many wonderful things.’ g. [þis question] wolde I knowe of you (Private Letters, Mull, I, 126.623) ‘I would like to know the answer to this from you.’

In summary, various types of subjects occur in the remaining subject-verb inversion cases found in the late 14th and the 15th century. In particular, in contrast to OE/EME, pronominal subjects also regularly invert with the finite verb in ME. Let us finally turn to the second question raised earlier in the context of Tables 2 and 3. As observed there, for example The Mirror of Saint Edmund (ms. Vernon), Purvey’s Prologue to the Bible and The New Testament all are texts from around 1390, but while the latter two texts already have a frequency of noninversion of 96% in Table 3, the first text only has a frequency of 21.3%. The question that arises then is why such differences in the frequency of noninversion occur in the different text samples studied here. For reasons of space, it is not possible to consider the status of each text with respect to its frequency of subject-verb inversion here. Instead, I will focus on a few texts which have a very low frequency of non-inversion and discuss three potential factors that may play a role in these low frequencies. The three factors are the following: (i) The grammatical conditions for the loss/decrease of subject-verb inversion are not met yet; (ii) a translation with a V2 source language; (iii) language contact. The details for option (i) depend on what factor can be determined causing the loss of subject-verb inversion in the history of English. If such a factor can be identified, it would of course be very likely that at least in some of the texts with low frequencies for non-inversion the relevant conditions for the loss/ decrease of inversion are not yet entirely met. This kind of explanation for certain low non-inversion frequencies is indeed possible if we adopt the analysis of the loss of subject-verb inversion in English which I proposed elsewhere (cf. Haeberli 2002). Since it would go beyond the scope of this article to discuss this approach in any detail, I simply give its main lines here, and the reader is referred to Haeberli (2002) for arguments in favor of this approach. The basic


proposal is that subject-verb inversion in contexts of non-operator fronting is possible in OE/EME because non-pronominal subjects can remain in a structurally low subject position to the right of the surface position of the finite verb and that this option is available because a higher subject position above the finite verb can be occupied by an empty expletive. Fronting of a non-operator therefore leads to ‘XP-V-S’ orders. During the ME period, empty expletives start being lost and, as a consequence, non-pronominal subjects cannot remain in a low subject position any more but have to move to the subject position to the left of the finite verb. Thus, we obtain ‘XP-S-V’ orders. The loss of subjectverb inversion therefore is the result of the loss of empty expletives in the history of English (cf. also Hulk and van Kemenade 1995: 249 for the observation that the loss of V2 and the loss of empty expletives coincide). As for the loss of empty expletives, the analysis in Haeberli (2002) is based on the standard assumption that the licensing of empty expletives depends on properties of the verbal inflectional morphology and it is therefore proposed that the loss of empty expletives is due to a change in the inflectional morphology in ME. More specifically, it is argued that it is the loss of the final /n/ in infinitives (cf. e.g. OE andswarian, EME ontswerien, late ME answere (‘to answer’)) which plays a crucial role here. In terms of such an analysis, the loss of subject-verb inversion can ultimately be reduced to the loss of the infinitival -n ending. Given this conclusion, we now can return to the ME texts studied earlier. In general, -n infinitives have become very rare in these texts. Among the 33 samples studied, there are only 11 in which the frequency of -n infinitives is still higher than 3% and, even in those, the -n infinitive is generally the clear minority form. The most striking exception to this observation can be found in The Mirror of St. Edmund (ms. Vernon). In this text sample, 382 out of 469 infinitives (81.4%) have an -n ending, which is by far the highest frequency among the texts studied here. Thus, the development towards n-less infinitives only seems to be in its initial stages in this text. What is interesting for our purposes now is that The Mirror of St. Edmund (ms. Vernon) is also the text which has by far the lowest frequency of non-inversion in Tables 2 and 3. Thus, the highest frequency of n-infinitives coincides with the lowest frequency for the absence of subject-verb inversion. From the point of view of the approach proposed in Haeberli (2002), this observation is not surprising because it relates the loss of subject-verb inversion to the loss of -n infinitives. Since the loss of -n infinitives is only in its initial stages, there are also no developments yet with respect to the loss of subject-verb inversion. The special status of The Mirror of St. Edmund (ms. Vernon) among the texts studied would thus be an illustration

264 Eric Haeberli

of factor (i) listed above for the variation with respect to subject-verb inversion, i.e. an illustration of a text which does not yet meet the necessary conditions for the loss of inversion.17 Let us now turn to the second factor that may be relevant for the inversion patterns found in some of the ME texts studied in Tables 2 and 3. The relevant text sample here is the one from William Caxton’s Reynard the Fox. This text sample has the second lowest frequency of non-inversion in Table 3, namely 34.8%. What is interesting now is that another text sample attributed to William Caxton shows a completely different picture. In Caxton’s Prologues and Epilogues, the absence of subject-verb inversion is the clear majority pattern with 82.6%. How can this contrast between two texts written by the same author be accounted for? A property of the first text is suggestive here. As observed in the text information of the Penn-Helsinki Corpus and as discussed in detail by Blake (1970), Caxton’s Reynard the Fox is a translation from a Dutch original. At that time, Dutch was on its way to becoming the relatively rigid V2 language it is today (cf. e.g. Weerman 1989: 183ff.) and it may therefore be that the Dutch source had an influence on the frequent use of inversion in Reynard the Fox. Although becoming marginal, inversion was still a grammatical option in late ME, and the frequency of its occurrence may thus sometimes have been influenced by a source text written in a language which makes frequent use of subject-verb inversion.18 Let us finally turn to a third factor which may play a role for variation in the frequency of subject-verb inversion in ME. Kroch and Taylor (1997) and Kroch, Taylor and Ringe (2000) show that a northern ME text from around 1400, The Northern Prose Rule of St. Benet, exhibits a fairly regular V2 syntax in which subject-verb inversion applies regardless of whether the subject is a pronoun or a full DP. Kroch et al. take this text as evidence for a dialect split with respect to the syntax of V2 in ME. They distinguish a northern dialect which has a regular V2 syntax with systematic subject-verb inversion from a southern dialect which is a continuation of the OE V2 system in which subject-verb inversion only occurs with non-pronominal subjects. In terms of this proposal, certain aspects of the subject-verb inversion syntax of ME could then be argued to be a manifestation of a language contact situation (cf. also Kroch, Taylor and Ringe 2000, Lightfoot 1997). In particular, properties of the regular V2 syntax of the north could have been introduced into the grammars of speakers of the south in a contact situation. Such a scenario would be particularly plausible for cases in which the OE/EME distinction between subject types is not maintained and pronominal and non-pronominal subjects have a similar status with respect to


subject-verb inversion. A text sample which has this property is The Mirror of St. Edmund (ms. Thornton), the text with the third lowest frequency of noninversion with full DPs in Table 3 (35.4%; non-inversion with pronouns 47.5%). For this text, it could be argued then that subject-verb inversion has not decreased in the same way as in most other ME texts because, due to northern influence, a different system has been introduced which derives (optional) subject-verb inversion orders. This scenario would not be implausible given that The Mirror of St. Edmund (ms. Thornton) is a text of northern origin (cf. Perry 1914).19 Thus, certain frequency variations in Tables 2 and 3 may be due to varying degrees of influence of the northern V2 syntax. In summary, we have seen in this section that by the 15th century subjectverb inversion in clauses with a fronted non-operator and a full DP subject has become the clear minority pattern in most of the ME text samples studied. However, the loss of subject-verb inversion is not yet completed at the end of the ME period and instances of inversion can still regularly be found in all ME texts (cf. also Bækken 1998 for a detailed discussion of the further developments concerning inversion in Early Modern English). As for the contexts in which the remaining cases of subject-verb inversion occur, it is relatively difficult to determine them very clearly at this point. I have shown, however, that certain contexts such as passivization may favor the occurrence of a subject in a position following the finite verb. With respect to the type of fronted element or the type of subject in inversion constructions, a wide range of elements can be found in inversion constructions. The most striking property of inversion in the later ME texts from a diachronic point of view is the fact that even pronominal subjects start occurring in inversion constructions fairly regularly in most texts. Finally, I discussed some possible explanations for the variation that can be found among the different ME texts with respect to the frequency of (non-)inversion in contexts of non-operator fronting. I proposed that low frequencies of non-inversion in certain texts may be the result of a situation in which the grammatical conditions for the loss of inversion are not yet met, of the influence of a source text in a translation, or of language contact. 5. Conclusion The aim of this article was to give a general overview of the loss of V2 or, more precisely, the loss of subject-verb inversion in clauses with a fronted nonoperator and a full DP subject in the history of English. Based on data taken from two parsed corpora, I have shown that the absence of subject-verb

266 Eric Haeberli

inversion is already fairly frequent in OE and that by the 15th century inversion has become the clear minority pattern in most texts although there is still considerable variation among different texts. The findings in this article raise several additional questions: (i) How can the situation in OE be analyzed in theoretical terms given that OE has a V2 syntax which is far less rigid than the one found in Modern Germanic? (ii) How can the loss or at least the drastic decrease of subject-verb inversion in the ME period be explained? (iii) How can the late ME inversion patterns be analyzed given that V2 also occurs with pronominal subjects? (iv) An issue which was mentioned in this article but which has not been dealt with conclusively here: Are there any factors which determine or at least favor the absence of inversion in OE or favor the occurrence of the remaining inversion cases in ME? Issues (i) to (iii) are addressed in Haeberli (2002). As for issue (iv), further research will be necessary which has to be based in particular on more detailed statistical evidence and on additional and larger text samples.

Appendix — Old and Middle English sources The data in this article are taken from the “Brooklyn-Geneva-Amsterdam-Helsinki Parsed Corpus of Old English” (Pintzuk et al. 2000) and the “Penn-Helsinki Parsed Corpus of Middle English” (Kroch and Taylor 1994). Below are the references to the texts from which the samples in these corpora are taken. For the page numbers of the samples see Pintzuk et al. 2000, Kroch and Taylor 1994.

Old English Ælfric’s Letters. Ælfric’s First and Second Letters to Wulfstan, Ælfric’s Letter to Wulfsige. In B. Fehr (ed.) (1914) Die Hirtenbriefe Ælfrics in altenglischer and lateinischer Fassung. Bibliothek der angelsächsischen Prosa IX, Verlag von Henri Grand, Hamburg. Ælfric’s Lives of Saints. Ælfric’s Lives of Saints. Ed. W. W. Skeat (1966 (1881–1900)), Early English Text Society (o.s. 76, 82, 94, 114), London. Apollonius. The Old English ‘Apollonius of Tyre’. Ed. P. Goolden (1958), Oxford University Press, Oxford. Bede. The Old English Version of ‘Bede’s Ecclesiastical History of the English People’. Ed. T. Miller (1959 (1890; 1891)), Early English Text Society (o.s. 95, 96), London. Boethius. King Alfred’s Old English Version of Boethius ‘De Consolatione Philosophiae’. Ed. W. J. Sedgefield (1899), The Clarendon Press, Oxford. Chronicle A. The Anglo-Saxon Chronicle, ms. A. In C. Plummer (ed.) (1965 (1892)) Two of the Saxon Chronicles Parallel, The Clarendon Press, Oxford.


Gregory (ms. C). Gregory the Great, Dialogues, ms. C. In H. Hecht (ed.) (1900) Bischofs Waerferth von Worcester Uebersetzung der Dialoge Gregors des Grossen. Bibliothek der Angelsächsischen Prosa V. Georg H. Wigands Verlag, Leipzig. Gregory (ms. H). Gregory the Great, Dialogues, ms. H. In H. Hecht (ed.) (1900) Bischofs Waerferth von Worcester Uebersetzung der Dialoge Gregors des Grossen. Bibliothek der Angelsächsischen Prosa V. Georg H. Wigands Verlag, Leipzig. Orosius. King Alfred’s Orosius, Part I. Ed. H. Sweet (1959 (1883)), Early English Text Society (o.s. 79), London. Wulfstan. The Homilies of Wulfstan. Ed. B. Bethurum (1957), The Clarendon Press, Oxford.

Middle English Book of Margery Kempe. The Book of Margery Kempe, Vol. I. Ed. S. B. Meech and H. E. Allen (1940), Early English Text Society (212), London. Brut. The Brut or the Chronicles of England, Part I. Ed. F. W. D. Brie (1960), Early English Text Society (o.s. 131), London. Capgrave’s Chronicle. John Capgrave’s Abbreuiacion of Chronicles. Ed. P. J. Lucas (1983), Early English Text Society (285), Oxford. Caxton, Prologues/Epilogues. Prologues and Epilogues of William Caxton. Ed. W. J. B. Crotch (1956 (1928)), Early English Text Society (176), London. Caxton, Reynard the Fox. The History of Reynard the Fox. Translated from the Dutch Original by William Caxton. Ed. N. F. Blake (1970), Early English Text Society (263), London. Chaucer. A Treatise on the Astrolabe; Boethius; The Tale of Melibee. In L. D. Benson (ed.) (1987) The Riverside Chaucer, Houghton Mifflin Company, Boston. Cloud of Unknowing. The Cloud of Unknowing. In P. Hodgson (ed.) (1958 (1944)) The Cloud of Unknowing and the Book of Privy Counselling, Early English Text Society (218), London. Cyrurgie of Guy de Chauliac. The Cyrurgie of Guy de Chauliac. Ed. M. S. Ogden (1971), Early English Text Society (265), London. Documents. Appeals; Petitions; Returns; Judgements; Testaments and Wills; Proclamations. In R. W. Chambers and M. Daunt (eds.) (1967 (1931)) A Book of London English 1384–1425, Clarendon Press, Oxford. Petitions: In J. H. Fisher, M. Richardson and J. L. Fisher (eds.) (1984) An Anthology of Chancery English, The University of Tennessee Press, Knoxville. Earliest English Prose Psalter. The Earliest Complete English Prose Psalter. Ed. K. D. Buelbring (1891), Early English Text Society (o.s. 97), London. Fitzjames, Sermo Die Lune. Richard Fitzjames. Sermo die lune in ebdomada pasche. Ed. F. Jenkinson (Facsimile). (1907) Cambridge University Press, Cambridge. Gregory’s Chronicle. Gregory’s Chronicle. In J. Gairdner (ed.) (1876) The Historical Collections of a Citizen of London in the Fifteenth Century, Camden Society (N. S. XVII), Westminster. Hilton, On Perfection. Walter Hilton’s Eight Chapters on Perfection. Ed. F. Kuriyagawa (1967), The Keio Institute of Cultural and Linguistic Studies, Tokyo.

268 Eric Haeberli

In Die Innocencium. In Die Innocencium. In J. G. Nichols (ed.) (1875) Two Sermons Preached by the Boy Bishop, at St. Paul’s Temp. Henry VII, and at Gloucester, Temp. Mary., Camden Society Miscellany VII, Camden Society (N. S. XIV), London. Julian of Norwich. Julian of Norwich’s Revelations of Divine Love. Ed. F Beer (1978), Carl Winter Universitätsverlag, Heidelberg. Life of St. Edmund. The Life of St. Edmund. In N. F. Blake (ed.) (1972) Middle English Religious Prose, Edward Arnold, London. Malory, Mort Darthur. Morte dArthur. In E. Vinaver (ed.) (1954) The Works of Sir Thomas Malory, Oxford University Press, Oxford. Mandeville’s Travels. Mandeville’s Travels. Translated from the French of Jean d’Outremeuse. Vol I. Ed. P. Hamelius (1919), Early English Text Society (o.s. 153), London. ME Sermons, ms. Royal. Middle English Sermons, Edited from British Museum Ms. Royal 18B. XXIII. Ed. W. O. Ross (1940), Early English Text Society (209), London. Mirk’s Festial. Mirk’s Festial: A Collection of Homilies, by Johannes Mirkus (John Mirk), Part I. Ed. T. Erbe (1905), Early English Text Society (e.s. 96), London. Mirror of St. Edmund, ms. Thornton. Mirror of St. Edmund (Thornton). In G.G. Perry (ed.) (1867 (1913)) Religious Pieces in Prose and Verse, Early English Text Society (26), London. Mirror of St. Edmund, ms. Vernon. Mirror of St. Edmund (Vernon). In C. Horstman (ed.) (1895) Yorkshire Writers: Richard Rolle of Hampole and His Followers, London. New Testament. The New Testament in English According to the Version by John Wycliffe About A. D. 1380 and Revised by John Purvey About A. D. 1388. Ed. J. Forshall and F. Madden (1879), Clarendon Press, Oxford. Old Testament. The Old Testament. In J. Forshall and F. Madden (eds.) (1850) The Holy Bible, Containing the Old and New Testaments, with the Apocryphal Books, in the Earliest English Versions Made from the Latin Vulgate by John Wycliffe and His Followers, Oxford University Press, Oxford. Phlebotomy. A Latin Technical Phlebotomy and Its Middle English Translation. Ed. L. E. Voigts and M. R. McVaugh (1984), Transactions of the American Philosophical Society 74, Part 2, Philadelphia. Private letters. Letters in: S. A. Moore (ed.) (1965 (1871)) Letters and Papers of John Shillingford, Mayor of Exeter 1447–1450, Camden Society (N. S. II), New York. N. Davis (ed.) (1971) Paston Letters and Papers of the Fifteenth Century, Part I, Clarendon Press, Oxford. C. L. Kingsford (ed.) (1919) The Stonor Letters and Papers, 1290–1483, Vols. I-II, Camden Society (Third Series, XXIX-XXX), London. A. Hanham (ed.) (1975) The Cely Letters 1472–1488, Early English Text Society (273), London. Purvey. John Purvey’s Prologue to the Bible. In J. Forshall and F. Madden (eds.) (1850) The Holy Bible, Containing the Old and New Testaments, with the Apocryphal Books, in the Earliest English Versions Made from the Latin Vulgate by John Wycliffe and His Followers, Oxford University Press, Oxford. Robert Reynes. The Commonplace Book of Robert Reynes of Acle. An Edition of Tanner Ms 407. Ed. C. Louis (1980), Garland Medieval Texts 1, Garland, New York/London. Rolle, Prose Treatises. Richard Rolle — Prose Treatises. In G.G. Perry (ed.) (1921 (1866)) English Prose Treatises of Richard Rolle de Hampole, Early English Text Society (o.s. 20), London. Siege of Jerusalem. The Siege of Jerusalem in Prose. Ed. A. Kurvinen (1969), Mémoires de la Société Néophilologique de Helsinki, XXXIV, Société Néophilologique, Helsinki.


Trevisa, Polychronicon. Polychronicon Ranulphi Higden, Monachi Cestrensis, Vols. VI, VIII. English Translations of John Trevisa and of an Unknown Writer of the Fifteenth Century. Ed. J. R. Lumby (1876, 1882), Rolls Series 41, London. Vices and Virtues. The Book of Vices and Virtues. A Fourteenth Century English Translation of the Somme le Roi of Lorens D’Orléans. Ed. W. N. Francis (1942), Early English Text Society (217), London. Wycliffite Sermons. English Wycliffite Sermons, Vol. I. Ed. A. Hudson (1983), Clarendon Press, Oxford.

Notes * Parts and earlier versions of the material discussed here were presented at the 6th Diachronic Generative Syntax Conference (University of Maryland, May 2000), the 15th Comparative Germanic Syntax Workshop (University of Groningen, May 2000) and at the 11th International Conference on English Historical Linguistics (University of Santiago de Compostela, September 2000). I would like to thank the audiences at these presentations for their valuable comments and suggestions. All remaining errors are my own responsibility. 1. If no secondary source is cited, the OE data are taken from the “Brooklyn-GenevaAmsterdam-Helsinki Parsed Corpus of Old English” (Pintzuk et al. 2000), a syntactically and morphologically annotated version of selected OE prose text samples from the Helsinki Corpus of English Texts. For the details of the Old English sources, see the Appendix. 2. One type of non-operator in clause-initial position, i.e. the subject, of course still frequently occurs in orders in which the finite verbal element occurs in second position in Modern English (e.g. John left). Similarly, many V2 clauses in OE are of the type ‘SU-V’. In this type of clause, we can therefore again not observe any developments in the surface word order patterns in the course of the history of English and they are therefore not directly relevant for our purposes here. Thus, the term ‘non-operator’ used in the text refers to non-subject non-operators. 3. The question that the examples in (3) raise is what the status of such multiple topics is in OE, i.e. whether they occur in specific contexts and how they can be analyzed in theoretical terms. I will return to these issues in future research. 4. As for the occurrence of constituents to the right of the finite verb, it could be analyzed in terms of extraposition, a process which has been postulated in many analyses of OE syntax (cf. e.g. van Kemenade 1987, Pintzuk 1991). 5. The figure without the Chronicle A is given because the number of relevant examples in this text is considerably higher than in the other ones that are considered in this table and this text (and its (non-)inversion pattern) therefore weighs more heavily in the totals than the other texts. This problem could also be avoided by calculating the average of the different percentages instead of calculating the percentages based on the total numbers. In this way, we would obtain the following results: average percentage for all ten texts: 32.3%; average percentage for the five early texts: 38.8%; average percentage for the five later texts: 25.7%.

270 Eric Haeberli

6. Dates of composition (before/after 950) based on Pintzuk (1991: 381ff). The figures for the Chronicle do not include sentences with clause-initial her since the high number of such sentences would lead to a certain distortion of the general picture (total of clauses of this type: 234; inversion: 80; non-inversion: 154). Clauses in which only an adjunct clause precedes the subject are not counted. Adjunct clauses generally do not trigger inversion in OE. 7. The exact percentages are the following: Bede 48.6% (vs. 50.0% in Table 1), Boethius 22.5% (27.5%), Chronicle A 17.9% (17.4%), Gregory C 55.9% (59.5%), Orosius 38.6% (39.5%), Ælfric’s Letters 13.3% (17.3%), Ælfric’s Lives of Saints 38.2% (33.3%), Apollonius 15.4% (15.2%), Gregory H 36.1% (39.5%), Wulfstan 27.6% (23.0%). 8. The average frequencies would be the following: all texts 31.4% (vs. 32.3%, cf. fn. 5 above); early texts 36.7% (vs. 38.8%, cf. fn. 5); later texts 26.1% (vs. 25.7%, cf. fn. 5). 9. These counts exclude cases with a clause-initial object and a resumptive element within the clause given that in such configurations subject-verb inversion also does not occur in the modern Germanic V2 languages. 10. The Penn-Helsinki Parsed Corpus of Middle English (Kroch and Taylor 1994) is a syntactically annotated version of ME prose text samples from the Helsinki Corpus of English Texts. The data in this article are based on the first version of the Penn-Helsinki corpus (PPCME1). For the details on the Middle English sources, see the Appendix. 11. All the text samples from PPCME1 which contain more than 25 main clauses with a constituent preceding the subject are included in the table. The dates given for the different texts are taken from the Helsinki Corpus manual (cf. Kytö 1993). Clauses in which only an adjunct clause precedes the subject are not counted. Furthermore, cases of subject-verb inversion in which the equivalents of OE þa/þonne (‘then’) or nu (‘now’) occur in initial position are not included in the figures in Table 2. These elements tend to behave like operators in OE and might still do so in ME. 12. Between the EME data discussed by Kroch & Taylor (1997) and the data in Table 2, there is a gap of around 100 years (i.e. between 1250 and 1350). This is due to the fact that prose texts are generally lacking from this period (cf. e.g. Allen 1995: 385 for the 14th century). It is therefore not possible to obtain an entirely coherent picture for the decrease of subjectverb inversion in the ME period. 13. Table 3 excludes inversions occurring in typical Modern English inversion contexts as shown in (6)/(7) (i.e.: fronted predicate with finite be (a/b); fronted locative with a subject following an unaccusative/passive predicate or be (c/d); clause-initial thus or point time adverbs with unaccusative verbs (e/f)). This list is not meant to be exhaustive for Modern English inversion contexts but it simply covers contexts which can be found fairly regularly in ME. It is therefore not impossible that the figures for inversion in Table 3 still contain some cases of inversion which are not entirely excluded in Modern English. One context which is not mentioned in the text is quotative inversion (“…”, said John). The reason for this is that I excluded quotative inversion already in Table 2 because the status of quotative inversion is not entirely straightforward. Consider for example the following sentence. (i) ‘Syre’, seide Moises, ‘Šif men aske how men clepeþ Šow, what schal I seye?’ (Vices, 101.88) ‘Lord’, said Moses, ‘if somebody asks what you are called, what should I say?’


At the surface, the inversion in (i) looks like a parenthetical V1 clause rather than like a genuine example in which subject-verb inversion occurs due to the fronting of a non-operator. Hence, it is not clear whether the status of quotative inversion is entirely on a par with the other clauses counted in Tables 2 and 3 and I already omitted this construction for Table 2. 14. The instances of inversion in Bede and the Chronicle and one out of the four inversions in Orosius involve fronted swa ‘so’. It may therefore be that swa can occasionally function as an operator like the adverbs þa/þonne ‘then’ or nu ‘now’. Yet, this conclusion has to remain speculative at this point and would have to be confirmed by a more extensive study of the syntactic behavior of swa. 15. The main exception here is Capgrave’s Chronicle in which the frequency of inversion is almost twice as high with pronominal subjects as with full DP subjects. I have to leave it open here how this surprising pattern can be explained. 16. Clauses with initial ‘then’ and ‘now’ are again not counted here (cf. note 11). 17. Other texts which are ranked low with respect to subject-verb non-inversion in Tables 2 and 3 and which still have relatively high frequencies of -n infinitives are the following: The Earliest English Prose Psalter (46.2% non-inversion, 45.8% -n), Chaucer (50% non-inversion, 44.9% -n), Reynes (51.9% non-inversion, 18.1% -n), Mandeville (62.5% non-inversion, 29.7% -n), Wycliffite Sermons (66.3% non-inversion, 15.4% -n). Given the approach discussed in the text, the fact that non-inversion is not more frequent yet in these texts could be related to the fact that -n infinitives still seem to be fairly productive, i.e. the syntactic development has not made more progress yet because the morphological development is still under way. It should be pointed out, however, that there are two texts which have similar frequencies of -n infinitives as Reynes/Wycliffite Sermons and one text which has a considerably higher frequency but they nevertheless also have relatively high frequencies of non-inversion. In The Brut and in Gregory’s Chronicle, the frequency of infinitival -n endings is 18.3% and 12.7% respectively whereas the rate of non-inversion is 81% and 80.8% respectively. Thus, the loss of subject-verb inversion is well advanced although there are still more than just some isolated cases of infinitival -n endings. A detailed investigation of this contrast between Reynes/Wycliffite Sermons and Brut/Gregory’s Chronicle would go beyond the scope of this article. Let us therefore simply mention two points which may be relevant in this context. First of all, it seems plausible that in a transitional phase of a morphological and a related syntactic change, the patterns of usage are not directly linked. In other words, it may be possible that the writings of two authors are similar with respect to their morphological properties but that one author uses the syntax more conservatively whereas the other one makes more frequent use of the new syntactic option. And secondly, a more general problem may arise here, namely the question whether for example the occasional occurrence of an infinitival -n ending really reflects a phonologically represented ending that is still available or whether it just reflects a conservative spelling. If it is the latter, no syntactic consequences would be expected. Whereas the contrast between Reynes/Wycliffite Sermons and Brut/Gregory’s Chronicle is relatively small and therefore could well be due to one of the factors mentioned before, there is a third text, The Book of Margery Kempe, which still has infinitival -n endings and also high frequencies of non-inversion. However, in this text -n infinitives are not simply a marginal option occurring with frequencies around 15%, but they occur with a rate of 66.7%. In terms

272 Eric Haeberli

of the correlation mentioned in the text, the high frequency of -n infinitives would lead us to expect a subject-verb inversion pattern which is still close to OE/EME, i.e. with a relatively low frequency of non-inversion. Instead, we find a frequency of non-inversion of 83.3%. This problem can be weakened a bit once we consider the history of this text. As observed in the text information of the Penn-Helsinki Corpus, Margery Kempe “was apparently illiterate and her book was actually written by two amanuenses, the first an Englishman long resident in Germany, and the second, a priest… The whole was then copied by a scribe.” (for more details cf. Meech 1940). Thus, both with respect to the syntax and with respect to the morphology/spelling it is uncertain whether we are really dealing with a unique source here. Instead, the various participants in the writing of this text (Margery Kempe and the writers/scribe) may have had an influence on its syntax and morphology. In this respect, it is interesting to observe that, although Meech (1940: ix) suggests that the first writers’s “spelling, inflections and style were freely changed by the second”, in the text sample from Book I which has been attributed to Margery’s first writer (cf. Meech 1940) the -n infinitive is much less frequent (51%) than in the text sample from Book II which has been attributed to Margery’s second writer (84.2%). Thus, the fact that the correlation between inversion and infinitival morphology suggested in the text does not seem to hold in The Book of Margery Kempe may be due to a situation in which the text actually does not reflect a single grammatical system. 18. The same scenario may also hold for example for The Book of Vices and Virtues (noninversion 40.5%). This text is a translation of the French work Somme le Roi, a text from the 13th century, i.e. from a period when French still showed V2 properties (cf. e.g. Roberts 1993, Vance 1997). Thus, we can find the following type of parallelisms between the French text and the ME text (examples taken from the passage given in Francis 1942: xlii): (i) a.

b.

[Ce] nous tesmoingne li roy, li prince, li conte et li empereur… [that us witnesses the king, the prince, the count and the emperor (Somme le roi) ‘The king, the prince, the count and the emperor are witnesses to this for us.’ And [þat] witnesseþ wel þe kynges. þe erles. þe princes. and þe emperoures… (Vices and Virtues)

19. A similar explanation could hold for Chaucer’s text sample which has the same properties as The Mirror of St. Edmund (ms. Thornton) (low frequency of non-inversion, similar frequencies of non-inversion with pronominal and non-pronominal subjects). The similarity between Chaucer’s syntax and the northern V2 system has already been observed by Kroch & Taylor (1997: 324, fn. 16). Assuming that the northern dialect reflects Scandinavian influence, they conclude that “Chaucer’s syntax may be of a piece with his East Midlands phonology, since the East Midlands were part of the Danelaw. His language may, therefore, indicate a certain conservative regionalism compared to the developing London standard.” Note that if a scenario along these lines can be maintained, the factors discussed in the text would account for the low frequency of non-inversion for all six texts in which noninversion is still 50% or lower in Table 3 (The Mirror of St. Edmund (ms. Vernon) and The Earliest English Prose Psalter: productive infinitival -n ending (cf. text and fn. 17); Reynard the Fox and Book of Vices and Virtues: translations (cf. text and fn. 18); The Mirror of St. Edmund (ms. Thornton) and Chaucer: language contact).

A structure-based analysis of morphosyntactic regularities in language contact* Eric Hoekstra Frisian Academy

1.

Introduction

Languages develop by language contact. There are always two or more competing language variants. A language changes by introducing new variants, and by ousting old variants. Sometimes a whole language gets replaced by another. We may then speak of an advancing language and a receding language. A receding language and an advancing language may together give birth to language varieties which are in an intuitive sense “intermediary” between the receding language and the advancing language. A receding language is usually the mother tongue of speakers whereas an advancing language typically advances as a second language. Speakers switch from the mother tongue to the second language. Due to imperfect acquisition, however, they impose syntactic structure of their mother tongue onto the imperfectly acquired second language.1 For their children, however, the imperfectly acquired second language will be the new mother tongue. The new mother tongue will be an intermediary between the advancing language and the receding language. Frisian is such a receding language, and in the process of recession before Dutch it has given birth to such language varieties as Town Frisian, Bildts and West-Frisian.2 Furthermore, in the process of recession before (Lower) Saxon dialects Frisian has given birth to the Hogeland dialect (including Westerkwartier) of Groningen. For ease of reference, I will refer to the Hogeland and Westerkwartier dialects as the Groningen dialect, though this is not, strictly speaking, correct.3 Thus, we find contact-based dialects in the area where once Frisian was spoken.

274 Eric Hoekstra

If we consider the contact dialects, we note that certain properties of Frisian survived in the contact dialects, whereas other properties did not survive. Why should that be so? In order to answer that question, we turn to theories of language contact, such as Thomason and Kaufman (1988) and Van Coetsem (1988) and subsequent work. Thomason and Kaufman point to the relevance of social factors for the process of language contact. Van Coetsem analyses the relevance of linguistic factors. Thus these two works are complementary, in my view. The purpose of this article is to be more specific about the mechanisms at work in dialect contact. To be able to do so, we will compare a number of dialects which are all to a greater or lesser extent based on language contact between Frisian and either Dutch or Saxon as a second language. This yields some interesting empirical results, which we will try to accommodate by a refinement of language contact theory. Such a refinement is called for not only because of the empirical considerations to be presented below, but also because of theory-internal considerations. The theory as it stands claims that the syntax of the mother tongue tends to be projected into the new language and the phonology of the second language. This presupposes that syntax and phonology are separate from each other. Clearly, they are not: agreement, a syntactic notion, is expressed phonologically in the form of affixes consisting of specific phonemes.

2. The theory According to the theory of Van Coetsem, syntactic structures are transferred from a mother tongue into a second language, in case a second language — usually a high status language — is inadequately acquired. Hence the syntax of contact dialects often betrays what was the original mother language when the contact dialect came into existence. The theory is based on insight from second language acquisition. Following Van Coetsem (1988) and others we distinguish between two cases: linguistic aspects which are below the level of consciousness of the language user and those which are above it. Van Coetsem proposed that elements of the grammar which are below the level of consciousness are hard to learn and therefore easily projected from the mother language into the target language. Research by Van Coetsem (1988), Van Bree (1997) and others indicates that the following elements of grammar are largely below the level of consciousness: phonetics, morphology of unaccented syllables, syntax. On the other hand, elements which are above the level of consciousness are easily

A structure-based analysis of morphosyntactic regularities in language contact 275

learned and taken from the target language. These include: content words, phonemic correspondences, morphology of accented syllables. In this way, a contact-based dialect is created by inadequate acquisition of the target language. Within the generative tradition, researchers have also noticed that people tend to be less conscious of syntax than of phonology.4 Below we will try to be more specific than Van Coetsem. Contextually determined flexion is automatic and subconscious, involving a relation between two nodes in the syntactic tree. From the minimalist point of view, we can formalize the contrast by using the term tree-extension. Instead of flexion, we can also speak of checking: (1) Contextual checking: necessitates tree-extension (move and join) to check off semantic features Inherent checking: phonological recognition, both regular and exceptional

Note that a specific ending may involve a mixture of contextual and inherent checking, in case after tree-extension look-up in the lexicon is needed in order to determine phonological shape. We will now make the following claim about imperfect second language acquisition: (2) Checking effects in language contact Imperfect second language acquisition involves the projection of contextual checking mechanisms of the first language into the contact dialect provided that the inherent checking mechanisms projected from the second language allows this.

Note that we are limiting ourselves to cases of language contact in which the first language or mother tongue is also the receding language, socio-historically speaking, and in which the second language is also the advancing language. Thus we keep the sociological factors constant in order to be able to get a clear view of the linguistic interaction. Thomason and Kaufman claimed that there is no linguistic system underlying language contact but they did not keep the sociological context fixed. A close look at Thomason and Kaufman would reveal that certain linguistic types of contact are systematically related to certain sociological states of affairs. In this article we study the case where an advancing second language replaces a receding first language. We will show that the proposed checking effects in language contact allow us to predict the actual shape of contact dialects. We will first introduce the historical background of the Frisian-Dutch and Frisian-Saxon contact dialects and then go on to analyse some of their linguistic

276 Eric Hoekstra

properties within the framework proposed above. We predict that the contact dialect will roughly exhibit the inherent flexion of Dutch and the contextual flexion of Frisian.

3. Frisian contact dialects 3.1 Three Frisian-Dutch contact-dialects 3.1.1 West-Frisian Nowadays, West-Frisian is a countryside dialect spoken in the North of the Dutch province of North Holland. It probably developed from being a Frisian dialect to being a Frisian-based contact dialect after the military defeat of the West-Frisians in 1253. The tax system of the County of Holland was imposed and Dutch naturally became a status language for all West-Frisians who pursued a career in army, trade or burocracy. Furthermore, the expansion of the Zuiderzee (nowadays IJsselmeer) in the preceding centuries would naturally have led to closer economic contacts of the West-Frisians with the County of Holland, and less economic contacts with the Frisians living on the other side of the water. The data presented below come from books written in West-Frisian around the middle of the twentieth century (Butter 1944). An all-round grammatical investigation of the similarities between Frisian and West-Frisian was conducted in Hoekstra (1994a, 1994b), reveiling many morphosyntactic similarities between Frisian and West-Frisian. Independent confirmation of the findings reported here comes from texts from the end of the nineteenth century presented in the dialect survey of Leopold and Leopold (1882). 3.1.2 Town Frisian Town Frisian (see Fokkema 1937 for a grammatical description) came into existence in the sixteenth and seventeenth centuries as a status dialect when speakers of Frisian attempted to speak Dutch under the influence of the prestige which Dutch had acquired at that time (Van Bree 1997, but cf. Jonkman 1993 for a different view). Its status decreased in the course of time, and now it has the same low prestige which most city dialects tend to have. However, it is still spoken by a part of the city population.


3.1.3 Bildts Bildts is a Frisian-based contact dialect spoken by approximately 10,000 speakers in the North of the Dutch province Fryslân (Friesland) (cf. Buwalda 1963, who also gives some grammatical information). The dialect came into existence after land had been reclaimed from the sea at the beginning of the sixteenth century. The land was subsequently settled by farmers from South and North Holland. Subsequently, there was large migration from the surrounding Frisian area. 3.2 A Frisian-Saxon contact dialect: Groningen It is uncontroversial that Frisian was once spoken in the Dutch province of Groningen, as is evident from the fact that documents drawn up in Old Frisian have been found there. Furthermore, the dialects of Groningen, especially those spoken in the West and the North of the province (Westerkwartier and Hogeland) are, morphosyntactically speaking, very similar to Frisian (Hoekstra 1998). Like the previous cases we discussed, the syntax of Groningen has many similarities with Frisian, but the phonology does not.

4. Linguistic properties and discussion 4.1 Inherent checking phenomena from Frisian which got eliminated from the contact dialects In this section, we will present an overview of the facts of two inherent checking phenomena in the contact dialects: the case of the conjugation of je-verbs and the irregular pasts and perfects of strong verbs. In both cases, it will turn out that the Frisian forms did not survive into the contact dialects. 4.1.1 The case of the conjugation of je-verbs Frisian has two verb classes judging from the infinitives, those having an ending in -e(n) and those having an ending in -je(n). Their conjugation in present and past tense, and the shape of the participle, is given below (Tiersma 1985: 68ff):5

278 Eric Hoekstra

(3) Frisian -e and -je conjugations INFINITIVE rûke ‘smell’ (-e) PRESENT 1sg rûk 2sg rûkst 3sg rûkt pl rûke PAST 1sg rûkte 2sg rûktest 3sg rûkte pl rûkten PAST PARTICIPLE rûkt

helje ‘fetch’ (-je) helje hellest hellet helje helle hellest helle hellen helle

Interestingly, neither Dutch nor Saxon has the -je verbs (nor the concomitant conjugation). What happened in the contact dialects? Neither the Dutch contact dialects (West-Frisian, Bildts, Town Frisian) nor the Saxon ones (the Groningen dialects) have -je verbs. (I use in the table the word ‘declension’ for a deviant conjugation class.) (4) je-declension? Dutch West Frisian Town Frisian Bildts Frisian Groningen Saxon

no no no no yes no no

Now, the je-flection itself is a class-feature. Hence it is an instance of inherent flexion: no element within the sentence selects a je-verb or a non je-verb, this is given in the lexicon. Language contact theory correctly predicts that this feature involving inherent flexion of the receding first language will not be projected into the contact dialect. 4.1.2 The irregular past and perfect of strong verbs Cor van Bree (1994) conducted an in-depth study of strong verbs in Town Frisian.6 He concluded that the past and perfect of strong verbs in Town Frisian was very much like Dutch and very much unlike Frisian. If we look at the other contact dialects, we note that the past and perfect forms are not reminiscent of Frisian. In the case of Town Frisian, Bildts and West Frisian, they look more like Dutch. In the case of Groningen, the irregular verb forms look very much like the adjacent Saxon dialects.


(5) strong past and perfects look like Frisian? Dutch no West Frisian no Town Frisian no Bildts no Groningen no Saxon no

Some examples from Town Frisian (Van der Burg 1991), Bildts (Buwalda et al. 1996), and Dutch vs. Frisian are given below: (6) ‘offer’ Town Frisian Bildts Dutch Frisian ‘break’ Town Frisian Bildts Dutch Frisian

INFINITIVE – PAST – PAST PARTICIPLE biede – boad – boaden biede – boad – boaden bieden – bood – geboden biede – bea – bean breke – brak – broken breke – brak – broken breken – brak – gebroken brekke – briek or bruts – brutsen

The irregular past and perfects are listed in the lexicon as idiosyncratic properties of the verbs involved. Hence we are dealing with inherent flexion. As such, it is correctly predicted that the irregular flexion of Frisian will not be projected in the contact dialects. Instead, the contact dialects take over the forms of the second language. 4.2 Contextual checking phenomena from Frisian which survived in the contact dialects In this section we will present an overview of three contextual checking phenomena in the contact dialects. Two of these involve leftward movement in the Kaynian framework. The prediction for leftward movement is that those phenomena will survive from Frisian into the contact dialects, since they involve contextual checking. 4.2.1 Ending of 2sg Verbs in Dutch have a -t in the second person singular in the present tense, and in inversion the -t is absent. In the past tense, there is no inflectional ending. In Frisian, 2sg verbs end in -st in present and past tense alike:

280 Eric Hoekstra

(7) Dutch je komt / kom je je kwam / kwam je

Frisian (do) komst / komst (o) (do) kaamst / kaamst (o)

Interestingly, the contact dialects all pattern with Frisian, with the exception of West-Frisian: (8) 2sg verb ending is -st? Dutch West Frisian Town Frisian Bildts Frisian Groningen Saxon

no no yes yes yes yes yes

Some examples of the -st ending from Town Frisian and Bildts are given below (the Groningen case is trivial, since Saxon dialects have a -s(t) ending like Frisian anyhow, details aside): (9) ‘have’ Town Frisian Bildts cf. Frisian cf. Dutch

PRESENT – PAST dou hest – dou hast dou hest – dou hast or haddest do hest – do hiest je hebt – je had

The -st suffix is a case of contextual flexion, since it involves a relation between the verb and the subject. The flexion is insensitive to lexical factors so we are dealing with contextual flexion exclusively. We now expect that the -st ending will be projected in the grammar of the contact dialect. This is what happens, since there is nothing in the second language to prevent this from happening. Let me explicate this. Dutch verbs have a -t in the second person singular present. The Frisian ending also ends in -t, hence there is no clash. Where Dutch does not have -t, there is no problem either; the Frisian -st forms are comparable to Dutch verbs having a stem ending in -st. My analysis seems to rely on a distinction between the creation of inflected forms and the recognition of inflected forms, the latter being a more superficial process. Language contact typically creates new contact dialects because of the mismatch between the creation of words and the recognition of words, put differently, because of the mismatch between grammar and parser.7 Hence the survival of -st into the contact dialects is exactly what we expect. Unexplained is the lack of -st in West-Frisian.


We can only suggest that -st got lost as a later development, though this suggestion is void in the absence of historical evidence confirming or disconfirming the hypothesis. 4.2.2 The naked verb-first construction (Imperativus-pro-Infinitivo or IPI) The Imperative-pro-Infinitive construction is a construction which, in spite of its name, has as its most characteristic property that the verb, a bare verb form or an infinitive, appears at the beginning of the clause: (10) Frisian Ik bin net fan doel [en praat der mei him oer] I am not of purpose [and talk it with him about “I do not intend to talk with him about it.”

Here the bare verb form praat ‘talk’ of the second conjunct appears at the beginning of the clause, instead of at the end. Normally, the embedded clause would have the infinitive verb form at the end: (11) Frisian Ik bin net fan doel [der mei him oer te praten] I am not of purpose [it with him about to talk-inf “I do not intend to talk with him about it”

In older Frisian, even at the beginning of this century, the verb appeared as an infinitive at the beginning of the clause. The development towards a bare verb form is very recent. The IPI-construction, with an infinitive, also occurs in mainly northern Dutch dialects, but not in Standard Dutch, neither written nor spoken. This construction is also found in Bildts, Town-Frisian and Groningen dialects, but not in West-Frisian, as shown below: (12) IPI-construction attested? Dutch West Frisian Town Frisian Bildts Frisian Groningen Saxon

no no yes yes yes yes yes

Examples from Town Frisian and Bildts are given in (13)–(14).

282 Eric Hoekstra

(13) Town Frisian Súst wel wizer weze en loop dy meid achterna would-2sg prt wiser be and walk that girl after “You wouldn’t be so foolish as to woo that girl.” (14) Bildts Dou mâgst d’r wel om dinke en doen soks nooit weer you may there prt about think and do such never again “You should take care never to do such a thing again.”

The absence of the construction in West-Frisian may be taken as an indication that the construction is not an ancient one. It also occurs in ‘Frisian Dutch’ (the Dutch of educated speakers from the province of Fryslân), which may be taken as further evidence that it is easily transferred from Frisian to Dutch in language contact. Movement is treated in generative grammar as in instance of contextual flexion. In the IPI-construction, the verb surfaces next to the complementizer en ‘and’ under the influence of that complementizer. It is not a lexical property of the verb that it surfaces in that position; this occurs to check off semantic features. Given the contextual nature of movement, we correctly predict that the grammar of the contact dialects exhibit the IPI-construction, seeing furthermore that no lexical or phonological property in the grammar of Dutch blocks this. Again we cannot explain why West-Frisian behaves differently in this respect. 4.2.3 Verb clustering The same argument can be made for verb clusters, which are head-final in Frisian and in the Frisian contact dialects but not in Dutch. Verb clustering is an instance of movement, and therefore we expect that Frisian word order will be projected in the contact dialects. This turns out to be correct.8 4.3 Contextual and inherent checking: The case of infinitival endings I will briefly summarize here the results of research reported elsewhere more fully (Hoekstra 2000). 4.3.1 Distribution of the two infinitives in Frisian One of the most noticeable characterics of the Frisian language family is the presence of two infinitival endings, as was already noted in Halbertsma (1865). This is not only a phonological matter but also a syntactic one since the distribution of each ending is syntactically determined. The two endings are -e


and -en. The latter is usually but not obligatorily pronounced as a syllabic nasal. The distribution of the two infinitives in Frisian is given below. (15) The infinitive in -e is found a. as a complement to modal auxiliaries Ik sil him opskilje I shall him up-phone “I will phone him up.” b. as a complement to the causative verb litte ‘let’ Lit him mar gewurde let him just become “Just let him be.” c. as a complement to the auxiliary for preposed verbal complements dwaan ‘do’ Boeken lêze docht er komselden books read does he rarely “Read books, he does rarely.” d. in unselected infinitival sentences Ik boeken lêze? I books read “Me read books?” (16) The infinitive in -en is found: a. as a complement to verbs of perception such as hearre ‘hear’ Ik hear Jarich al laitsjen I hear Jarich already laugh “I already hear Jarich laugh.” b. as a complement to the verb gean ‘go’ Ik gean sitten / stean / lizzen / oan ’e rekstôk hingjen I go sit / stand / lie / on the bar hang “I’ll sit down / I’ll stand up / I’ll lie down / I’ll hang from the bar.” c. as a complement to the verb bliuwe ‘stay’ Ik bliuw sitten / stean / lizzen / oan ’e rekstôk hingjen I remain sit / stand / lie / on the bar hang “I’ll remain sitting / standing / lying / hanging from the bar.” d. as a complement to the verb komme ‘come’ plus a particle of directional movement Dêr komt Jan oan rinnen there comes Jan along walk “John comes walking along.”

284 Eric Hoekstra

e.

as a nominal complement, e.g. to various determiners such as it ‘it’, dat ‘that’ It rinnen docht my deugd the walking does me virtue “To walk is good for me.” f. as a complement to the infinitival marker te ‘to’ Om te winnen in order to win “In order to win.” g. as complement to the verb hawwe ‘have’ Hy hat it boek op tafel lizzen he has the book on table lie “He has the book lying on the table.”

4.3.2 The distribution of the infinitives in the contact dialects The following table shows the distribution of the infinitives in -e and -en in the contact dialects: (17) contexts (15) (16)

West-Frisian -e -en

Town Frisian -e -en

Bildts Groningen -e -n -en -n

Saxon -n -n

As can be seen, the Frisian system is retained in the Dutch-Frisian contact dialects, but not in the Saxon-Frisian contact dialects.

Survival of the Frisian system in the Dutch contact dialects and its death in the Saxon contact dialects Consider first the case of Dutch-Frisian contact dialects as they developed. There is no phonological barrier for interpreting the Frisian endings as Dutch, as Dutch exhibited arbitrary phonological variation between its infinitival endings -en and -e. Thus Frisian speakers could impose the syntactic conditioning of the choice between -e and -en on the contact dialect, an instance of contextual checking. Hence we find the Frisian system of infinitives in the Dutch-Frisian contact dialects: West-Frisian, Town Frisian and Bildts. Consider next the case of Saxon-Frisian contact dialects. Saxon required the -en, Frisian required either -e or -en. The Frisian -e now no longer can survive, since it violates the Saxon phonological requirement that -en be present. Therefore, the -e gets lost and only -en survives. Under this view, speakers recognize a phonological requirement in the second language, Saxon, such that 4.3.3


infinitives always end in -n. Thus inherent checking of the infinitival ending for -n blocked the contextual imposition of the -e ending in the Saxon contact dialects.

5. Conclusion The advantage of language contact theory is, first, that it is consistent with what we know of history. History offers many known examples of speakers of one language adopting a second language which, imperfectly learned and hence influenced by the dying mother tongue, changed into a first language. That is, for example, how Latin spread into France at the cost of Celtic, changing into French in the process. On a smaller scale, the same process must have taken place at the level of dialects. We have attempted to spell out the mechanism of the creation of new contact dialects under the influence of language contact. It seems as if the grammar of the mother tongue survives into the contact dialect if it can superficially look like the second language. Thus the syntactic distribution of infinitives in Frisian survived into the contact dialects involving Dutch because the endings are phonologically similar to what is found in Dutch. By the same line of reasoning, the Frisian infinitives did not survive into contact dialects involving Saxon. The same applies to the -st ending, which survived into the contact dialects because it is superficially similar to the Dutch system having either -t or ø. Word order phenomena survive trivially since they do not involve phonological or lexical conditions in the second language; they pattern with contextual flexion. Our account implies that inadequate second language learning primarily targets the phono-lexical component. Hence irregular morphology is not projected from the mother tongue into the contact dialect. We have presented a complex typological body of facts involving phenomena from contact dialects, which not only serve as a test for the theory proposed here but which may also serve as a test for theories proposed by others.

Notes * I would like to thank the participants of the 15th Workshop on Comparative Germanic Syntax for comments and discussion. 1. The study of second language acquisition is thus relevant for explaining why a contact dialect has a specific shape and not another shape, seeing that the contact dialect came into existence as an example of imperfect second language acquisition. Put differently, second language acquisition is relevant for historical linguistics and dialectology.

286 Eric Hoekstra

2. The international community refers as “West-Frisian” to what is referred to as “Frisian” within the Netherlands, that is, the official second language of the province of Fryslân. Within the Netherlands, the term “West-Frisian” is reserved for the dialect spoken in WestFriesland, in the North of the province of North Holland, that is, on the West side of the IJsselmeer. In this article, I will reserve the term West-Frisian for that dialect. 3. The recession of Frisian in the province of Groningen may be due to the advance of the Saxon dialect of Drente or to the advance of the Plattdeutsch dialect of adjacent Germany, or both. Plattdeutsch and Drente are closely related Saxon dialects whose influence may well have combined to oust Frisian from Groningen. Many (rich) people from Drente emigrated to Groningen (Naarding 1961), and when the East of the province was turned into fertile land, there was migration from adjacent Germany (see Schmitt 1942). 4. Eric Reuland proposed to explain differences in the behavior of anaphors in this vein. He related their distribution to PF and LF licensing. The type of licensing depended on the morphophonological shape, PF-licensing being associated with conscious effort and LF-licensing with unconscious application. 5. Johannes Hoekstra (1993) analyses the process of linguistic change involving je-verbs within Frisian. Contrary to e.g. Thomason and Kaufman (1988), he claims that in spite of the large heterogeneity of the material linguistic systematicity can nevertheless be observed. This bears out our earlier claim that linguistic systematicity can only be observed if the sociological context is kept constant. 6. Van Bree (1994, 1997) analysed Town Frisian within the framework of Van Coetsem, and reached the conclusion taken over and defended here: Town Frisian arose as a contact dialect with Frisian as a mother tongue. 7. The actual mental situation is more complicated: we have two grammars, a Frisian and a Dutch one, and two parsers, a Frisian and a Dutch one. I am implying that Frisian grammar is checked off by a Dutch parser. The contact dialect is fed by the grammar of the first language as filtered by the parser of the second. The parser of the second language is much more developed than the grammar of the second language, since comprehension is better developed than production when people do not speak a language perfectly yet. This fact is also well-known from first language acquisition: children’s passive knowledge is greater than their active knowledge. 8. It has been objected, correctly, by Rolf Bremmer (1997) that Middle Frisian did not have verb-final order. Note, however, that this applies to written Middle Frisian. If spoken Middle Frisian did not have verb-final order either, then it is accidental that we nowadays find verbfinal order in the area where Frisian was once spoken. On the other hand, we know that written language may be completely different from spoken language. My guess is therefore that Middle Frisian shows head-initial orders under the influence of the Dutch written standard.

Syntax and Semantics

Swiping in Germanic* Jason Merchant University of Chicago

Establishing the level of representation or the point in a derivation at which movement takes place has never been a trivial matter, and as such remains an topic of substantial ongoing interest. For overt movement, this question is complicated by the availability in principle of two components in which movement could take place with indistinguishable effects on word order: in the derivation leading to Spell-Out, or in the mapping from Spell-Out to PF. To a great extent, the reasoning brought to bear on this question has been concentrated on A- and A¢-movement and their properties; head-movement, in contrast, has remained a distant third. In this article, I show that a little-studied peculiarity of elliptical wh-questions in Germanic can cast new light on this question, providing evidence that there is indeed head-movement which takes place late in the derivation of PF, after Spell-Out. The peculiarity in question comes from a range of data found only under sluicing in a subset of the Germanic languages. In particular, it is found in sluices involving certain prepositions, in which the [+wh] object of the preposition appears not after the preposition in the usual head-complement order, but before it, as in (1). (1) Peter went to the movies, but I don’t know who with.

I will call this kind of exceptional inversion of the usual order of the preposition and its argument swiping, for sluiced wh-word inversion with prepositions (in Northern Germanic). Any examination of swiping or other curiosities of sluicing must, of course, begin with the necessary background on sluicing itself, which is given in Section 1 with particular attention to the facts across Germanic. Section 2 lays out the data on swiping and establishes the two generalizations that will be the

290 Jason Merchant

focus of attention. The analytical challenges raised by these data are taken up in Section 3, which explores the ramifications of swiping for our understanding of pied-piping, ellipsis, and head-movement in particular. Section 4 concludes by re-examining the crosslinguistic situation in light of these ramifications and showing that swiping can perhaps shed light on the internal structure of wh-pronouns as well.

1.

Background on sluicing

Sluicing is a variety of ellipsis widely found in the world’s languages, and illustrated for English in (2) and (3). (2) a. Jack bought something, but I don’t know what. b. Someone called, but I can’t tell you who. c. Beth was there, but you’ll never guess who else. (3) a. Jack called, but I don’t know {when/how/why/where from}. b. Sally’s out hunting — guess what! c. A car is parked on the lawn — find out whose.

Analyses of sluicing have varied over the years (see Ross 1969, van Riemsdijk 1978, Levin 1982, Lobeck 1995, Chung, Ladusaw, and McCloskey 1995, Giannakidou and Merchant 1998, Romero 1998, Lasnik 1999, and Merchant 2001), but I will follow Ross (1969) and Merchant (2001) in assuming that sluicing consists of the movement of a wh-phrase to SpecCP in a constituent question followed by the deletion of the sentential portion, given schematically in (4). (4)

CP XP[+wh]

C' C[+wh, +Q]

IP ... t ...

One of the strongest pieces of evidence for adopting such an analysis over analyses that posit no movement and sometimes no structure at all in the ellipsis site, is the close parallel between the availability of stranded prepositions under regular wh-movement (when no ellipsis is involved) and under sluicing. Generally

Swiping in Germanic 291

speaking, a language L will allow preposition-stranding under sluicing if and only if L allows preposition-stranding under regular wh-movement. Thus in languages such as English, Frisian, Swedish, Norwegian, Danish, and Icelandic, which all allow regular argument wh-phrases such as who to strand a preposition under wh-movement (the (b) sentences in examples (5)–(10)), we also find the possibility of omitting a preposition that corresponds to a preposition marking the correlate of the wh-phrase in the antecendent to the sluice, as shown in the (a) sentences in (5)–(10). (5) English a. Peter was talking with someone, but I don’t know (with) who. b. Who was he talking with? (6) Frisian a. Piet hat mei ien prutsen, ma ik wyt net (mei) wa. Piet has with someone spoken but I know not (with who b. Wa hat Piet mei prutsen? (7) Swedish a. Peter har talat med någon; jag vet inte (med) vem. Peter has talked with someone I know not (with who b. Vem har Peter talat med? (8) Norwegian a. Per har snakket med noen, men jeg vet ikke (med) hvem. Per has talked with someone but I know not (with who b. Hvem har Per snakket med? (9) Danish a. Peter har snakket med en eller anden, men jeg ved ikke Peter has talked with one or another but I know not (med) hvem. (with who b. Hvem har Peter snakket med? (10) Icelandic a. Pétur hefur talað við einhvern en ég veit ekki (við) hvern. Peter has spoken with someone but I know not (with who b. Hvern hefur Pétur talað við?

In other Germanic languages, such as German, Swiss German, Yiddish, and Dutch which generally do not allow preposition-stranding under wh-movement, retention of the preposition under sluicing is obligatory:

292 Jason Merchant

(11) German a. Anna hat mit jemandem gesprochen, aber ich weiß nicht, Anna has with someone spoken but I know not *(mit) wem. *(with who b. *Wem hat sie mit gesprochen? (12) Swiss German (Glarus dialect)1 a. Der Ruedi hät ds ganz Läbe vumene Land träumt, aber ich wäiss the Ruedi has the whole life of.a land dreamt, but I know nüd *(vu) welem. not *(of which-dat ‘Ruedi has dreamt his whole life of some country, but I don’t know of which.’ b. *Welem Land häsch ds ganz Läbe vu träumt? (13) Yiddish a. Zi hot mit emetsn geredt, ober ikh veys nit *(mit) vemen. she has with someone spoken but I know not *(with who b. *Vemen hot zi mit geredt? (14) Dutch2 a. Anna heeft met iemand gesproken, maar ik weet niet Anna has with someone spoken but I know not ??/?/%(met) wie. ??/?/%(with who b. */??/?Wie heeft zij mee gesproken?

This correlation is not limited to the Germanic family, of course, though Germanic does almost seem to have a monopoly on productive prepositionstranding under wh-movement. In a number of non-Germanic languages (viz. Greek, Russian, Czech, Polish, Slovene, Serbo-Croatian, Bulgarian, French, Catalan, Spanish, Italian, Persian, Hindi, Hebrew, Arabic, and Basque), we find this correlation as well: they allow preposition-stranding neither in nonelliptical wh-questions nor under sluicing (with some qualifications irrelevant here; see Merchant 2001). The correlation illustrated above finds its most natural explanation in the theory of sluicing that takes it to consist of the usual operation of wh-movement, subject to the usual language-particular constraints, followed by deletion of the IP out of which extraction has taken place. For example, in English, both the derivation represented in (15a) and that in (15b) are allowed: in the first,


the option of pied-piping has been taken, while in the second, the option of stranding the preposition has been taken.3 Both result in grammatical outputs, just as in their non-elliptical counterparts. (15) Peter was talking with someone, but I don’t know a. [with who]1 [IP he was talking t1]. b. who2 [IP he was talking [with t2]].

In a language such as German on the other hand, which does not allow preposition-stranding, wh-movement must pied-pipe a selecting preposition.4 This yields (16a) as the only possibility; the wh-movement in (16b) violates whatever constraint it is that rules out preposition-stranding in non-elliptical structures in German. (16) Anna hat mit jemandem gesprochen, aber ich weiß nicht, Anna has with someone spoken but I know not a. [mit wem]1 [IP sie t1 gesprochen hat]. [with who she spoken has b. *wem2 [IP sie [mit t2] gesprochen hat]. who she [with spoken has ‘Anna spoke with someone, but I don’t know with who.’

While this state of affairs is expected under the approach to sluicing adopted here, it does have the perhaps slightly disappointing result that sluicing can shed little interesting light on the mechanisms that regulate preposition-stranding cross-linguistically, a topic of long-standing debate (see Takami 1992 for a recent overview). Only one minor conclusion relevant to this debate can be drawn from this range of facts: whatever requires pied-piping of prepositions is not solely a constraint operative at the PF interface on the stranded item (as mooted in a generalized form by Chomsky 1995): since in (16b) the illicitly stranded preposition mit is deleted via ellipsis, it cannot trigger a PF interface crash. The force of these data from preposition-stranding under sluicing in the Germanic languages is to lend strong support to the approach adopted here, that sluicing is the result of wh-movement followed by deletion of IP at PF.

2. The empirical extent of swiping Swiping, though rarely discussed, has not entirely escaped notice in the literature. Ross 1969, for example, in his original investigation of sluicing, noted the

294 Jason Merchant

phenomenon, and other researchers since have remarked on it (see Section 3.1 below). But examples have been few and far between, and little in the way of a sustained empirical investigation has ever been carried out. In this section, therefore, I lay out the fundamental facts and establish the two generalizations which will form the basis for the theorizing in the following section. The two major descriptive questions that arise with respect to swiping are the following: 1. With what kinds of wh-phrases does swiping occur? 2. In what environments does swiping occur? These questions will be taken up in the following two sections respectively. I begin, however, with a brief overview of the basic facts, giving some historical and crosslinguistic context. Swiping in English is illustrated in the following examples. (17) a. Lois was talking, but I don’t know who to. b. They were arguing; God only knows what about. c. This opera was written by someone in the 19th century, but we’re not sure who by. [Chung et al. 1995: (4d)] d. He was shouting, but it was impossible to tell who at. e. A: She got a package in the mail.B: Really? Who from? f. He’ll be at the Red Room, but I don’t know when till. g. Bees are getting into the house, but we can’t figure out where from. h. He sold his farm and moved away, but no-one knows where to. i. Abby quit and got a new job — guess what as! j. She bought a robe, but God knows who for. k. Howard shares the apartment with someone, but I have no idea who with. [Rosen 1976: (16)] l. She fixed it, but she wouldn’t let us in on what with. m. Gordon thought the information had been leaked, but he wouldn’t say who to. n. The tapes indicate that Gordon thought the information had been leaked, but they don’t indicate who by. o. Lisa wants Bart to get involved, but what in isn’t exactly clear. p. Tests indicate the megalith was constructed, but not what of. q. Although we don’t yet know who from, we know she received a package last Monday with instructions on bomb assembly.

Furthermore, speakers accept the following swipes to varying degrees, with judgments best for (18a), slightly worse for (18b), and worst of all, though still marginally possible, for (18c). (These gradations hold for speakers who accept


all three; some speakers reject all three, still others accept only (18a), and still others accept only (18a) and (18b), yielding an inter-speaker implicational hierarchy that maps directly onto intra-speaker variability.) (18) a. %He’s been living in Arizona, but I don’t know how long for. b. %She bought it all right, but don’t even ask how much for! c. %There’s a lot of cities on her list, so she’ll be traveling a lot, but I don’t know how many to.

An attested example is given in the recent American comic strip below (published 22 June 2000 in the Minneapolis Star-Tribune and syndicated newspapers):

But swiping is not a phenomenon limited to recent, perhaps casual, American speech; it is attested also in the careful, British upper-class speech of the characters of Austen, for example: (19) ‘But I will quiz you with a great deal of pleasure, if you will tell me what about.’ [Jane Austen (1775–1817), Mansfield Park, Ch. 5]

Although the previous literature on swiping has dealt with it only in its English manifestation, it is also found in a proper subset of the other preposition-stranding Germanic languages (though not in the non-prepositionstranding Germanic languages, nor to my knowledge outside of Germanic at all). In particular, it is found in Danish and in some varieties of Norwegian (determination of which varieties will have to await further research; of four Norwegian informants, two rejected (20b) while two accepted it). (20) a.

Per er gået i biografen, men jeg ved ikke hvem med. Per is gone to cinema but I know not who with

Danish

296 Jason Merchant

b. %Per gikk på kino, men jeg veit ikke hvem med. Per went to cinema but I know not who with ‘Per went to the movies but I don’t know who with.’

Norwegian

Swiping is not, however, found in Swedish, Icelandic, or Frisian. (21) a. *Per gick på bio, men jag vet inte vem med. Per went to cinema but I know not who with b. *Pétur fór í bíó, en ég veit ekki hverjum með. Per went to cinema but I know not who with c. *Per is nei de bioskoop gien, mar ik wyt net wa mei. Per is to the cinema gone but I know not who with ‘Per went to the movies, but I don’t know who with.’

Swedish Icelandic Frisian

This cross-linguistic distribution will be taken up again in Section 4 below. For now, we return to the two important empirical questions about swiping that began this section: with what kinds of wh-phrases does swiping occur and in what environments?5 The next two sections answer these questions respectively, and lay the empirical groundwork for the analysis that follows. 2.1 The range of wh-elements in swiping The variety of wh-expressions that occurs in swiping in English at least is quite limited. We have, in fact, already seen examples of every kind possible. All other types of wh-phrases lead to ungrammaticality, as seen in (22) and (23).6 (22) a. *She bought a robe for one of her nephews, but God knows which (one) for. b. *They were arguing about animals, but we couldn’t figure out what kind about. c. *This opera was written by an Italian composer in the 19th century, but we’re not sure which (composer/one) by. d. *He was shouting to one of the freshmen Republican senators supporting the bomber program, but it was impossible to tell exactly which (senator) to. e. *He’ll be at the Red Room, but I don’t know what time till. f. *She’s driving, but God knows what town to. g. *She fixed it, but she wouldn’t let us in on what tool with. h. *They were riding in somebody’s car, but I don’t know whose in. i. *He’s renting an apartment with a rich guy, and wait till you hear how rich (of a guy) with!


(23) a. *He’s been living in Arizona, but I don’t know how much time for. b. *She bought it all right, but don’t even ask how much money for! c. *She’s going to a lot of States, but I don’t know how many cities to.

The data in (17), (18), (22), and (23) are summarized in Table 1. Table 1.Possible and impossible wh-elements in swiping Swiping possible

Swiping impossible

% who

* which * which one * which composer * whose * how rich * how rich of a guy * what kind * what time * what town * how much time * how much money * how many cities

% what % when % where % how long % how much % how many

It is fairly simple to state the generalization concerning wh-expressions in swiping, which I codify in the following condition. (24) The minimality condition: Only ‘minimal’ wh-operators occur in swiping

‘Minimal’ here means heads (X0s). We notice that they are all monomorphemic (the last three presumably subject to varying degrees of reanalysis across speakers, as reflected in the gradations in speaker judgments noted above).7 All complex (polymorphemic, phrasal: XP) wh-operators, on the other hand, are in the right column. This leaves which and whose; the former is a head on all analyses, while the latter at least on some. I return to their status below. 2.2 In what environments does swiping occur? Swiping, as codified in the acronym itself, occurs only in sluicing. The following examples show that the wh-preposition inversion found in swiping is not possible in any of the other environments in which prepositions select wh-elements: nonelliptical questions, embedded or matrix (a–b), in situ (c), in

298 Jason Merchant

embedded questions with VP-ellipsis (d), in (wh-) pseudoclefts (e), in (it-)clefts (f), in headed relative clauses, finite (g) or infinitival (h), or in free relatives (i). (25) a. b. c. d. e. f. g. h. i.

*I don’t know [who to] Lois was talking. *[Who to] was Lois talking? *Who talked [who to] [what about]? *We know when she spoke, but we don’t know [what about] she did. *[What about] she was talking was Buddenbrooks. *It was Thomas Mann [who about] she was speaking. *I finally met the guy [who about] she won’t shut up. *The officer [who to] to make such complaints is out of the office today. *I always hate [who with] he goes out.

This generalization is equally simple to state: (26) The sluicing condition: Swiping only occurs in sluicing

This condition, and its minimality companion above, form the major explicanda for any theory of swiping.

3. Accounting for swiping We are now in a position to consider the theoretical import of the data established so far. After reviewing all extant accounts of swiping, I argue that the most satisfactory approach to swiping assimilates it to a kind of head-movement at PF, yielding a better data fit than the alternatives, and capturing the minimality and sluicing conditions in a straightforward way. 3.1 Previous accounts The earliest account of swiping is that of Ross 1969, who himself recognized its unsatisfying ad hoc character (Rosen 1976, who follows Ross in all essentials, calls it a ‘feeble expedient’). Ross proposed a transformation that would delete a variable string; the portion of interest here is given in (27) in current notation. (27) … [CP XP[+wh] [IP 1 1

X 2 0

—

(P) — (3 (3

Y] 4 0

… fi OPT


This account suffers from three shortcomings. First, it must countenance non-constituent deletion, an otherwise undesirable and perhaps unattested option. Second, it offers no account of the minimality condition: since it is sensitive only to a stranded preposition, not to the nature of the moved wh-element, all of the sentences in (22) and (23) above can be generated (Rosen 1976: 209 fn 1 notes this problem, but leaves it as an unresolved difficulty). Third, though least damaging, without further qualification it predicts that any language that allows P-stranding should also allow swiping, a predication falsified by Frisian, Icelandic, and Swedish (of course, the force of this objection is somewhat blunted pending a more complete understanding of what exactly regulates the distribution of swiping cross-linguistically to begin with). An alternative account to Ross’s is that proposed in van Riemsdijk 1978 (followed by Lobeck 1995 and Chung et al. 1995 as well), which assimilates swiping to R-pronoun inversion in German and Dutch.8 As is well known, in these languages certain pronouns (known as R-pronouns) appear before their selecting preposition (van Riemsdijk argues persuasively that they are in SpecPP); examples are given below. (28) a. Womit wolltest du es tun? b. Waarmee wou je het doen? wherewith wanted you it do ‘What did you want to do it with?’

German Dutch

Under this approach, English swiping is simply a parochial version of R-pronoun inversion. As interesting as the parallel to R-pronouns is, however, this analysis suffers from a number of shortcomings. Again there is little hope of capturing the minimality condition by applying usual analyses of R-pronoun inversion in Dutch and German to the English and Scandinavian cases (see Section 4.2): the R-pronouns are distinguished by a feature [+R], not by ‘minimality’ in the sense needed here. In other words, it is mysterious why swiping in English and Scandinavian should allow a wider range of operators than occur as R-pronouns in Dutch and German. Likewise, there is no account of the sluicing condition — R-pronouns are not limited to ellipsis environments in Dutch and German, and it is unclear how the sluicing condition could be captured. Finally, Richards 2001: 139–140 suggests that swiping is a case of stranding in an intermediate SpecFP:

300 Jason Merchant

CP

(29) DP1

C

FP PP2 t1

F

IP ... t2 ...

This account, based on a number of principles regulating which element in a chain is pronounced, supports an ingenious explanation of the sluicing condition: essentially, this stranding in an intermediate specifier is possible only because the lower trace t2 is inside an ellipsis site (roughly, the hypothesized algorithms that determine pronunciation will crash if t2 must be considered as well as t1 and DP1; ellipsis of a constituent containing t2 avoids this conflict). As appealing as this aspect of the analysis is, we are again left with no account of the minimality condition: there is no reason not to expect both phrasal and ‘minimal’ wh-expressions to be extractable from the PP in SpecFP in (29). Furthermore, allowing preposition-stranding in intermediate landing sites raises a different specter of ‘Postal’s problem’ (Postal 1972), the problem of ruling out examples like (30), given successive-cyclic movement: (30) *Who1 do you think [CP [PP for t1]2 [IP she bought it t2]]?

Most commonly, the solution to Postal’s problem has been sought in the conditions on preposition-stranding in general: Kayne 1981 and Hornstein and Weinberg 1981 pursue conceptually similar analyses in which a stranded preposition must incorporate into or be ‘reanalyzed’ with a governing V or possess certain government properties. If these analyses are on the right track, then examples like (30) will be ruled out independently (though see Takami 1992 and Baltin and Postal 1993 for criticism of the ‘reanalysis’ approach). Even if Richards’ approach to (30) is adopted, however, Postal’s problem reemerges when ellipsis targets a constituent containing the origin site for movement, as in (31). (31) a.

We need to know which bills Wellstone said he’d vote for, and *which bills1 he said [CP [PP for t1]2 [IP he wouldn’t [VP vote t2]]]. b. *Wellstone voted for more bills than OP1 he said [CP [PP for t1]2 [IP he would [VP vote t2]]].


A swiping incarnation of this problem also arises. Although wh-extraction is allowed out of VP-ellipsis sites (especially under certain conditions of contrast, satisfied here; see Hardt 1993 for discussion), nevertheless wh-phrase inversion remains impossible in such cases, as we saw above in (25d), repeated here. (32) *We know when she spoke, but we don’t know [what about] she did. (cf. We know which books she read, but we don’t know which magazines she did.)

Parallel to the reasoning above, the ellipsis of the VP containing the trace of the moved PP should have the effect of voiding the PF crash otherwise caused by such structures; PP movement to SpecFP creates a legitimate PF object on this theory, because tPP is elided. Finally, since this account rests on additional functional structure in the clausal architecture, we would like to find independent evidence for the existence of the posited FP, or at least some explication of its role besides hosting the intermediate movement necessary for swiping. In sum, all extant accounts of swiping stumble on the minimality condition, the sluicing condition, or both. 3.2 Swiping as head movement at PF In this section, I propose to analyze swiping as involving head movement of the wh-word to its selecting preposition. Such head movement accounts directly for the minimality condition, and is discussed in Section 3.2.1. I then show in Section 3.2.2 that the sluicing condition can be captured if this head movement occurs at PF, after Spell-Out and after elliptical deletion. 3.2.1 Head movement and the minimality condition Recall from Table 1 above that the set of elements that occur in swiping are exactly those that are heads (who, what, when, and where being the core cases; how long, how much, and how many being subject to variable reanalysis as heads). If swiping is head movement of the wh-word to its selecting P, we expect exactly this group to be able to ‘incorporate’ into P, as in the derivation of (33b) from (33a). (Here and below I represent the wh-pronominals as Ds, without projection: I assume that they are here ‘minimal maximal’ elements, like clitics, in Chomsky’s 1995: 249 terms.)

302 Jason Merchant

(33) a.

PP

→ whoD

toP

b.

PP whoD+toP

twho

By using head movement to derive the observed inversion, we correctly rule out the possibility that phrases will participate in swiping, since structure preservation rules out XP-movement to an X0 as in (34). (34) a.

PP toP

b. * DP

PP DP+toP

whichD {oneN/composerN}

twhich one

which one

Likewise, we can now see why which does not occur in swiping, despite the fact that it is a head. Unlike the wh-pronouns, which must select a complement (which may be null due to NP-ellipsis; see Lobeck 1995). As such, headmovement of which to P as in (35a,b) would result in D-excorporation, disallowed in English (see Baker 1988). Compare also (35c), disallowed for the same reason. (35) a.

PP toP

b. * DP

whichD

PP whichD+toP

N

DP

twhich

N

c. *She’s driving (to some town), but God knows what to town.

The account extends as well to the absence of whose in swiping, under any analysis of this element. If whose is analyzed as consisting of who in the specifier of a DP headed by the genitive ’s (i.e., [DP who [’sD [NP …]]), then its absence in swiping is explained by structure preservation as in (34). If instead whose is analyzed as a wh-pronoun (and hence potentially a head) in the genitive case (whoGEN, assigned genitive case by a null DGEN as in Abney 1987), then the derivation (36a,b) will be ruled out by whatever rules out possessor raising in English in general. Compare again the overt counterpart in (36c).


(36) a.

PP toP

b. * DP

PP

whose+toP

whose

DP

twhose DGEN

N

DGEN

N

c. *They were riding in somebody’s car, but I don’t know whose in car.

The advantage of this account is that by positing head movement as the mechanism that drives the inversion found in swiping, we account directly for the minimality condition. But if the only evidence for treating the wh-pronouns as heads were from swiping, we might be uneasy. Fortunately, however, there is independent evidence that these elements can behave like heads. This evidence comes from the behavior of modifiers which attach either to heads or to phrases, but not to both. One kind of modifier of heads is the series of elements used to mark ‘aggressive non-D-linking’ such as the hell, on earth, etc. (see Pesetsky 1987: 111 (40a) and Den Dikken and Giannakidou (2002) for English, and Hoekstra 1993 for Frisian).9 These markers attach only to heads as in (37), never to phrases as in (38).10 (37) a.

Who the hell was he talking to? What the hell was he talking about? {When/where/why} the hell was he talking? b. What the hell book was he reading!? c. What the hell kind of a doctor is she, anyhow!?

(38) a. *What book the hell was he reading!? b. *What kind the hell of a doctor is she, anyhow!? c. *What kind of a doctor the hell is she, anyhow!?

Crucially, the wh-pronouns retain their head status as diagnosed by the attachment of these modifiers even in swiping: (39) a. He was talking, but God knows who the hell to. b. He was talking, but God knows what the hell about.

The second kind of evidence comes from modifiers with the opposite property. These include exactly and for example, which attach only to phrases as in (41) and not to heads as in (40).

304 Jason Merchant

(40) a. *Which exactly train did they take? b. *What for example books should he be reading? c. *What exactly kind of a doctor is she? (41) a. {Exactly} Which train {exactly} did they take? b. What books for example should he be reading? c. {Exactly} What kind of a doctor {exactly} is she?

Again we find this distribution reflected under swiping as well: (42) a. *He was talking about something, but God knows what exactly about. b. A: You should talk to somebody in marketing for help with that. B: *Could you tell me who for example to?

These diagnostics thus pick out exactly the desired class of wh-pronouns, in perfect agreement with the data from swiping.11 The minimality condition, then, leads us to assume naturally that the movement behind the exceptional order of wh-word and preposition in swiping is head-movement. 3.2.2 Prosodically conditioned head movement and the sluicing condition To account for the sluicing condition we must examine the interaction of headmovement and ellipsis. Though doing so requires us in part to rely on systems whose workings are at present not fully understood, the potential rewards are great, since this goes directly to the question of whether some or all headmovement occurs after Spell-Out (see Chomsky 1995, 2001, Zwart 2001a for the timing of head-movement(s) in particular, and Aoun and Benmamoun 1998, Bobaljik 2002, and Sauerland 1999 for proposals relating to XP movement after Spell-Out). If we can establish that the head-movement involved in swiping must take place after deletion, itself a PF operation as argued in Section 1 above, we have powerful new evidence for taking head-movement in at least one of its manifestations to be a late process, occurring in the mapping from Spell-Out to PF (in common but slightly misleading terminology, ‘at PF’), and perhaps essentially phonological in nature. Although certain details of the algorithms for assigning prosodic prominence will be irrelevant here, the logical structure of the argument should be clear. Essentially, it rests on three premises that have already been established: 1. that swiping involves (a kind of) head movement (of D to P), 2. that swiping occurs only in sluicing, and 3. that sluicing is IP-deletion at PF. From these we can conclude that swiping must be linked to the absence of prosodic material in


IP, i.e., to IP-deletion having occurred. Since IP-deletion occurs after Spell-Out, so must the head-movement involved in swiping. Conditioning the head-movement on its immediately following prosodic environment may allow us to build an account of another fact of swiping: exceptionally, in these cases the main stress falls on the preposition, not on its complement. The facts are in (43), where []φ represents the prosodic phrase formed by the CP. (43) Ben was talking, but I don’t know a. [who TO]φ b. *[WHO to]φ c. [to WHO]φ d. *[TO who]φ

In both grammatical cases, the prosodic prominence falls on the final element of φ, a result of general head-final prominence algorithms operative in English, however encoded (cf. the nuclear stress rule and its descendents); for our purposes, it will suffice to call this constraint HeadFinal (i.e., the prosodic head of φ should be final in φ). These prominences interact in crucial ways with the phrasing algorithms for focus as well (see Schwarzschild 1999, Büring 2001), and in particular with a set of facts discussed in Rosen 1976. Rosen claims that swiping is possible only when there is no antecedent for the preposition. She points out that the best cases of swiping are of adjunct PPs with no antecedent, as in most of the cases examined so far, and identifies two important classes of antecedents that disallow swiping entirely: when the PP is the predicate as in (44a) and when the PP is an idiom chunk as in (44b) (similar to this second case are cases like (44c), where a predicate selects a preposition which combines with the predicate in an idiosyncratic way). (44) a. We were with somebody. I forget who (*with). b. Smersh intends to do away with someone. Find out who (*with). c. She got involved in something over her head, but I don’t remember what (*in).

She immediately qualifies this blanket claim, however, by noting that there are many cases of swiping where a preposition does occur in the antecedent nonetheless: (45) She fixed it with something, but God only knows what with.

306 Jason Merchant

She claims that examples like (45) represent interference from performance factors. In terms of the theory of sluicing presented in Merchant (2001), the possibility of swiping in (45) but not in (44) comes from the possibility of taking either a VP or the IP in the antecedent sentence to be the licenser for deletion of IP. (That such an option is necessary independently of the cases considered here is argued in Merchant (2001): Section 5.4.3.) In swiping, because of HeadFinal, the preposition is focused, and hence, ceteris paribus (as here), the content of the preposition should not be GIVEN, in Schwarzschild’s 1999 sense. This can be formally captured by applying Schwarzschild’s constraint AvoidF — here we will only be interested in AvoidF violations triggered by the preposition, not those potentially triggered by the wh-word (the computation for the latter is significantly more involved; see Romero 1998). For a usual, antecedentless case of swiping, the analysis is as in (46): (46) She fixed it, but God only knows … a. what WITH. HeadFinal –, AvoidF – b. *WHAT with. HeadFinal *, AvoidF –

If the preposition is obligatorily part of the antecedent (marked IP:A), AvoidF will be violated: (47) [IP:A We were with somebody]. I forget a. WHO. HeadFinal –, AvoidF – b. *who WITH. HeadFinal –, AvoidF *

With adjunct PPs, on the other hand, the lower VP segment, excluding the PP, can be taken as the antecedent to satisfy the identity condition on ellipsis (see Merchant 2000 for independent evidence for this). When the lower VP-segment is taken, the preposition in the sluicing clause will not have an antecedent for deletion, and hence must be retained; otherwise, the result will violate the conditions on deletion (as in *She fixed it, but God only knows what [IP she fixed it with t]). In (48), both HeadFinal and AvoidF are satisfied. (48) She [VP [VP:A tshe fixed it] with something], but God only knows what WITH.

With argument PPs, this way out is not possible: the PP will always form part of the antecedent, be it IP or VP, making F-marking on the preposition under swiping an unavoidable violation of AvoidF. Let us assume, therefore, that the account presented thus far is correct in its essentials: swiping is derived by head-movement of a wh-pronoun to a preposition


in SpecCP after IP-deletion has occurred, after Spell-Out.12 The question that these conclusions do not address is the question of what, in featural terms, permits or triggers the movement itself. Of course, the answer to this latter question is independent of the conclusions reached so far: whatever is triggering the movement does not affect the fact that this movement is head-movement, nor that it must occur after Spell-Out. In fact, answering this question in full cannot be done in isolation from an answer to the question of what permits preposition-stranding to begin with. It is clear, despite the prosodic sensitivity noted above, that the head-movement in swiping must also be linked in some way to the presence of the [+wh] feature that drives wh-movement in the first place. As such it is related to the mechanisms for pied-piping, possibly affording us a rare look into the mechanism of feature-passing (or ‘percolation’). Although usually feature-passing is not visible in the overt syntax (i.e., there is no obvious component of the lower X (from which the passed feature F originates) on the higher Y (to which F is passed)), swiping perhaps presents just such a case. One way to analyze featurepassing is as feature movement from a lower head onto a higher one. Usually this feature movement has the property of only moving the relevant feature, with the observed effects of making the higher Y a target for movement sensitive to F (as in PP pied-piping under wh-movement). Note that this kind of pied-piping is the exception, not the rule: usually, pied-piping proceeds from the specifier position of the pied-piped YP, where feature-passing can be handled with the regular mechanisms for spec-head agreement and head-XP projection (see Webelhuth 1992, Aissen 1996, Kennedy and Merchant 2000 for recent discussion and references). In contrast to specifier pied-piping, piedpiping from a complement position as in wh-PPs has remained more mysterious. Take (49) for example: (49) [PP with what] did she fix it?

We must assume that the [+wh] feature F that originates on what is passed on to the PP, marking the PP as [+wh] and hence a possible attractee for movement to SpecCP, where it checks F against a compatible F in C. The simplest mechanism in current terms for implementing this ‘passing on’ is feature movement of F from what to with in (49). Usually, feature movement that occurs before Spell-Out, as must happen in (49) for the PP to be displaced, must pied-pipe the originator X of F, due to ‘PF convergence requirements’ (as Chomsky 1995 discusses for I-to-C movement; cf. the lack of I-to-C movement in matrix sluices, as independently noted by Lasnik 1999 and Merchant 2001).

308 Jason Merchant

Seen from this perspective, the swiping cases are in fact the expected ones, since the originator X has been moved along with F, adjoining to the target Y, in this case a preposition. The mystery that remains is why cases like (49) are wellformed, in fact almost the only possible option in languages that displace wh-phrases. Here again we run up against the current lack of any sufficiently deep understanding of the mechanisms of preposition-stranding and its flip side, pied-piping of prepositions. Essentially, this reduces to the larger question of how degrees of analyticity are to be encoded in the grammar, both within and across languages. Unfortunately, the parametrization of analyticity is in its infancy. Apparently in most languages, the connection between D and P is stronger than in English, Frisian, and the Scandinavian languages. If we think of degree of analyticity in terms of whether or not head-to-head feature movement is obligatory, then we can say that in most languages, feature movement from D to P must take place, while in the preposition-stranding languages it is only the case that such movement may take place. It is important, of course, to remember that we are talking about analyticity only in the domain of prepositions and their objects; all other domains can and will require independently determined degrees of analyticity. We might then think of swiping in English, for example, as a manifestation of the fact that English stands at the extreme end of the scale of analyticity in this domain; apparently certain elements of the grammar are somewhat unstable at such extremes. The instability in English is manifested in the ability of feature movement, in most languages without ramifications for the relative position of D and P, to pied-pipe its originator. For some reason, when feature movement becomes optional, pied-piping also becomes an option, subject to the constraints on head movement observed above; a possible way to think about this rare correlation is that by making feature movement optional, we make room for economy — that is, as opposed to when a movement is obligatory, in which case economy considerations play no role. Compare the three possibilities below, where strikethrough indicates a trace of movement. (50) a.

b. P

c. P[F]

D[F]

D[F]+P D[F]

D[F]

The option in (50a), where no feature movement has occurred, is the most economical, since no operations apply. (This fact is irrelevant for languages like


German etc., since in these languages feature movement is obligatory by hypothesis.) This leaves options (50b) and (50c). I would contend that a grammar that supports such optionality is in imbalance; there is no reason for the less economical options ever to be taken, but since they must be generated, some kind of grammatical patch is kludged together to permit the attested structures despite economy considerations (cf. Slobin’s 1997 ‘grammatical viruses’). Specifically, a movement operation must be posited that is not subject to economy considerations (since otherwise the availability of (50a) would block it). Once such a movement operation is added to the grammar, given that it is by hypothesis uneconomical, or immune to economy, suddenly the comparison between ‘mere’ feature movement in (50b) and the usually more costly head-movement in (50c) becomes moot: the movement can effect either of these, leading to the results we have seen. Whatever the results of these extended considerations, however, we are left with a fairly persuasive argument that the sluicing condition reduces to a sensitivity to IP-ellipsis, detectable only after Spell-Out, and which must therefore be captured by mechanisms, prosodic, featural or otherwise, which operate in the mapping from Spell-Out to PF. Combined with the conclusion of the previous section then, an argument emerges for a kind of head-movement at PF.

4. The cross-linguistic distribution of swiping We are now in a better position to consider again the at first glance somewhat puzzling distribution of swiping across the Northern Germanic languages (English, Frisian, and the Scandinavian languages), seen above in (20) and (21) and repeated here (further English-specific questions regarding swiping will have to await further research).13 (51) a.

Per er gået i biografen, men jeg ved ikke hvem med. Per is gone to cinema but I know not who with b. %Per gikk på kino, men jeg veit ikke hvem med. Per went to cinema but I know not who with ‘Per went to the movies but I don’t know who with.’

(52) a. *Per gick på bio, men jag vet inte vem med. Per went to cinema but I know not who with

Danish Norwegian

Swedish

310 Jason Merchant

b. *Pétur fór í bíó, en ég veit ekki hverjum með. Per went to cinema but I know not who with c. *Per is nei de bioskoop gien, mar ik wyt net wa mei. Per is to the cinema gone but I know not who with ‘Per went to the movies, but I don’t know who with.’

Icelandic Frisian

While English, Danish, and some varieties of Norwegian allow swiping, Frisian, Icelandic, and Swedish do not. Though a detailed explanation of the presence or absence of swiping in these languages remains to be constructed, some preliminary conclusions can be drawn. It is clear that these data rule out simply assimilating swiping to preposition-stranding across the board, as early researchers did; the absence of swiping in Swedish, Icelandic, and Frisian (and for certain speakers of Norwegian, who nonetheless allow preposition-stranding in general) indicates that merely being able to strand a preposition is not a sufficient predictive criterion for swiping. Ideally, however, these data do open another window into the mechanisms of preposition-stranding in general. In terms of the analysis presented above, we would like to see if the above distribution correlates with other, perhaps only very subtle, differences in wh-movement among these languages, in particular differences between head and phrasal movements. If swiping is an instance of head movement, and if the difference among the Northern Germanic languages with respect to swiping is to be traced to properties of head movement, we might expect to find independently differing properties of head movement among these languages. Another possibility is that the availability of swiping is to be traced to the Janus-faced nature of pronominals in the swiping languages, behaving as heads for some purposes and as phrases for others, similar to clitics; by hypothesis, Frisian and Swedish wh-pronominals must be phrasal (must project to XP) at all levels of analysis. While positing such lexical differences in the category of the pronominals in question fits with the analysis to correctly capture the distribution seen above, we would like to uncover some independent evidence that such a distinction is necessary. Perhaps surprisingly, there are some suggestive indications that at least English and Frisian differ in exactly the way required. Hoekstra 1993 discusses several strands of evidence relating to certain pecularities in the distribution of Frisian wat ‘what’ which he argues can be explained by assuming that wat does not behave like a simple pronominal (i.e., as a (projecting) D with no complement as with English [DP what]). Rather, he shows that wat occurs only as a D that selects an NP: [DP wat [NP e]] (cf. Eng. which).14 Because of this obligatory selection of a complement (even though null), a branching structure will result,

Swiping in Germanic

requiring projection to DP. The primary evidence comes from an intricate set of facts concerning what Hoekstra analyzes as the clitic complementizer dat, too involved to go into here. Though wat does not share all its properties with the other simplex wh-words in Frisian, we can note that wa ‘who’ also has a relevant property that contrasts with its English congener — unlike in English, wa can select an NP directly: Wa minkse docht soks no (lit. who human.being does such now) ‘What kind of a person does something like that now?’ Hoekstra argues further that wa in at least some of its uses must be like wat in selecting a null NP complement. Although this evidence from Frisian is only partial, it is highly suggestive, and if the conclusions drawn from the differences between English and Frisian in this domain are correct, it may be indicative that the absence of swiping in Frisian can be traced to independent aspects of the projection properties of Frisian wh-pronominals. It remains of course to be seen whether similar evidence can be adduced for the Scandinavian languages. It also remains an important fact that swiping is found only in languages that allow preposition-stranding under usual wh-movement, surely no accident. Although many authors have sought to relate the availability of prepositionstranding to the nature of the V-P relation (via incorporation or ‘reanalysis’), the evidence from swiping seems to indicate that we should instead concentrate on the relation of P to D in these languages, searching for a property of this latter relation to build an analysis of preposition-stranding on, as mooted in the previous section.

5. Conclusion The nature and timing of movement operations have been central themes in linguistic theory since Aspects and continue to be so; as such, any source of illumination bearing on them is welcome. In this article, I have argued that one such source comes from a rather modest and largely overlooked set of data hidden away in the nooks and crannies of the grammars of certain Germanic languages. Based on the generalization regarding the nature of the wh-elements that occur in swiping, I concluded that swiping must instantiate a kind of headmovement, of a wh-word to a preposition. Combining this with the evidence that sluicing structures are derived by deletion at PF, largely based on the correlation between preposition-stranding in nonelliptical interrogatives and under sluicing, I argued that the head-movement involved in swiping is most

311

312

Jason Merchant

naturally understood as taking place at PF as well, that is, in the mapping from Spell-Out to PF. Though certain details remain unfortunately unclear, due to our incomplete understanding both of the mechanisms that regulate this mapping and of pied-piping in particular, I think it is safe to conclude that swiping does provide an argument that at least one kind of head-movement occurs after Spell-Out, with effects on word order. If the reasoning laid out here is correct, it is also an indication that even the most unassuming curiosity gathering dust in the grammar’s attic may potentially be a source of unexpected insight into the design and construction of the whole.

Notes * Thanks to Marcel den Dikken, Anastasia Giannakidou, Jarich Hoekstra, Eric Potsdam, and Peter Svenonius for comments and questions that furthered my thinking on these issues. Thanks also to the audience at the 15th Comparative Germanic Syntax Workshop at the University of Groningen. I am also grateful to the many informants and native speaker linguists who took the time to answer my queries and render judgments on swiping, in particular Line Mikkelsen (Danish), Ger de Haan, Jarich Hoekstra, Oebele de Vries (Frisian), Øystein Nilsen, Peter Svenonius, Tarald Taraldsen, Øystein Vangsnes (Norwegian), Anders Holmberg (Swedish), and Gunnar Hannson (Icelandic). Unless otherwise stated, other sluiced examples in Dutch, German, Swiss German, Icelandic, and Yiddish are taken from Merchant (2001), and I remain indebted to the informants thanked there. This work was supported in part by the Netherlands Organization for Scientific Research (NWO). 1. E. Haeberli (p.c.) informs me that the indicated judgments hold for other dialects of Swiss German as well. 2. The situation in Dutch appears to be the most fluid of the languages examined here, which is reflected in the variety of stigmata used in both in the (a) and (b) examples in (14), representing the variety of responses received from informants. Many informants (6 of 10) accepted the sluiced version without the preposition (all accepted it with the preposition); some informants (2 of 10) also accepted preposition-stranding in the non-elliptical question in (14b), though the correspondence was less than perfect (normative factors perhaps influencing the judgment in non-elliptical test cases). In general Dutch appears to be a transition case, allowing preposition-stranding in a wider range of cases than German, for example, but in a more restricted set than Frisian. Normatively rejected preposition-stranding under A¢-movement is attested even in carefully edited literature; compare the examples in (i) (from J. J. Voskuil, Meneer Beerta: Het Bureau I, Van Oorschot, Amsterdam, 1996, p.65): (i) Onrechtvaardigheid wind ik me over op. injustice work I refl about up ‘Injustice, I get worked up about.’

Swiping in Germanic

From the same work comes the following bare sluiced wh-phrase whose correlate is in a PP (op. cit., p.31): (ii) De jongen leek op iemand, maar hij kon niet bedenken wie. the boy seemed on someone but he could not think who ‘The boy looked like someone, but he couldn’t think who.’ So given the fluidity of the situation in current Dutch usage, the variety of judgments found for the sluicing examples does not seem entirely unexpected. 3. A small caveat applies to the English data here: in the (rather restrcited and unnatural, highly normative, primarily written) register that allows pied-piping, the accusative form whom is preferred, hence the example in the text with the preposition might have for some speakers the slight feeling of a register clash, which I abstract away from. 4. The only exception to this comes from R-pronouns (also in Dutch and Swiss German), which can strand prepositions under wh-movement. These, however, can never be the sole remnant under sluicing, as seen in (i), because the wh-R-pronoun cannot be focused in this environment for independent reasons (i.e., in (i), the stress must fall on von, not on wo, though either stress is possible in other environments; see Gussenhoven 1983, Reis 1985, Hoekstra 1995), which conflicts with the requirement imposed by sluicing that the operator, not the preposition, bear primary focus (due to the interaction of the presence of the correlate preposition in the antecedent and the identity condition on sluicing); see Merchant 2001: Ch. 2 for more data and discussion. (i) Kepler hat von etwas geträumt, aber ich weiß nicht, {wovon /*wo}. Kepler has of something dreamt but I know not {whereof/*where ‘Kepler dreamt of something, but I don’t know what.’ 5. A third relevant question, but one which I will not attempt to address here, concerns the kinds of prepositions that swiping occurs with; it appears on brief inspection that most ‘simplex’ prepositions can occur in swiping, while many ‘complex’ ones cannot. At a first approximation, for example, swiping appears to be allowed with about, after, as, at, by, for, from, in, near(?), of, on, till, to, under(?), and with, but not with above, because of, before, between, despite/in spite of, during, instead of, into, on top of, regarding, or underneath. As Marcel den Dikken (p.c.) points out, this distribution might very well lend support to the analysis presented here, if complex prepositions prohibit incorporation on independent grounds, and if the ‘simplex’ vs. ‘complex’ distinction can be independently motivated. 6. I do not give full data for [−wh] elements selected by prepositions as in (i) and (ii), displaced or in situ, in elliptical environments or not: in no such case is the inversion characteristic of swiping found. (i) a. *Andy bought a present {[Beth for] / [her for] / [himself for]}. b. *{[Beth for] / [her for] / [himself for]} Andy bought a present. (ii) a. *Andy voted for Gore and Beth [Nader for]. b. *Andy voted for Bush because Beth did [Nader for]. 7. The wh-elements how and why do not occur in swiping, since they are never selected by a preposition.

313

314 Jason Merchant

8. The structural part of van Riemsdijk’s analysis (namely, that the wh-element is in SpecPP) seems to be endorsed in Culicover 1999 as well, though Culicover argues in general that swiping is ‘sui generis’, and that the PP is exhaustively dominated by a CP node (“radically violating standard X¢-theory”, p.138). 9. Cf. the distribution of the free relative morpheme -ever (whoever, whatever, wherever, whenever, however, whichever), thanks to T. Stowell (p.c.) for bringing this parallel to my attention. (i) a. b.

How{ever} potent {*ever} this symbol is, … How {the hell} potent {*the hell} do you think this is?!?

10. As J. McCloskey (p.c.) notes, the distribution of the hell in (i) supports the first treatment of whose above, as who + ’s, and argues against Abney’s analysis of whose as a genitive-marked head. (i) a. Who the hell’s car is parked on the lawn?! b. *Whose the hell car is parked on the lawn?! 11. I note here that this phenomenon bears certain similarities to D-to-P incorporation in other languages as well (see Baker 1988). A similarly limited form of D-to-P movement is found in Latin with the preposition cum ‘with’: while the argument of cum generally appears to the right as in (i), personal pronouns appear to the left. (i) cum femina ‘with {the/a} woman’ (ii) mecum

‘with me’

12. An alternative that comes to mind is that the head-movement takes place before wh-movement, and that this head-movement somehow conditions the later IP-ellipsis, once the amalgamated PP has reached SpecCP. It is not clear to me, however, how to understand the mechanisms that would be involved in such an account, nor whether such an account could successfully be combined with the independently needed trigger for sluicing in nonswiping cases. (Another possibility is that there is a feature E on D that must be checked against the feature E on C which triggers IP-deletion (see Merchant 2001). Movement of D to P would allow this E, like the feature [+wh], to be checked in a spec-head configuration with C[+wh, E].) A nontrivial complication for any account analyzing the head-movement in swiping as occurring prior to Spell-Out is that such head-movement must presumably be undone or otherwise fixed by LF, in order for the wh-element to be interpreted in the usual way (typically taken to be in SpecCP). 13. For example, one interesting puzzle that remains concerns the distribution of what … for in its idiomatic meaning synonymous with why, as in (ia). The puzzle arises from the fact that in this meaning, the preposition for cannot be pied-piped, as in seen in (ib). (i) a. What did you do that for? b. *For what did you do that? ‘Why did you do that?’ Given this state of affairs, one might expect that what for would not occur in swiping, if the initial pied-piping of for into SpecCP that is necessary to feed the derivation of swiping is ruled out in principle. This expectation is incorrect, however; (iia) is possible, which must have the structure in (iib).

Swiping in Germanic

(ii) a. b.

He did it, but I don’t know what for. … [CP [PP what1+for t1]2 [IP he did it t2]]

Thus whatever rules out pied-piping in (ib) must be repaired either by the deletion of IP or by the head-movement in the PP in (ii). Another interesting puzzle comes from a further curiosity to be found in sluicing, and its intersection (null, as it turns out), with swiping. The phenomenon in question is known as ‘multiple sluicing’ and concerns examples like (iii), which are possible for some speakers in English, and in which apparently more than one wh-phrase has survived IP-ellipsis (see Merchant 2001: Section 4.1 for discussion and references). (iii) Everyone brought something different to the potluck, but I can’t remember who what. The question that arises is whether swiping can occur with either or both of the multiple remnants in ‘multiple sluicing’. The answer is no, as we see for the first remnant in (iv) and for the second in (v) (though see Richards 1997, who reports related data with slightly differing judgments): (iv) Jack lived with different people each year of college, but I don’t know a. [with who] when b. *[who with] when. (v) Everyone was talking (with someone different), but it was impossible to tell a. who [with who]. b. *who [who with]. 14. Note that Frisian wat, like English what, can also take an overt NP complement (Hoekstra 1993: (26)): Wat lju binne dat? lit. what people are that, i.e., ‘What kind of people are they?’; cf. What fool left the lights on? and What thinking person would agree to that? (this what, like the determiner which, does not appear in swiping, as expected). Similarly, if wat is used to question a previous utterance, it can select the ‘dummy’ nouns ding ‘thing’ or guod ‘stuff’: Wat ding?, wat guod? (meaning roughly ‘What did you say?’ or ‘Beg your pardon?’; cf. the identical usage of Greek Ti pragma?, Italian Che cosa?, both lit. what thing).

315

The ambiguity of weak reflexive pronouns in English and German* Markus Steinbach University of Mainz

1.

Introduction

It is a well-known fact that sentences containing a (weak) reflexive pronoun are ambiguous in many languages, cf. e.g. Kemmer (1993), Geniušiene˙ (1987), or Dobrovie-Sorin (1998). In German, transitive sentences with an accusative reflexive pronoun in the position of the direct object have multiple readings. The anticausative, middle, and reflexive interpretation of transitive reflexive sentences are illustrated in (1) (‘RP’ stands for reflexive pronoun).1 (1) a.

Die Tür öffnet sich the-nom door opens RP-acc ‘The door opens’ b. Das Buch liest sich leicht the-nom book reads RP-acc easily ‘This book reads easily’ c. Peter rasiert sich Peter-nom shaves RP-acc ‘Peter is shaving (himself)’

(anticausative)

(middle)

(reflexive)

That is, the reflexive pronoun in (1) can but need not be linked to a semantic argument of the verb. Only in (1c) is the reflexive pronoun both a syntactic and a semantic argument of the verb. By contrast, it is only a syntactic but not a semantic argument in (1a) and (1b), i.e. it is c-selected but not s-selected. We call the reflexive pronoun in (1c) argument reflexive and the one in (1a) and (1b) non-argument reflexive. The latter indicates valency reduction of the first semantic argument. The subject is linked to the second semantic argument in these examples.

318

Markus Steinbach

As opposed to transitive reflexive sentences, intransitive sentences do not involve reduction of the first semantic argument. The subject of intransitive sentences is always linked to the first semantic argument of the verb, which is either an unergative or unaccusative one-place predicate or a two-place predicate with an implicit second semantic argument.2 It is generally assumed that the sole semantic argument of unergatives is external (i.e. it has typical proto-agent properties) and the one of unaccusatives is internal (i.e. it has typical proto-patient properties). (2) a.

Peter schläft/erwacht Peter sleeps/awakes b. Peter liest Peter reads

(unergative/unaccusative) (impl. second argument)

Non-reflexive transitive sentences do not involve valency reduction at all. In (3) both the subject and the direct object must be linked to a semantic argument of the two-place verb. The subject is linked to the first semantic argument and the direct object to the second one. (3) Peter liest dieses Buch Peter reads this book

The interpretation of intransitive and transitive reflexive sentences in German is systematically illustrated in the following table. Intransitive sentences might involve reduction of the second semantic argument, whereas transitive reflexive sentences might involve reduction of the first semantic argument. Hence only in transitive reflexive sentences can the subject be linked to the second semantic argument of the predicate. We will see in Section 4 that this is exactly the grammatical function of non-argument reflexives. In (4) ‘Ø’ stands for the semantic argument which is deleted in anticausatives and the parentheses indicate an implicit semantic argument. (4) Intransitive and transitive reflexive sentences in German Syntax

Example

Semantics

a. intransitive: Subject + Verb

Peter schläft Peter erwacht Peter liest

unergative V ·xÒ unaccusative V ··yÒÒ V ·x ·(y)ÒÒ impl. sec. argument

V ·Ø ·yÒÒ anticausative b. transitive reflexive: Die Tür öffnet sich Subject + Verb + RP Das Buch liest sich leicht V ·(x) ·yÒÒ middle construction Peter rasiert sich V ·x ·xÒÒ reflexive

The ambiguity of weak reflexive pronouns in English and German 319

At first sight English seems to differ form most Indo-European languages. Unlike German, English does not indicate reduction of the first semantic argument by means of a reflexive pronoun. Transitive reflexive sentences do not involve valency reduction in English, as can be seen in (5a). In (5a), as well as in the corresponding non-reflexive sentence (5b), both syntactic arguments are obligatorily linked to a semantic argument. As opposed to the corresponding German sentences in (1), transitive reflexive sentences in English always get a reflexive interpretation. (5) a. Peter is shaving himself b. Peter is reading this book

(reflexive)

Anticausative and middle constructions are syntactically intransitive. Thus, in English intransitive sentences might involve reduction of both the first and the second semantic argument of the verb, cf. (6b), (6c), and (6d). Moreover, intransitive sentences can also get a reflexive interpretation with some verbs, as can be seen in (6e). That is, intransitive sentences in English get all three interpretations of transitive reflexive sentences in German. Besides, unergatives and unaccusatives are also intransitive in English. (6) a. b. c. d. e.

Peter sleeps/awakes Peter is reading The door opens The book reads easily Peter is shaving

(unergative/unaccusative) (impl. second argument) (anticausative) (middle) (reflexive)

Table (7) summarizes the interpretations of intransitive and transitive reflexive sentences in English. In addition to the syntactic differences between English and German anticausatives and middle constructions, both languages differ in another respect. Unlike German, English has two constructions that receive a reflexive interpretation: transitive reflexive sentences like (5a) and intransitive sentences like (6e). By contrast, in German the reflexive interpretation is restricted to transitive reflexive sentences.3

320 Markus Steinbach

(7) Intransitive and transitive reflexive sentences in English Syntax

Example

Semantics

a. intransitive: Subject + Verb

Peter sleeps Peter awakes Peter reads The door opens The book reads easily Peter shaves

V ·xÒ V ··yÒÒ V ·x ·(y)ÒÒ V ·Ø ·yÒÒ V ·(x) ·yÒÒ V ·x ·xÒÒ

unergative unaccusative impl. sec. argument anticausative middle construction reflexive

Peter shaves himself

V ·x ·xÒÒ

reflexive

b. transitive reflexive: Subject + Verb + RP

Because middle constructions in English are not reflexive, they are often analyzed as a kind of unmarked passive. Syntactic theories, for example, assume that the first semantic argument (the external theta-role) is linked to an empty pronominal element. The second semantic argument (the internal theta-role) is regularly linked to the D-structure object, which moves to the subject position for reasons of case, cf. e.g. Stroik (1992, 1999) or Hoekstra and Roberts (1993). According to lexical theories, a lexical rule suppresses the first semantic argument and promotes the second one. As a consequence, the latter is linked to the subject, cf. e.g. Fagan (1992) or Ackema and Schoorlemmer (1994, 1995). Note, however, that crosslinguistically non-reflexive middle constructions are a great exception. Most languages use (weak) reflexive pronouns to indicate valency reduction in middle and related constructions. Moreover, intransitive sentences in English show the same ambiguity as their reflexive counterparts in languages like German.4 Therefore we want to argue that we gain greater insight into the regularities of valency reduction, if the analysis of intransitive sentences in English is embedded in the context of corresponding reflexive constructions in other languages. We will see that the ambiguity of intransitive sentences in English is as regular as the ambiguity of transitive reflexive sentences in German. We confine ourselves to the anticausative, middle, and reflexive interpretation illustrated in (1) for German and (6c–e) for English and ignore further aspects of intransitive and transitive reflexive sentences in English and German. This article is organized as follows. The next section discusses briefly the morphosyntactic properties of weak reflexive pronouns in Indo-European languages. In Section 3 we turn to German. First, we argue that German differs from most Indo-European languages, including English, in having only one type of reflexive pronoun. German does not tell weak from strong reflexive pronouns. Second, we discuss the relevant morphosyntactic properties of the

The ambiguity of weak reflexive pronouns in English and German

reflexive pronoun in transitive reflexive sentences, which we call a middle marker. Section 4 shows that all three interpretations of transitive reflexive sentences illustrated in (1) can be uniformly derived from the same syntactic representation.5 Section 5 deals with the ambiguity of intransitive sentences in English. We discuss to what extent the analysis of transitive reflexive sentences in German can be applied to the corresponding intransitive sentences in English. Section 6 summarizes the main findings of this article.

2. Weak reflexive pronouns Most Indo-European languages use weak reflexive pronouns to indicate the reduction of the first semantic argument.6 Moreover, a weak reflexive pronoun is always ambiguous between an argument reflexive and a non-argument reflexive interpretation. Hence we observe the same regularities as in the German examples in (1). The reflexive pronoun can but need not be linked to a semantic argument of the verb. Consider, for example, the Russian examples in (8), which correspond to the German examples in (1). The weak reflexive from -sja, a verbal affix, also induces the anticausative, the middle, and the reflexive interpretation. Additionally, -sja also induces the passive interpretation, which is not available for transitive reflexive sentences in German. Dom stroitsja (plotnikami) (passive)7 house build-sja (by the carpenters ‘The house is being built (by the carpenters)’ b. Dver’ (legko) otkrylas’ (middle constr./anticausative) door-nom (easily opened-sja ‘the door opened (easily)’ c. Ivan moetsja (reflexive) Ivan-nom washes-sja ‘Ivan is washing himself ’

(8) a.

The same ambiguity can be found in Italian or French. Middle constructions in Italian and French, unlike their counterparts in German, need not receive a generic interpretation. They can refer to particular events. In this case the interpretation of (reflexive) middle constructions is nearly identical to the interpretation of (periphrastic) passives in Italian and French.8 We turn to the differences between passives and middle constructions in Section 4.

321


(9) a.

un veston de laine se lave facilement a jacket of wool RP washes easily b. la branche s’ est cassée the branch RP is broken c. Pierre se rase Pierre RP shaves

(middle construction) (anticausative) (reflexive)

Crosslinguistically, weak reflexive pronouns have quite different morphosyntactic properties. They can be verbal affixes, verbal clitics, or independent words, cf. (10). Nevertheless, they seem to be always ambiguous between an argument and a non-argument interpretation and the latter usually induces valency reduction. Following Kemmer (1993), we call the reflexive pronouns in (1), (8), and (9) ‘middle markers’. Consequently, the sentences in (1), (8), and (9) can be subsumed under the notion of middle voice. (10) Possible middle markers –verbal inflection –verbal affix –verbal clitic –accusative reflexive pronoun

Language –Greek –Russian –Italian –German

To summarize, weak reflexive pronouns are generally ambiguous. This ambiguity of weak reflexive pronouns seems to be universal.9 However, weak reflexive pronouns are also subject to language specific restrictions, which result from the specific morphosyntactic properties of different languages. We discuss first the general morphosyntactic properties of weak reflexive pronouns. Then we turn to the interaction between these universal properties of weak reflexive pronouns and language specific restrictions on the middle voice (i.e. transitive reflexive sentences) in German. How can we account for the difference between argument and nonargument reflexives? Note that German, like many other languages, does not morphologically distinguish reflexive from personal pronouns in the first and second person. Although German does not always have two morphologically distinct forms, it discriminates between reflexive pronouns and personal pronouns in principle. This observation can be connected to the principle of morphological economy: a NP that is locally bound must be maximally underspecified, cf. Burzio (1992) and Gärtner (1991). Reflexive pronouns are less specified than personal pronouns (and personal pronouns are less specified than R-expressions). Hence, whenever a language discriminates between reflexive and personal pronouns, they must be used in different (syntactic) contexts.


On the other hand, if there is only one pronominal form, it can be used in every context a pronominal form can occur in, cf. e.g. the first and second person object pronouns in German. The logic of this argument can be illustrated as follows. (11) X is locally bound

Æ

X is maximally underspecified (i.e. there is no Y that is less specified than X)

(11) does not imply that every maximally underspecified element must be locally bound. In German the first and second person pronominal forms are maximally underspecified (there is no comparable pronominal form that is less specified), but they need not be locally bound. In sum, only maximally underspecified pronominal expressions can be locally bound. This observation relates to a second property of maximally underspecified elements we want to argue for in this article. Pronominal expressions which are maximally underspecified can be locally bound and they are not lexically specified for the morphosyntactic feature [R]. According to Reinhart and Reuland (1993) and Reuland and Reinhart (1995), the feature [R] is a morphosyntactic feature, that depends on the inherent specification of a lexical item. All nominal expressions are intrinsically specified for Φ-features. In addition, only elements the Φ-features of which are not maximally underspecified are lexically specified as [+R]. Consider, for example, reflexive pronouns and personal pronouns in German. (12) and (13) illustrate the specification of the Φ-features number, person, gender, and case for personal pronouns and reflexive pronouns respectively (homonymous forms within one paradigm are written in italics and homonymous forms in both paradigms are underlined). (12) Personal pronouns in German person

gender

nominative accusative

dative

singular

1 2 3

Ø Ø masc. fem. neuter

ich du er sie es

mich dich ihn sie es

mir dir ihm ihr ihm

plural

1 2 3

Ø Ø Ø

wir ihr sie

uns euch sie

uns euch ihnen


(13) reflexive pronouns in German person

nominative

accusative

dative

singular

1 2 3

mich dich sich

mir dir sich

plural

1 2 3

uns euch sich

uns euch sich

Reflexive pronouns in Modern German differ from personal pronouns at least in two dimensions: they are not specified for gender and they lack nominative case. Furthermore, reflexive pronouns distinguish accusative from dative forms only in first and second person singular, whereas personal pronouns also distinguish third person accusative forms from third person dative forms. Consider, for example, the personal pronoun ihn and the corresponding reflexive pronoun sich.10 (14) a. ihr ·[+s, −p], [−1.p, −2.p, +3.p], [+m, −f, −n], [−nom, +acc, −dat]Ò b. sich ·[Ø], [−1.p, −2.p, +3.p], [Ø], [+acc, +dat]Ò

(14) shows clearly that the reflexive pronoun in (14b) is less specified than the personal pronoun in (14a). The specification of a pronoun depends on its morphological properties and its position in the respective paradigm. Pronominal elements with similar feature specification are compared to each other and only the expression with less specified Φ-features is not lexically specified for [R].11 By contrast, a pronominal element that is not maximally underspecified is always [+R]. Thus only the personal pronoun in (14a) is [+R]. In sum, weak reflexive pronouns are not lexically specified for the feature [R]. As a consequence, they can be either [+R] or [−R] depending on the (syntactic) context. We will argue in the Section 4 that this underspecification is responsible for the ambiguity of reflexive pronouns in German. First, we take a closer look at the correlation between reflexive pronouns and middle voice in German.


3. The middle marker in German German has two specific properties: (i) German is a so-called one-form language and (ii) the middle marker in German is a NP that must be assigned structural case. We turn to the distinction between one-form and two-form languages first. Unlike most Indo-European languages, German does not morphologically distinguish between weak and strong reflexive pronouns. Instead, German has only one form of reflexive pronoun that must be used in all contexts.12 Let us compare German and English, which are the two languages we are interested in. Recall that weak reflexive pronouns are used as middle markers. Note furthermore that the reflexive interpretation of weak reflexive pronouns is limited to verbs, which are very likely to describe reflexive events, as, for example, verbs of grooming like shave or wash. In contrast to that, verbs like hear do not permit a weak reflexive pronoun in object position. Similar differences can be found in all languages that distinguish weak from strong reflexive forms, cf. Kemmer (1993). The examples in (15) show that English is a two-form language. The weak form of the reflexive pronoun, which is morphologically empty, is systematically ambiguous between an argument and a non-argument interpretation and can only trigger a reflexive interpretation with verbs like shave. (15) a. b. c. d.

The door opens This book reads easily Peter is shaving (himself) Peter is hearing *(himself)

As opposed to English, German has only one form, which is grammatical with all kinds of verbs, as can be seen in (16) and (17). It must be used in anticausatives and middle constructions, with verbs of grooming and with verbs like hören (‘hear’).13 Note that the reflexive pronoun in (16d) need not be stressed. Thus, the distinction between stressed and unstressed reflexive pronouns in German does not correspond to the distinction between weak and strong forms in two-form languages.14 We put the focus on some other constituent to exclude a ‘strong’ (i.e. stressed) reading of sich in (16d) and (17). (16) a.

Die Tür öffnet *(sich) the door opens *(RP) b. Dieses Buch liest *(sich) gut this book reads *(RP) well


c.

Peter rasiert *(sich) Peter shaves *(RP) d. Hat sich der Peter nur gehört oder auch gesehen? has RP the Peter only heard or also seen? Er hat sich nur [F GEHÖRT] he has RP only heard

Both examples in (17) illustrate the same point. The presence of the unstressed simple reflexive pronoun sich is not restricted to verbs that are likely to be reflexive. (17) a.

… weil sogar [F HANS] sich nun nicht mehr belügt because even Hans RP now no more tells-a-lie b. … weil sie sich nur [F GEStern] im Fernsehen gesehen hat because she RP only yesterday in-the TV seen has

As opposed to two-form languages like English, German is not sensitive to whether a predicate is likely to be reflexive or not. The following figure illustrates this difference between English and German. (18) Example: English German:

anticausatives/middles Ø sich (−focus)

shave Ø & x-self sich (±focus)

hear x-self sich (±focus)

We conclude that German, unlike English, does not distinguish weak from strong reflexive pronouns. One form corresponds to both the weak and the strong form in two-form languages. Let us turn to the middle marker in German now. We already saw that the reflexive pronoun is an independent word.15 Note first that German has personal and impersonal middle constructions. The latter are derived from one-place predicates. Impersonal middle constructions are also transitive reflexive sentences. The reflexive pronoun is bound by an impersonal subject, the third person neuter personal pronoun es ‘it’. Both unergative and unaccusative verbs are grammatical in impersonal middle construction. This is illustrated in (19a) and (19b). (19) a.

Es schläft sich gut in diesem Bett it sleeps RP well in this bed b. Gesundheitsstudie: In welchem Bezirk stirbt es sich am frühesten study-on-health in which district dies it RP at-the earliest

The impersonal subject es is not linked to a semantic argument of the verb and the reflexive pronoun indicates valency reduction again. The only argument of


the one-place verb is suppressed and cannot be linked to syntax, because the subject position is already occupied by es. Example (19b) illustrates another property of MCs in German. The suppressed argument need not be the actor or agent of the verb/event. The only argument of unaccusative verbs has typical patient properties.16 Another crucial property of the middle constructions in German is that the reflexive pronoun must be assigned accusative case. This is illustrated in (20) and (21). The subject of a middle construction cannot correspond to a dative object in the ‘active’ counterpart, cf. Fagan (1992), Abraham (1995a), and Steinbach (2002). Sentence (20b) is ungrammatical if the reflexive pronoun is assigned accusative and it does not yield a middle interpretation if it is assigned dative. A verb selecting a dative object is only grammatical in impersonal middle constructions like (20c). In this case, the dative object preserves its case and the accusative reflexive pronoun indicates again reduction of the first semantic argument. (20) a.

Wir helfen einem Obdachlosen we help a-dat homeless-person-dat b. Ein Obdachloser hilft sich leicht a-nom homeless-person-nom helps RP-*acc/dat easily c. Einem Obdachlosen hilft es sich leicht a-dat homeless-person-dat helps it RP-acc easily (impersonal middle constr.)

Hence, dative reflexive pronouns cannot indicate valency reduction. The same holds for accusative and dative reflexive pronouns assigned case by a preposition. Both sentences in (21) can only get a reflexive interpretation. The reflexive pronoun, which is coreferent with the subject of the sentence, must be linked to a semantic argument of the verb. (21) a.

Peter1 ist sehr von sich1 überzeugt Peter is very of himself-dat convinced ‘Peter is very sure of himself.’ b. Er1 hat mich auf sich1 aufmerksam gemacht he has me to himself-acc attentive made ‘He called my attention to himself.’

Figure (22) illustrates the possible interpretations of reflexive pronouns in German. The accusative reflexive pronoun in the position of the direct object can be called a morphosyntactic middle marker. Figure (22) shows a clear


correlation between reflexivity and middle voice. However, it also illustrates that reflexivity cannot be reduced to middle voice and vice versa. (22) Possible interpretations of the middle marker and reflexive marker in German middle marker inherent reflexive anticausative middle reflexive marker reflexive morpheme

reflexive

P+RP-acc, RP-dat, P+RP-dat acc-RP (direct obj)

4. The Interpretation of argument and non-argument reflexives in German So far, we saw that the middle marker in German has two essential properties: (i) it must be a reflexive pronoun and (ii) it must be assigned structural case. Recall from Section 2 that the former follows from the morphosyntactic specification of (weak) reflexive pronouns in general. The latter is a specific property of the German case system, which distinguishes structural from oblique case. Dative case differs in several respects from accusative and nominative, cf. e.g. Vogel and Steinbach (1998) or Bayer et al. (2001).17 As a consequence, nominative and accusative are structural cases whereas dative case is oblique. We follow Vogel and Steinbach (1998), who argue that in German the term syntactic argument should be defined on the basis of structural case. (23) Only NPs that are assigned structural case (or have a structural case feature to check) are syntactic arguments

Hence, our analysis is based on the following two assumptions: (i) (weak) reflexive pronouns are not lexically specified for [R] and (ii) only nominative and accusative case are structural cases. Both assumptions enable us to derive the first ambiguity of transitive reflexive sentences (i.e. the ambiguity of the reflexive pronoun) illustrated in the second collumn in (24). The second ambiguity is due to two different semantic operations, which are discussed at the end of this section.


(24) The interpretation of the reflexive pronoun and the implicit argument in transitive reflexive sentences in German interpretation of the reflexive pronoun

interpretation of the implicit argument

argument-reXexive RP

non-argument reXexive

a. reXexive interpr. saturation

b. middle interpretation

reduction

c. anticausative interpr.

Reinhart and Reuland (1993) and Fox (1993) argue that a well-formed maximal A-chain is a sequence of coindexed syntactic arguments that contains exactly one [+R]-expression, which must be its head. The definition of maximal A-chains is given in (25). (25) a.

Maximal A-chains: A maximal A-chain is any sequence of coindexation of syntactic arguments that satisfies antecedent government. b. General Condition on A-chains (GCC): A maximal A-chain (a1, …, an) contains exactly one link — a1 — that is both [+R] and case-marked.

According to (25), every syntactic argument that is specified as [+R] must head its own maximal A-chain. By contrast, syntactic arguments that are specified as [−R] must be contained in a maximal A-chain, which is headed by another syntactic argument specified as [+R]. Consider the non-reflexive transitive sentence (26a). Let us assume that both the subject and the direct object move to the specifiers of AgrSP and AgrOP respectively to check their structural case features. A-movement creates two maximal A-chains. The semantic interpretation of structure (26) is written in italics. For sake of simplicity we omit CP and verb movement.


(26) a. Peter read ‘War and Peace’ AgrSP b. NP-NOM1

AgrOP

NP-ACC2 1 2

(lx (ly R 〈x, y〉) (w&p)) (p)

VP t1 p

V′

(ly R 〈x, y〉) (w&p)

t2

V

w&p

R 〈x, y〉

Both chains are headed by [+R]-expressions, which are linked via their VP-internal base positions to a semantic argument of the verb. The first chain is interpreted in SpecVP. According to the linking-principles in (27), SpecVP is always linked to the first semantic argument of the verb. The second chain is interpreted in the complement position of V0. This position is linked to the second semantic argument. (27) a. SpecVP is linked to the first argument of the verb. b. The complement of V0 is linked to the second argument of the verb.

The simplified interpretation of (26) is given in (28). (28) R ·p, w&pÒ

Recall that reflexive pronouns, unlike R-expressions, are not lexically specified for [R]. Therefore they can be either [+R] or [−R]. According to the General Condition on A-Chains (GCC), the latter must be a proper part of a maximal A-chain, which is headed by another A-element specified as [+R]. As a consequence, the [−R] reflexive pronoun is forced to form a maximal A-chain with the subject of the sentence in order to meet the GCC. This complex A-chain must be interpreted in its base position and is thus linked to the second semantic argument of the verb. By contrast, a [+R] reflexive pronoun must head its own chain, because the GCC excludes A-chains which contain two [+R]-expression. Hence, the sentence contains two distinct chains and the chain headed by a [+R] reflexive pronoun is linked to the second argument of the verb itself. The ambiguity of the accusative reflexive pronoun is illustrated in lines b and c of the following table.

The ambiguity of weak reflexive pronouns in English and German

(29) A-chains and [+/−R]-expressions in German

a. simple chain

syntax

semantics

[+R, NP] – [−R, trace]

1 argument

b. complex chain [+R, NP] – [−R, RP] – [−R, trace] – [−R, trace] 1 argument c. two chain d. two chains

[+R, NP] – [−R, trace]; [+R, RP] – [−R, trace]

2 arguments

[+R, NP] – [+R, trace]; [+R, NP] – [−R, trace] 2 arguments

Let us take a closer look at [+R] reflexive pronouns first. The syntactic representation in (30) equals the one in (26). Both structures contain two maximal A-chains, which are each linked to a semantic argument. The reflexive pronoun is linked to the second semantic argument of the two-place predicate and the subject to the first one. (30) a.

Peter wäscht sich Peter washes RP-acc AgrSP

b. [+R]1

AgrOP x, RP ) → Wp〈)p, RP〉 VP (lx (W ly〈W 具x,〉y)典)(p(RP))(

[+R]2

(ly W 具〈x, y典) 〉) (RP) → W 〈x, RP〉 V′ (ly

t1 t2

V

RP

W 〈x, y〉

A reflexive pronoun is usually locally bound.18 Following Pollard and Sag (1994), the reflexive pronoun in (31a) is bound by the less oblique co-argument p. After λ-abstraction on the antecedent we get the final semantic representation in (31b). All arguments coindexed with the antecedent are converted into variables bound by the λ-operator, cf. Reinhart (1983b). (31) a. W ·pi, RPiÒ b. (λx W ·x, xÒ) (p)

The [−R] reflexive pronoun, unlike its [+R] counterpart, must be included in an A-chain the head of which must be a [+R] expression. Hence the [−R]

331


reflexive and the subject must be coindexed in syntax. The complex A-chain is linked again in its base position, the complement position of V0. We already saw that this position is linked to the second semantic argument of the verb, cf. the linking-principles in (27).19 (32) a.

Die Tür öffnet sich the door opens RP-acc AgrSP

b. [+R]1

AgrOP [-R]2

VP t1

V′

(ly O 〈x, y〉) (t)

t1

V

t

O 〈x, y〉

A non-argument interpretation is impossible for dative reflexive pronouns, because dative case is oblique. Recall from (23) that the definition of the term syntactic argument is based on the notion of structural case. Therefore, dative reflexive pronouns are not subject to the GCC. As a consequence, they cannot indicate valency reduction and verbs selecting a dative object can only occur in impersonal middle constructions. SpecVP cannot be linked to the first semantic argument in (32), because this position is not the base of the A-chain. Hence, complex A-chain formation prevents the first semantic argument from linking. The final semantic representation in (33) contains an unbound argument variable. (33) O ·x, tÒ

Following Chierchia (1989b) and Reinhart (1996), we assume that two operations on unbound semantic argument variables are available. They can either be bound by a quantifier or completely removed from the semantic representation. The former operation is called argument saturation, the latter argument reduction, cf. Steinbach (2002). Both operations are illustrated in (34). (34) Saturation: (λy P ·x, yÒ) (a) Reduction: (λy P ·x, yÒ) (a)

Æ Æ

(λy OP x P ·x, yÒ)) (a) (λy P ·yÒ) (a)


Argument reduction is much more restrictive than argument saturation and depends on the lexical meaning of the verb. Reduction changes a two-place predicate (a relation between two arguments) into a one-place predicate (a property). Saturation, on the other hand, introduces a quantifier that binds the free argument variable. Argument reduction is responsible for the anticausative and inherent reflexive interpretation, whereas argument saturation yields the middle interpretation.20 The class of verbs that permit reduction is a subclass of the verbs that permit saturation. The first argument of zeichnen ‘draw’ in (35a), for example, can only be saturated but not reduced, whereas öffnen ‘open’ in (35b) forms a perfectly grammatical anticausative. Öffnen permits both saturation and reduction. (35) a.

Das Bild zeichnet sich (leicht) (middle interpretation only) the picture draws RP (easily/*a little) b. Die Tür öffnet sich (leicht) (anticausative and middle interpr.) the door opens RP (easily/a little)

Reinhart (1996) argues that zeichnen and öffnen differ in one crucial respect. The subject of both verbs is lexically specified for the basic semantic feature [+ causing-change]. But only the subject of zeichnen in (35) is additionally specified for [+ mental state involved]. Therefore, öffnen permits all kinds of [+ causing-change] subjects, e.g. agents, instruments, or other causing entities, because the verb does not obligatorily select the feature [+ mental state involved]. As opposed to öffnen, the very similar verb zeichnen only select agents, which fulfill both specifications, [+ causing-change] and [+ mental state involved].21 (36) a.

Peter/dieser Schlüssel/der Sturm hat die Tür geöffnet Peter/this key/the storm has the door opened b. Peter/*dieser Stift/*der Wind hat dieses Bild gezeichnet Peter/this pencil/the wind has this picture drawn

The basic intuition is that the anticausative interpretation of basically two-place predicates requires that the event described by the verb can be conceptualized as taking place without an explicitly mentioned cause. The opening of a door, for example, can be perceived as if it takes place all by itself. This restrictions on the anticausative interpretation of two-place predicates is part of the lexical meaning of a verb. Reduction of the first argument is only possible if it is not specified as [+ mental state involved] (i.e. if the verb does not entail volitional involvement for this argument). This can be stated in the following way, cf. Reinhart (1996: 19).


(37) Argument reduction P ·x, yÒ & ¬ (x = [+ mental state involved]) Æ

P ·yÒ

As opposed to argument reduction, argument saturation applies to nearly all kinds of verbs that select at least one semantic argument.22 The implicit semantic argument can be either existentially or generically bound. Both quantifiers unselectively bind any free variable in their scope. Unlike other Indo-European languages, German has two different forms for passives and middle constructions, as can be seen in (38). Thus, German shows a division of labor with respect to the interpretation of the first implicit argument. (38) a.

Der Bericht schreibt sich schnell the report writes RP quickly b. Der Bericht wurde schnell geschrieben the report was quickly written

Passives are responsible for existential quantification and middle constructions for generic quantification.23 We follow Krifka et al. (1995) and Cohen (1996), who analyze the generic quantifier as a dyadic semantic operator that relates two open formulas, the restrictor and the nuclear scope, and binds any free variable in its scope.24 Consider the following example. (39) a.

Wie liest sich ‘Krieg und Frieden’? how reads RP ‘war and peace? ‘How can War and Peace be read?’ b. ‘Krieg und Frieden’ liest sich [leicht]F ‘war and peace reads RP [easily c. GENS,X,Y; [y = war-and-peace] [read(s, x, y) & easy(s)]

We already saw that the subject of the middle construction is linked to the second argument position of the verb. In addition, it is mapped to the restrictor in (39c). The adverb leicht ‘easily’ is the focus and the rest of the sentence can be called the background. The focus and the background are mapped together to the nuclear scope, cf. Cohen (1996: 151f). The generic operator takes sentential scope. It binds the ‘topic’ War and Peace, the implicit argument, and the situation variable. The middle construction in (39) is thus a characterizing or habitual sentence, cf. Krifka et al. (1995). Simplifying somewhat the generic operator can be analyzed as the phonologically empty counterpart to the frequency adverb usually. The generic sentence in (39) is evaluated with respect to a set of (salient) alternatives. According to Cohen, this set of alternatives is usually determined by the focus. Recall that the focus semantic value of a


sentence is a set of alternatives. The alternatives to the focus in (39c) are for example {hard, badly, well, quickly, excellently, …}. Hence, sentence (39) would be true if a situation in which someone reads War and Peace is very likely to be a situation in which War and Peace is easy to read for this person.25 This analysis accounts for the fact that middle constructions can be modified by temporal adverbials and adverbs of quantification.26 Furthermore, it correctly predicts that the individual-level predicates in (40) cannot undergo middle formation, because they do not supply a situation variable, cf. Kratzer (1995).27 (40) a. *So wie mein Vater heißt es sich nicht so leicht like my father names it RP not that easily b. *Vom Gorilla stammt es sich nicht so leicht ab from.the gorilla be-descended it RP not that easily prt

So far, we proposed a uniform analysis of argument and non-argument reflexives in German. Our analysis is based on the following two assumptions, which are both independently motivated: (i) (weak) reflexive pronouns are not lexically specified for [R] and (ii) German distinguishes between structural and oblique case. Our analysis accounts for the two essential features of the middle marker in German. A middle marker must be assigned structural case and it must be a reflexive pronoun. Thus we can derive the following observation: (41) Only an accusative reflexive pronoun in the position of the direct object is a middle marker in German

Besides, we argued that the implicit first argument can either be bound by a generic quantifier or deleted. These semantic operations are responsible for the middle and the anticausative interpretations. In the final section, we discuss whether our analysis of the middle voice in German can also be applied to English.

5. Intransitive Sentences in English Recall from Section 3 that English is a two-form language. The weak reflexive pronoun in English is morphologically empty. Consequently, English does not have an overt morphosyntactic middle marker. Intransitive sentences in English correspond to intransitive and transitive reflexive sentences in German. Our analysis of German can be directly applied to English, if we assume that English has an invisible weak reflexive pronoun in object position.28 Like its German


counterpart, this weak reflexive pronoun is not lexically specified for [R] and can thus be either [+R] or [−R]. The [+R] reflexive pronoun in (42) induces the reflexive interpretation of intransitive sentences in English. It heads its own chain, which is linked to the second semantic argument of the verb. Besides, the argument reflexive in (42) is bound by a less oblique co-argument. (42) a. Peter is shaving [+R]-RP AgrSP b. [+R]1

AgrOP [+R]2

VP t1 p

(lx S 〈x, RP〉) (p) → S 〈pi, RPi〉 V′

(ly S 〈x, y〉) (RP) → S 〈x, RP〉

t2

V

RP

S 〈x, y〉

Recall from Section 3 that the reflexive interpretation of intransitive sentences in English is restricted to certain verbs. The distribution of weak and strong reflexive pronouns in two-form languages is subject to additional constraints. Strong reflexive pronouns seem to be intrinsically contrastive.29 Therefore, they must be used with verbs like hear, which are unlikely to describe reflexive events. This seems to block the reflexive interpretation of weak reflexive pronouns with these verbs. German, unlike English, has only one reflexive pronoun, which is responsible for the reflexive interpretation in all contexts. The [−R] reflexive pronoun, on the other hand, must be included in a complex A-chain, which is linked to the second semantic argument of the verb. This is illustrated by the anticausative in (43). Again, the first semantic argument can either be saturated or reduced. The anticausative interpretation in (43) results from argument reduction. In middle constructions, the implicit argument is bound by the generic quantifier parallel to middle constructions in German.


(43) a. The door opens [−R]-RP AgrSP b. [+R]1

AgrOP [–R]2

VP t1

V′

(ly O 〈x, y〉) (d ) → S 〈x, d〉

t1

V

d

O 〈x, y〉

This analysis of intransitive sentences in English relies on the same linking principles we proposed for German in (27) above. In English, intransitive sentences might be either genuine intransitive or basically transitive with an invisible reflexive pronoun in object position. As a consequence, intransitive sentences in English have more interpretations than their German counterparts. The subject of intransitive sentences in English need not be linked to the first semantic argument. Therefore they also receive an anticausative, a middle, and a reflexive interpretation. Note finally that one might argue for an alternative analysis of English. Instead of assuming an invisible reflexive pronoun in object position, we can also modify the linking principles proposed in (27) for German. Consequently, the subject of an intransitive sentence in English can not only be linked to the first semantic argument but also to the second one. This is stated in (44). (44) SpecVP is linked either to the first or second semantic argument of a predicate.

This proposal attributes the ambiguity of intransitive sentences directly to the principles of argument linking. By contrast, the first proposal attributes the ambiguity to different underlying syntactic representations (i.e. presence or absence of an invisible reflexive pronoun in object position). So far, both alternatives seem to account for the ambiguity of intransitive sentences in English. We leave this issue for further research. 6. Concluding remarks In this article we argued for a unified analysis of middles and related constructions in German and English. The ambiguity of intransitive sentences in English


is quite regular if we look at it from the point of view of reflexive constructions in other languages. Our analysis of the ambiguity of transitive reflexive sentences in German and intransitive sentences in English does not rely on special syntactic or lexical operations. It is based on the assumption that weak reflexive pronouns can but need not be linked to a semantic argument of the verb. This seems to be a universal property of weak reflexive pronouns, because weak reflexive pronouns are generally the pronominal elements which are morphologically maximally underspecified. On the other hand, weak reflexive pronouns have quite different morphosyntactic properties across languages. A crosslinguistic study should also consider these language specific properties of weak reflexive pronouns. Therefore the postsyntactic analysis we developed for German and English does not necessarily hold for other languages. Middle voice might be lexical in some languages and syntactic in others.

Notes * I would like to thank Werner Abraham, Artemis Alexiadou, Marie Christine Erb, HansMartin Gärtner, Juliane Möck, Marga Reis, Ralf Vogel, Jan-Wouter Zwart and the audience and the organizers of the Comparative Germanic Syntax Workshop. 1. In addition to these interpretations, transitive reflexive sentences can also get the inherent reflexive interpretation illustrated in (i), which equals the anticausative interpretation. But unlike anticausative verbs like öffnen (‘open’) in (ii), inherent reflexive verbs do not have a corresponding causative variant, cf. (ib). They seem to be basic two-place predicates, the first semantic argument of which is obligatorily deleted. We exclude inherent reflexive verbs from the following discussion, cf. Steinbach (2002) for more details. (i) a.

Peter erkältet sich Peter-nom catches-a-cold RP-acc b. *Hans erkältet Peter Hans-nom catches-a-cold Peter-acc

(ii) a. b.

Die Tür öffnet sich the door-nom opens RP-acc Hans öffnet die Tür Hans-nom opens the door-acc

2. The only exception is the impersonal subject in (i), which cannot be linked to a semantic argument, because verbs like regnen (‘rain’) are zero-place predicates, cf. also footnote 20 below. (i) Es regnet it rains 3. Note that in German intransitive sentences containing verbs like rasieren (‘shave’) do not get a reflexive interpretation. They are interpreted parallel to (2b) above, i.e. the second


semantic argument of the verb is not linked to syntax and it is not coreferent to the first semantic argument (which is linked to the subject). 4. Apart from the fact that most syntactic and semantic theories do not offer a uniform analysis of the systematic ambiguity of intransitive sentences in English, they are all faced with additional empirical and conceptual shortcomings, which are discussed in Fagan (1992), Ackema and Schoorlemmer (1995), and Steinbach (2002, Chapter 3). Besides, there is no morphosyntactic evidence for A-movement or a lexical rule of argument suppression and syntactic as well as lexical analyses cannot account for the presence of a (weak) reflexive pronoun in the corresponding constructions in many other languages. 5. Fagan (1992) and Steinbach (2002) show in detail that all three interpretations of transitive reflexive sentences are related to the same underlying syntactic structure. Both the argument and non-argument reflexive do not differ in syntax, cf. also footnote 14. 6. We are mainly interested in the following four interpretations induced by weak reflexive pronouns: anticausative, middle, passive, and reflexive. We do not discuss language specific restrictions of these constructions and we also ignore further interpretations that are only possible in some languages. 7. In the perfective aspect Russian has a second kind of passive formation. This periphrastic form consists of the auxiliary ‘be’ and the passive participle (dver’ byla zakryta nami, ‘the door was closed by us’, cf. Miller 1993: 238). 8. The examples are from Miller (1993). For a more detailed survey of French middle constructions see e.g. Grimshaw (1982), Fagan (1992), and Dobrovie-Sorin (1998). 9. Middle constructions in Standard Dutch are an interesting exception. Although Dutch has weak reflexive pronouns, Dutch middle constructions, unlike anticausatives, are not reflexive. Note that southern dialects of Dutch have reflexive middle constructions. 10. A discussion of the correct morphological specification of individual forms within these paradigms would lead us beyond our topic. There are at least two issues that require further discussion: (i) homonymous forms within one paradigm (e.g. the personal pronoun sie). (ii) homonymous forms in two paradigms (e.g. the 2nd person reflexive and personal pronouns dich and dir). 11. The following definition accounts for this generalization. (i) A pronominal element X is not lexically specified for [R] iff a. there is another pronominal element Y and X and Y are equally specified for the features ·F1, …, FnÒ and b. only Y is additionally specified for a feature ·Fn+1Ò. (ii) A pronominal element Y is lexically specified as [+R] iff a. there is another pronominal element Y and Y and X are equally specified for the features ·F1, …, FnÒ and b. only Y is additionally specified for a feature ·Fn+1Ò. 12. Note that the complex form sich selbst is formed by adjunction of the adnominal focus particle selbst to the NP sich, cf. e.g. Abraham (1995b), König and Siemund (1996a, 1996b) and Steinbach (2002, Chapter 4). In German, the expression sich selbst is not grammaticalized, as opposed to Dutch zichzelf or English him-/her-/itself. It can always be replaced by the


stressed simple reflexive pronoun sich, although some speakers seem to prefer sich selbst in some contexts. In other languages, weak reflexive pronouns cannot be stressed at all. They are the intrinsically weak counterparts of the corresponding strong reflexive pronouns. 13. Sentence (16c) is grammatical without the reflexive pronoun but does not receive a reflexive interpretation, cf. note 4. 14. Unlike the argument reflexive in (16c) and (16d), the non-argument reflexive in (16a) and (16b) must not be stressed. The distribution of stressed and unstressed reflexive pronouns in German depends solely on the interpretation of the reflexive pronoun itself. Stressed reflexive pronouns are interpreted as the focus of the sentence, which requires that they are linked to an argument variable in semantics. Thus only argument reflexives can be stressed. See Fagan (1992) and Steinbach (2002) for a discussion of the differences between argument and non-argument reflexives in German. 15. In the following we ignore the adverbial modification that middle constructions usually require, cf. Ackema and Schoorlemmer (1994), Iwata (1999), and Steinbach (1999a, 2002), who argue that the presence of the adverbial is a matter of informativity. Further properties of middle constructions are discussed in Section 4. 16. Further examples are two-place verbs like finden (‘find’) or vergessen (‘forget’). Both verbs select neither an actor nor an agent. Besides, examples (i) illustrates that middle constructions are not restricted to present tense. (i) … ein Telefonbuch fand sich nicht a phonebook found RP not ‘… a phonebook could not be found’ 17. The distinction between structural and oblique case is independently motivated. As opposed to accusative and nominative case, dative case is morphologically marked. Movement into case positions is an exclusive property of accusative (A.c.I.) and nominative (passive). Moreover, dative objects cannot be promoted to subject in middle constructions, werden-passives and in tough-movement constructions and they are asymmetrically bound by accusative objects and nominative subjects. Furthermore, German has so-called free and multiple datives but no free and multiple accusatives or nominatives. Further differences between structural nominative and accusative case and oblique dative case concern constraints on word order, extraction, sentential complements, idioms, free relatives and processing asymmetries, cf. Steinbach (2002) for a detailed discussion. 18. Reinhart and Reuland (1993) and Pollard and Sag (1994) argue that reflexive pronouns need not necessarily be locally bound. Reflexive pronouns that cannot be bound by a (less oblique) co-argument are exempt from the structural conditions on binding. Note that our binding theory consists of three parts: (i) syntactic binding (A-chain formation, restricted to [−R] reflexive pronouns that are assigned structural case). (ii) semantic binding (o-binding, restricted to [+R] reflexive pronouns that are locally o-commanded, cf. Pollard and Sag (1994)). (iii) logophoric binding (restricted to reflexive pronouns that are exempt from (ii)).


19. Impersonal middle constructions can be derived the same way. The reflexive pronoun is bound by a [+R]-expression, the impersonal subject es ‘it’. The complex A-chain cannot be linked to a semantic argument, because es is only a quasi-argument, which is not interpreted in semantics. Note that the impersonal subject, which bears the default values third person singular neuter, is the nominative counterpart to the reflexive pronoun. Within the nominative paradigm, the third person singular neuter pronoun is the ‘weakest’ element, which is used as a pleonastic element that fulfills quite different syntactic functions in German. 20. Argument saturation also yields the passive interpretation in languages that have reflexive passives, cf. Section 2 and below. 21. Note that Reinhart’s specification corresponds to Dowty’s (1991) contributing properties for the agent proto-role. Zeichnen has the following two entailments for its subject: (i) volitional involved in the event or state and (ii) causing an event or change of state in another participant, cf. Dowty (1991: 572). Öffnen, on the other hand, has only the second entailment for its subject. 22. Individual-level predicates seem to be the only class that is generally excluded form argument saturation in middle constructions, cf. below. 23. This seems to be a construction specific property. As far as we can tell the generic quantification in middle constructions can be attributed neither to the reflexive pronoun nor to the adverbial modification. In many languages reflexive constructions can also receive a passive interpretation and adverbials do not necessarily trigger a generic interpretation. Besides, anticausatives and simple reflexives are not generic statements, although they are reflexive. In this respect middle constructions equal conditional sentences and (embedded) V2-clauses, which have construction specific semantic properties as well. 24. Krifka et al. call the nuclear scope ‘matrix’. 25. The exact definition of the meaning of the generic quantifier and the set of contextually salient alternatives is a difficult task. We refer the reader to Krifka et al. (1995: 43f) and Cohen (1996). 26. This is illustrated in (i) and (ii). (i) Der Bach hat sich gestern Abend ausnahmsweise mal ganz gut gespielt the Bach has RP yesterday evening exceptionally once quite well played ‘Exceptionally, this piece by Bach played well yesterday evening.’ (ii) a. b. c.

Dieses Buch liest sich immer wieder gut this book reads RP again and-again well Dieses Auto fährt sich normalerweise ganz gut this car drives RP normally quite well Hier pflegte es sich gut zu schlafen here used it RP well to sleep

Besides, locative adverbials in middle constructions are ambiguous between NP-modification and VP-modification, cf. (iii). (iii) … weil sich die Bücher in diesem Laden gut verkaufen because RP the books in this shop well sell


Hence, middle constructions clearly differ from individual-level predicates. The latter do not permit temporal adverbials, adverbs of quantification and the VP-modification of locative adverbials, cf. e.g. Kratzer (1995). Note that Fagan (1992) and Ackema and Schoorlemmer (1994) argue that ‘middle verbs’ are individual-level predicates. The examples in (i), (ii), and (iii) contradict this proposal. By contrast, our analysis correctly predicts that middle constructions are characterizing (habitual) sentences. 27. We can also account for the observation that middle constructions usually attribute a property to their subject. The tripartite structure in (39c) predicates the nuclear scope over the restrictor. That is, for the restrictor of a generic sentence generally holds what is described in the nuclear scope. The ‘responsibility’ of the subject for the event described by the verb seems to be closely related to this, cf. e.g. Fagan (1992) or Greenspon (1996). Finally, the ‘arbitrary’ interpretation of the implicit subject and the modal meaning of middle constructions can also be derived from the semantic representation in (39c), cf. Steinbach (2002) for further discussion. 28. See Keyser and Roeper (1984) and Zwart (1998) for a similar idea. Note, however, that we do not assume that the invisible reflexive pronoun absorbs the external argument of the verb. According to our analysis, it links the subject to the second semantic argument. Besides, Keyser and Roeper argue that the invisible reflexive pronoun is a syntactic clitic, cf. also Keyser and Roeper (1992). 29. In English the strong form consists of a pronominal part and the adnominal intensifier self. Historically, the adnominal intensifier was a focus particle. This can still be observed in German. Therefore it is very likely that strong reflexive pronouns intrinsically form a contrast to other type-identical salient alternatives. The minimal alternative to the reflexive interpretation is always the nonreflexive interpretation, which is only expected with verbs like hear, which are unlikely to be reflexive.

‘Binominal each-constructions’ (BECs) in German and English Malte Zimmermann University of Amsterdam

1.

Introduction

In this article, I propose a unified analysis of ‘Binominal each-constructions (BECs)’ in English and German. BECs are sentences that contain a binominal element like each in English and je(weils) in German.1 Examples are given in (1a,b): (1) a. The boys bought two books each. b. Die Jungen haben jeweils zwei Bücher gekauft. the boys have each two books bought

(German)

The overall problem addressed in this article is twofold. What is the proper interpretation of binominal elements? And why are they located where they are? I assume the two questions to be related. In particular, I will argue that the syntactic structure of BECs follows from the lexical meaning of the binominal elements. The structure of the article is as follows. In the remainder of Section 1, I motivate a unified analysis of BECs in English and German by introducing a number of syntactic and semantic properties shared by BECs of both languages. However, in Section 1.2 I point out some interesting differences in the syntactic distribution of binominal elements in the two languages. In Section 2, I argue that an LF-movement account à la Stowell and Safir (1988) (S&S in what follows) is empirically inadequate. I then propose my own in situ analysis of BECs in Section 3, and show that it accounts for all the properties observed with BECs in English and German. In Section 4, I attribute the observed distributional differences between English and German to general structural differences between the two languages. I conclude in Section 5.

344 Malte Zimmermann

1.1 Motivating a unified analysis: Shared properties BECs in English and German are very similar in their syntactic and semantic properties. The distinction into ‘syntactic’ and ‘semantic’ properties is used for expository reasons only. It will turn out in the course of the discussion that some restrictions on BECs that look like syntactic restrictions on the surface may well be semantic in nature. Therefore, the two terms should not be taken too seriously at this point. Semantically, the binominal element (each, jeweils) establishes a distributive relation between the denotations of two nominal expressions in both languages, hence the label ‘binominal’. In (1a,b) (repeated here with clarifying subscripts), elements of the Share-NP denotation (sets or groups of two books) are distributed in a 1:1-fashion over the atomic members of the Range-DP denotation (the plural individual denoted by the boys/die Jungen in (1a,b)) (cf. Choe 1987, S&S, Moltmann 1991, 1997, Sakaguchi 1998 among many others).2 (1) a. [Range-DP The boys] bought [Share-NP two books] each. b. [Range-DP Die Jungen] haben jeweils [Share-NP zwei Bücher] gekauft. the boys have each two books bought German = ‘Every boy bought two (different) books.’

Second, the Share-NP must be an indefinite expression, preferably introduced by a cardinal, or by other quantifying adjectives (2a,b). The Share-NP cannot be introduced by a ‘real’ determiner like the or every, nor can it be a proper noun (cf. (3a,b)). (2) a. The men read two books / several books each. b. Die Männer lasen jeweils zwei Bücher/einige Bücher. the men read-past each two books /some books

German

(3) a. *The men read the book / every book / ‘War and Peace’ each. b. #Die Männer lasen jeweils the men read-past each das Buch /jedes Buch /‘Krieg und Frieden’ German the book/every book/‘War and Peace’

This ‘Indefiniteness Requirement’ suggests that the binominal element stands adjacent to an NP, not a DP (hence the notation Share-NP). Third, the Range-DP must denote a pluralic entity (4a,b). BECs with a singular Range-DP are ungrammatical in both languages (5a,b):3

‘Binominal each-constructions’ (BECs) in German and English 345

(4) a. The men / five men / many men / John and Phil read two books each. b. Die Männer/fünf Männer/viele Männer/Bodo und Gert lasen the men /five men /many men /Bodo and Gert read-past jeweils zwei Bücher. each two books (5) a. *Bill / the boy read two books each. b. #Konstantin /der Junge las jeweils zwei Bücher. Konstantin/the boy read-past each two books

Of those shared properties that could be taken to be syntactic in nature, the following is the most important for the analysis to be presented here: the binominal element seems to form a constituent with the Share-NP. It is displaced together with the Share-NP under A- and A-bar-movement, as in the passive sentences (6a,b), and the wh-questions (7a,b): (6) a. One interpreter (each) was assigned (*each) to the visiting diplomats. b. (Jeweils) ein Mädchen wurde von den Männern (#jeweils) erwartet. (each one girl was by the men (#each expected ?‘One girl each was expected by the men.’ (7) a. How many books (each) did the boys buy (*each)? b. (Jeweils) wieviele Bücher haben die Jungen (#jeweils) gekauft? =(7a) (each how.many books have the boys (#each bought

Second, the distributive relation in BECs is subject to a locality condition, Choe’s (1987) ‘Clausemate Condition’ (cf. also S&S): the binominal element cannot distribute over (the denotation of) a Range-DP in a higher clause. (8) a. *The boys said that the girl had seen one balloon each. b. #Die Jungen sagten, dass das Mädchen jeweils einen Ballon the boys said that the girl each one balloon gesehen hatte. seen had … ‘Each of the boys said that the girl had seen a different balloon.’

Finally, as pointed out by Choe (1987), S&S, and others, the Share-NP has to take narrow scope w.r.t. the Range-DP. It can never be interpreted with widest scope such that it receives a referential or specific interpretation. This is captured under the label ‘narrow scope requirement’. (9) a.

Five men love two women each. … ‘There are two women x,y such that each of the five men loves x,y.’


b. Fünf Männer lieben jeweils zwei Frauen. five men love each two women … ‘There are two women x,y such that each of the five men loves x,y.’

This list of shared properties suggests that there is indeed good reason for venturing out in search of a unified analysis for BECs in English and German (and possibly in other languages that have BECs with the above-mentioned properties as well). Such a unified analysis must, of course, account for all the properties of BECs listed. Furthermore, some of the properties of BECs make it clear that binominal each cannot be analyzed as an instance of floated, or adverbial each. The two each’es differ in important respects as shown in detail by Sakaguchi (1998: 60–70). In particular, adverbial each does never form a constituent with VP-internal NPs/DPs. Also, adverbial each does not impose an indefinite requirement on the object NP, nor does it require an object NP to take narrow scope w.r.t. the Range-DP over which it distributes.4 The latter two properties are illustrated in (10a,b): (10) a. The students have each seen the teacher/Professor Higgins. b. The men each wrote to two women. (to Hilary Clinton and to Gro Harlem Brundland) ‘There are two women such that each of the men wrote to them.’

The German binominal jeweils also differs from the adverbial distributive element in morphological shape. (11) is an example featuring adverbial jeder ‘each’. (11) Die Männer haben jeder ein Buch gelesen. the men have each a book read ‘The men have each read a book.’

Hence, binominal each differs from the better-researched adverbial each, and must be analyzed differently. This being said, let us turn back to the enterprise of finding a unified analysis for BECs. Before we can do so, we need to look at a number of differences between English and German BECs, though. 1.2 Differences between German and English BECs The German binominal jeweils differs from English each in three respects. First, jeweils has to precede the Share-NP, whereas each follows it (cf. (1a,b)). In Section 3, I argue that the German word order derives from the underlying English word order through predicate inversion of jeweils.5


The second difference concerns the syntactic distribution of binominal elements in English and German. In English, binominal each is excluded from all underlying subject positions, e.g. in Small Clauses (SCs) (cf. (12a)), ECM-constructions (cf. (12b)), and in finite sentences (cf. (12c)). S&S treat this as a Subject Condition effect. In German, jeweils can occur in underlying subject position (cf. (13a,b)). (12) a. *The boys painted [SC two houses each red]. b. *The professors expected [two students each to be late]. c. *One woman each saw/loves John and Bill. (13) a.

Die Jungen haben [jeweils zwei Häuser rotgestrichen]. (SC) the boys have [each two houses red-painted *‘The boys have painted two houses each red.’ b. …, weil jeweils zwei Bewunderer die Tenöre (vor der because each two admirers the tenors (outside the Oper) erwarten. (finite) opera await ‘because each of the tenors is being awaited by two different admirers.’

It is clear that a unified analysis of BECs in English and German should account for this difference in syntactic distribution. In Section 4, I will account for the distributional differences on the base of general structural (i.e. syntactic!) differences between English and German. The final difference concerns the meaning of each and jeweils. It has been observed (cf. Link 1998, Moltmann 1991, 1997) that jeweils has a wider range of meanings than each. Apart from distributing solely over individuals, it can also distribute over contextually salient (cf. (14a)) or overtly expressed (cf. (14b)) sets/groups of events/situations. On this adverbial reading, it functions as the counterpart of English at a time/each time/respectively. (14) a.

Peter hat jeweils seinen Lieblingskandidaten vorgeschlagen. Peter has each-time his favourite-candidate suggested ‘Peter suggested his favourite candidate each time.’ b. Peter lobte und kritisierte Maria aus jeweils zwei Gründen. P. praised and criticized M. for each two reasons ‘Peter praised and criticized Mary for two reasons respectively.’

I take this second, adverbial reading of jeweils to modify full propositions. It does not impose any restrictions of locality or otherwise on the other members of the clause. Therefore, German sentences containing jeweils will be grammatical


(on the adverbial reading), even if some of the restrictions in 1.1 are violated. It is this property of jeweils, which makes it sometimes hard to check if a binominal reading is present or not. Contrary to Moltmann (1997), who treats both uses of jeweils as instances of the same lexical item, I assume genuine lexical ambiguity for jeweils. This assumption is supported by the fact that other languages have two morphologically different forms for the two readings: (15) a. Dutch: binominal elk b. French: binominal chacun(e)

vs. vs.

adverbial telkens, steeds adverbial chaque fois

In what follows, I will not be concerned with adverbial jeweils. In this section it has been shown that English and German BECs share a number of properties. I argue that these common properties suggest a unified analysis of the construction in both languages. Such an analysis must be able to account for certain differences in the distribution of binominal elements in the two languages.

2. The ‘Extraction Analysis’ (S&S) and its problems Before I present the in situ analysis of BECs, it will be instructive to look at S&S’s analysis in terms of LF-movement of the binominal element. In Section 2.1, I present S&S’s analysis, which I will refer to as the ‘Extraction Analysis (EA) henceforth. In 2.2, I discuss a number of problems for the EA both from a language-internal (English) and from a cross-linguistic perspective. 2.1 The Extraction Analysis (EA) It is important to note that S&S do not attempt to give a cross-linguistic analysis of BECs, but look only at English. They assign sentence (16a) the surface structure in (16b), given as a tree-diagram in (16c) for convenience. (16) a. The boys bought two books each. b. [IP the boysj [VP bought [NP [NP two booksi] [QP PROi each-ej]]]]


c.

IP VP the boysj bought

NP

NP two booksi

QP PROi

Q′ each -[e]j

The EA is based on the following three assumptions. i.

Binominal each is assumed to be a postnominal binary quantifier. It projects a QP with a PRO-specifier and a covert complement [e], the counterpart of overt one, which is coindexed with the Range-DP. PRO is coindexed with the Share-NP. The assumption of a covert complement [e] is supported by the existence of complex binominal elements in other languages, e.g. French chac-un(e) ‘each-one’ and German je-weils ‘each-while’. ii. The QP is in postnominal position because it is a complex nominal modifer, and because complex nominal modifiers always occur postnominally in English. (17) a. the [NP [NP man] [AP proud of his son]] b. *the [NP [AP proud of his son [NP man]]]

iii. The each-QP has to raise at LF so that the covert complement e is locally A-bar-bound by the Range-DP. This gives us the following LF for (16a). (18) a.

[IP [the boysj]1 [IP [QP PROi each-ej]2 [IP t1 [VP bought two booksi t2]]]]


IP

b.

IP the boysj,1

QP2

PROi each-[e]j

IP t1

VP V bought

NP NP

t2

two booksi

The EA seems to account for those BEC-properties that were labelled ‘syntactic’ in 1.1. It accounts for the Subject Condition effects observed given that overt subextraction out of underlying subject positions is impossible in English (cf. (19a,b)). (19) a. *[Of which book]1 do you consider [a review t1] important? b. *Who1 did [a sister of t1] get twins?

Since the each-QP in (12c), repeated here as (20), has to raise out of its embedding NP for the covert complement [e] to be licensed, the absence of binominal each from subject position is accounted for. Subextraction is blocked from subject position. (20) *One girl each saw / loved Bill and John

The position that extraction from subject position is also blocked at LF appears at first sight problematic in the light of data like (21a) from May (1985). (21) a. [Somebody from everyi city] despises iti b. [NP everyi city]1 [NP somebody from t1] despises iti

In the ‘inverse linking construction’ in (21a), the pronoun it can be interpreted as bound by the universal quantifier inside the subject NP. This seems to show that the quantified phrase every city must LF-extract out of the subject NP in violation of the ban on extraction from subjects. The binding facts in (21a) can be reconciled with the ban on extraction from subjects, however, if one assumes with May (1985) that the quantified phrase adjoins to NP (or better, that it moves to the specifier position of DP) at LF. From here, it can c-command the

‘Binominal each-constructions’ (BECs) in German and English

pronoun (cf. Kayne 1994, who discusses similar cases) without having to extract from subject position. The EA also accounts for the ‘Clausemate Condition’, which requires the Range-DP to be a clausemate of binominal each, with the extra stipulation that LF-movement of the each-QP has to be clause bound. This stipulation seems to be supported by the general clause boundedness of Quantifier Raising (QR) (cf. Reinhart 1997), which blocks an inverse scope reading for (22): (22) Somebody believes that some professors have praised every student. … ‘For every student y there is some (different) person x who believes y to have been praised by a specific group of professors.’

Third, the ‘narrow scope requirement’ follows trivially from the fact that the Range-DP has to c-command the raised each-QP at LF. Hence, it will obligatorily take wide scope over the rest of the sentence.6 Thus, the EA seems to have something to it. Nevertheless, it also faces a number of problems, which we will turn to now. 2.2 Problems for the Extraction Analysis There are two different kinds of problems with the EA (apart from the general question of if and how LF-structures like (18a,b) can be interpreted at all). First, the analysis can be shown to undergenerate and overgenerate for English. Second, it can be shown to undergenerate for German. Strictly speaking, problems with the German data cannot be taken as criticism against S&S’s EA per se. After all, the EA never made any claims about German in the first place. Nevertheless, the problematic German data suggest that the EA may not be the best candidate to serve as the base for a unified, cross-linguistic analysis of BECs. Since this is what I am aiming at here, I take it that the German data do pose a problem for the EA. 2.2.1 The EA undergenerates for English Let us look at the problematic English data first. The EA makes an incorrect prediction for binominal each in adjunct position. Adjunct positions are generally assumed to block subextraction out of them (cf. Huang’s 1982 Condition on Extraction Domains). Nevertheless, binominal each is licensed in adjunct position in English, as witnessed by (23a).7 (23b) shows the impossibility of overt subextraction out of this position.

351


(23) a. The students prepared for the finals with two text books each. b. *[On which topic]1 did the students prepare for the finals [with two books t1]?

2.2.2 The WCO-problem Second, the EA would incorrectly predict (24) to be grammatical. I will refer to this overgeneralization as the ‘WCO-problem’ for reasons that will become clear below.8 (24) *Which boys1 did she tell two men each that they should arrest t1?

The problem with (24) is the following. The attentive reader may already have wondered why the BEC-structure assumed by S&S does not give rise to a Weak Crossover (WCO) violation. Let us look at (25) and (26) for illustration. (25) a. The boys ate two apples each. b. LF: [the boys]1,i [PROj each-ei]2 [t1 ate two applesj t2] |________________|____| (26) a. Which boys ate two apples each? b. LF: [CP [which boys]1,i [PROj each-ei]2 [t1 ate two applesj t2]] |__________________|____|

In the LF-structures, a nominal expression c-commands first a coindexed pronominal element, and then a trace. This is a configuration which might possibly give rise to a WCO-violation, as is usually assumed for structures like (27). (27) *[Which boys]1,i did she tell theiri mother that the men would arrest t1 soon? |________________|_____________________________|

However, the LFs for the BECs in (25b) and (26b) do not seem to lead to a WCO-violation. This would be unproblematic, were it not for the ungrammatical (24). (24) should have the LF-structure in (28) on the EA. (28) [CP[which boys]1,i [PROj each-ei]2 [did she tell two menj t2 that they should arrest t1]]

Now, if (26b) does not give rise to a WCO-violation, then (28) should not either. If this is correct, though, we have no explanation for the ungrammaticality of (24). Consider the following examples. (29) a. We told two men each that we would leave a.s.a.p. b. [Which boys]1 did she tell us that the men would arrest t1 soon?


(29a) shows that each is possible with the indirect object of the verb to tell. (29b) shows that long extraction of the embedded object which boys is possible in principle. Both factors cannot be responsible for the ungrammaticality of (24). Also, (26a) shows that a which-phrase can function as Range-DP for binominal each. Thus, it is tempting to attribute the ungrammaticality of (24) to the same factor that makes (27) ungrammatical, a WCO-violation. But if (24) constitutes a WCO-violation, we would expect (25a) and (26a), which have an analogous LF-structure, to be ungrammatical, too, contrary to fact. Thus, either there is another, independent reason for the ungrammaticality of (24), or we are forced to conclude that S&S’s Extraction Analysis is inconsistent w.r.t. WCO-effects. Given the WCO-problem, it seems clear that an analysis without LF-movement of the eachQP is preferable since it does not create WCO-configurations for the BECs in (25a) and (26a) in the first place. Let us now look at those data from German which are problematic for the EA as a unified analysis of BECs. 2.2.3 Counterevidence from German There are two kinds of data from German that argue against the EA. First, German allows for binominal jeweils in subject position even in cases where overt subextraction out of the subject is impossible (cf. Section 1.2).9 The following minimal pair may serve as an example. (30) Jeweils zwei Verkäufer stürzen den Kunden entgegen. each two salesclerks rush the customers towards ‘The customers are being approached by two salesclerks each (31) a. *[Von was]1 stürzen [zwei Verkäufer t1] den Kunden entgegen. [of what rush [two salesclerks the customers towards b. Zwei Verkäufer von Kosmetikprodukten stürzen den Kunden two salesclerks of cosmetic-products rush the customers entgegen. towards ‘Two vendors of cosmetic products fling themselves at the customers.’

The grammaticality of (30) is unexpected on the EA, since jeweils should not be able to raise out of the embedding subject. As (31a) shows, overt subextraction from this position is impossible. Again, it is clear that an in situ analysis, which does not postulate LF-movement of jeweils, encounters no such problem with instances of subject jeweils. The second argument against extending the EA to German regards the supposed status of the postnominal each-QP as a complex nominal modifier. In


German, jeweils precedes its Share-NP even though head-initial complex modifiers cannot occur prenominally.10 This casts doubt on the analysis of the binominal as a complex modifier (though not necessarily against the EA as a whole). (32) *der [stolz(e) auf seinen Sohn] Mann the [proud of his son man

The existence of prenominal jeweils in the absence of prenominal head-initial complex modifiers argues against the analysis of the binominal as (part of) a complex modifier. Summing up this section, we have seen that S&S’s Extraction Analysis is empirically inadequate for both English and German. In particular, the existence of binominal elements in adjunct and subject islands argues against an account in terms of LF-extraction. Instead, these data seem to support an anlysis that leaves the binominal element in situ. In the next section, I will present such an in situ analysis.

3. The ‘In-Situ Analysis (ISA)’ of BECs In the preceding sections it was shown (i) that the binominal element forms a constituent with the Share-NP at S-structure, and (ii) that moving the binominal at LF runs into problems. Let us take the opposite route, then, and assume that the binominal element does not move at LF, but stays in its surface position, where it is also interpreted. In Lipták and Zimmermann (2000) (henceforth L&Z), we proposed such an ‘In Situ-Analysis (ISA)’, which I will essentially adopt here. In the remainder of this section, I will show that the ISA accounts for BECs in a satisfactory manner. In order to check the viability of the ISA we must show two things. First, that we can derive the correct readings of BECs from their surface structure in a compositional fashion. Second, we must show that the ISA accounts for the properties of BECs listed in 1.1. My claim is that both requirements are met by the ISA. In Section 3.1, I present the ‘In Situ-Analysis’. In 3.2, I account for the different position of the binominal element w.r.t. the Share-NP in English (postnominal) and German (prenominal). In 3.3, I (briefly) show how the semantics of the construction works. It is shown that a type-driven interpretation of the surface structure of BECs yields the correct truth conditions. The core assumption of the semantic analysis is that the binominal element is (semantically) the main functor of the sentence, and takes the denotation of all


other material in the clause (including the verb denotation) as its semantic arguments. Finally, I show in 3.4 that the typical properties of BECs follow directly from the semantics of the binominal element. 3.1 The syntactic structure of BECs The ISA assigns the following syntactic structure to the English BEC in (1) (cf. L&Z) (33) a.

[IP [R-NP The boysj [VP bought [DP D [SC [Sh-NP two books] P [eachecj]]]]]. IP

b. NP the boysj

VP

V bought

DP Spec

D′ D

SC NP two books

P

QP each [e]j

The ISA is based on the following assumptions: i.

The each-QP is not a complex modifier of the Share-NP, but a syntactic predicate in a prepositional Small Clause (SC). This predicate is related to its SC-subject (the Share-NP) via an empty prepositional head P. P is sometimes realized overtly in phrases like two books for each person or the German zwei Bücher pro Kopf ‘two books per head’. ii. With S&S, I assume that each has a null complement ec, coindexed with the Range-DP. In German, this complement is expressed overtly as -weils.11 Unlike S&S, I do not assume the presence of a PRO in each-QP. iii. The SC is the complement of an empty D-head. That a D-head is present can be seen from the fact that BECs can be passivized (cf. (34a,b)), as opposed to ‘bare’ resultative SCs without a D-head (cf. (35a,b)):

355


(34) a. [DP One girl each]1 was expected t1 by the men. b. *One girl was expected each by the men. (35) a. *The house red was painted by the men. b. [DP the house]1 was painted t1 red by the men.

iv. The each-QP can receive a proper interpretation in situ. Hence, it does not have to move at LF for interpretive reasons. I will come back to this point in Section 3.3. v. The observed word order variation between English and German reduces to predicate fronting of the jeweils-QP into a position preceding Share-NP. This is schematized in (36). (36) [XP jeweils1 X [SC zwei Bücher P t1]]

According to the ISA, the prenominal position of binominal elements is derived from the postnominal ‘base’ position. In this sense, English BECs display the base generated order overtly. In the next section, I turn to the question to which position binominal jeweils moves in German. 3.2 On the landing site of jeweils-fronting If binominal jeweils moves into prenominal position through predicate fronting, what is its landing site? I assume that the landing site of jeweils is an A-baror operator-position. Indeed, it may be the case that jeweils occupies the same surface position as genitive possessor phrases like Peters ‘Peter’s’. Such a position is supported by the fact that jeweils carries the same morphological -s-marking as possessives, as witnessed by the independent occurrence of the stem jeweil- in the adjective jeweil-ig ‘respective’. The -s could be a functional head in D (cf. Abney 1987), or the spell out of a (combination of) functional head(s) (cf. Kayne 1994, Den Dikken 1995, 1998).12 In L&Z, we assumed that jeweils moves to the specifier of the SC-embedding DP. This view is concurrent with analyses of the German possessive genitive in the spirit of Abney (1987), cf. e.g. Olsen (1991). Thus, a DP with preposed jeweils would be assigned the following structures for the two analyses of the -s respectively. (37) a.

[DP jeweil1 [D -s] [SC zwei Bücher PE t1]] each two books b. [DP jeweil1 [D+P -s] [SC zwei Bücher tP t1]]


A second option is that the binominal element jeweil- moves to the specifier of a functional projection (FP) below D. The existence of such functional projections has been argued for repeatedly in recent years (cf. Kayne 1994, Den Dikken 1995, 1998, Szabolcsi 1994, Corver 1998 among many others; for an overview of the proposed functional projections within DP cf. Coene 1999, Ch. 1). Movement to SpecFP results in the following structure. (38) [DP D [FP jeweil1 [F+P -s] [SC zwei Bücher tP t1]]

Contrary to L&Z, I would like to adopt the ‘Movement to SpecFP’-analysis here. The reason for this is that jeweils can cooccur with an overt D-head as in (39a), which presumably has the structure in (39b). (39) a.

Ich mag diese Autos mit den/ihren jeweils zwei Spoilern. I like these cars with the/their each two spoilers ‘I like these cars with their two spoilers each.’ b. Ich mag diese Autos [PP mit [DP [D den/ihren] [FP jeweil [F+P -s] [SC zwei Spoilern tP t1]]]]

Given the existence of sentences like (39a), I tentatively conclude that the prenominal position of the binominal element in German is the result of predicate fronting to the specifier of an FP below D. This might also explain the ‘Indefiniteness Requirement’ on the Share-NP that was observed in 1.1. Recall that the Share-NP must never be headed by a ‘true’ determiner like the, every, or this. Assume that the D-head serves to encode the argumenthood of a nominal expression (cf. Szabolcsi 1994). Then argumenthood should not be expressed twice on the embedding DP and on the Share-NP. In other words, the indefiniteness requirement follows from a ban on DP-recursion in argument position (cf. Den Dikken 1995, 1998). Before I conclude this section, I would like to briefly discuss yet another treatment of binominal jeweil(s). Semantically, it would also make sense to treat jeweils as a modifier of the adjectival cardinal expression itself, giving a surface structure roughly as in (40). (40) [DP D [SC [NP [AP jeweils [A zwei]] [NP Bücher]]]] each two books

The question is if and how a structure like (40) is related to the English pattern where the binominal element and the cardinal expression are not string adjacent: (41) two books each


One way to relate (40) and (41) would be to assume that the underlying SC-predicate in the two constructions is the cardinal AP each two/jeweils zwei, as schematized in (42): (42) [SC books P [each two]]

The two surface patterns observed for German and English in (40) and (41) would then be the result of moving the cardinal alone in the English case, whereas in German the entire SC-predicate moves to prenominal position.13 Since, I have not been able to work out the details of such an approach yet, I will tentatively stick to the Movement-to-SpecFP hypothesis here. Hopefully, the future will show which analysis is empirically and conceptually more adequate. 3.3 Interpreting BECs Leaving purely syntactic matters behind for a moment, let us now look at the interpretation of BECs. In this section, I show that BECs can be interpreted directly off their surface structure, using a type-driven semantics à la Heim and Kratzer (1998). The upshot of this discussion will be that LF-movement is not required for the analysis of BECs, at least not for interpretive reasons. My semantic treatment of BECs is based on the assumption that the SC-predicate, i.e. the binominal element is (semantically) the main functor of the clause. This means that it takes the denotations of all other material in the clause, including that of the main verb, as semantic arguments. This makes each-[e]/jeweils on their basic (transitive) use to expressions of type ·ett·eet, etÒÒ. The binominal element takes an indefinite containing a variable (provided by the Share-NP), and a transitive relation (provided by the verb) as its first two arguments and maps them onto an expression of type ·etÒ, i.e. an ordinary VP-denotation.14 The resulting expression takes the pluralic Range-DP and maps it onto a truth-value. The structure in terms of semantic types is illustrated in (43). (43)

sentence 〈t〉 Range-DP 〈e〉

VP 〈et〉 V 〈eet〉

SC 〈eet, et〉 Share-NP 〈ett〉

binominal 〈ett 〈eet, et〉〉


For ditransitive sentences (cf. (44)), each-[e]/jeweils are of the corresponding higher type ·ett, ··eeetÒ, ·eetÒÒÒ. (44) Peter gave the girls four roses each.

Given the semantic structure in (43), we can formulate the following lexical entry for transitive each-[e]/jeweils: (45)

each-[e]trans = λQλRλx*. "x [i-parti(x, x*) Æ $Y [Q(Y) & R(Y)(x)]]15

The entire interpretive procedure is illustrated in (46). IP

(46) a. R-NP the boysi

V bought lylx. bought’(y)(x)

∀x [atom(x, the_boys’) → ∃Y [two books’(Y) & bought’ (Y)(x)]] VP lx*. ∀x [atom(x, x*) → ∃Y [two books’(Y) & bought’ (Y)(x)]] SC (= PP) Sh-NP two books lY. two books’(Y)

lRlx*. ∀x [atom(x, x*) → ∃Y[two books’(Y) & R(Y)(x)]]

eachQP

each [e]i lQlRlx*. ∀x [atom(x,x*) → ∃Y [Q(Y) & R(Y)(x)]]

b. c.

IP = "x [atom(x, the_boys’) Æ $Y [two books’(Y) & bought’ (Y)(x)]] For each atomic member x of the group denoted by the boys, there is a set Y such that Y consists of two books and x bought Y.

The reader may verify that the truth-conditions in (46c) correctly capture the meaning of (46a). It can be shown that the same semantic result is obtained for the German counterpart of (46a) with prenominal word order (cf. Lipták and Zimmermann 2000). Based on the lexical entry for each-[e]/jeweils in (45), it is easy to write similar entries for binominal elements in other syntactic configurations. These are given in (47) (‘dobj’ = direct object, ‘iobj’ = indirect object, ‘pobj’ = prepositional object). (47) a.

The boys gave Mary three roses each. each-[e]ditrans,dobj1 = λQλRλzλx*. "x[i-parti(x, x*) Æ $Y [Q(Y)& R(Y)(z)(x)] b. Peter gave the girls two roses each. each-[e]ditrans,dobj2 = λQλRλz*λx. "z[i-parti(z, z*) Æ $Y [Q(Y) & R(Y)(z)(x)] c. The boys gave three women each a flower. each-[e]ditrans,iobj = (45)


d. The jury members sentenced the defendant to five years each. each-[e]ditrans,pobj1 = (47a) e. The judge sentenced the defendants to five years each. each-[e]ditrans,pobj2 = (47b)

To conclude, I have shown that a compositional interpretation of binominal jeweils/each in-situ is possible, using a type-driven semantics. If we assume that semantic interpretability should be a restricting factor on syntactic structures (cf. for example Partee 1976), the semantics given here supports the ISA. The ISA features only minimal (overt) movement, i.e. predicate fronting of the binominal element in the case of German BECs. 3.4 Accounting for the properties of BECs In this section, I show that the ISA accounts for all the properties of BECs listed in 1.1. First, it follows directly from the lexical meaning of the binominal element in (45) that BECs express a distributional relationship between the denotations of two nominal expressions. Second, the indefiniteness requirement on the Share-NP was shown to follow from the structure of the SC-embedding DP in Section 3.2. The binominal element is located below the (empty) determiner and cannot occur with a DP as its subject, presumably due to a ban on recursive embeddings of DPs. Third, the requirement that the Range-DP denote a pluralic individual also follows from the semantics of each/jeweils, more precisely from the individual subpart relation i-part (x,x*). If there is no pluralic Range-DP that can serve as antecedent, this subpart condition cannot apply and the expression will not be well-formed. Note that proper quantified expressions introduced by the quantifiers every or no are excluded for the same reason: (48) a. *No boy bought two books each. b. *Every boy bought two books each.

If we look at the semantic expressions that result from combining the quantified phrase with the VP-denotation of bought two books each, we see that the variable that has to serve as input for the i-part relation is singular. ¬$z [boy’(z) & "x[i-part(x,z) Æ $Y [two books’(Y) & bought’ (Y)(x)]] b. "z [boy’(z) & "x[i-part(x,z) Æ $Y [two books’(Y) & bought’ (Y)(x)]]

(49) a.


In (49a,b), there is no pluralic entity over which the binominal could distribute, and the entire sentence fails to receive a well-formed interpretation.16 Turning to those properties that were labelled ‘syntactic’ in 1.1, we can observe the following. That the Share-NP and the binominal element form a syntactic constituent turns out to be a direct consequence of the semantics of each/jeweils. After all, the Share-NP denotes the first semantic argument of the binominal, hence it should also form a syntactic unit with the latter. The ‘Clausemate Constraint’, which requires the Range-DP to be a clausemate of the binominal element, also follows — perhaps somewhat surprisingly — from the semantics of the construction. I repeat the relevant sentences for convenience. (50) a. *The boys said that the girl had seen one balloon each. b. The boys said that the girls had seen one balloon each. π ‘Each of the boys said that the girls (as a group) had seen one balloon.’

In (50a) there is no appropriate (i.e. plural denoting) Range-DP available, and the sentence is ungrammatical. In (50b), each obligatorily has to distribute over the clausemate DP the girls, and the sentence is unambiguous. Both facts follow straightforwardly from the lexical meaning of the binominal in (45). In (50a), the interpretation proceeds until the VP-denotation and the subject’s denotation have to be combined. Since the subject denotation is not of the right kind (a singular entity), it cannot serve as semantic argument and the semantic computation crashes. In (50b) the appropriate (plural) DP the girls has to serve as Range-DP because the outermost argument slot of the binominal needs to be filled at that point in the semantic computation. Perhaps this failure in number agreement (and the subsequent failure to serve as a semantic argument for each) can be likened to other instances of number agreement mismatches, as in (51). (51) *The boys runs.

In (51), too, the only nominal argument, which could potentially serve as subject, has the wrong number specification. If we follow Chomsky (2000), and much of the generative literature, in treating agreement as a syntactic mechanism (even if it is semantically motivated, since it governs the assignment of the correct arguments to a given functor, most often a verb), we can still treat the clausemate condition with BECs as a syntactic restriction. It is important to keep in mind, though, that the restriction


follows from the lexical meaning of the binominal elements, which require arguments of a particular sort. Hence, the clausemate condition derives from the meaning of the binominal element plus a restriction on agreement, and not from a ban on (covert) cross-sentential movement of the binominal, as assumed by S&S.17 Finally, let us look at the ‘narrow scope requirement’, which says that the Share-NP can never take wide scope over the Range-DP. I propose to treat the indefinite Share-NP as merely introducing a (choice-function or individual) variable (cf. Heim 1982, Reinhart 1995, 1997, Winter 1997, Kratzer 1998, Matthewson 1999) which needs to be bound locally (in the afterclause of the material implication) by an existential operator. I assume the local existential closure of the variable to be a consequence of the presence of jeweils/each. It follows that indefinite cardinal NPs with jeweils or each are by necessity interpreted as referentially dependent on some other DP. This means that binominal Share-NPs would form the counterpart of those DPs in Straits Salish, discussed by Matthewson (1999), whose variable is obligatorily bound nonlocally, and which are always interpreted referentially independent. On the other hand, the constituent Share-NP + binominal would be surprisingly similar to Hungarian NPs with reduplicated cardinals, which are also obligatorily construed as referentially dependent (cf. Farkas 1997). More work in this area needs to be done. Summing up, in this section I have shown that the in-situ analysis (ISA) proposed in this article can account for the properties of BECs with one additional stipulation to the effect that the presence of the binominal element results in local existential closure of the variable contained in the Share-NP. Furthermore, most of the properties of BECs have been shown to follow directly from the semantics of the binominal element as given in (45). All in all, Section 3 has shown that the ISA (i) generates surface structures for BECs which can be interpreted in a compositional fashion, and (ii) accounts for the syntactic and semantic properties of BECs. Since the ISA yields these results entirely without recourse to LF-movement, I take it to be a more economical account of BECs than S&S’s EA.

4. The subject problem In the preceding section it was shown that an in situ analysis accounts for the properties of English and German BECs in a natural and elegant way. The problem remains, however, of how to explain the different syntactic distribution


of binominal elements in the two languages. Recall from Section 1.2 that English does not allow for binominal each in subject position, whereas German does. The relevant data are repeated in (52) and (53) for convenience: (52) a. *The boys painted [SC two houses each red]. b. *The professors expected [two students each to be late]. c. *One woman each saw/loves John and Bill.

(SC) (ECM) (finite)

(53) a.

Die Jungen haben [jeweils zwei Häuser rotgestrichen]. (SC) the boys have [each two houses red-painted *‘The boys have painted two houses each red.’ b. …, weil jeweils zwei Verkäufer auf die Kunden zustürzen. because each two salesclerks on the customers rush-towards ‘because each of the customers is approached by two different salesclerks.’ (finite)

Clearly, the difference between (52) and (53) poses a problem for the ISA, since so far we have assumed that BECs have essentially the same structure in English and German.18 In this section, I argue that the ISA can be maintained in the face of (53a,b), and that the different distribution of each/jeweils follows from general structural (i.e. syntactic) differences between English and German. In Section 4.1, I look at binominals in the subject position of resultative SCs (cf. (52a) and (53a)). In Section 4.2, I turn to each/jeweils in subject position of finite active sentences. 4.1 Subject-jeweils in resultative Small Clauses That binominal each is impossible in subject position of English resultative SCs follows directly from the lexical meaning of the binominal, repeated in (54). (54)

each-[e]trans = λQλRλx*. "x [i-parti(x,x*) Æ $Y [Q & R(Y)(x)]]

Recall that the binominal element requires its second argument to denote at least a binary relation, e.g. a transitive verb. Now, as shown in (55), there is no such relation-denoting expression available in the resultative SCs in (52a), but only the property-denoting SC-predicate red. (55) a. [VP paint [SC [two houses each] red] b. eet,et et

As a consequence of this type mismatch, the interpretation cannot proceed and the entire SC will not receive a proper interpretation. Hence, binominal each is


banned from subject position of resultative SCs for interpretive reasons. German resultative SCs differ from their English counterparts in that the SC-predicate (rot ‘red’ in 53a) stands adjacent to the matrix verb. I assume that reanalysis of SC-predicate and matrix verb into a complex transitive verb is possible under adjacency. Hence, the underlying SC in (53a) has the following surface structure. (56) a.

[VP [DP jeweils1 [NP zwei Häuser t1]] [V rot+gestrichen]] each two houses red-painted b. [ eet,et ] [ eet ]

After reanalysis, the sister of jeweils+Share-NP, i.e. the complex verb rotgestrichen ‘painted red’, denotes a relation and meets the semantic requirements of binominal jeweils. There is no type mismatch and the interpretation can proceed. This accounts for the occurrence of jeweils in what looks like resultative SCs in German. This line of thinking gets support from the fact that binominal each is substantially better in resultative SCs in English, if the constituent ShareNP+each does not intervene between the matrix verb and the SC-predicate, e.g. in Heavy-NP-shift configurations. This is illustrated in (57): (57) ?The boys painted red two houses of friends in the neighbourhood each.

Assuming that English also allows for reanalysis under adjacency, the grammaticality of (57) is accounted for. This observation reduces the different behaviour of each and jeweils w.r.t. resultative SCs to the different linear order of the verb and its complements in the two languages, namely VO (English) vs. OV (German). 4.2 Subject-jeweils in finite active sentences Binominal each is excluded from the subject position of active transitive sentences for the same reason that bans it from occurring in SCs, namely a type mismatch. To see this more clearly, let us look at a simple transitive sentence with subject each and its type-logical structure in (58). VP

(58) one girl each 〈eet, et〉

〈et〉 loves 〈eet〉

the boys 〈e〉


The structure in (58) shows the VP of the sentence with the subject in its base position inside VP (cf. Koopman and Sportiche 1991) before it moves to SpecIP.19 We see that the subject is hierarchically higher than the direct object even before moving to SpecIP. It follows that the latter combines first with the transitive verb to yield an expression of type ·etÒ. Now, as in the SC-case, the subject constituent Share-NP+each requires a sister argument of type ·eetÒ. As we see from (58), this requirement is not fulfilled in English transitive sentences, a type-mismatch occurs, and (58) receives no well-formed interpretation. It is important to note that this state of affairs holds both for the base-generated structure as well as for surface structure with the subject in SpecIP. Therefore, subject-each will always be excluded irrespective of the question of whether the subject can reconstruct into its base position or not. Why is jeweils licensed in subject position in German? I assume this to be a direct consequence of a different VP-structure in German. In particular, I assume that German transitive sentences contain “flat” VPs as shown in (59). Such a flat syntactic structure is proposed for German (albeit at the S-level) in Haider (1986d).20 VP

(59)

jeweils zwei Verkäufer 〈eet, et〉

den Kunden 〈e〉

entgegenstürzen 〈eet〉

Under the plausible assumption that an expression in a ternary structure can semantically combine with its two sister complements in any order (type permitting!), the denotation of the subject constituent jeweils+ShareNP can first combine with the verb denotation before it combines with the object denotation. The lexical meaning of subject jeweils differs minimally from that of object jeweils, in that it reverses the order of arguments of the relation in the afterclause of the material implication. The lexical entry is given in (60), and the derivation of the meaning of (53b) with the (underlying) structure in (59) is shown in (61): (60)

jeweils subj fi λQλRλx*. "x[i-parti(x,x*) Æ $Y [Q(Y) & R(x)(Y)]

(61) a.

[jeweilssubj zwei Verkäufer entgegenstürzen] fi λx*. "x[i-parti(x,x*) Æ $Y [*zwei Verkäufer’(Y) & entgegenstürzen’(x)(Y)]]


b. [jeweilssubj zwei Verkäufer den Kunden entgegenstürzen] fi "x[i-parti(x,[den Kunden]) Æ $Y[*zwei Verkäufer’(Y) & entgegenstürzen’(x)(Y)]]

Although it satisfies the semantic requirements, the VP-structure in (59) is not in line with Kayne’s (1994) claim that syntactic structures universally have to be binary branching, a consequence of his Linear Correspondence Axiom (LCA). The LCA predicts VPs always to have the structure in (58). The subject then moves to SpecIP at surface structure for reasons of Case (checking or assignment). (59) also differs from the VP-structure assumed for German in Haider (1993). There, Haider argues that — unlike in English — the subject remains inside VP at surface structure, but in a position structurally higher than the object. Again, this would yield a structure like (58) for the German VP. Haider’s assumption for a hierarchically layered VP seems to follow from two conceptual considerations. First, a layered VP would account for a number of subjectobject asymmetries, e.g. with binding facts (cf. (62a,b)), without having to fall back on linear precedence as a defining factor for binding domains (cf. Barss and Lasnik 1986, Jackendoff 1990 for accounts that assume linear precedence to be of grammatical relevance). However, as soon as we allow for linear precedence as a grammatically relevant factor, the case for binary branching is not so strong anymore. (62) a.

…, weil jede Mutteri ihreni Sohn verehrt. because every mother her-acc son admires Mutteri verehrt. b. *…, weil ihri Sohn jede because her son every-acc mother admirers

Second, the assumption of a hierarchic VP allows Haider to treat the mapping from argument structure to syntactic structure as nonambiguous. The “designated” (or external) argument will always be mapped onto a unique syntactic position inside VP. On the other hand, a flat VP will only give rise to an ambiguous mapping if linear precedence is ignored as a syntactically relevant factor. In addition, German has been argued to belong to those languages that license their arguments through morphological Case (e.g. Haider 1993:122). So perhaps, morphological Case marking with nominative is sufficient to keep the external argument from accidentally being interpreted as the internal argument in a flat structure.21 Independent of the question whether the external argument always has to precede the object, or whether both word orders can be freely generated (cf. Haider 1986d), morphological Case marking may ensure nonambiguous mapping from


argument structure to syntactic structure. After all, an argument carrying nominative will be identifiable as the subject regardless of its syntactic position. Sentences like (63) pose a serious problem for the flat VP hypothesis. Here, the direct object and the non-finite verb are preposed to sentence-initial position to the exclusion of the subject. On the standard assumption that only heads or maximal projections can move, (63) seems to suggest that the German VP is indeed hierarchic as in Haider (1993). (63) Einen Hund gebissen hat der Mann. a-acc dog bitten has the-nom man ‘What the man has done is bite a dog.’

A way out of this dilemma is to assume, contra Haider (1993), that German subjects have to move to SpecIP (or one of its minimalist successors) overtly for reasons of Case (checking or assignment) like their English counterparts. On such a view, the flat-VP hypothesis could be upheld even in the face of VP-preposing data like (63). The constituent that moves in (63) is a remnant VP which contains the trace of the subject as in (64): (64) [VP t1 einen Hund gebissen]2 [hat der Mann1 t2]

If these assumptions are correct, the surface structure of an ordinary transitive sentence will be as in (65). (65) [IP subj1 [VP t1 obj V]

The important fact about the structure in (65) w.r.t. the interpretation of BECs is that the subject trace is a syntactic sister of V. Assuming that this trace is of the same semantic type as its antecedent (i.e. ·eet,etÒ), it can first take the verb denotation and then the object denotation as argument. The further interpretation, involving λ-abstraction over the trace index, and functional application of the resulting expression to the raised subject in SpecIP gives the correct truthcondition, as the gentle reader may verify for herself. Note that a subject trace of type ·eÒ does not lead to a converging interpretation in sentences containing subject jeweils.22 This is important for it excludes the possibility that the subject trace is mistakenly interpreted as an internal argument, which would give (53b) an incorrect reading on which the customers rush towards two salesclerks each. The flat VP-structure proposed here would allow for this possibility in principle. Since the indefinite plural Share-NP zwei Verkäufer ‘two salesclerks’ is also not unambiguously marked for nominative, the accusative plural being isomorphic, we can also not resort to morphological Case marking in order to


exclude the unwanted reading. With ordinary subjects, I assume that unambiguous nominative marking prevents the inverse interpretation on which the external nominative argument is falsely interpreted as an internal argument. Summing up these considerations, a flat VP can be maintained for German in the face of VP-topicalization data like (63), on the assumption that the subject has to move to SpecIP overtly in German. It was suggested (though not worked out in the last detail) that the resulting structure can be correctly interpreted, given that the subject trace of jeweils+ShareNP is of type ·eet,etÒ. Let us conclude, then, that the assumption of a flat VP-structure for German is a feasible way to meet the requirements of the semantics of BECs, even if such a move encounters a number of empirical problems and is at odds with recent developments in generative grammar, namely with antisymmetric approaches to natural language. I would like to stress again, that the assumption of a flat VP is forced by the particular semantic framework of type-driven semantics, which I adopted here. The biggest benefit of such a position is that it allows us to maintain compositionality in its strongest form in the interpretation of BECs. One of the (syntactic) consequences of this position is that we seem to be forced to take linear precedence relations as grammatically relevant.23 However this may be, those who would like to keep the assumption of strict binary branching for German, could choose for either of the following three options: (i) abandon a unified analysis of BECs in English and German; (ii) abandon strict compositionality in the interpretation of BECs; or (iii) maintain a unified analysis and compositionality and develop a new semantics which is more in line with the demands of recent minimalist and antisymmetric analyses of English and German. Hopefully, the future will show what such ‘antisymmetric semantics’ should look like. Summing up, in this section I have argued that the different distribution of binominal jeweils and each follows from general structural differences. Jeweils is licensed with SC-subjects because the SC-predicate is adjacent to the sentencefinal matrix verb and reanalysis of the two to a complex transitive verb can go through. It has been shown that, whenever a SC-predicate comes to be adjacent to the matrix verb in English (licensing reanalysis), binominal each in the SC-subject is possible as well. Finally, the licensing of jeweils in subject position of active transitive sentences was argued to be the result of a flat VP-structure in German.


5. Conclusion In this article, I have presented a unified analysis of ‘binominal each constructions’ (BECs) in English and German. Motivation for a unified analysis comes from the overall similarity of BECs in the two languages. First, I showed that an LF-extraction account à la Stowell and Safir (1988) does not hold for English, nor does it account for the German facts, namely the grammaticality of binominal jeweils ‘each’ in subject position. I went on to present my ‘In situanalysis (ISA)’. On the ISA, LF-movement is not required for interpretive reasons, since BECs can be interpreted directly off their surface structure. Syntactically, the binominal element is the predicate of a prepositional small clause (SC). Semantically, it serves as the main functor of the entire sentence, taking the denotations of all other material in the clause as its semantic arguments. I showed that the ISA allows for deriving the correct truth conditions for BECs in a compositional fashion. The ISA also accounts for all the syntactic and semantic properties of BECs. Finally, I discussed a difference in syntactic distribution between English binominal each and German binominal jeweils. I argued that the grammaticality of jeweils in underlying subject position in small clauses and active sentences can be reduced to general structural differences between German and English: the possibility of reanalysis of the SC-predicate and the adjacent sentence final verb into a complex transitive verb in German (licensing jeweils in SCs), and the existence of ternary branching, ‘flat’ VPs in German.

Notes 1. Unlike Link (1998), I take je to be a shortened version of binominal jeweils. The claims made in this article for jeweils are meant to hold for je as well. I agree with Link (1998: 128) that unlike je, jeweils can have an additional eventive reading. Cf. Section 1.2. 2. The terms ‘Range-DP’ and ‘Share-NP’ are to be understood as convenient labels, rather than as theoretically significant categories. The term ‘Range-DP’ is adopted from S&S’s ‘Range-NP’, ‘Share-NP’ is taken from Stowell and Beghelli (1997) and replaces their earlier ‘D(istributive)-NP’. 3. Actually, the German sentences are only ungrammatical on the binominal reading, which interests us here. The impossibility of the binominal reading is marked by ‘#’. Cf. Section 1.2 below for more on the different readings of jeweils. 4. Yet another difference between the two instances of each concerns the licensing of crosssentential anaphora. The object of adverbial each licenses these, whereas the object of binominal each does not (cf. Sakaguchi 1998: 69):


(i) The boys each played with a ballooni. Iti was a bright, yellow one. (ii) The boys played with a balloon each. #Iti was a bright, yellow one. 5. This position is supported by languages or dialects that optionally allow the binominal element to occur in both positions. This is the case in French (cf. (i)), and in some Dutch dialects (cf. (ii)): (i) Les hommes ont achetés (chac-un) deux livres (chac-un). the men have bought (each-one two books (each-one ‘The men have bought two books each.’

French

(ii) De jongens hebben (elk) twee boeken (%elk) gekocht. the boys have (each two books (each bought ‘The boys have bought two books each.’

Dutch

6. There is a slight problem with this account for — as far as I can see — nothing would keep the Share-NP from raising at LF and adjoining to a position above Range-DP, from where it could take scope over the latter. 7. S&S (1988: 447) acknowledge that the grammaticality of binominal each in adjunct position poses a problem for their analysis. 8. I thank Daniel Büring (p.c.) for suggesting (24) as a potential problem for the EA. 9. This careful phrasing gives credit to the fact that German is laxer w.r.t. subextraction from subject position than English. This is witnessed by (i), the English counterpart of which is ungrammatical (cf. (ii)). (i)

(ii)

[Von welcher Firma]1 hat [ein Angestellter t1] am Flughafen auf [from which company has [an employee at-the airport for mich gewartet? me waited *?From which company did an employee wait for me at the airport?

The grammaticality of (i) does not invalidate the point made in the main text, since binominal jeweils also occurs with subjects that clearly do not allow for overt extraction, like (31a) in the main text. 10. I assume that S&S’s each-QP is head-initial, since the quantifier each precedes its covert complement. 11. Unlike S&S, I do not assume the complement of each, je- to be pronominal in nature. Rather, it denotes an i-part relation, which splits up the Range-DP denotation into its individual subparts (more on this in Section 3.3). 12. Note that if this line of reasoning is correct, the binominal element in German would be jeweil-, rather than jeweils. Nevertheless, I shall continue to use jeweils whenever I refer to the German binominal element. 13. I thank Norbert Corver (p.c.) for suggesting an approach along these lines to me. 14. Throughout, I treat indefinite NPs as containing a (choice-function or individual) variable which needs to be bound by an existential quantifier (cf. Heim 1982, Reinhart 1995, 1997, Winter 1997, Matthewson 1999).


15. The i-part relation is taken from Link (1983). Unlike L&Z, I analyse the indefinite ShareNP not as a G(eneralized) Q(uantifier), but as simply containing a variable, as e.g. in Heim (1982). Cf. also the end of Section 3.4 below. 16. The simultaneous application of universal quantification over a set plus applying the i-part relation to the individual elements resulting from universal quantification might be responsible for the pleonastic flavor of the expression in (48a). 17. There is cross-linguistic evidence that agreement is involved in BECs. Note that French shows obligatory gender agreement between the Range-DP and the overt complement of binominal chac- (cf. i). Similarly, Modern Irish shows animacy agreement in a structure which resembles BECs both syntactically and semantically (cf. ii). I thank Jim Mc Closkey (p.c.) for the Modern Irish data. (i) Les hommes ont achetés trois livres chac-un-(*e). the men-masc have bought three books each-one-(*fem) ‘The men bought three books each.’ (ii) a.

b.

Cheannaigh siad teach an duine bought three house the person ‘They bought three houses per person.’ Chosuuigh na tilhe dh’e ch’eod mile puut an ceann. cost the houses two hundred thousand pounds the one ‘The houses cost two hundred thousand pounds a piece.’

French

Modern Irish

18. Recall that (53b) also poses a problem for S&S’s EA, given that subextraction out of the subject is impossible (cf. (31a,b) above). 19. Here and beyond, ‘IP’ may also be read as a cover term for the functional projections proposed in the Split IP hypothesis of Pollock (1989) and Chomsky (1991). 20. Cf. also Barss and Lasnik (1986) and Jackendoff (1990) for arguments that the VP of English double object constructions(to the exclusion of the subject) is flat. 21. This raises the nontrivial question of how Case marking can be visible to the type-driven semantics employed here. I will leave this question open for further research. 22. The interpretation would proceed as follows. First, the verb (type ·eetÒ) combines with the subject trace (·eÒ) and the object (·eÒ) to yield a VP-denotation of type ·tÒ (this is the normal type of VPs on the VP-Internal Subject Hypothesis). λ-abstraction over the index of the trace opens up this argument position, yielding an expression of type ·etÒ. This expression cannot combine with the denotation of the subject jeweils + ShareNP (type ·eet,etÒ), and interpretation fails. 23. This position is exactly opposite to the one found in Kayne (1994), where linear precedence was deconstructed to be a by-product of hierarchic relations between elements.

371

References

Key: BLS = Proceedings of the annual meeting of the Berkeley Linguistics Society. Berkeley, California: Berkeley Linguistics Society. CLS = Papers from the annual regional meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society. ConSole = Proceedings of the Conference of the Student Organization of Linguistics in Europe. Leiden: SOLE. NELS = Proceedings of the North East Linguistic Society. Amherst: GLSA (from vol. 11), Montreal: Montreal Working Papers in Linguistics (vol. 6). WCCFL = The Proceedings of the West Coast Conference on Formal Linguistics. Stanford: Stanford Linguistics Association.

Abney, S. 1987. The English noun phrase in its sentential aspect. Dissertation, MIT. Abraham, W. 1995a. “Diathesis: the middle, particularly in West-Germanic”. In Discourse Grammar and Typology — Papers in Honor of John W. M. Verhaar, W. Abraham and T. Givon (eds), 3–48. Amsterdam: John Benjamins. Abraham, W. 1995b. Deutsche Syntax im Sprachvergleich: Grundlagen einer typologischen Syntax des Deutschen. Tübingen: Niemeyer. Ackema, P, Neeleman, A. and Weerman, F. 1993. “Deriving functional projections”. NELS 23: 17–31. Ackema, P. and Schoorlemmer, M. 1994. “The middle construction and the syntax-semantics interface.” Lingua 93: 59–90. Ackema, P. and Schoorlemmer, M. 1995. “Middles and nonmovement.” Linguistic Inquiry 26: 173–197. Aissen, J. 1996. Pied-piping, abstract agreement, and functional projections in Tzotzil. Natural Language and Linguistic Theory 14: 447–491. Aissen, J. and Perlmutter, D. 1976. “Clause reduction in Spanish”. BLS 2: 1–30. Aissen, J. and Perlmutter, D. 1983. “Clause reduction in Spanish”. In Studies in Relational Grammar 1, D. Perlmutter (ed), 360–403. Chicago: The University of Chicago Press. Alexiadou, A. 1994. Issues in the syntax of adverbs. Dissertation, University of Potsdam. Alexiadou, A. 1997. Adverb placement: a case study in antisymmetric syntax. Amsterdam/ Philadelphia: John Benjamins. Alexiadou, A. 1999. “On the properties of some Greek word order patterns”. In Studies in Greek Syntax, A. Alexiadou, G. Horrocks and M. Stavrou (eds), 45–65. Dordrecht: Kluwer Academic Publishers. Alexiadou, A. and Anagnostopoulou, E. 1998. “Parametrizing Agr: word order, verbmovement and EPP-checking”. Natural Language and Linguistic Theory 16: 491–539.

374 References

Alexiadou, A. and Anagnostopoulou, E. 2001a. “The Subject In-situ Generalization, and the role of Case in driving computations”. Linguistic Inquiry 32: 193–231. Alexiadou, A. and Anagnostopoulou, E. 2001b. “Covert feature movement and the placement of arguments”. In The Minimalist Parameter, G.M. Alexandrova and O. Arnaudova. (eds). Amsterdam: John Benjamins. Alexiadou, A. and Fanselow, G. This volume. “On the correlation between morphology and syntax: the case of V-to-I.” Allen, C. 1990. “Review of A. van Kemenade, Syntactic case and morphological case in the history of English”. Language 66: 146–152. Allen, C. 1995. Case marking and reanalysis. Grammatical relations from old to early modern English. Oxford: Clarendon Press. Altmann, H. 1981. Formen der “Herausstellung” im Deutschen. Linksversetzung, Rechtsversetzung, Freies Thema und verwandte Konstruktionen. Tübingen: Niemeyer. Aoun, J. and Benmamoun, E. 1998. “Minimality, reconstruction, and PF-movement”. Linguistic Inquiry 29: 569–597. Aoun, J., Benmamoun, E. and Sportiche, D. 1994. “Agreement, word order and conjunction in some varieties of Arabic”. Linguistic Inquiry 25: 195–220. Aoun, J. E. and Li, A. Y-H. 1993. The Syntax of Scope. Cambridge, Mass.: MIT Press. Ariel, M. 1990. Accessing noun-phrase antecedents. London/New York: Routledge. Arnold, S. 2000. “Existential ‘amount’ relatives.” MA thesis, University of California, Santa Cruz. Baker, M. 1988. Incorporation. Cambridge, Mass: MIT Press. Baptista, M. 1999. “Functional projections and parametrization”. UPennWPL 6.1: 17–35. Barbiers, S. 2000. “The right periphery in SOV languages: English and Dutch”. In The derivation of OV and VO, P. Svenonius (ed), 181–218. Amsterdam: John Benjamins. Barnes, M. 1987. “Some remarks on subordinate clause word order in Faroese”. Scripta Islandica 38: 3–35. Barss, T. and Lasnik, H. 1986. “A note on anaphora and double objects”. Linguistic Inquiry 17: 347–354. Battye, A. and Roberts, I. (eds) 1995. Clause structure and language change. New York: Oxford University Press. Bayer, J. 1996. Directionality and Logical Form: On the scope of focusing particles and wh-insitu. Dordrecht: Kluwer. Bayer, J., Bader, M. and Meng, M. 2001. “Morphological underspecification meets oblique case: syntactic and processing effects in German.” Lingua 111: 465–514. Bækken, B. 1998. Word order patterns in early modern English. Oslo: Novus Press. Bean, M. 1983. The development of word order patterns in Old English. London: Croom Helm. Belletti, A. 1990. Generalized verb movement: aspects of verb syntax. Torino: Rosenberg and Sellier. Bennis, H. and Hoekstra, T. 1989. Generatieve Grammatica. Dordrecht: Foris Publications. Bernstein, J. 1993. Topics in the syntax of nominal structure across Romance. Dissertation, CUNY. Bhatt, R. M. 1999. Verb movement and the syntax of Kashmiri. Dordrecht: Kluwer. Bianchi, V. 1999. Consequences of antisymmetry: headed relative clauses. Berlin: Mouton de Gruyter. Blake, N. F. 1970. “Introduction”. In The History of Reynard the Fox, N. F. Blake (ed). London: Early English Text Society.

References 375

Blancquaert, E. and Pée, W. (eds) 1930–1982. Reeks Nederlandse Dialect-atlassen (RND). Antwerp: De Sikkel. Blom, A. 1984. “Enkele opmerkingen over te en om.” Spektator 14: 170–176. Bobaljik, J. D. 1995. Morphosyntax: The syntax of verbal inflection. Dissertation, MIT. Bobaljik, J. D. 1997. “If the Head fits..: on the morphological determination of Germanic syntax”. Linguistics 35–6: 1029–1056. Bobaljik, J. D. 2000. “The rich agreement hypothesis in review.” Ms., McGill University. Bobaljik, J. D. 2002. “A-chains at the PF interface: copies and ‘covert’ movement.” Natural Language and Linguistic Theory 20: 197–267. Bobaljik, J. D. and Jonas, D. 1996. “Subject positions and the role of TP”. Linguistic Inquiry 27: 195–236. Bobaljik, J. D. and Thráinsson, H. 1998 “Two heads aren’t always better than one”, Syntax 1: 37–71. Bolinger, D. D. 1968. “Postposed main phrases: an English rule for the Romance subjunctive.” Canadian Journal of Linguistics 14: 3–30. Bolinger, D. D. 1972. That’s that. The Hague/Paris: Mouton. Borer, H. 1989. “Anaphoric AGR”. In The Null Subject Parameter, O. Jaeggli and K. Safir (eds), 69–109. Dordrecht: Kluwer Academic Publishers. Boškovic´, Ž. 1996. “Selection and the categorial status of infinitival complements”. Natural Language and Linguistic Theory 14: 269–304. Boškovic´, Ž. 1997. The syntax of nonfinite complementation: an economy approach. Cambridge, Mass: MIT Press. Boškovic´, Ž. and Takahashi, D. 1998. “Scrambling and Last Resort.” Linguistic Inquiry 29: 347–366. Bouchard, D. 1984. On the content of empty categories. Dordrecht: Foris. Brandner, E. 1994. Specifiers and functional projection. Dissertation, University of Stuttgart. Brandt, M. 1990. Weiterführende Nebensätze. Zu ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist and Wiksell. Brandt, M. 1997. “Der ‘redesituierende’ wie-Satz”. Sprache & Pragmatik 44: 1–35. Bremmer, R. 1997. “Het ontstaan van het Fries en het Hollands.” In Negen eeuwen FrieslandHolland. Geschiedenis van een haat-liefde verhouding, P. H. Breuker and R. Bremmer (eds), 67–76. Zutphen: Walburg Pers. Bresnan, J. 1972. Theory of complementation in English syntax. Dissertation, MIT. Bresnan, J. 1982. “Control and complementation”. In The mental representation of grammatical relations, J. Bresnan (ed), 282–390. Cambridge, Mass: MIT Press. Bresnan, J. 1994. “Locative inversion and the architecture of universal grammar”, Language 70: 72–131. Browning, M. 1987. Null operator constructions. Dissertation, MIT. Büring, D. 2001. “Let’s phrase it! Focus, word order, and prosodic phrasing in German double object constructions”. In Competition in syntax, G. Müller and W. Sternefeld (eds), 69–105. Berlin: Mouton de Gruyter. Burzio, L. 1992. “On the morphology of reflexives and impersonals.” In Theoretical analyses in Romance linguistics, Chr. Laeufer and T. Morgan (eds), 399–414. Amsterdam: John Benjamins. Butter, F. 1944 Trienke Boôd. De roman van een boeredochter. Hoorn: West-Friesland.

376 References

Buwalda, H. S. 1963 Hoe skrive wy ’t in ’t Bildts? ’n Handlaiding foar de spelling en inkele dingen út ’e spraakkûnst. St.-Annaparochie: Drukkerij Van Leer and De Jong. Buwalda, H. S., Buwalda, S. H. and A. C. B. van der Burg. 1996. Woordeboek fan ’t Bildts (en list fan topnimen). Leeuwarden: Frisian Academy. Cardinaletti, A. 1990 Impersonal constructions and sentential arguments in German. Padova: Unipress. Cardinaletti, A. and Roberts, I. 1991. “Clause Structure and X-Second.” Ms., University of Venice and University of Geneva. Cardinaletti, A. and Starke, M. 1995. “The Typology of Structural Deficiency. On the Three Grammatical Classes.” FAS Papers in Linguistics 1: 1–55. Carlson, G. 1977. “Amount Relatives”. Language 53: 520–542. Chafe, W. and Nichols, J. (eds). 1986. Evidentiality: the linguistic coding of epistemology. Norwood, New Jersey: Ablex. Cheng, L. 1991. On the typology of wh-questions. Dissertation, MIT. Chierchia, G. 1983. “Outline of a semantic theory of (obligatory) control”. WCCFL 2: 19–31. Chierchia, G. 1984a. “Anaphoric properties of infinitives and gerunds”. WCCFL 3: 28–39. Chierchia, G. 1984b. Topics in the syntax and semantics of infinitives and gerunds. Dissertation, University of Massachusetts at Amherst. Chierchia, G. 1989a. “Anaphora and attitudes de se”. In Semantics and contextual expression, R. Bartsch, J. van Benthem and P. van Emde Boas (eds), 1–32. Dordrecht: Foris. Chierchia, G. 1989b. “A semantics for unaccusatives and its syntactic consequences.” Ms., Cornell University. Chierchia, G. 1995. Dynamics of meaning: anaphora, presupposition, and the theory of grammar. Chicago: University of Chicago Press. Choe, J.-W. 1987. Anti-Quantifiers and a Theory of Distributivity. Dissertation, University of Massachusetts at Amherst. Chomsky, N. 1957. Syntactic structures. Den Haag: Mouton. Chomsky, N. 1970. “Remarks on nominalization.” In Readings in Transformational Grammar, R. Jacobs, and P. Rosenbaum (eds), 184–221. Waltham, Mass: Ginn. Chomsky, N. 1977. “On WH-movement”. In Formal Syntax, P. Culicover, T. Wasow and A. Akmajian (eds), 71–133. New York: Academic Press. Chomsky, N. 1982. Some concepts and consequences of the theory of government and binding. Cambridge, Mass.: MIT Press. Chomsky, N. 1986b. Barriers. Cambridge, Mass.: MIT Press. Chomsky, N. 1991. “Some notes on economy of derivation and representation.” In Principles and parameters in comparative grammar, R. Freidin (ed), 417–454. Cambridge, Mass.: MIT Press. Chomsky, N. 1993. “A Minimalist Program for Linguistic Theory”, in K. Hale and S. Keyser (eds), The View from Building 20. Cambridge, Mass: MIT Press, 1–52. Chomsky, N. 1994. Bare phrase structure. MIT Occasional Papers in Linguistics 5. Cambridge, Mass: MIT. Also appeared in Government and Binding Theory and the Minimalist Program. Webelhuth, G. (ed), 1995, Oxford: Basil Blackwell. Chomsky, N. 1995. The Minimalist Program, Cambridge, Mass.: MIT Press.

References 377

Chomsky, N. 2000. “Minimalist inquiries: the framework”. In Step by step: essays on minimalist syntax in honor of Howard Lasnik, R. Martin, D. Michaels, and J. Uriagereka (eds), 89–155. Cambridge, Mass: MIT Press. Chomsky, N. 2001. “Derivation by phase.” In Ken Hale: a life in language, M. Kenstowicz (ed), 1–52. Cambridge, Mass: MIT Press. Chomsky, N. and Lasnik, H. 1977. “Filters and control”. Linguistic Inquiry 8: 425–504. Chomsky, N. and Lasnik, H. 1993. “The theory of principles and parameters”. In Syntax: an international handbook of contemporary research, J. Jacobs, A. von Stechow, W. Sternefeld and T. Vennemann (eds), 506–569. Berlin: Walter de Gruyter. Chung, S., Ladusaw, W. and McCloskey, J. 1995. Sluicing and logical form. Natural Language Semantics 3: 239–282. Cinque, G. 1993. “On the evidence for partial N-movement in the Romance DP”. University of Venice Working Papers in Linguistics 3.2: 21–40. Cinque, G. 1997. “The interaction of passive, causative, and ‘restructuring’ in Romance”. Ms., University of Venice. To appear in The syntax of Italian dialects, C. Tortora (ed), New York: Oxford University Press. Cinque, G. 1999. Adverbs and functional heads. A Crosslinguistic Perspective. New York: Oxford University Press. Cinque, G. 2001. “Restructuring and the order of aspectual and root modal heads.” In Current studies in Italian syntax: essays offered to Lorenzo Renzi, G. Cinque and G. Salvi (eds), 137–155. Amsterdam: Elsevier. Coene, M. 1999. Definite null nominals in Romanian and Spanish. A generative approach to the internal syntax of DP. Dissertation, University of Antwerp. Coetsem, F. van 1988. Loan phonology and the two transfer types in language contact. Dordrecht: Foris. Cohen, A. 1996. Think generic: the meaning and use of generic sentences. Dissertation, Carnegie Mellon University. Comrie, B. 1984. “Subject and object control: syntax, semantics and pragmatics. BLS 10: 450–464. Corbett, G. 1995. “Agreement”. In Handbuch der erklärenden Syntax, J. Jacobs, A. von Stechow, W. Sternefeld, and T. Vennemann (eds), 1235–1244. Berlin: Mouton de Gruyter. Corver, N. 1994. “Parenthetical clauses: their nature and distribution”. Paper presented at the University of Tilburg, October. Corver, N. 1998. “Predicate movement in pseudopartitive constructions”. In Possessors, Predicates and Movement in the Determiner Phrase, A. Alexiadou and Ch. Wilder (eds), 215–257. Amsterdam: Benjamins. Culicover, P. 1999. Syntactic nuts: Hard cases, syntactic theory, and language acquisition. Oxford: Oxford University Press. Dabrowska, E. 1997. “The LAD goes to school: a cautionary tale for nativists”. Linguistics 35: 735–767. Dayal Srivastav, V. 1994. “Scope marking as indirect wh-dependency”. Natural Language Semantics 2: 137–170. Dayal Srivastav, V. 1996. “Scope marking: in defence of indirect dependency”. In Lutz and Müller (eds), 107–130.

378 References

Dayal Srivastav, V. 2000. “Scope marking: cross linguistic variation in indirect dependency”. In Lutz, Müller and von Stechow (eds), 157–193. De Bakker, C. 1997. Germanic and Romance inversion in French: A diachronic study, Dissertation, University of Amsterdam. Debrabandere, F. 1986. “Westvlaams en algemeen Nederlands”. Taal and Tongval 23: 88–93. Den Besten, H. 1983. “On the interaction of root transformations and lexical deletive rules”. In On the formal syntax of the Westgermania: Papers from the third Groningen Grammar Talks, Abraham, W. (ed), 47–131. Amsterdam/Philadelphia: John Benjamins. Den Dikken, M. 1995. “Copulas.” Ms., Free University Amsterdam. Den Dikken, M. 1998. “Predicate Inversion in DP.” In Possessors, Predicates and Movement in the Determiner Phrase, A. Alexiadou and Ch. Wilder (eds), 177–214. Amsterdam: John Benjamins. Den Dikken, M. and Giannakidou, A. 2002. “From hell to polarity: Aggressively nonD-linked wh-phrases as polarity items”. Linguistic Inquiry 33: 31–61. De Rooy, J. 1965. Als – of – dat. Een semantisch-onomasiologische studie over enkele subordinerende conjuncties in het ABN, de Nederlandse dialecten en het Fries, vergelijkendsynchronisch beschouwd. Assen: Van Gorcum. Di Sciullo, A.M. and Williams, E. 1987. On the definition of Word. Cambridge, Mass: MIT Press. Dobrovie-Sorin, C. 1998, “Impersonal se Constructions in Romance and the Passivization of Unergatives.” Linguistic Inquiry 29: 399–437. Dowty, D. 1991, “Thematic roles and argument selection.” Language 67: 547–619. Egerland, V. 1998. “On verb-second violations in Swedish and the hierarchical ordering of adverbs”. Working Papers in Scandinavian Syntax 61: 1–22. Emonds, J. E. 1976. A transformational approach to English syntax: Root, structure-preserving and local transformations. New York: Academic Press. Ernst, T. 1998. “The scopal basis of adverb licensing”. NELS 28: 127–142. Erteschik-Shir, N. 1984. “Der”. Nordic Journal of Linguistics 8: 131–148. Espinal, M. T. 1991. “The representation of disjunct constituents”. Language 67: 726–762. Evers, A. 1982. “Twee functionele principles voor de regel ‘Verschuif het werkwoord’”. GLOT 5: 11–30. Eythórsson, T. 1996. “Functional categories, cliticization, and verb movement in the Early Germanic languages.” In H. Thráinsson, S. Epstein, and S. Peter (eds), 109–139. Fagan, S. 1992. The syntax and semantics of middle constructions. Cambridge: Cambridge University Press. Falk, C. 1993. Non-referential subjects in the history of Swedish. Dissertation, University of Lund. Fanselow, G. 2001. “Features, θ-roles, and free constituent order”. Linguistic Inquiry 32: 405–437. Fanselow, G. and Mahajan, A. 1996. “Wh-expletive constructions in German and Hindi”. In Lutz and Müller (eds), 131–161. Farkas, D. 1988. “On obligatory control”. Linguistics and Philosophy 11: 27–58. Farkas, D. 1997. “Dependent indefinites”. In Empirical Issues in Formal Syntax and Semantics. Selected Papers from the Colloque de Syntaxe et de S’emantique de Paris (CSSP 1995), F. Corblin, D. Godard and J.-M. Marandin (eds), 243–267. Berne: Lang. Felser, C. 2001. “The syntax of wh-expletives in German.” Journal of Germanic Linguistics 1: 5–38.

References 379

Fokkema, K. 1937. Het Stadsfries; een bijdrage tot de geschiedenis en de grammatica van het dialect van Leeuwarden. Assen: Van Gorcum. Fokkema, K. 1948. De Liwwadder an ’t wood: zeven eeuwen taalgebruik in Leeuwarden. Assen: Van Gorcum. Fox, D. 1993. “Chain and binding: a modification of Reinhart and Reuland’s ‘reflexivity’.” Ms., MIT. Fox, D. 1995. “Economy and scope”. Natural Language Semantics 3: 283–341. Francis, W. N. 1942. “Introduction”. In The Book of Vices and Virtues, W. N. Francis (ed). London: Early English Text Society. Fuss, E. 1998. “Zur Diachronie von Verbzweit. Die Entwicklung von Verbstellungsvarianten im Deutschen und Englischen.” M. A. thesis, Johann-Wolfgang-Goethe-Universität Frankfurt. Gärtner, H.-M. 1991. “Erweiterungen der Bindungstheorie.” Sprachwissenschaft in Frankfurt 3. Geniušiene, E. 1987. The typology of reflexives. Berlin: Mouton de Gruyter. Giannakidou, A. and Merchant, J. 1998. “Reverse sluicing in English and Greek”. The Linguistic Review 15: 233–256. Givón, T. 1976. “Topic, pronouns, and grammatical agreement”. In Subject and Topic C. N. Li (ed), 149–188. New York: Academic Press. Goossens, J. 1970. “Belgisch beschaafd Nederlands en Brabantse expansie”. De Nieuwe Taalgids: Van Haeringennummer 54–70. Greenspon, M. D. 1996. A closer look at the middle construction. Dissertation, Yale University. Grewendorf, G. 1984. “Reflexivierungsregeln im Deutschen”. Deutsche Sprache 1: 14–30. Grewendorf, G. 1988. Aspekte der deutschen Syntax. Tübingen: Narr. Grimshaw, J. 1982. “On the lexical representation of romance reflexive clitics.” In The Mental Representation of Grammatical Relations, Bresnan, J. (ed), 87–148. Cambridge, Mass.: MIT. Grinder, J. T. 1970. “Super equi-NP deletion”. CLS 6: 297–317. Gussenhoven, C. 1983. On the grammar and semantics of sentence accents. Dordrecht: Foris. Haeberli, E. and Haegeman, L. 1995. “Clause structure in old English: Evidence from negative concord”. Journal of Linguistics 31: 81–108. Haeberli, E. 2000. “Adjuncts and the syntax of subjects in old and middle English”. In Diachronic Syntax: Models and Mechanisms, S. Pintzuk, G. Tsoulas and A. Warner (eds), 109–131. Oxford: Oxford University Press. Haeberli, E. 2002. “Inflectional morphology and the loss of V2 in English”. In Syntactic effects of morphological change, D. Lightfoot (ed), 89–106. New York: Oxford University Press. Haegeman, L. (ed). 1997. The new comparative syntax. London: Longman. Haegeman, L. 1999. “Embedded verb positions and remnant movement in the West Germanic OV languages”. Ms., University of Geneva. Haider, H. 1986a. Deutsche Syntax — generativ. Habilitationsschrift, University of Vienna. Haider, H. 1986b. “Fehlende Argumente: vom Passiv zu kohärenten Infinitiven”. Linguistische Berichte 101: 3–33. Haider, H. 1986c. “Nicht-sententiale Infinitive”. Groninger Arbeiten zur germanistischen Linguistik 28: 73–114.

380 References

Haider, H. 1986d. “Configurationality in disguise. Word order and the V-2 property”. In Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks Groningen 1984. W. Abraham and S. de Meij (eds), 39–64. Amsterdam: Benjamins. Haider, H. 1991a. “Fakultativ kohärente Infinitivkonstruktionen”. Arbeitspapiere des Sonderforschungsbereichs 340 “Sprachtheoretische Grundlagen für die Computerlinguistik” Nr. 17. University of Stuttgart. Haider, H. 1991b. “PRO-Bleme?” In Strukturen und Merkmale syntaktischer Kategorien, G. Fanselow and S. Felix (eds), 121–143. Tübingen: Narr. Haider, H. 1993. Deutsche Syntax, generativ: Vorstudien zur Theorie einer projektiven Grammatik. Tübingen: Narr. Haider, H. 2000. “OV is more basic than VO.” In The derivation of VO and OV, P. Svenonius (ed), 45–67. Amsterdam: John Benjamins. Halbertsma, J. 1865. Review of Die Nordfriesische Sprache nach der Mooringer Mundart, by B. Bendsen, Leiden: Brill, 1860. De Vrije Fries 10: 345–438. Hale, K. and Keyser, S. J. 1993. “On argument structure and the lexical expression of syntactic relations”. In The view from Building 20: essays in honor of Sylvain Bromberger, K. Hale and S.-J. Keyser, 53–109. Cambridge, MA: MIT Press. Hand, M. 1993. “Parataxis and parentheticals”. Linguistics and Philosophy 16: 495–507. Hardt, D. 1993. Verb phrase ellipsis: Meaning, form, and processing. Dissertation, University of Pennsylvania. Heim, I. 1982 The semantics of definite and indefinite noun phrases. Dissertation, University of Massachusetts at Amherst. Heim, I. 1987. “Where does the definiteness restriction apply? Evidence from the definiteness of variables”. In The Representation of (In)definiteness, E. J. Reuland and A. G. B. ter Meulen (eds), 21–42. Cambridge, MA: MIT Press, Heim, I. and Kratzer, A. 1998. Semantics in generative grammar. Oxford: Blackwell. Heine, B. 1992. “Grammaticalization chains”. Studies in Language 16: 335–368. Heine, B. 1993. Auxiliaries: cognitive forces and grammaticalization. Oxford: Oxford University Press. Herburger, E. 1994. “A semantic difference between full and partial wh-movement in German”. Paper presented at the LSA Annual Meeting, Boston, January 7. Hinterhölzl, R. 2000. “Licensing movement and stranding in the West Germanic OVlanguages”. In The derivation of VO and OV, P. Svenonius (ed), 293–326. Amsterdam: John Benjamins. Hoekstra, E. 1993a. “Dialectal variation inside CP as parametric variation”. In Dialektsyntax, W. Abraham and J. Bayer (eds), 161–179. Opladen: Westdeutscher Verlag. Hoekstra, E. 1993b. “On the parametrisation of functional projections in CP.” NELS 23, 191–204. Hoekstra, E. 1994a. “Positie- en bewegingsaspect bij selectie van de infinitief op -e of -en in het Westfries en het Fries”. Taal en Tongval 46: 66–73. Hoekstra, E. 1994b. “Oer de oerienkomst tusken de dialekten fan Noard-Hollân en it Frysk”. In Philologia Frisica anno 1993, P. H. Breuker, S. Dyk, D. Gorter, L. G. Jansma and W. Visser (eds), 81–103. Leeuwarden: Frisian Academy.

References 381

Hoekstra, E. 1998. “Oer de oerienkomst tusken de dialekten fan Grinslânsk en it Frysk.” In Philologia Frisica anno 1996, P. H. Breuker, S. Dyk, L. Jansma, W. Visser and J. Ytsma (eds), 117–137. Leeuwarden: Frisian Academy. Hoekstra, E. 2000. “The case of the two infinitives, language contact theory and the question of Frisian substrate in the Netherlands.” Ms., Leeuwarden: Frisian Academy. Hoekstra, J. 1992. “Fering tu-infinitives, North Sea Germanic syntax and Universal Grammar”. In Friesische Studien I, V. Faltings, A. Walker and O. Wilts (ed), 99–142. Odense: Odense University Press [= NOWELE Supplement vol. 8]. Hoekstra, J. 1993. “The split CP hypothesis and the Frisian complementizer system.” Ms., Leeuwarden: Frisian Academy. Hoekstra, J. 1995. “Preposition stranding and resumptivity in West Germanic”. In Studies in comparative Germanic syntax, H. Haider, S. Olsen, and S. Vikner (eds), 95–118. Dordrecht: Kluwer. Hoekstra, Joh. 1993. “It feroaringsproses by je-tiidwurden”. Tydskrift foar Fryske Taalkunde 8, 43–51. Hoekstra, T. and Roberts, I. 1993, “Middle constructions in Dutch and English.” In Knowledge and Language. Vol. 2: Lexical and Conceptual Structure, E. Reuland and W. Abraham (eds), 183–220. Dordrecht: Kluwer. Höhle, T. N. 1996. “The w…w-construction: appositive or scope-indicating?” In Lutz and Müller (eds), 37–58. [also in Lutz, Müller and von Stechow (eds), 249–270] Holmberg, A. 1983. “Performer-parentheticals in Swedish, English and French”. In Papers from the Seventh Conference on Scandinavian Linguistics, F. Karlsson (ed), 60–79. Helsinki: Dept. of Linguistics. Holmberg, A. 1986. Word order and syntactic features in the Scandinavian Languages and English. Dissertation, University of Stockholm. Holmberg, A. 1997. “Scandinavian Stylistic Fronting: movement of phonological features in the Syntax”. Working Papers in Scandinavian Syntax 60: 81–124. Holmberg, A. 1999. “Remarks on Holmberg’s Generalization”. Studia Linguistica 53: 1–39. Holmberg, A. 2000. “Scandinavian stylistic fronting: how any category can become an expletive”. Linguistic Inquiry 31: 445–483. Holmberg, A. and Platzack, C. 1988. “On the role of inflection in Scandinavian syntax”. Working Papers in Scandinavian Syntax 42: 25–42. Holmberg, A. and Platzack, C. 1995. The role of inflection in scandinavian syntax. New York/ Oxford: Oxford University Press. Hornstein, N. 1994. “An argument for minimalism: the case of antecedent-contained deletion.” Linguistic Inquiry 25: 455–480. Hornstein, N. 1999. “Movement and control”. Linguistic Inquiry 30: 69–96. Hornstein, N. and Weinberg, A. 1981. “Case theory and preposition stranding”. Linguistic Inquiry 12: 55–91. Horrocks, G. and Stavrou, M. 1987. “Bounding theory and Greek syntax: evidence for whmovement in NP”. Journal of Linguistics 23: 79–108. Huang, J. 1982. Logical relations in Chinese and the theory of grammar. Dissertation, MIT. Hulk, A. and Van Kemenade, A. 1995. “Verb second, pro-drop, functional projections and language change”. In Battye and Roberts (eds), 227–256.

382 References

Hulk, A. and Van Kemenade, A. 1997. “Negation as a reflex of clause structure”. In Negation and Polarity: Syntax and Semantics, D. Forget, P. Hirschbuehler, F. Martineau and M.-L. Rivero (eds), 183–207. Amsterdam: John Benjamins. Iwata, S. 1999. “On the status of implicit arguments in middles.” Journal of Linguistics 35: 527–553. Jackendoff, R. 1990. “On Larson’s treatment of double object constructions”. Linguistic Inquiry 21: 427–456. Jacobsen, B. and Jensen, P.-A. 1982. “Some remarks on Danish weakly stressed der”. Ms., Aarhus Business School. Jonas, D. 2002. “Residual V-to-I”. In Syntactic effects of morphological change, D. Lightfoot (ed). New York: Oxford University Press. Jonkman, R. 1993. It Leewarders. In taalsosjologysk ûndersyk nei it Stêdsk yn ferhâlding ta it Nederlânsk en it Frysk. Leeuwarden: Frisian Academy. Julien, M. 2000. Syntactic heads and word formation. Dissertation, University of Tromsø. Kawasaki, N. 1993. Control and arbitrary interpretation in English. Dissertation, University of Massachusetts at Amherst. Kayne, R. S. 1975. French syntax: The transformational cycle. Cambridge, Mass.: MIT Press. Kayne, R. S. 1980. “Extensions of binding and case-marking”. Linguistic Inquiry 11: 75–96. Kayne, R. S. 1981. “On certain differences between French and English”. Linguistic Inquiry 12: 349–371. Kayne, R.S. 1993. “Toward a modular theory of auxiliary selection”. Studia Linguistica 47: 3–31. Kayne, R. S. 1994. The antisymmetry of syntax. Cambridge, Mass: MIT Press. Kayne, R. S. 1998. “Overt versus covert movement”. Syntax 1: 128–191. Kayne, R. S. 1999. “Prepositional Complementizers as Attractors”. Probus 11: 39–73. Kemmer, S. 1993. The Middle Voice. Amsterdam/Philadelphia: John Benjamins. Kennedy, C. 1997. “Antecedent-contained deletion and the syntax of quantification.” Linguistic Inquiry 28: 662–688. Kennedy, C. and Merchant, J. 2000. “Attributive comparative deletion”. Natural Language and Linguistic Theory 18: 89–146. Keyser, S. J. and Roeper, T. 1984. “On the middle and ergative constructions in English.” Linguistic Inquiry 15: 381–416. Keyser, S. J. and Roeper, T. 1992. “Re: The abstract clitic hypothesis.” Linguistic Inquiry 23: 89–125. Kiparsky, P. 1995. “Indo-european origins of Germanic syntax.” In Battye and Roberts (eds), 140–169. Kiparsky, P. and Kiparsky, C. 1970. “Fact”. In Papers in linguistics, M. Bierwisch and K. E. Heidolph (eds), 143–173. The Hague: Mouton. Klamer, M. 1998. “Verbs becoming conjunctions in Malayo-Polynesian.” Ms., Free University Amsterdam/KNAW. [appeared as “How report verbs become quote markers and complementizers”, Lingua 110 (2000): 69–98] Koeneman, O. 1997. “On V to I movement and flexible syntax.” ConSole 5: 183–198. Koeneman, O. 2000. The flexible nature of verb movement. Dissertation, Utrecht University. Koeneman, O. and Neeleman, A. 2001. “Verb movement, predication and the distribution of expletives”. Lingua 111: 189–233.

References 383

König, E. and Siemund, P. 1996a. “Emphatische Reflexiva und Fokusstruktur: zur Syntax und Bedeutung von selbst”. Sprache und Pragmatik 4: 1–42. König, E. and Siemund, P. 1996b. “Intensifiers and reflexives: a typological perspective”. In Reflexives: form and function, Z. Frajzyngier and T. S. Curl (eds), 41–74. Amsterdam: Benjamins. Koopman, H. 1995. “On verbs that fail to undergo Verb-Second.” Linguistic Inquiry 26: 137–163. Koopman, H. and D. Sportiche. 1991. “The position of subjects.” Lingua 85: 211–258. Koopman, H. and Szabolcsi, A. 2000. Verbal complexes, Cambridge, Mass.: MIT Press. Koopman, W. 1995. “Verb-Final main clauses in Old English prose”. Studia Neophilologica 67: 129–144. Koopman, W. 1998. “Inversion after single and multiple topics in old English”. In Advances in English Historical Linguistics, J. Fisiak and M. Krygier (eds), 135–149. Berlin: Mouton de Gruyter. Kosmeier, W. 1986. “The status of the finite inflection in Icelandic and Swedish”. Working Papers in Scandinavian Syntax 26: 1–41. Koster, J. 1984. “On binding and control”. Linguistic Inquiry 15: 417–459. Koster, J. 2000. “Pied piping and the word orders of English and Dutch.” NELS 30: 415–426. Kratzer, A. 1986. “Conditionals”. CLS 22.2: 1–15. Kratzer, A. 1994. The event argument and the semantics of voice. Ms. University of Massachusetts, Amherst. [to appear with MIT Press, Cambridge, Mass.] Kratzer, A. 1995, “Stage-level and individual-level predicates.” In The Generic Book, G. N. Carlson and F. J. Pelletier (eds), 125–175. Chicago. Kratzer, A. 1996. “Severing the external argument from its verb”. In Phrase Structure and the Lexicon, J. Rooryck and L. Zaring (eds), 109–137. Dordrecht: Kluwer. Kratzer, A. 1998. “Scope or pseudo-scope? Are there wide-scope indefinites?” In Events in Grammar, S. Rothstein (ed), 163–196. Kluwer: Dordrecht. Krifka, M. et al. 1995, “Genericity: An introduction.” In The Generic Book, G. N. Carlson and F. J. Pelletier (eds), 1–124. Chicago. Krisch, Th. 1997. “Delbrücks Arbeiten zur Wortstellung aus heutiger Sicht”. In Bertholt Delbrück y la sintaxis indoeuropea hoy, E. Crespo and J.-L. Garcia-Ramón (eds), 283–309. Madrid: Ediciones UAM and Wiesbaden: Dr. Ludwig Reichert Verlag. Kroch, A. 1989. “Reflexes of grammar in patterns of language change”, Language Variation and Change 1: 199–244. Kroch, A. 2000. “Syntactic change”. In Handbook of contemporary syntactic theory, M. Baltin and C. Collins (eds), 629–739. Malden, Mass.: Blackwell. Kroch, A. and Taylor, A. 1994. The Penn-Helsinki parsed corpus of Middle English. Philadelphia: Department of Linguistics. Kroch, A. and Taylor, A. 1997. “Verb movement in old and middle English: dialect variation and language contact”. In Van Kemenade and Vincent (eds), 297–325. Kroch, A., Taylor, A. and Ringe, D. 2000. “The middle English verb-second constraint: a case study in language contact and language change”. In Textual Parameters in Older Language, S. Herring, P. van Reenen, and L. Schoesler (eds), 353–391. Amsterdam: John Benjamins. Kuno, S. 1975. “Super-Equi-NP deletion is a pseudo-transformation”. NELS 8: 29–44.

384 References

Kytö, M. 1993. Manual to the diachronic part of the Helsinki corpus of English texts. Helsinki: Department of English. Laan, K. ter 1953. Proeve van een groninger spraakkunst. Winschoten: Van der Veen. Landau, I. 1999. Elements of control. Dissertation, MIT. Landau, I. 2000. Elements of control: structure and meaning in infinitival constructions. Dordrecht: Kluwer Academic Publishers. Larson, R. 1988. “On the Double Object Construction”. Linguistic Inquiry 19: 335–391. Lasnik, H. 1981 “Restricting the theory of transformations: a case study”. In Explanation in linguistics: The logical problem of language acquisition, N. Hornstein and D. Lightfoot (eds), 152–173. London: Longmans. Lasnik, H. 1999. “On feature strength: Three Minimalist approaches to overt movement”. Linguistic Inquiry 30: 197–217. Lebeaux, D. 1984. “Anaphoric binding and the definition of PRO”. NELS 14: 253–274. Lebeaux, D. 1985. “Locality and anaphoric binding”. The Linguistic Review 4: 343–363. Leopold, J. A. and Leopold, L. 1882. Van de Schelde tot de Weichsel. Nederduitse dialecten in dicht en ondicht. Groningen: Wolters. Levin, L. 1982. “Sluicing: A lexical interpretation procedure”. In The mental representation of grammatical relations, J. Bresnan (ed), 590–654. Cambridge, Mass: MIT Press. Lightfoot, D. 1995. “Why UG needs a learning theory: triggering verb movement”. In Battye and Roberts (eds), 31–52. Lightfoot, D. 1997. “Shifting triggers and diachronic reanalysis”. In Van Kemenade and Vincent (eds), 253–272. Link, G. 1983. “The logical analysis of plural and mass terms: A lattice-theoretical approach”. In Meaning, Use, and Interpretation of Language, R. Bäuerle et al. (eds), 302–323. Berlin: Walter de Gruyter. Link, G. 1998. “Je drei Äpfel — three apples each’: Quantification and the German ‘je’.” In Algebraic Semantics in Language and Philosophy, by G. Link, 117–132. Stanford: CSLI Publications. Lipták, A. and Zimmermann, M. 2000. “A unified analysis of binominal each-constructions in English, dutch, and German”. SKY journal of linguistics 13: 123–154 Lobeck, A. 1995. Ellipsis: Functional heads, licensing, and identification. Oxford: Oxford University Press. Lutz, U. 2001. “Wh-scope marking and partial movement in German was…w-constructions”. In Linguistic form and its computation, Chr. Rohrer, A. Rossdeutscher, and H. Kamp (eds), 65–107. Stanford: CSLI Publications. Lutz, U. and Müller, G. (eds). 1996. Papers on wh-scope marking. Stuttgart and Tübingen: Arbeitspapiere des SFB 340, Bericht Nr. 76. Lutz, U., Müller, G. and Stechow, A. von (eds). 2000. Wh-scope marking. Amsterdam: Benjamins. Manzini, R. 1983. “On control and control theory”. Linguistic Inquiry 14: 412–446. Manzini, R. and Roussou, A. 2000. “A minimalist theory of A-movement and control”. Lingua 110: 409–447. Martin, R. A. 1996. A minimalist theory of PRO and control. Dissertation, University of Connecticut, Storrs. Matthews, G. H. 1965. Hidatsa syntax. The Hague: Mouton.

References 385

Matthewson, L. 1999. “On the interpretation of wide-scope indefinites”. Natural Language Semantics 7: 79–134. May, R. 1985. Logical Form: its structure and derivation. Cambridge, Mass: MIT Press. McCloskey, J. 1996. “Subjects and subject positions in Irish”. In The Syntax of Celtic Languages, B. Borsley and Ian Roberts (eds), 241–283. Cambridge: Cambridge University Press. McCloskey, J. and Hale, K. 1984. “On the Syntax of Person Number Inflection in Modern Irish”, Natural Language and Linguistic Theory 2: 487–533. McDaniel, D. 1989. “Partial and multiple wh-movement”. Natural Language and Linguistic Theory 7: 565–604. Meech, S. B. 1940. “Introduction”. In The Book of Margery Kempe, S. B. Meech and H. E. Allen (eds). London: Early English Text Society. Merchant, J. 2000. “Antecedent-contained deletion in negative polarity items”. Syntax 3: 144–150. Merchant, J. 2001. The syntax of silence: Sluicing, islands, and identity in ellipsis. Oxford: Oxford University Press. Miller, D. G. 1993. Complex verb formation. Amsterdam and Philadelphia: John Benjamins. Milsark, G. L. 1979. Existential sentences in English. New York: Garland Publishers. Mitchell, B. 1985 Old English syntax. Oxford: Clarendon Press. Moltmann, F. 1991. “On the syntax and semantics of binary distributive quantifiers”. NELS 22: 279–292. Moltmann, F. 1997. Parts and wholes in semantics. NewYork/Oxford: Oxford University Press. Moore, J. 1990. “Spanish clause reduction with downstairs cliticization”. In Grammatical relations: a cross-theoretical perspective, K. Dziwirek, P. Farrell and E. Mejias-Bikandi (eds), 319–334. Stanford: Center for the Study of Language and Information. Moore, J. 1994. “Romance cliticization and relativized minimality”. Linguistic Inquiry 25: 335–344. Moro, A. 1997. The raising of predicates. Cambridge: Cambridge University Press. Müller, G. 1995. A-bar syntax: A study in movement types. Berlin: Mouton de Gruyter. Müller, G. 1997. “Partial wh-movement and Optimality Theory”. The Linguistic Review 14: 249–306. Müller, G. and Sternefeld, W. 1996. “A-chain formation and economy of derivation”. Linguistic Inquiry 27: 480–511. Müller, G. and Sternefeld, W. 2001. “The rise of competition in syntax: a synopsis”. In Competition in syntax, G. Müller and W. Sternefeld (eds), 1–68. Berlin: Mouton de Gruyter. Naarding, J. 1961. “Stad en Ommelanden”. In De Ontfriesing van Groningen, K. Heeroma en J. Naarding (ed), 45–62. Zuidlaren: J. K. Hertz. Neeleman, A. 1994. Complex Predicates. Dissertation, Utrecht University. Nishigauchi, T. 1998. ‘Multiple sluicing’ in Japanese and the functional nature of whphrases. Journal of East Asian Linguistics 7: 121–152. Noonan, M. 1992. A grammar of Lango. Berlin: Mouton de Gruyter. Nuñez, J. 1995. The copy theory of movement and linearization of chains in the minimalist program. Dissertation, University of Maryland.

386 References

O’Neil, J. H. 1997. Means of control: deriving the properties of PRO in the Minimalist Program. Dissertation, Harvard. Olsen, S. 1991. “Die deutsche Nominalphrase als ‘Determinansphrase’.” In DET, COMP und INFL: zur Syntax funktionaler Kategorien und grammatischer Funktionen, S. Olsen and G. Fanselow (eds), 1–14. Tübingen: Niemeyer. Ouhalla, J. 1988. The syntax of head movement: a case study of Berber. Dissertation, University College London. Ouhalla, J. 1994. “The syntactic representation of arguments”. Ms. Berlin: Max Planck. Partee, B. 1976. “Some transformational extensions of Montague Grammar”. In Montague Grammar, B. Partee (ed), 51–76. New York: Academic Press. Penner, Z. and Bader, T. 1991. “Main clause phenomena in embedded clauses: the licensing of embedded V2-clauses in Bernese Swiss German”. The Linguistic Review 8: 75–95. Perry, G. G. 1914. “Preface”. In Religious Pieces in Prose and Verse, G. G. Perry (ed). London: Early English Text Society. Pesetsky, D. 1987. “Wh-in-situ: Movement and unselective binding”. In The representation of (in)definiteness, E. Reuland and A. ter Meulen (eds), 98–129. Cambridge, Mass: MIT Press. Pesetsky, D. 1995. Zero syntax. Cambridge, Mass: MIT Press. Pesetsky, D. 2000. Phrasal movement and its kin. Cambridge, Mass: MIT Press. Petter, M. 1998. Getting PRO under control. Dissertation, Free University Amsterdam. Picallo, C. 1985. Opaque domains. Dissertation, CUNY. Picallo, C. 1990. “Modal verbs in Catalan”. Natural Language and Linguistic Theory 8: 285–312. Pili, D. 2000. “Some considerations on the structure of the VP in Italian.” Linguistik online 6.2. [http://viadrina.euv-frankfurt-o.de/~wjournal/2_00/] Pintzuk, S. 1991. Phrase structures in competition: variation and change in old English word order. Dissertation, University of Pennsylvania. [appeared 1999, New York: Garland] Pintzuk, S., Haeberli, E., Van Kemenade, A., Koopman, W., and Beths, F. (eds). 2000. The Brooklyn-Geneva-Amsterdam-Helsinki parsed corpus of Old English. Department of Language and Linguistic Science, University of York. [http://www-users.york.ac.uk/ ~sp20/corpus.html] Pittner, K. 1995. “Zur Syntax von Parenthesen”. Linguistische Berichte 156: 85–108. Platzack, C. 1985. “Narrative inversion in Old Icelandic”, Íslenskt mál 7: 127–144. Platzack, C. 1988. “The emergence of a word order difference in Scandinavian subordinate clauses”. McGill Working Papers in Linguistics, special issue on comparative Germanic syntax: 215–238. Platzack, C. 1995. “The loss of verb second in English and French”. In Battye and Roberts (eds), 200–226. Platzack, C. and Holmberg, A. 1989. “The role of AGR and finiteness in the Germanic VO languages”. Working Papers in Scandinavian Syntax 43: 51–76. Pollard, C. and Sag, I. 1994. Head-driven Phrase Structure Grammar. Stanford: CSLI. Pollock, J. Y. 1989. “Verb movement, universal grammar and the structure of IP”, Linguistic Inquiry 20: 365–424. Postal, P. 1972. “On some rules that are not successive cyclic”. Linguistic Inquiry 3: 211–222. Postal, P. 1993. “Some defective paradigms”. Linguistic Inquiry 24: 347–364.

References 387

Postma, G. 1999. “De negatief-polariteit van het syntactische ghe-partikel in het Middelnederlands.” Nederlandse Taalkunde 4: 310–329. Postma, G. 1998. “SAY and the complementizer.” Paper presented at a miniconference on COMP, quotes and narrative style, Leiden University, November 20. Rapp, I. and Von Stechow, A. 1999. “Fast ‘almost’ and the visibility parameter for functional adverbs”. Journal of Semantics 16: 149–204. Reinhart, T. 1981. “A second COMP position.” in Theory of markedness in generative grammar, A. Belletti et al. (eds), 518–557. Pisa: Scuola Normale Superiore. Reinhart, T. 1983a. “Point of view in language: the use of parentheticals”. In Essays on deixis, G. Rauh (ed), 169–194. Tübingen: Narr. Reinhart, T. 1983b. Anaphora and semantic interpretation. London: Croom Helm. Reinhart, T. 1987. “Specifier and operator binding”. In The Representation of (In)definiteness, E. J. Reuland and A. G. B. ter Meulen (eds), 130–67. Cambridge, Mass.: MIT Press. Reinhart, T. 1995. “Interface strategies.” Utrecht: OTS Working Papers. Reinhart, T. 1996. “Syntactic effects of lexical operations: reflexives and unaccusatives”. Utrecht: OTS Working Papers. Reinhart, T. 1997. “Quantifier Scope. How labor is divided between QR and choice functions”. Linguistics and Philosophy 20: 335–397. Reinhart, T. and Reuland, E. 1993. “Reflexivity.” Linguistic Inquiry 24: 650–720. Reis, M. 1985. “Satzeinleitende Strukturen im Deutschen. Über COMP-, Haupt- und Nebensätze, W-Bewegung und die Doppelkopfanalyse”. In Erklärende Syntax des Deutschen, W. Abraham (ed), 271–311. Tübingen: Narr. Reis, M. 1995. “Wer glaubst du hat recht? On so-called extractions from Verb-Second clauses and Verb-First parenthetical constructions in German”. Sprache & Pragmatik 36: 27–83. Reis, M. 1996a. “Extractions from Verb-Second clauses in German?” In On extraction and extraposition in German, U. Lutz and J. Pafel (eds), 45–88. Amsterdam: Benjamins. Reis, M. 1996b. “On was-parentheticals and was…w-constructions in German”. In Lutz and Müller (eds), 257–288. Reis, M. 1997. “Zum syntaktischen Status unselbständiger Verbzweit-Sätze”. In Sprache im Fokus. Festschrift für Heinz Vater zum 65. Geburtstag, C. Dürscheid, K.-H. Ramers, and M. Schwarz (eds), 121–144. Tübingen: Niemeyer. Reis, M. 2000a. “On the parenthetical features of German was…w-constructions and how to account for them”. In Lutz, Müller and von Stechow (eds), 359–408. Reis, M. 2000b. “Integrierte Parenthesen und w-Konstruktionen. Weitere Überlegungen zum Grammatik-Pragmatik-Verhältnis”. Paper presented at the “Sprache & Pragmatik Symposium”, May. [to appear in Sprache & Pragmatik] Reker, S. 1991. “Vormleer”. In Groninger Grammatica, by S. Reker. Veendam: Staalboek. Reuland, E. and Reinhart, T. 1995. “Pronouns, anaphors and case.” In Studies in Comparative Germanic Syntax, H. Haider, S. Olsen and S. Vikner (eds), 241–268. Dordrecht. Richards, N. 1999. “Subject extraction without subjects”. Paper presented at the Workshop on Grammatical Functions, University of Illinois, July 10. Richards, N. 2001. Movement in language: Interactions and architectures. Oxford: Oxford University Press. Rivero, M. L. 1994. “Verb movement and the structure of IP in the languages of the Balkans”. Natural Language and Linguistic Theory 12: 63–120.

388 References

Rizzi, L. 1978. “A restructuring rule in Italian syntax”. In Recent transformational studies in European languages, S. J. Keyser (ed), 113–158. Cambridge, Mass: MIT Press. Rizzi, L. 1982. Issues in Italian syntax. Dordrecht: Foris. Rizzi, L. 1990. Relativized minimality. Cambridge, Mass.: MIT Press. Rizzi, L. 1996. “Residual verb second and the Wh-criterion”. In Parameters and functional heads, A. Belletti and L. Rizzi (eds), 63–90. New York: Oxford University Press. Rizzi, L. 1997. “The fine structure of the left periphery”. In Elements of grammar, L. Haegeman (ed), 289–330. Dordrecht: Kluwer Academic Publishers. Roberts, I. 1993 Verbs and diachronic syntax: a comparative history of English and French, Dordrecht: Kluwer. Roberts, I. 1997. “Have/Be Raising, Move F and Procrastinate”. Linguistic Inquiry 29: 113–125. Roberts, I. and Roussou, A. To appear. “The extended projection principle as a condition on the tense-dependency”. In Subjects, expletives and the EPP, P. Svenonius (ed). New York: Oxford University Press. Rochette, A. 1988. Semantic and syntactic aspects of Romance sentential complementation. Dissertation, MIT. Rochette, A. 1990. “On the restructuring classes of verbs in Romance”. In Binding in Romance: essays in honour of Judith McA’Nulty, A. M. Di Sciullo and A. Rochette (eds), 96–128. Ottawa: Canadian Linguistic Association. Rochette, A. 1999. “The selection properties of aspectual verbs”. In Beyond principles and parameters: essays in memory of Osvaldo Jaeggli, K. Johnson and I. Roberts (eds), 145–165. Dordrecht: Kluwer Academic Publishers. Rohrbacher, B. 1994. The Germanic languages and the full paradigm: a theory of V to I raising, Dissertation, University of Massachusetts at Amherst. Rohrbacher, B. 1999. Morphology-driven Syntax. A theory of V-to-I raising and pro-drop. Amsterdam: John Benjamins. Romero, M. 1998. Focus and reconstruction effects in wh-phrases. Dissertation, University of Massachusetts at Amherst. Rooryck, J. 1997. “On the interaction between Raising and Focus in sentential complementation”. Studia Linguistica 50: 1–49. Rosen, C. 1976. “Guess what about?” NELS 6: 205–211. Rosen, S.T. 1989. Argument structure and complex predicates. Dissertation, Brandeis University. Rosen, S. T. 1990. “Restructuring Verbs are Light Verbs”. WCCFL 9: 477–491. Ross, J. 1969. “Guess who?” CLS 5: 252–286. Ross, J. R. 1973. “Slifting”. In The formal analysis of natural languages, M. Gross, M. Halle, and M.-P. Schützenberger (eds), 133–169. The Hague: Mouton. Sabel, J. 1996. Restrukturierung und Lokalität: Universelle Beschränkungen für Wortstellungsvariationen. Berlin: Akademie Verlag. Safir, K. and Stowell, T. 1988. “Binominal each”. NELS 18: 426–450. Sakaguchi, M. 1998. Distributivity in English and Japanese. Dissertation, UCLA. Sauerland, U. 1999. “Scope reconstruction without reconstruction”. WCCFL 17: 582–596. Schmidt, D. A. 1980. A history of inversion in English, Dissertation, The Ohio State University. Schmitt, L. E. 1942. “Die Stadt Groningen und die Mundarten zwischen Laubach und Weser”. Zeitschrift für Mundartforschung 18: 134–170.

References 389

Schönfeld, M. 1970. Historische Grammatica van het Nederlands, A. van Loey (ed). Zutphen: W. J. Thieme. Schwarzschild, R. 1999. “GIVENness, AVOID-F, and other constraints on the placement of accent”. Natural Language Semantics 7: 141–177. Sigurðsson, H.Á. 1989. Verbal syntax and case in Icelandic. Dissertation, University of Lund. Sigurðsson, H.Á. 1990. “V1 declaratives and verb raising in Icelandic”. Syntax and Semantics 24: Modern Icelandic Syntax, J. Maling and A. Zaenen (eds), 41–69. San Diego: Academic Press. Sigurðsson, H.Á. 1991. “Icelandic Case-marked PRO and the licensing of lexical arguments”. Natural Language and Linguistic Theory 9: 327–363. Slobin, N. 1997. “Agreement, default rules, and grammatical viruses”. Linguistic Inquiry 28: 318–343. Snyder, W. 1995. Language acquisition and language variation: the role of morphology. Dissertation, MIT. Sportiche, D. 1988. “A theory of floating quantifiers and its corollaries for constituent structure”. Linguistic Inquiry 19: 425–449. Steinbach, M. 1999a. “Argument structure and reflexivity: middle constructions in German.” Paper presented at the Texas Linguistics Society conference on Argument Structure, March 6. Steinbach, M. 1999b. “Notes on parenthetical constructions.” Arbeitspapiere des Sonderforschungsbereichs 340. Bericht Nr. 144. University of Stuttgart/Tübingen. Steinbach, M. 2002. Middle voice: A comparative study in the syntax-semantics interface of German. Amsterdam: John Benjamins. Sternefeld, W. 2000. “Grammatikalität und Sprachvermögen. Anmerkungen zum Induktionsproblem in der Syntax.” In Von der Philologie zur Grammatiktheorie: Peter Suchsland zum 65. Geburtstag, J. Bayer and Chr. Römer (eds), 15–44. Tübingen: Niemeyer. Stockwell, R. 1984. “On the history of the Verb-second rule in English”, in Historical Syntax, J. Fisiak (ed), 575–592. Berlin: Mouton de Gruyter. Stowell, T. 1982. “The tense of infinitives”. Linguistic Inquiry 13: 561–570. Stowell, T. 1981. Origins of phrase structure. Dissertation, MIT. Stowell, T. 1984. “Null operators and the theory of proper government.” Ms., UCLA. Stroik, T. 1990. “Adverbs as V-sisters.” Linguistic Inquiry 21: 654–661. Stroik, T. 1992. “Middles and movement.” Linguistic Inquiry 23: 127–137. Stroik, T. 1999. “Middles and reflexivity.” Linguistic Inquiry 30: 119–131. Strozer, J. R. 1976. Clitics in Spanish. Dissertation, UCLA. Strozer, J. R. 1981. “An alternative to restructuring in Romance syntax”. Papers in Romance (PRom) 3: 177–184. Sturm, A. 1986. Primaire syntactische structuren in het Nederlands. Dissertation, University of Nijmegen. Szabolsci, A. 1983. “The possessor that ran away from home”. The Linguistic Review 3, 89–102. Szabolcsi, A. 1987. “Functional categories in the noun phrase”. In Approaches to Hungarian 2, I. Kenesei (ed), 167–190. Szeged: JATE. Takami, K. 1992. Preposition stranding: From syntactic to functional analyses. Berlin: Mouton de Gruyter.

390 References

Tappe, H. T. 1981. “Wer glaubst du hat recht? Einige Bemerkungen zur COMP-COMPBewegung im Deutschen”. In Sprache: Formen und Strukturen. Akten des 15. Linguistischen Kolloquiums Münster 1980. Vol. 1, M. Kohrt and J. Lenerz (eds), 203–212. Tübingen: Niemeyer. Taraldsen, K. T. 1986. “Som and the binding theory”. In Topics in Scandinavian Syntax, L. Hellan and K. K. Christensen (eds), 149–184. Dordrecht: Reidel. Taraldsen, K. T. 1991. “Two arguments for functional heads”. Lingua 84: 85–108. Tasmowski, L. 1989. “A Look at Sembler + Infinitive from Different Angles”. In Sentential complementation and the lexicon. Studies in honour of Wim de Geest, D. Jaspers et al. (eds), 403–421. Dordrecht/Providence: Foris Publications. Thiersch, C. 1978. Topics in German syntax. Dissertation, MIT. Thomason, S. S. and T. Kaufman. 1988. Language contact, creolization, and genetic linguistics. Berkeley: University of California Press. Thornton, R. 1990. Adventures in long-distance movement. Dissertation, University of Connecticut, Storrs. Thráinsson, H. 1993. “On the structure of infinitival complements”. Harvard Working Papers in Linguistics 3: 181–213. Thráinsson, H. 1996. “On the (non-)universality of functional categories”. In Minimal ideas: syntactic studies in the minimalist framework, W. Abraham, S. D. Epstein, H. Thráinsson, and C. J. W. Zwart (eds), 253–281. Amsterdam: John Benjamins. Thráinsson, H., Epstein, S. D., and Peter, S. (eds) 1996. Studies in Comparative Germanic Syntax II. Dordrecht: Kluwer. Tiersma, P. M. 1985. Frisian reference grammar. Dordrecht: Foris. Travis, L. 1984. Parameters and effects of word order variation. Dissertation, MIT. Trissler, S. 1993. “P-stranding im Deutschen”. In Extraktion im Deutschen I, F.-J. d’Avis et al. (eds), 247–291. Tübingen/Stuttgart: Arbeitspapiere des Sonderforschungsbereichs 340, Bericht 34. Van Bree, C. 1994. “Het probleem van het ontstaan van het ‘Stadsfries’ in verband met nieuwe talen in contact-theorieën”. Acta Universitatis Wratislaviensis 1651, Handelingen Regionaal Colloquium Neerlandicum. Van Bree, C. 1997. “Die Syntax des Stadtfriesischen im Vergleich mit Niederländisch und Friesisch”. In Friesische Studien 3, V. F. Faltings, A. G. Walker and O. Wilts (eds), 1–48. Odense: Odense University Press. [= NOWELE supplement vol. 18] Van Craenenbroeck, J. 2000. “Complementerend van: een voorbeeld van syntactische variatie in het Nederlands”. Nederlandse Taalkunde 5: 133–163. Van der Burg, A. C. B. 1991. Woardeboek fan ut Leewarders. Leeuwarden: Frisian Academy. Van Ginneken, J. 1935. “Kleine woorden wortelen diep”. Onze Taaltuin 4: 193–211. Van Kampen, J. 1997. First steps in wh-movement. Dissertation, University of Utrecht. Van Kemenade, A. 1987. Syntactic case and morphological case in the history of English. Dordrecht: Foris. Van Kemenade, A. 2000. “Jespersen’s cycle revisited: formal properties of grammaticalization”. In Diachronic Syntax: Models and Mechanisms, S. Pintzuk, G. Tsoulas and A. Warner (eds). Oxford: Oxford University Press. Van Kemenade, A. and Vincent, N. 1997. Parameters of morphosyntactic change. Cambridge: Cambridge University Press.

References 391

Van Riemsdijk, H. 1978. A case study in syntactic markedness: The binding nature of prepositional phrases. Dordrecht: Foris. Van Riemsdijk, H. 1982. “Correspondence effects and the empty category principle”. Tilburg Papers in Language and Literature 12: 1–19. [Also in Studies in Generative Grammar and Language Acquisition, Y. Otsu et al. (eds), 5–16. Tokyo: International Christian University, 1983.] Vanacker, V. F. 1985. ’t Is van + hoofdzin. Naamkunde 17: 347–357. Vance, B. 1989. Null subjects and syntactic change in medieval French. Dissertation, Cornell University. Vance, B. 1997. Syntactic change in medieval French. Verb second and null subjects. Dordrecht: Kluwer. Vanden Wyngaerd, G. and J. W. Zwart. 1991. “Reconstruction and Vehicle Change.” In Linguistics in the Netherlands 1991, F. Drijkoningen and A. van Kemenade (eds), 151–160. Amsterdam: John Benjamins. Vangsnes, Ø.A. 1999. The identification of functional architecture. Dissertation, University of Bergen. Vangsnes, Ø.A. 2001. “Icelandic expletive constructions and the distribution of subject types”. To appear in Subjects, Expletives, and the Extended Projection Principle, P. Svenonius (ed). New York: Oxford University Press. Verkuyl, H. J. 1977. “Het performatieve van”. Spektator 6: 481–484. Verkuyl, H. J. 1979. “Ik meen van niet”. Onze Taal 48, 61–62. Vikner, S. 1991 “Relative der and other C0 elements in Danish”. Lingua 84: 109–136. Vikner, S. 1994. “Finite verb movement in Scandinavian embedded clauses”. In Verb Movement, D. Lightfoot and N. Hornstein (eds), 117–147. Cambridge: Cambridge University Press. Vikner, S. 1995a. Verb movement and expletive subjects in the germanic languages. New York: Oxford University Press. Vikner, S. 1995b. “V0-to-I0 movement and inflection for person in all tenses”. Working Papers in Scandinavian Syntax 55: 1–27. Vogel, R. and Steinbach, M. 1998. “The dative — an oblique case”. Linguistische Berichte 173: 65–90. Von Stechow, A. 1996. “Partial wh-movement and Logical Form: an introduction”. In Lutz and Müller (eds), 1–36. Von Stechow, A. and Sternefeld, W. 1988. Bausteine syntaktischen Wissens. Opladen: Westdeutscher Verlag. Walker, A. 1995. “Frisian.” In The Dialects of Modern German, Ch.V. J. Russ (ed), 1–30. London: Routledge. Webelhuth, G. 1992. Principles and parameters of syntactic saturation. Oxford: Oxford University Press. Weerman, F. 1989. The V2 conspiracy: a synchronic and a diachronic analysis of verbal positions in Germanic languages. Dordrecht: Foris. Williams, E. 1980. “Predication”. Linguistic Inquiry 11, 203–238. Winter, Y. 1997. “Choice-functions and the scopal semantics of indefinites”. Linguistics and Philosophy 20: 399–467.

392 References

Woordenboek der Nederlandsche Taal, XVIII 1958, F. de Tollenaere (ed). ’s-Gravenhage: Martinus Nijhoff. Wurmbrand, S. 1998a. “Downsizing infinitives”. MIT Working Papers in Linguistics 25: 141–175. Wurmbrand, S. 1998b. Infinitives. Dissertation, MIT. Wurmbrand, S. 1998c. “Restructuring control”. UPenn Working Papers in Linguistics 5(1): 175–188. Wurmbrand, S. 1999. “The size of infinitives”. Proceedings of the CLA Annual Conference 1998, 443–454. Wurmbrand, S. 2000. “A-movement in German: AGREE and ATTRACT”. Paper presented at Harvard University, Cambridge, Mass, May. Wurmbrand, S. 2001. Infinitives: restructuring and clause structure. Berlin/New York: Mouton de Gruyter. Zagona, K. 1982. Government and proper government of verbal projections. Dissertation, University of Washington, Seattle. Zamparelli, R. 2000. “Distributive conjunction and sentence reduction.” Ms., University of Bergamo. Zimmermann, I. 1997. “So und wie in satzadverbiellen Phrasen”. Sprache & Pragmatik 44: 36–68. Zwart, C.J.W. 1993. Dutch syntax: A minimalist approach. Dissertation, University of Groningen. Zwart, C. J. W. 1997. Morphosyntax of verb movement: a minimalist analysis of the syntax of Dutch. Dordrecht: Kluwer Academic Publishers. Zwart, C. J. W. 1998. “Nonargument middles in Dutch.” GAGL 42: 109–128. Zwart, C. J. W. 2000. “A head raising analysis of relative clauses in Dutch.” In The syntax of relative clauses, A. Alexiadou, P. Law, A. Meinunger, and Chr. Wilder (eds), 349–385. Amsterdam: John Benjamins. Zwart, C. J. W. 2001a. “Syntactic and phonological verb movement.” Syntax 4: 34–62. Zwart, C. J. W. 2001b. “Object shift with raising verbs”. Linguistic Inquiry 32: 547–554.

Name index A Abney, S. P. 222, 302, 314, 356 Abraham, W. 37, 327, 338, 339 Ackema, P. 176, 320, 339, 340, 342 Aissen, J. 103, 307 Alexiadou, A. ix, 88, 177, 220, 224, 230, 231, 338 Allen, C. 249, 270 Altmann, H. 38 Anagnostopoulou, E. 88, 224, 230 Aoun, J. 224, 235, 304 Ariel, M. 195 Arnold, S. 87 B Bader, T. 26 Baker, M. C. 302, 314 Baptista, M. 232 Barbiers, S. 60, 65 Barnes, M. 241 Barss, A. 366, 371 Bayer, J. 40, 328 Bækken, B. 265 Bean, M. 249 Belletti, A. 184, 229 Benmamoun, E. 235, 304 Bennis, H. 42, 66 Berman, A. 81 Bernstein, J. 223 Bhatt, R. M. 237 Bianchi, V. 86–87, 91 Blake, N. F. 264 Blancquaert, E. 42 Blom, A. 42 Bobaljik, J. D. 88, 124, 178, 185, 198, 200, 226, 227–229, 230, 231, 241, 304 Bolinger, D. D. 22, 38

Borer, H. 100 Boškovic´, Ž. 94, 225 Bouchard, D. 100 Brandner, E. 40 Brandt, M. 38 Bremmer, R. 286 Bresnan, J. 94, 209, 253 Browning, M. 81, 85–86, 87 Büring, D. 305, 370 Burzio, L. 322 Butter, F. 276 Buwalda, H. S. 277, 279 C Cardinaletti, A. 173, 185, 246 Carlson, G. 87 Chafe, W. 47 Cheng, L. 65, 225 Chierchia, G. 94, 101, 102, 109, 114, 116–118, 127, 195, 332 Choe, J.-W. 344, 345 Chomsky, N. 27, 36, 38, 69, 73, 93, 131, 156, 159, 169, 173, 175, 176, 178, 181, 188, 189, 197, 198, 199, 205, 212, 217, 221, 293, 301, 304, 307, 361, 371 Chung, S. 290, 299 Cinque, G. 126, 132, 152, 153, 159, 167, 173, 220, 223, 231, 241 Coene, M. 357 Cohen, A. 334, 341 Comrie, B. 124 Corbett, G. 234 Corver, N. 26, 38, 357, 370 Cremers, C. 65 Culicover, P. 314

394 Name index

D Dabrowska, E. 36 Dayal, V. 6, 33, 34, 36, 40 De Bakker, C. 200 De Haan, G. 312 De Rooij, J. 41 Debrabandere, F. 61 Den Besten, H. 181, 186, 194 Den Dikken, M. x, 303, 312, 313, 356, 357 Di Sciullo, A. M. 125 Dobrovie-Sorin, C. 317, 339 Donhauser, K. 241 Dowty, D. 341 E Egerland, V. 152–154 Emonds, J. E. 177, 184, 220 Erb, M. C. 338 Ernst, T. 136 Erteschik-Shir, N. 69, 77, 78, 91 Espinal, M. T. 22, 38 Evers, A. 178 Eythórsson, T. 249 F Fagan, S. 320, 327, 339, 340, 342 Falk, C. 226, 227, 239 Fanselow, G. ix, 40, 177, 225 Farkas, D. 95, 362 Felser, C. 40 Féry, C. 241 Fischer, S. 241 Fokkema, K. 276 Fox, D. 36, 329 Francis, W. N. 272 Fuss, E. 246 G Gärtner, H.-M. 322, 338 Geniušiene, E. 317 Giannakidou, A. 290, 303, 312 Givón, T. 234 Goossens, J. 61 Greenspon, M. D. 342 Grewendorf, G. 112

Grimshaw, J. 339 Grinder, J. T. 100 Gussenhoven, C. 313 H Haeberli, E. ix, 246, 249, 262, 263, 266, 312 Haegeman, L. 141, 219, 221, 226, 249 Haider, H. 125, 132, 136, 230, 365, 366, 367 Halbertsma, J. 282 Hale, K. 73, 235 Hand, M. 40 Hardt, D. 301 Heim, I. 87, 193, 196, 358, 362, 370, 371 Heine, B. 61 Herburger, E. 31 Hinterhölzl, R. ix, 141 Hoekstra, E. viii, x, 41, 52, 276, 277, 282 Hoekstra, J. 303, 310, 311, 312, 313, 315 Hoekstra, Joh. 286 Hoekstra, T. 42, 66, 320 Höhle, T. N. 39, 40 Holmberg, A. x, 22, 39, 151, 157, 159, 167, 173, 177, 205, 217, 221, 226, 239, 312 Hornstein, N. 99, 100, 138, 146, 300 Horrocks, G. 222 Huang, J. 351 Hulk, A. 246, 263 I Iwata, S. 340 J Jackendoff, R. 366, 371 Jacobsen, B. 69, 70, 78, 80, 83 Jensen, P.-A. 69, 70, 78, 80, 83 Jonas, D. 88, 227, 241 Jonkman, R. 276 Julien, M. 167 K Kaufman, T. 274, 275, 286 Kawasaki, N. 95

Name index 395

Kayne, R. S. viii, x, 41, 51, 63, 66, 86, 88, 91, 131, 133, 147, 148, 152, 154, 157, 161, 164, 167, 169, 171, 300, 351, 356, 357, 366, 371 Kemmer, S. 317, 322, 325 Kennedy, C. 146, 307 Keyser, S. J. 73, 342 Kiparsky, C. 47 Kiparsky, P. 47, 249 Klamer, M. 61, 67 Koeneman, O. ix, 175–178, 185, 199, 200, 201, 226 König, E. 339 Koopman, H. ix, 152, 167, 173, 230, 365 Koopman, W. 249, 250, 251 Kosmeijer, W. 177, 221 Koster, J. ix, x, 100 Kratzer, A. 73, 193, 335, 342, 358, 362 Krifka, M. 334, 341 Krisch, Th. 237 Kroch, A. 238, 246, 247, 249, 252, 259, 264, 266, 270, 272 Kromann, M. 91 Kuno, S. 100 Kytö, M. 270 L Ladusaw, W. 290 Landau, I. 95–97, 99, 100, 102, 124, 125, 127 Larson, R. 134 Lasnik, H. 69, 93, 188, 290, 307, 366, 371 Lebeaux, D. 99, 100 Leopold, J. A. 276 Leopold, L. 276 Levin, L. 290 Li, A. Y.-H. 224 Lightfoot, D. 246, 264 Link, G. 347, 369, 371 Lipták, A. 354, 359 Lobeck, A. 290, 299, 302 Lutz, U. 33, 40 M Mahajan, A. 40

Manzini, R. 99, 100 Martin, R. A. 93, 95 Matthews, G. H. 47 Matthewson, L. 362, 370 May, R. 350 McCloskey, J. 91, 222, 232, 235, 290, 314, 371 McDaniel, D. 40 Meech, S. B. 272 Merchant, J. x, 290, 292, 306, 307, 312–315 Mikkelsen, L. viii, 312 Miller, D. G. 339 Milsark, G. L. 84, 85 Mitchell, B. 249 Möck, J. 338 Moltmann, F. 344, 347, 348 Moore, J. 126 Moro, A. 52 Müller, G. 31, 35, 36, 40 N Naarding, J. 286 Neeleman, A. 143, 185, 199 Nichols, J. 47 Nilsen, Ø. ix Noonan, M. 233 Nuñez, J. 150 O O’Neil, J. H. 100 Olsen, S. 356 Ouhalla, J. 221, 222, 235 P Partee, B. 360 Pée, W. 42 Penner, Z. 26 Perlmutter, D. 103 Perry, G. G. 265, 268 Pesetsky, D. 133, 134, 146, 149, 303 Petter, M. 95 Picallo, C. 125 Pili, D. 145 Pintzuk, S. 246, 247, 249, 250, 266, 269

396 Name index

Pittner, K. 38 Platzack, C. 39, 151, 173, 177, 195, 205, 221, 226, 239, 246 Pollard, C. 331, 340 Pollock, J. Y. 177, 184, 200, 220–222, 371 Postal, P. 88, 300 Postma, G. 65, 67 Potts, C. 91 Pullum, G. 91 R Rapp, I. 153 Reinhart, T. 11, 12, 22, 38, 87, 173, 323, 329, 331, 332, 333, 340, 341, 351, 362, 370 Reis, M. vii, viii, 17, 20–22, 25, 29, 32, 33, 36–40, 313, 338 Reuland, E. 173, 286, 323, 329, 340 Richards, N. 88, 299, 300, 315 Ringe, D. 264 Rivero, M. L. 230 Rizzi, L. 21, 70, 88, 103, 152, 154, 158, 161, 193 Roberts, I. 177, 194, 195, 226, 242, 246, 272, 320 Rochette, A. 94, 126 Roeper, T. 342 Rohrbacher, B. 177, 178, 226 Romero, M. 290, 306 Rooryck, J. 41, 51–53, 65 Rosen, C. 298, 299, 305 Rosen, S. T. 126 Ross, J. R. 19, 22, 290, 293, 298, 299 Roussou, A. 100, 194, 195 S Sabel, J. 126 Safir, K. 343, 369 Sag, I. 331, 340 Sakaguchi, M. 344, 346, 369 Sauerland, U. 304 Schmidt, D. A. 253 Schmitt, L. E. 286 Schönfeld 38 Schoorlemmer, M. 320, 339, 340, 342

Schwarzschild, R. 305, 306 Siemund, P. 339 Sigurðsson, H. Á. 195, 215, 224 Slobin, N. 309 Snyder, W. 226 Sportiche, D. 73, 235, 365 Starke, M. 173 Stavrou, M. 222 Steinbach, M. x, 12, 21, 38, 327, 328, 332, 338–340, 342 Sternefeld, W. 29, 33, 36, 40 Stockwell, R. 245, 246, 253 Stowell, T. 81, 178, 209, 314, 343, 369 Stroik, T. 134, 320 Strozer, J. R. 125 Sturm, A. 194, 201 Svenonius, P. 173, 217, 312 Szabolcsi, A. ix, 152, 167, 222, 357 T Takahashi, D. 225 Takami, K. 293, 300 Taraldsen, K. T. 78, 89, 91, 312 Tasmowski, L. 65 Taylor, A. 246, 249, 252, 259, 264, 266, 270, 272 Thiersch, C. 23, 27, 222 Thomason, S. S. 274, 275, 286 Thornton, R. 22 Thráinsson, H. vii, 185, 206–208, 227–229, 241 Tiersma, P. M. 277 Travis, L. 220 V Van Bree, C. 274, 276, 278, 286 Van Coetsem, F. 274, 275, 286 Van Craenenbroeck, J. viii, 65 Van der Burg, A. C. B. 279 Van der Meulen, I. 65 Van Ginneken, J. 65 Van Kampen, J. 22 Van Kemenade, A. 67, 245–247, 249, 252, 259, 263, 269 Van Koppen, M. 65

Name index 397

Van Riemsdijk, H. 40, 290, 299, 314 Vanacker, V. F. 57 Vance, B. 200, 272 Vanden Wyngaerd, G. 65, 146 Vangsnes, Ø. A. ix, 211, 215, 217, 312 Veenstra, T. 241 Verkuyl, H. J. 66 Vikner, S. viii, 69–71, 75, 77–79, 82, 83, 84, 88, 89, 91, 131, 175, 184, 199, 200, 219, 222, 226, 230, 231, 232, 241, 242 Vogel, R. 328, 338 Von Stechow, A. 29, 33, 40, 153 Vries, O. 312

W Webelhuth, G. 307 Weerman, F. 194, 264 Weinberg, A. 300 Williams, E. 94, 98, 99, 125, 199 Winter, Y. 362, 370 Wurmbrand, S. viii, 95, 97, 105, 125, 126 Z Zagona, K. 125 Zamparelli, R. 160, 161 Zimmermann, I. 38 Zimmermann, M. x, 354, 359 Zwart, C. J. W. 37, 39, 41, 52, 53, 126, 131, 146, 175, 193, 194, 199, 230, 304, 338, 342

Subject index A A-chain 81, 329f General condition on A-chains (GCC) 329 accessibility 181, 199 acquisition, second language 285 adjective 223 adjunct 97, 138f, 233 distribution 139 fronting 257 island effect 354 Larsonian analysis of 134 PP adjunct 257, 305 adverb 132ff, 152, 340 degree modification 159 event related 132ff fronting 159, 257 marking verb phrase boundary 137, 177, 184, 189, 204, 207, 219, 224 in middle construction 340, 342 order 132, 136 position in English 134, 241 in Icelandic 214, 239 of quantification 193, 342 violating verb second 151ff aggressive non-D-linking 303 Agree 197, 199 agreement 106, 177, 215, 274, 328, 361, 371 agreement feature 211 AGR-parameter 205 person and number subject-verb agreement 203f, 235 rich 177f, 184, 205 correlation with verb movement 177, 203f, 219ff strong 205

analogy 36 Antecedent Contained Deletion (ACD) 145f ACD-resolution 145 anticausative 317f, 333 argument licensing, non-canonical 19 reduction 332 shift see (i) movement > weak pronoun (ii) object > shift saturation 332 syntactic 318 aspect 230, 339 auxiliary 103, 154, 167, 189, 200, 217, 227, 247 B binding 38, 39, 109, 127, 133, 322, 331, 340 theory 93 binominal (element) 343ff compositional interpretation of 360 each 343ff construction 343ff extraction analysis of 348f in-situ analysis of 354f jeweils 343ff fronting 356 blending 37 C Case 212, 225 in A¢-movement 88 morphological case 212 structural case 104, 328, 340 checking 205f, 275 clausal architecture 71, 134, 161, 207, 227 infinitives 105, 123, 207f

400 Subject index

left periphery 70, 152, 160 second position 151ff split CP see functional > projection > CP split IP see functional > projection > IP clause mate condition 345, 351, 361 clause union 103 cleft 298 pseudocleft 298 clitic 154, 234 subject pronoun 234 command 199 c-command 134, 178 COMMAND 178 m-command 178 comparison 51f complementizer 70, 78, 170, 206 complementizer trace effect 78f, 82 infinitival 42, 67 order 79 prepositional 41ff, 51, 66 say 61, 67 van 41ff conjunction, distributive 160 control 93ff, 203ff arbitrary 95 control infinitival 93, 203f control shift 95 exhaustive 95 imperfect 95 implicit 95, 125 non-obligatory 94f, 102f, 119f, 124 obligatory 94f, 101f, 116, 119f, 124, 125 partial 95, 98, 100, 102, 124 semantic 99f, 115 shift see variable control split 95 syntactic 99f variable control 95, 97, 100 control verb 95, 97, 100, 210 D definiteness effect 84f

deixis 212, 215 determiner 36, 344 dissociation 41ff, 54, 59, 63, 64 E each adverbial each 346 binominal each see binominal economy 203 ellipsis 290 equidistance see locality event 178, 210 event interpretation 208 event related adverb 134 evidentiality 47, 54 expletive 52, 70, 72, 83, 201 expletive construction 85f transitive (TEC) 88f, 185, 187, 229 empty expletive 185 loss of empty expletives 263 subject 69ff, 326 wh 29 Extended Projection Principle (EPP) 71f, 80f, 224 Extension Condition 159, 197 extraction construction see movement > A¢ movement > out of embedded clauses extraposition 269 F factive 9, 13, 29, 47, 58 feature 141, 212 agreement see agreement checking 140, 175, 198, 205 lexical 212 phi-features see person, number, gender strong vs. weak 175, 205, 221 focus 225, 305, 334 focus particle 151 Focus Phrase see functional > projections > FocP focus-scrambling see scrambling fronting 193, 225, 245

Subject index 401

adverb see adverb argument vs. adjunct 251 of non-operator 246, 252 operator 246 predicate 253 stylistic 239 VP see verb phrase functional categories 203 categorial vs. extensional 211 semantics of 208f head 175, 198, 208 AGRS, yielding event interpretation 208 D 355 projection 198, 208 AGROP 145 AGRP 178 CP 70 split 21, 154 DP 311 DP hypothesis 222 FinP 152, 157f, 171 FocP 152, 158, 171 ForceP 158, 161 IP 71, 371 split 227 TopP 152, 157f, 171

impersonal construction 69, 83f, 239, 338, 341 infinitive distribution (Frisian) 283 inflection 177, 188, 277 inflectional morphology 2, 232 inflectional paradigm 205, 220, 226, 229f infinitival 203, 263, 282 irregular verbs 278f loss of 238 rich 200, 203, 226, 232 Split Infl 222, 227f integrated parenthetical construction see parenthetical intraposition 137 inverse linking 350 inversion locative 253 narrative 21, 193f, 201 predicate inversion 346 quotative inversion 38, 67, 270 subject-verb 152, 158, 199, 245ff absence of 249f, 260f, 270f decline of 252 loss of 246ff with R-pronouns 299 it-anaphor 116f

G gender 223, 235, 323 gerund 126, 209 Government and Binding 69, 93 grammaticalization 41, 61f, 163

J je-verbs, conjugation of 277f

H head 176, 297 Holmberg’s generalization 151ff, 157, 167 I identification 211f generalized strategy 215 Identification Principle 211 preferred identifier 213 imperativus-pro-infinitivo (IPI) 281

L language change 32, 237, 246ff language contact 61, 262, 264, 273 contact-based dialect 273ff Lexical Functional Grammar 94 Linear Correspondence Axiom (LCA) 366 locality 99, 106, 141, 156, 181, 322 domain extension 169 equidistance 169, 173 islands 27, 30, 172 Logical Form (LF) 138 LF-movement see movement

402 Subject index

LF-structure 352f M meaning postulate 117f middle 317f, 334 marker 322, 325f, 335 Minimal Link Condition 156 minimality condition see swiping relativized 70 morphology see agreement, see inflection movement 275 A-movement 80, 84, 104, 108, 125, 138, 143, 345 A¢-movement 3ff, 66, 70, 78, 84, 88, 143, 225, 312, 345 out of embedded clauses 3ff, 15, 78, 300 out of V2 clauses (EV2) 4, 23ff, 39 covert see LF-movement DP-movement 224 feature movement 307f head movement 151f, 304 at PF 289, 301f leftward 152 LF-movement 138, 152, 205, 229, 349 and morphology 131f, 178, 203f, 219f N-movement 222 operator movement 72 remnant 141, 152 roll up movement 136 trigger for 177 verb movement 131, 151f, 175 to C 152, 160, 177, 179, 192, 237, 246 to I 88, 177, 184, 203ff, 219ff, 230, 233 of weak pronouns 155 wh-movement 152, 159, 290f, 310 partial 4, 27ff direct vs. indirect dependency approach to 29, 31, 33, 34 N negation 155, 200, 214

negative polarity 134, 138 noun phrase 222, 235 DP hypothesis 222 number 73, 75, 203, 205, 211, 323 O object 318 indefinite 162 shift 131, 151, 228f operator 91 invariant 70, 73f null 81, 194, 201 wh 75f, 193, 297 optimality theory (OT) 35 order 79 OV/VO 131, 139, 186f output condition 176, 178, 200 P parametric variation 131, 140f parasitic gaps 135 parenthetical constructions integrated parenthetical (IP) 3ff, 33, 37, 55 vs. extraction construction 15ff, 27f structure 19f V1 parenthetical constructions (VIP) 11f, 19f, 23 was-IP 7f, 22, 37 wie/so-IP 11, 38 unintegrated parenthetical construction 5f, 12 particle 239 passive 103, 126, 320f, 334 long 103, 124, 126 percolation 307 person 203, 205, 211, 226, 323 first person 17, 25, 226, 322 second person singular 226, 279, 322 Pesetsky’s paradox 133 PF operation 304 phrase structural object 211 pied piping 71, 73, 75, 163, 293, 307f precedence, linear 366f

Subject index 403

predicate collective 98, 121 negated 10, 200 used in integrated parentheticals 9f, 13, 24, 29f, 40 preposition 289, 313 phrase (PP) 49 stranding 75, 290f, 308 PRO 93ff, 110, 206, 215, 217, 349 PRO-theorem 93 Projection Principle 93 pronoun 323 bound 134, 138, 148 R-pronoun 313 reflexive pronoun 317ff, 342 weak pronoun 151, 155f, 165 wh-pronoun 302, 310 prosody 240, 304 Q quantification existential 334 generic 334, 341 quantifier 332 binary 349 quantifier phrase (QP) 349 quantifier raising (QR) 145f quotative van 54f inversion see inversion R Range DP 344, 369 reanalysis 237, 364 reference set 36 reflexive 113, 317f, 325, 339 relative clause 70ff, 146 operator 71, 86 see also operator raising analysis 86 restructuring 103f, 113 long passive property of 103, 124 restructuring/non-restructuring distinction 106, 123f, 125 root/non-root asymmetry 181

S scope 18, 25, 132, 136, 145 inverse 148, 150 narrow 345, 362 scrambling 107, 133, 141ff, 225 extension property of 142 focus-scrambling 142 permutation property of 142 silent 137, 142f Share NP 344, 369 Shortest Move 156 slifting 19, 38 sluicing 289f sluicing condition 298 multiple sluicing 315 Small Clause 355, 363 som see operator > invariant Spell Out 140f, 304f stray affix filter 188 strict/sloppy interpretation 99, 116f subject 318 agreement see agreement expletive see expletive full DP subject 247 in infinitives 101f, 113, 115 mass denoting indefinite subject 214 non-pronominal 260 pronominal 234, 246, 260 subject condition 347, 350, 363 subject position 72f, 90, 224, 228, 239, 263, 365 VP-internal see verb phrase subject pronoun clitic 234 subject-verb inversion see inversion subordination 36, 39, 207 subordinator 158 superiority 142 swiping 289ff minimality condition on 297 syntax/semantics mismatch 102, 114f, 122 T tense 97, 177, 190, 205, 207, 211 tense condition 178

404 Subject index

temporal anchoring through 208f ternary structure 365 topic 269 drop 21 Topic Phrase see functional > projections > TopP topicalization 140, 152, 159, 225 transitivity 84, 87f, 110 transparency properties 107 type-driven interpretation 354 U underspecification 322 universal base hypothesis 131 V valency reduction 317f verb verb clustering 282 control verb see control epistemic verb 42f, 51 verb final 249, 286 verb first 193f, 201 see also parenthetical modal verb 189, 206, 210 verb movement see movement participial verb 167f, 217, 224 biclausal structure 169 verb phrase (VP) ellipsis 298

flat vs. hierarchic structure 365, 371 fronting 134 intraposition 137 layered (Larsonian) shell structure 133f remnant 157f, 367 VP internal subject hypothesis 72f, 365 raising verb 42f, 51 verb second (V2) 151ff, 222, 237, 245f, 269 effect 237 embedded 170, 180, 203f, 341 loss of see inversion > subject-verb residual 247 root 158 as XP-movement 151ff W Weak Crossover (WCO) 138, 142, 147, 352 whwh-construction was-w-construction 27f, 34 w-copy construction 34, 40 w-daß-construction 15, 29, 34 wh-elements 296 wh-expletive see expletive wh-movement see movement wh-operator see operator

In the series LINGUISTIK AKTUELL/LINGUISTICS TODAY (LA) the following titles have been published thus far, or are scheduled for publication: 1. KLAPPENBACH, Ruth (1911-1977): Studien zur Modernen Deutschen Lexikographie. Auswahl aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene Malige-Klappenbach. 1980. 2. EHLICH, Konrad & Jochen REHBEIN: Augenkommunikation. Methodenreflexion und Beispielanalyse. 1982. 3. ABRAHAM, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983. 4. ABRAHAM, Werner & Sjaak De MEIJ (eds): Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks, Groningen, 1984. 1986. 5. GREWENDORF, Günther and Wolfgang STERNEFELD (eds): Scrambling and Barriers. 1990. 6. BHATT, Christa, Elisabeth LÖBEL and Claudia SCHMIDT (eds): Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences. 1989. 7. ÅFARLI, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. 8. FANSELOW, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. 9. GELDEREN, Elly van: The Rise of Functional Categories. 1993. 10. CINQUE, Guglielmo and Guiliana GIUSTI (eds): Advances in Roumanian Linguistics. 1995. 11. LUTZ, Uli and Jürgen PAFEL (eds): On Extraction and Extraposition in German. 1995. 12. ABRAHAM, W., S. EPSTEIN, H. THRÁINSSON and C.J.W. ZWART (eds): Minimal Ideas. Linguistic studies in the minimalist framework. 1996. 13. ALEXIADOU Artemis and T. Alan HALL (eds): Studies on Universal Grammar and Typological Variation. 1997. 14. ANAGNOSTOPOULOU, Elena, Henk VAN RIEMSDIJK and Frans ZWARTS (eds): Materials on Left Dislocation. 1997. 15. ROHRBACHER, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and pro-drop. 1999. 16. LIU, FENG-HSI: Scope and Specificity. 1997. 17. BEERMAN, Dorothee, David LEBLANC and Henk van RIEMSDIJK (eds): Rightward Movement. 1997. 18. ALEXIADOU, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. 19. JOSEFSSON, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. 20. LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. 21. KLEIN, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. 22. ALEXIADOU, Artemis and Chris WILDER (eds): Possessors, Predicates and Movement in the Determiner Phrase. 1998. 23. GIANNAKIDOU, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. 24. REBUSCHI, Georges and Laurice TULLER (eds): The Grammar of Focus. 1999. 25. FELSER, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999.

26. ACKEMA, Peter: Issues in Morphosyntax. 1999. ° 27. RUZICKA, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. 28. HERMANS, Ben and Marc van OOSTENDORP (eds): The Derivational Residue in Phonological Optimality Theory. 1999. 29. MIYAMOTO, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 1999. 30. BEUKEMA, Frits and Marcel den DIKKEN (eds): Clitic Phenomena in European Languages. 2000. 31. SVENONIUS, Peter (ed.): The Derivation of VO and OV. 2000. 32. ALEXIADOU, Artemis, Paul LAW, André MEINUNGER and Chris WILDER (eds): The Syntax of Relative Clauses. 2000. 33. PUSKÁS, Genoveva: Word Order in Hungarian. The syntax of È-positions. 2000. 34. REULAND, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. 35. HRÓARSDÓTTIR, Thorbjörg. Word Order Change in Icelandic. From OV to VO. 2000. 36. GERLACH, Birgit and Janet GRIJZENHOUT (eds): Clitics in Phonology, Morphology and Syntax. 2000. 37. LUTZ, Uli, Gereon MÜLLER and Arnim von STECHOW (eds): Wh-Scope Marking. 2000. 38. MEINUNGER, André: Syntactic Aspects of Topic and Comment. 2000. 39. GELDEREN, Elly van: A History of English Reflexive Pronouns. Person, ‘‘Self’’, and Interpretability. 2000. 40. HOEKSEMA, Jack, Hotze RULLMANN, Victor SANCHEZ-VALENCIA and Ton van der WOUDEN (eds): Perspectives on Negation and Polarity Items. 2001. 41. ZELLER, Jochen : Particle Verbs and Local Domains. 2001. 42. ALEXIADOU, Artemis : Functional Structure in Nominals. Nominalization and ergativity. 2001. 43. FEATHERSTON, Sam: Empty Categories in Sentence Processing. 2001. 44. TAYLAN, Eser E. (ed.): The Verb in Turkish. 2002. 45. ABRAHAM, Werner and C. Jan-Wouter ZWART (eds): Issues in Formal German(ic) Typology. 2002. 46. PANAGIOTIDIS, Phoevos: Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. 2002. 47. BARBIERS, Sjef, Frits BEUKEMA and Wim van der WURFF (eds): Modality and its Interaction with the Verbal System. 2002. 48. ALEXIADOU, Artemis, Elena ANAGNOSTOPOULOU, Sjef BARBIERS and HansMartin GAERTNER (eds): Dimensions of Movement. From features to remnants. 2002 49. ALEXIADOU, Artemis (ed.): Theoretical Approaches to Universals. 2002. 50. STEINBACH, Markus: Middle Voice. A comparative study in the syntax-semantics interface of German. 2002. 51. GERLACH, Birgit: Clitics between Syntax and Lexicon. 2002. 52. SIMON, Horst J. and Heike WIESE (eds): Pronouns – Grammar and Representation. 2002.

53. ZWART, C. Jan-Wouter and Werner ABRAHAM (eds): Studies in Comparative Germanic Syntax. Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26-27, 2000)(Workshop). 2002. 54. BAPTISTA, Marlyse: The Syntax of Cape Verdean Creole. The Sotavento varieties. 2002. 55. COENE, M. and Yves D'HULST (eds): From NP to DP. Volume 1: The syntax and semantics of noun phrases. n.y.p. 56. COENE, M. and Yves D'HULST (eds.): From NP to DP. Volume 2: The expression of possession in noun phrases. n.y.p. 57. DI SCIULLO, Anna-Maria (ed.): Asymmetry in Grammar. Volume 1: Syntax and semantics. n.y.p. 58. DI SCIULLO, Anna-Maria (ed.): Asymmetry in Grammar. Volume 2: Morphology, phonology, acquisition. n.y.p. 59. DEHÉ, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002. 60. TRIPS, Carola: From OV to VO in Early Middle English. 2002. 61. SCHWABE, Kerstin and Susanne WINKLER (eds.): The Interfaces. Deriving and interpreting omitted structures. n.y.p. 62. CARNIE, Andrew, Heidi HARLEY and Mary WILLIE (eds.): Formal Approaches to Function in Grammar. In honor of Eloise Jelinek. n.y.p.

Studies in Comparative Germanic Syntax: Proceedings from the 15th Workshop on Comparative Germanic Syntax

Portuguese Syntax: New Comparative Studies (Oxford Studies in Comparative Syntax)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

Comparative Studies in Romanian Syntax

The Syntax of Anaphora (Oxford Studies in Comparative Syntax)

Verb Movement and Expletive Subjects in the Germanic Languages (Oxford Studies in Comparative Syntax)

The Syntax of Anaphora (Oxford Studies in Comparative Syntax)

The Syntax of Anaphora (Oxford Studies in Comparative Syntax)

Scrambling, Remnant Movement, and Restructuring in West Germanic (Oxford Studies in Comparative Syntax)

Parameters and Functional Heads: Essays in Comparative Syntax (Oxford Studies in Comparative Syntax)

Particles: On the Syntax of Verb-Particle, Triadic and Causative Constructions (Oxford Studies in Comparative Syntax)

The Syntax of Ellipsis: Evidence from Dutch Dialects (Oxford Studies in Comparative Syntax)

Particles: On the Syntax of Verb-Particle, Triadic and Causative Constructions (Oxford Studies in Comparative Syntax)

Germanic

Subjects, Expletives, and the EPP (Oxford Studies in Comparative Syntax)

Comparative Syntax of the Balkan Languages

Clause Structure and Word Order in Hebrew and Arabic: An Essay in Comparative Semitic Syntax (Oxford Studies in Comparative Syntax)

Theoretical Comparative Syntax: Studies in Macroparameters (Routledge Leading Linguists)

Clause Structure and Word Order in Hebrew and Arabic: An Essay in Comparative Semitic Syntax (Oxford Studies in Comparative Syntax)

Parameters and Universals (Oxford Studies in Comparative Syntax)

Parameters of Slavic Morphosyntax (Oxford Studies in Comparative Syntax)

Movement and Silence (Oxford Studies in Comparative Syntax)

Comparisons and Contrasts (Oxford Studies in Comparative Syntax)

Negation and Clausal Structure: A Comparative Study of Romance Languages (Oxford Studies in Comparative Syntax)

The Germanic Languages

The Higher Functional Field: Evidence from Northern Italian Dialects (Oxford Studies in Comparative Syntax)

The Syntax of the Celtic Languages: A Comparative Perspective

The Grammar of Identity: Intensifiers and Reflexives in Germanic Languages (Routledge Studies in Germanic Linguistics)

The syntax and semantics of split constructions: a comparative study

The Oxford Handbook of Comparative Syntax (Oxford Handbooks)

Jacob Wackernagel, Lectures on Syntax: With Special Reference to Greek, Latin, and Germanic

Studies in Comparative Germanic Syntax: Proceedings from the 15th Workshop on Comparative Germanic Syntax

Portuguese Syntax: New Comparative Studies (Oxford Studies in Comparative Syntax)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

Comparative Studies in Romanian Syntax

The Syntax of Anaphora (Oxford Studies in Comparative Syntax)

Verb Movement and Expletive Subjects in the Germanic Languages (Oxford Studies in Comparative Syntax)

The Syntax of Anaphora (Oxford Studies in Comparative Syntax)

The Syntax of Anaphora (Oxford Studies in Comparative Syntax)

Scrambling, Remnant Movement, and Restructuring in West Germanic (Oxford Studies in Comparative Syntax)

Parameters and Functional Heads: Essays in Comparative Syntax (Oxford Studies in Comparative Syntax)

Particles: On the Syntax of Verb-Particle, Triadic and Causative Constructions (Oxford Studies in Comparative Syntax)

The Syntax of Ellipsis: Evidence from Dutch Dialects (Oxford Studies in Comparative Syntax)

Particles: On the Syntax of Verb-Particle, Triadic and Causative Constructions (Oxford Studies in Comparative Syntax)

Germanic

Subjects, Expletives, and the EPP (Oxford Studies in Comparative Syntax)

Comparative Syntax of the Balkan Languages

Clause Structure and Word Order in Hebrew and Arabic: An Essay in Comparative Semitic Syntax (Oxford Studies in Comparative Syntax)

Theoretical Comparative Syntax: Studies in Macroparameters (Routledge Leading Linguists)

Clause Structure and Word Order in Hebrew and Arabic: An Essay in Comparative Semitic Syntax (Oxford Studies in Comparative Syntax)

Parameters and Universals (Oxford Studies in Comparative Syntax)

Parameters of Slavic Morphosyntax (Oxford Studies in Comparative Syntax)

Movement and Silence (Oxford Studies in Comparative Syntax)

Comparisons and Contrasts (Oxford Studies in Comparative Syntax)

Negation and Clausal Structure: A Comparative Study of Romance Languages (Oxford Studies in Comparative Syntax)

The Germanic Languages

The Higher Functional Field: Evidence from Northern Italian Dialects (Oxford Studies in Comparative Syntax)

The Syntax of the Celtic Languages: A Comparative Perspective

The Grammar of Identity: Intensifiers and Reflexives in Germanic Languages (Routledge Studies in Germanic Linguistics)

The syntax and semantics of split constructions: a comparative study

The Oxford Handbook of Comparative Syntax (Oxford Handbooks)

Jacob Wackernagel, Lectures on Syntax: With Special Reference to Greek, Latin, and Germanic

Recommend Documents