Eurotyp: 5 Clitics in the Languages of Europe [Reprint 2010 ed.] 9783110804010, 9783110157512

"..., this is a valuable book, ... Van Riemsdijk and the entire EUROTYP working group are to be congratulated on a

253 56 47MB

English Pages 1038 [1048] Year 1999

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Contributors
Abbreviations
Clitics: A state-of-the-art report
Part I Area studies
Pronouns in Germanic and Romance languages: An overview
Clitics in the Slavic languages
Pronouns in Scandinavian languages: An overview
Part II Theory
A. Feature articles
The typology of structural deficiency: A case study of the three classes of pronouns
Peer comments by
Josef Bayer
Molly Diesing
Carmen Dobrovie-Sorin
Liliane Haegeman
Anders Holmberg
Juan Uriagereka
Reply to comments by Anna Cardinaletti and Michal Starke
How clitics license null phrases: A theory of the lexical interface
Peer comments by
Mark C Baker
Marcel den Dikken
Frank Drijkoningen
Hubert Haider
Marco Haverkort
Tarald Taraldsen
Reply by Joseph Emonds
1. Clitic clusters and the Wackernagel position
“Clitic third” in Croatian
Clitics and Bulgarian clause structure
Clitics across categories: The case of Romanian
2. Functional categories and the position of clitics
Italian/Romance clitics: Structure and derivation
The internal structure of AgrS and subject clitics in the Northern Italian dialects
Agreement marking in Welsh and Romance
Clitics, subjects and tense in European Portuguese
Pronominal clitic dependencies
3. Clitics and scrambling
On the nature and distribution of Zurich German pronominal clitics
On the semantics and syntax of Scandinavian pronouns and object shift
4. Semantic features
Conditions on clitic doubling in Greek
On the nature of pronoun movement
5. Phonological aspects
The phonology of clitic groups
Appendix
Clitic Questionnaire
Index of languages
Index of names
Index of subjects
Recommend Papers

Eurotyp: 5 Clitics in the Languages of Europe [Reprint 2010 ed.]
 9783110804010, 9783110157512

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Clitics in the Languages of Europe

m 1749

I

1999

I

Empirical Approaches to Language Typology EUROTYP

Editors Georg Bossong Bernard Comrie

Mouton de Gruyter Berlin · New York

20-5

Clitics in the Languages of Europe

edited by Henk van Riemsdijk

w G DE

Mouton de Gruyter Berlin · New York

1999

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

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

Library of Congress

Cataloging-in-Publication-Data

Clitics in the languages of Europe / edited by Henk van Riemsdijk. p. cm. — (Empirical approaches to language typology ; 20-5) "The present volume is one of a series of nine volumes in which the results of the European research project "Typology of Languages in Europe" (EUROTYP) are published" — General pref. Includes bibliographical references (p.) and index. ISBN 3-11-015751-9 1. Europe — Languages — Clitics. I. Riemsdijk, Henk C. van. II. Typology of Languages in Europe (Project) III. Series. P380.C5 1999 415—dc21 99-31443 CIP

Die Deutsche Bibliothek —

Cataloging-in-Publication-Data

Clitics in the languages of Europe / ed. by Henk van Riemsdijk. — Berlin ; New York : Mouton de Gruyter, 1999 (Empirical approaches to language typology ; 20 : EUROTYP ; 5) ISBN 3-11-015751-9

© Copyright 1999 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin. All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Typesetting and printing: Arthur Collignon GmbH, Berlin. Binding: Lüderitz &C Bauer, Berlin. Printed in Germany.

General preface

The present volume is one of a series of nine volumes in which the results of the European research project "Typology of Languages in Europe" (EUROTYP) are published. The initiative for a European project on language typology came from a proposal jointly submitted to the European Science Foundation (ESF) by Johannes Bechert (University of Bremen), Claude Buridant (University of Strasbourg), Martin Harris (University of Salford, now University of Manchester) and Paolo Ramat (University of Pavia). On the basis of this proposal and following consultations with six experts the Standing Committee for the Humanities of the ESF decided to organize a workshop (Rome, January 1988), in which this idea was further explored and developed. The results of this workshop (published by Mouton, 1990) were sufficiently encouraging for the Standing Committee to appoint a preparatory committee and entrust it with the tasks of drawing up a preliminary proposal, of securing interest and participation from a sufficiently large number of scholars and of finding a suitable programme director. The project proposal formulated and sent out by Simon Dik (University of Amsterdam) as chair of this committee met with very supportive and enthusiastic reactions, so that the Standing Committee for the Humanities recommended the funding of a planning stage and the General Assembly of the ESF approved a year zero (1989) for an ESF Programme in Language Typology. During this planning phase all major decisions concerning the management structure and the organisation of the work were taken, i.e., the selection of a programme director, the selection of nine focal areas around which the research was to be organized, the selection of a theme coordinator for each theme and the selection of the advisory committee. The first task of the programme director was to draw up a definitive project proposal, which was supplemented with individual proposals for each theme formulated by the theme coordinators, and this new proposal became the basis of a decision by the ESF to fund the Programme for a period of five years (1990-1994). Language typology is the study of regularities, patterns and limits in cross-linguistic variation. The major goal of EUROTYP was to study the patterns and limits of variation in nine focal areas: pragmatic organiza-

vi

General preface

tion of discourse, constituent order, subordination and complementation, adverbial constructions, tense and aspect, noun phrase structure, clitics and word prosodic systems in the languages of Europe. The decision to restrict the investigation to the languages of Europe was imposed for purely practical and pragmatic reasons. In the course of the project an attempt was made, however, to make as much sense of this restriction as possible, by characterizing the specific features of European languages against the background of non-European languages and by identifying areal phenomena (Sprachbünde) within Europe. More specifically, the goals of the EUROTYP project included the following: - to contribute to the analysis of the nine domains singled out as focal areas, to assess patterns and limits of cross-linguistic variation and to offer explanations of the patterns observed. - to bring linguists from various European countries and from different schools or traditions of linguistics together within a major international project on language typology and in doing so create a new basis for future cooperative ventures within the field of linguistics. More than 100 linguists from more than 20 European countries and the United States participated in the project. - to promote the field of language typology inside and outside of Europe. More specifically, an attempt was made to subject to typological analysis a large number of new aspects and domains of language which were uncharted territory before. - to provide new insights into the specific properties of European languages and thus contribute to the characterization of Europe as a linguistic area (Sprachbund). - to make a contribution to the methodology and the theoretical foundations of typology by developing new forms of cooperation and by assessing the role of inductive generalization and the role of theory construction in language typology. We had a further, more ambitious goal, namely to make a contribution to lingustic theory by uncovering major patterns of variation across an important subset of languages, by providing a large testing ground for theoretical controversies and by further developing certain theories in connection with a variety of languages. The results of our work are documented in the nine final volumes: Pragmatic Organization of Discourse in the Languages of Europe (edited by G. Bernini) Constituent Order in the Languages of Europe (edited by A. Siewierska) Subordination and Complementation in the Languages of Europe (edited by N. Vincent)

General preface

vii

Actance et Valence dans les langues d l'Europe (edited by J. Feuillet) Adverbial Constructions in the Languages of Europe (edited by J. van der Auwera) Tense and Aspect in the Languages of Europe (edited by Ö. Dahl) Noun Phrase Structure in the Languages of Europe (edited by F. Plank) Clitics in the Languages of Europe (edited by H. van Riemsdijk) Word Prosodic Systems in the Languages of Europe (edited by H. van der Hulst) In addition, the EUROTYP Project led to a large number of related activities and publications, too numerous to be listed here. At the end of this preface, I would like to express my profound appreciation to all organizations and individuals who made this project possible. First and foremost, I must mention the European Science Foundation, who funded and supported the Programme. More specifically, I would like to express my appreciation to Christoph Mühlberg, Max Sparreboom and Genevieve Schauinger for their constant and efficient support, without which we would not have been able to concentrate on our work. I would, furthermore, like to thank my colleague and assistant, Martin Haspelmath, and indeed all the participants in the Programme for their dedication and hard work. I finally acknowledge with gratitude the crucial role played by Johannes Bechert and Simon Dik in getting this project off the ground. Their illness and untimely deaths deprived us all of two of the project's major instigators. Berlin, September 1995

Ekkehard König, Programme Director

Preface When I was first approached with the suggestion that I coordinate one of the groups in the EUROTYP project, I was surprised on two counts. First, I had approached the European Science Foundation myself in 1983 with the suggestion to start a project on the typology of European languages by the name of Eurogrammar. This proposal was rejected out of hand at the time. But now, not even six years later, I was gratified to note that the time was ripe for such a project. The second surprise was that the group was to be called 'The Principles and Parameters Approach to Typology.' Thereby this group would have been out of tune with respect to all the others which bore names referring to specific phenomena or subareas of grammar such as 'Subordination' or 'The Structure of Noun Phrases.' And while it struck me as perfectly reasonable that the principles and parameters approach should be well represented in the EUROTYP project, that fact did not seem to justify devoting a group to it. Most of the other groups were also going to be characterized by a specific, ideologically rather coherent approach to linguistic typology. These considerations led to the suggestion that this group be devoted to the study of the typology of clitics in the European languages, a topic which had received considerable attention within generative grammar but which so far had failed to attain a theoretically stable status comparable to, say, NP-movement or Wh-movement. This suggestion was accepted and the clitics group was born. The rules of the game within the ESF-culture were that the group had to be limited to 10 participants. This seemed too small a number because even within the narrow field referred to as the principles and parameters approach there were many more colleagues with interesting ideas on the topic of clitics. We therefore decided to introduce a distinction between 'permanent or core members' and 'variable members.' We started with five core members, Guglielmo Cinque, Lars Hellan, Christer Platzack, Luigi Rizzi and myself. In the course of the group's activities, Lars Hellan brought in Mila Dimitrova-Vulchanova as his co-worker, and Luigi Rizzi was to be represented more and more by Anna Cardinaletti, later also seconded by Michal Starke. Furthermore, when Joseph Emonds assumed his position at the University of Durham, he joined the core group, as did Lida Veselovskä. The variable positions were put to good use, however. Many other colleagues participated in the project on a less permanent basis, and many of them are represented by their contribution in the present volume. In the course of the five official years we had for the project, the clitics group met ten times: in Tilburg (February 1990), Geneva (November 1990), II Ciocco, Tuscany (May 1991, in connection with the plenary meeting of EUROTYP held there), Vienna (October 1991), Lund (May 1992), San Sebastian/Donostia (September 1992, in connection with the second plenary meeting), Trondheim (June 1993), Durham (October 1993), Strasbourg (May 1994, jointly with the plenary meeting there), and in Venice (November 1994). In May 1995, after the

χ

Preface

formal end of the project, the group met yet one more time, this time in Milan. My fellow core members, who took it upon themselves to host and organize a meeting at their home universities are to be thanked for the wonderful workshops we enjoyed there. Special thanks go to Martin Prinzhorn and Andrea Moro for hosting the Vienna and Milan meetings respectively. A particular word of thanks is in order to the original core members, and most particularly to Lars Hellan, for ensuring the smooth and productive continuation of the group's work when illness prevented me from attending the San Sebastian meeting and, more generally, from participating in the group's activities in the fall of 1992. In the course of our work, especially in the early years when putting the final volume together still seemed far away, we presented preliminary results in a series of working papers. These were very helpful in intensifying the discussion about the status of clitics within linguistic theory. Therefore, they deserve to be mentioned separately here: Vol. VIII, 1 (edited by Henk van Riemsdijk and Luigi Rizzi): Clitics and Their Hosts. Vol. VIII, 2 (by Marco Haverkort): Clitics and Parametrization — Case Studies in the Interaction of Head Movement Phenomena. Vol. VIII, 3 (edited by Luigi Rizzi): Clitics in Romance and Germanic. Vol. VIII, 4 (edited by Lars Hellan): Clitics in Germanic and Slavic. Vol. VIII, 5 (edited by Lars Hellan and Henk van Riemsdijk): Clitic Doubling and Clitic Groups. Vol. VIII, 6 (edited by Henk van Riemsdijk and Lars Hellan): Clitics Their Origin, Status and Position. One thing had been clear from the start: really broad coverage of a close to complete number of European languages was going to be impossible to achieve given the principles and parameters approach adopted in this group. The compromise we finally settled for was to aim at a broad and not too theoretical presentation of the main properties of clitics in some of the major language groups (Romance, Slavic, Scandinavian, and (the remaining) Germanic languages). Alongside this, we decided to present a survey of a larger sample of languages, 29 to be exact, for which respondents were available who would know what our questions intended to establish. A first pilot investigation along these lines was carried out by Riet Vos. A slightly modified version of the questionnaire was then used to collect data on all of the 29 languages. This part as well as the compact presentation of the results presented in this volume was accomplished by Lida Veselovskä. In the overall organization of the present volume, we call this the 'descriptive' part. It should be clearly understood, however, that we are talking a type of description which is heavily guided by the theoretical concerns that are prevalent within the principles and parameters approach to linguistic theory. Similarly, the second major part of this book, which is called 'theory,' relies in all its aspects on sound description. The distinction between description and theory, then, is a

Preface

xi

problematic one, but we have decided to maintain it in the interest of a transparent organization of the many contributions to this book. The second part ('theory') is itself composed of two major subparts. The first of these contains two major and representative approaches to the grammar of clitics. It was clear from the outset, given the current state of the art in the research concerning clitics (see my introductory article to this volume), that anything like a coherent picture was not going to be achievable, let alone anything even approaching a consensus. We have tried to overcome this problem by selecting two rather diverging approaches for more comprehensive presentation, and furthermore by presenting them in the manner of feature articles together with peer comments and a reply by the authors, very much in the manner of the journal Behavioral and Brain Sciences. These feature articles are supplemented in a second subpart by 13 articles on more specific topics in the grammar of clitics. These have been grouped together under five headings, though there is, of necessity, a considerable overlap among the topics and subareas that these articles pertain to. A special remark is in order here with regard to the bibliographical resources concerning clitics. We might have considered including a large bibliography on clitics in the present volume, had it not been for the fact that we knew that a comprehensive bibliography was going to appear before this volume could possibly be completed. This bibliography, entitled Clitic: A Comprehensive Bibliography 1892-1991 and compiled by Joel A. Nevis, Brian D. Joseph, Dieter Wanner and Arnold Zwicky did indeed appear in 1994 with the John Benjamins Publishing Company (Amsterdam). It is a highly recommended research tool for anyone working on clitics. In view of its existence, we decided not to opt for a common bibliography at the end of the present volume, but instead to present each individual contribution with its own bibliography, thereby making them easier to use separately. Many more people than those mentioned by name here have contributed in various ways. Since a comprehensive listing would not only be excessively long but also by necessity incomplete in view of my memory limitations, they will, I hope, forgive me for thanking them collectively and anonymously here. I do wish to make an exception, however, for the late Simon Dik, who was absolutely instrumental in getting this project off the ground, for Ekkehard König and Martin Haspelmath, who were the nicest Central Committee anyone could have hoped for, for Max Sparreboom at the European Science Foundation office, to whose quiet and unobtrusive administrative guidance the project owes a lot, and finally for Mischa Corsius, who valiantly tackled the gargantuan task of editing the manuscript of this volume. Goirle, February 1996

Henk van Riemsdijk

Contents

Contributors Abbreviations

xvii xxi

Henk van Riemsdijk Clitics: A state-of-the-art report

1

Part I Area studies Anna Cardinaletti Pronouns in Germanic and Romance languages: An overview . . .

33

Mila Dimitrova- Vulchanova Clitics in the Slavic languages

83

Lars Hellan and Christer Platzack Pronouns in Scandinavian languages: An overview

123

Part Π Theory A. Feature articles Anna Cardinaletti and Michal Starke The typology of structural deficiency: A case study of the three classes of pronouns

145

Peer comments by Josef Bayer Molly Diesing Carmen Dobrovie-Sorin Liliane Haegeman Anders Holmberg Juan Uriagereka

235 243 249 257 263 267

Reply to comments by Anna Cardinaletti and Michal Starke

....

273

Joseph Emonds How clitics license null phrases: A theory of the lexical interface . 291 Peer comments by Mark C. Baker

369

xiv

Contents

Marcel den Dikken Frank Drijkoningen Hubert Haider Marco Ηανerkort Tarald Taraldsen

375 381 385 391 399

Reply by Joseph Emonds

411

B. Topics 1. Clitic clusters and the Wackernagel position Damir Cavar and Chris Wilder "Clitic third" in Croatian

429

Mila Dimitrova- Vulchanova and Lars Hellan Clitics and Bulgarian clause structure

469

Carmen Dobrovie-Sorin Clitics across categories: The case of Romanian

515

2. Functional categories and the position of clitics Adriana Belletti Italian/Romance clitics: Structure and derivation

543

Cecilia Poletto The internal structure of AgrS and subject clitics in the Northern Italian dialects 581 Ian Roberts Agreement marking in Welsh and Romance

621

Alain Rouveret Clitics, subjects and tense in European Portuguese

639

Dominique Sportiche Pronominal clitic dependencies

679

3. Clitics and scrambling Kathrin Cooper On the nature and distribution of Zurich German pronominal clitics 711 Gunlög Josefsson On the semantics and syntax of Scandinavian pronouns and object shift 731

Contents

xv

4. Semantic features Elena Anagnostopoulou Conditions on clitic doubling in Greek

761

Norbert Corver and Denis Delfitto On the nature of pronoun movement

799

5. Phonological aspects Marina Nespor The phonology of clitic groups

865

Appendix Riet Vos and Ludmila Veselovskä Clitic Questionnaire Index of languages Index of names Index of subjects

891 1011 1014 1020

List of contributors Elena Anagnostopoulou Dept. of Language and Literature Tilburg University P.O. Box 90153 NL - 5000 LE - Tilburg [email protected] Mark C. Baker Department of Linguistics McGill University 1001 Sherbrooke St. W Montreal, PQ H3A 1G5 Canada mbaker@ langs.lan.mcgill.ca Josef Bayer Institut für Germanistische Sprachwissenschaft Friedrich-Schiller-Universität Jena Ernst-Abbe-Platz 4 D-07743 Jena [email protected] Adriana Belletti Facoltä di Lettere e Filosofia Universitä degli Studi di Siena Via Roma 47 I-53100-Siena [email protected] Anna Cardinaletti Dipartimento di Studi Linguistici e Letterari Universitä di Venezia Palazzo Cosulich Dorsoduro 1405, Zattere 1-30123-Venezia [email protected]

Damjr Cavar Universität Potsdam Institut für Linguistik - Allgemeine Sprachwissenschaft Postfach 60 15 53 D-14415 Potsdam cavar @ rz.uni-potsdam.de Kathrin Cooper Seminar für Allg. Sprachwissenschaft Universität Zürich Plattenstrasse 54 CH-8032-Zürich [email protected] Norbert Corver Dept. of Language and Literature Tilburg University P.O. Box 90153 NL - 5000 LE - Tilburg [email protected] Denis Delfitto Vakgr. Romaanse Talen en Culturen Universiteit Utrecht Kromme Nieuwegracht 29 3512 - HD - Utrecht [email protected] Molly Diesing Department of Linguistics Morrill Hall Cornell University Ithaca, NY 14853, USA diesing @ comell.edu

xviii

List of contributors

Marcel den Dikken Holland Institute of Generative Linguistics Vakgroep Taalkunde (ATW) Vrije Universiteit Amsterdam De Boelelaan 1105 NL - 1081 HV - Amsterdam [email protected] Mila Dimitrova-Vulchanova Department of English, NTNU Ν - 7055 - Dragvoll, Trondheim [email protected] Carmen Dobrovie-Sorin Universite de Paris 7, Case 7003, Linguistique, 2, place Jussieu F-75005 Paris [email protected] Frank Drijkoningen Department of Linguistics University of Utrecht Trans 10, NL - 3512 JK - Utrecht frank.drijkoningen @ let.ruu.nl Joseph E. Emonds Department of Linguistics and English Language University of Durham Elvet Riverside, Durham United Kingdom, DH1 3JT [email protected] emonds @ nias.knaw.nl Liliane Haegeman Department of Linguistics University of Geneva CH-1211-Geneva 4 haegeman @ lettres.unige.ch

Hubert Haider Institut für Sprachwissenschaft Universität Salzburg Mühlbacherhofweg 6 A - 5020 - Salzburg hubert.haider@ sbg.ac.at Marco Haverkort Linguistics, Boston University 718 Commonwealth Avenue Boston, MA 02215, USA [email protected] Lars Hellan Department of Linguistics, NTNU Ν - 7055 Dragvoll, Trondheim lars .hellan @ hf. ntnu. no Anders Holmberg University of Troms0 / ISL Ν - 9037 - Troms0 andersh @ isl.uit.no Gunlög Josefsson Lunds Universitet Institutionen för Nordiska Spräk Helgonabacken 14 S - 223 62 Lund [email protected] Marina Nespor Department of Italian / HIL University of Amsterdam Spuistraat 210 NL - 1012 VT - Amsterdam [email protected]

List of contributors Christer Platzack Lunds Universitet Institutionen för Nordiska Sprak Helgonabacken 14 S - 223 62 Lund [email protected] Cecilia Poletto Dipartimento di Linguistica Universitä di Padova Palazzo Maldura Via B. Pellegrino 1 I-35137-Padova poletto@uxl .unipd.it Henk van Riemsdijk Tilburg University Department of Language and Literature P.O. Box 90153 NL - 5000 LE - Tilburg [email protected] Ian Roberts Institut für Linguistik: Anglistik Universität Stuttgart Keplerstraße 17, D-70174-Stuttgart [email protected] Alain Rouveret Departement de Linguistique Universite de Paris VIII St. Denis Dominique Sportiche Department of Linguistics, UCLA 405, Hilgard Avenue Los Angeles CA 90095-1543, USA sportiche @ ucla.edu

xix

Michal Starke Department of Linguistics University of Geneva CH-1211-Geneva 4 [email protected] Knut Tarald Taraldsen ISL, Universitetet i Troms0 N-9037 Troms0 [email protected] Juan Uriagereka Department of Linguistics, Rm 1401C, Marie Mount Hall, University of Maryland College Park MD 20742-7505, USA [email protected] Ludmila Veselovska English Department, Phil. Faculty Palacky University Krizkovskeho 10 772 00 Olomouc Czech Republic [email protected] [email protected] Riet Vos Dept. of Language and Literature Tilburg University P.O. Box 90153 NL - 5000 LE - Tilburg [email protected] Chris Wilder Zentrum für Allgemeine Sprachwissenschaft (ZAS) Berlin Jägerstrasse 10-11 D-10117 Berlin [email protected]

List of abbreviations IP 2P 3P A Ä/A' ABS AC ACC ACCP ACL AGR AR ARG ART AuxCl C, CP

CAUS CED

CL CL-ARG CL-AUX CL-GEN CL-Q CLP COMP COND COP D, DP D-POSS DAT DCL DEF

first person second person third person argument non-argument absolutive Accidental Coreference accutive Accusative phrase accusative clitic agreement Alternative Realization argument article auxiliary clitic complementizer, complementizer phrase causative Condition on Extraction Domains clitic argument clitic auxiliary clitic genitive clitic Wh-clitic clitic phrase complementizer conditional copula determiner, determiner phrase dative possessive clitic dative dative clitic definite

DEM DIM DL

demonstrative diminutive Deep Lexicalization DO direct object ? DTE DUR durative EP European Portuguese ECM Exceptional Case Marking ECP Empty Category Principle Extended EPP Projection Principle ergative ERG EST Extended Standard Theory exclusive EXC FEM/F feminine FOC focus FP maximal projection of some functional category FUT future GEN genitive; gender GER gerund GT Generalized Transformation HAB habitual HLC hostless clitics HMC Head Movement Constraint Invisible ICP Category Principle IMPERF/IMPF imperfect IMPR imperative INF infinitive; infinite

xxii

List of abbreviations

INFL IO LF LOC LSC MASC/M MOD MPLT NEG NEUT/NT/N NIDs NOM NONCL-AUX NP OBJ Ρ PP

P-stranding PART/PRT PF PI PL

POSS/POS PP PRES/PRS pro, PRO

inflectional host indirect object Logical Form locative Logical Subject Condition masculine mood ?...framework negation marker neuter Northern Italian dialects nominative non-clitic auxiliary noun phrase object preposition; phrase prepositional phrase; projection principle preposition stranding particle; participle Phonological Form prosodic inversion plural; Phonological Lexicalization possessive past participle present pronominal null categories

Q QCL QP RECP REFL SCL SF SG SM SOV SP Spec/SPEC SSC SUBJ/SBJ SVO T, TP T.-M. UG UTER V2

vos VP VS

wco WF WP LGB

Wh-marker Wh-clitic quantifier phrase reciprocal reflexive subject clitic singuler, feminine singular singular, masculin subject-objectverb strong pronouns specifier Specified Subject Condition subject; subjunctive subject-verbobect tense, tense phrase Tobler-Musafia Universal Grammar —.neuter verb second verb-objectsubject verb phrase verb-subject weak cross-over West-Flemish weak pronouns Lectures on Government and Binding

Henk van Riemsdijk

Clitics: A state-of-the-art report

1.

Introductory remarks1

There is considerable agreement among linguists that alongside the major types of grammatical constructs such as lexical heads, phrases, stems, derivational affixes, etc. there is a separate class of grammatical elements, commonly called clitics. On the negative side, these are characterized by the fact that they are not easily classifiable in terms of the morphological or syntactic notions listed above. On the positive side, however, their definition is a topic of much debate. As is to be expected, this uncertainty is the result of the fact that developments in the construction of the theory of syntax and morphology have not, so far, resulted in a framework within which clitics are straightforwardly accommodated and which would thereby yield clear-cut decisions about what is and what is not a clitic. In a much cited dictum,2 Edward Sapir has characterized clitics as follows: Enclisis is neither true suffixation nor juxtaposition of independent elements. It has the external characteristics of the former (including strict adherence to certain principles of order), the inner feeling of the latter. (Sapir 1930: 70-71)

In most accounts, this ambivalence in status is primarily related to the question whether clitics should be treated as affixes or as independent (free) syntactic forms. Theoretical developments in syntax and morphology have produced instruments of considerable sophistication which lead to further questions as to the status of clitics. For example, if clitics are syntactically independent elements, are they head-like or phrase-like? When clitics partake in movement-type dependencies, are these dependencies more like Α-movement or like Ä-movement? On the other hand, if clitics are affixes, which type of affixes do they resemble most? Such more fine-grained classificatory criteria are helpful in formulating the question in a precise way, even if they have not necessarily resulted yet in a definitive theoretical characterization of the notion clitic. As will become clear below, the attempt to fit clitics into a framework defined in terms of such notions has thus far resulted in the impression that clitics constitute a rather hybrid category. There are two logical alternatives. Either the notion of clitic is a spurious concept, or they constitute a truly independent grammatical concept with properties that are largely sui generis. Both lines of thought are amply represented in the work of the past two decades as well as in the present volume.

2

Henk van Riemsdijk

Basically, then, the syntax of clitics has been providing us with a considerable challenge and still is. We are faced with a solid body of properties, many of them pretheoretical in nature, that hold of various subclasses of the category of clitics. If this means that a general theory of clitics is still lacking, it should be kept in mind, however, that the ability to ask precise questions about a syntactic or morphological phenomenon is an encouraging sign of theoretical progress.

2.

Some relevant properties

2.1.

Kayne's analysis of French clitics

One of the first major contributions to the analysis of clitics in generative grammar is found in Kayne (1975), which is based on his unpublished 1969 dissertation. Kayne's central concern was establishing a coherent syntactic movement theory of clitics in which general constraints would do most of the work in explaining the major properties of French clitics. In order to do so, the basic analytical tools had to be developed. Many of the criteria for clitic status proposed by Kayne still underpin most of the work on clitics today. Let us summarize the main criteria established by Kayne.3 Identificational criteria for clitics (i) Special position Clitics appear in a position where the corresponding full phrase does not occur. Direct object clitics, for example, occur to the left of the finite verb, while full direct objects are on the right. (1)

a.

II a lu tous les articles. he has read all the articles b. *Ila lu les (them) c. II les a lus.4

(ii) Obligatoriness As (lb) shows, the appearance in the special position is obligatory, as opposed to, e.g. the behavior of certain quantifiers. (2)

a. b.

II a lu tout (everything). II a tout lu.

(iii) Adjacency to V The clitic must be adjacent to the verb, which constitutes its "host".

Clitics: A state-of-the-art report (3)

3

a.

Elle να les apprecier beaucoup. she will them appreciate much b. *Elle va les beaucoup apprecier c. Elle να beaucoup les apprecier.

(iv) Obligatory presence of the verb The clitic is dependent on the presence of its host verb, it cannot survive without it. (4)

Qui as-tu vu? — who did-you see?

*Le/*la/ *les him her them

Similarly, when the verb disappears under gapping, the clitic cannot be maintained.5 (5)

a.

Jean a pris le couteau et Marie a pris la fourchette. John has taken the knife and Mary has taken the fork b. *Jean I'a pris et Marie la

(v) No modification Unlike other elements that can move to a pre-verbal position, such as the quantifier tout, clitics can never be modified. (6)

a. b.

(7)

II a compris presque tout. he has understood almost everything II a presque tout compris.

a.

II ne connait que nous deux. he knows only us two b. *Il nous deux connait

(vi) No stress Clitics cannot be stressed. (8)

a.

Jean prefere celle-la. John prefers that one b. Jean la prefere. c. *Jean LA prefere

The property listed in (iv) above may well be related to this fact as well, cf. note 5.

4

Henk van Riemsdijk

(vii) No conjunction Clitics cannot be conjoined. (9)

a.

Je connais Jean et Marie. I know John and Mary b. *Je le et la connais

(viii) Fixed and special order Clitics occur in a fixed order, and very often this order deviates from the order that the corresponding full phrases would occur in. (10)

(11)

a.

Jean donnera le livre ά moi seul. John will-give the book to me alone b. *Jean donnera a moi le livre *Jean le me donnera Jean me le donnera.

For many researchers, the eight properties listed above constitute the point of departure for their own work on clitics. This may have given rise to a certain emphasis on the properties of clitics in the Romance languages and in French in particular. To give just one example, if we compare the French quantitative en, which is undoubtedly a clitic in terms of Kayne's critera, with its Dutch equivalent, the quantitative er, we see that er is a clitic by most criteria, but it need not be adjacent to the verb.6 Nevertheless, the occurrence of er is contingent on the presence of the finite verb in the clause, as is shown in the gapping example below. (12)

a. b. c.

(13)

II en a lu un. he there(of) has read one Hij heeft er gisteren een gelezen. he has there yesterday one read ... dat hij er gisteren een gelezen heeft. that he there yesterday one read has

Hij heeft er gisteren een gelezen en (*er) eergisteren he has there yesterday one read and the-day-before-yesterday twee. two

Many of the issues addressed in the present volume will be of this nature. Elements are found in various languages that display a sufficient number of

Clitics: A state-of-the-art report

5

properties thought to be characteristic of clitics but that also deviate in certain ways. How can these be accommodated? Ultimately, the question is whether it is possible to establish a coherent and unified concept of clitic that allows for some epiphenomenal cross-linguistic variation in the way that clitics manifest themselves and behave under various syntactic conditions. Alternatively, it might turn out that there is no unified category of clitics, On this view, the apparent convergence of a number of properties would ultimately be a kind of mimicry in which fundamentally different creatures (for example affixes and syntactically independent functional categories) behave alike in a number of ways.

2.2.

Some major phenomena

The property of clitics that has attracted the special interest of syntactic theorists is the fact that "they do not occur where we would expect them", that is, property (i) above. Within the generative paradigm, this translates into the question of whether some kind of movement is involved. From this question, a number of other questions immediately arise: is it really movement, is there always a trace, does the movement obey the general constraints on movement, is long distance movement possible, etc. Three issues of this type played a particularly prominent role in Kayne's book, and we will summarize them here. (i) The clause mate restriction The relative locality of movement processes has been a central issue in the establishment of a set of constraints on movement in general. Dependencies of the type found in passive constructions, for example, are known to be limited, by and large, to the simplex clause. Wi-movement-like processes, on the other hand, can apparently span long distances, but they can be broken down into series of successive cyclic steps that each span the domain of a single clause only. With one class of exceptions, to be discussed under (ii) below, the relation between the position of a clitic and the position that the corresponding full phrase would occur in is tightly clause-bound. Consider the following examples: (14)

a.

II croyait que les etudiants liraient ces livres. he believed that the students would-read these books b. II croyait que les etudiants les liraient. c. *// les croyait que les etudiants liraient

(15)

a.

Nous avons essaye de comprendre ces livres. we have tried to understand these books b. Nous avons essaye de les comprendre. c. *Nous les avons essaye de comprendre

6

Henk van Riemsdijk

Even with the still somewhat primitive movement theory of the mid seventies, it was a rather straightforward task to ensure this domain restriction. On the whole, the Tensed S Constraint and the Specified Subject Constraint could be relied upon to yield this result.7 (ii) Clitic climbing In the examples given under (i), the clitic shows up on the main verb of its own clause, or on the auxiliary accompanying that main verb. The French causative construction with faire shows a somewhat different pattern in that the clitic can, and often must, be attached to the causative verb rather than to the complement verb that it is a dependent of. This process, called clitic climbing,8 is illustrated in (16). (16)

a.

Elle fera manger ce gäteau a son enfant. she will-make eat this cake to her child b. *Ellefera le manger a son enfant. c. Elle lefera manger ä son enfant.

In French, clitic climbing is essentially limited to causatives but in other Romance languages such as Italian, the phenomenon is more wide-spread. Rizzi (1978) discusses a series of verbs, in particular modals, aspectuals and motion verbs, that have this property of allowing the clitic to climb to the higher verb. Some examples are given in (17). (17)

a. b. c.

Gianni gli ha dovuto parlare personalmente. G. to-him has had-to speak personally Mario gli continua a raccontare stupide storie. M. him continues to tell stupid stories Piero li venne a chiamare alia stazione. P. them came to call at-the station

Rizzi calls these verbs restructuring verbs because, in conformity with other researchers studying similar phenomena,9 he argues that there is a process, restructuring, which turns a bi-clausal structure into one that is, at least for the purposes of clitic placement, mono-clausal. Hence, these analyses are compatible with the assumption that the placement of clitics is fundamentally clause-bound. (iii) Clitic doubling The complementary distribution of clitics and the corresponding full phrase, as found in French, constitutes a strong indication that movement is involved. Take the case of direct object clitics. Whenever the nominal element fulfilling the role of direct object happens to be a weak pronoun of the clitic type it must move to the verb.

Clitics: A state-of-the-art report (18)

7

a.

II respecte ton pere. he respects your father b. II ne respecte que toi. he respects only you c. *// respecte te d. II tet respecte [e];

This complementarity, together with the observation that the relation between the clitic and the corresponding full phrase position is subject to locality constraints in much the way other movement processes are, led Kayne to develop a movement analysis for clitics. It turns out, however, that there are cases in which clitics cooccur with an overt phrase in the other position. This phenomenon, called "clitic doubling" by Jaeggli (1982), constitutes a considerable problem for the assumption that all clitic constructions are derived by means of movement. An example is given in (19). (19)

Miguelitole regalo un caramelo α Mafalda. Μ. her gave a candy to M.

In many cases, doubling is even obligatory: (20)

a.

Le duele la cabeza α Mafalda. her hurts the head to M. b. *Duele la cabeza α Mafalda

The analysis of clause-boundedness, climbing and doubling constitutes a considerable challenge that any attempt at improving the theory of syntax must face. Most of the contributions in this volume address these issues in some way or another.

2.3.

Zwicky's tests for affix status

In Kayne's work, the assumption that clitics are syntactically independent elements is not in itself a matter of debate. And to the extent that the theory of syntactic movement can be successfuly invoked to account for the properties of clitics, this position is strongly supported. Nevertheless, there are good reasons to investigate the relation between clitics and affixes in more detail. Zwicky (1977) presents a number of criteria for deciding whether a given grammatical formative is, or is not, an affix. Five such criteria are presented (Zwicky 1977: 2-3).

8

Henk van Riemsdijk

Ci) Ordering While many languages allow considerable variation in surface ordering among syntactic constituents, within the word ordering is strict. Whenever two affixes occur in alternative order, a precise difference in cognitive meaning is associated with the two orderings. As an example, consider the interaction of the causative suffix chi- and the reciprocal suffix riaku- in Quechua as described in Muysken (1981: 305ff). (21)

a.

b.

maqa- chi- naku- rka- η beat CAUS RECP PL 3 'They let each other be beaten.' maqa- naku- ya- chi- η beat RECP DUR CAUS 3 'He is causing them to beat each other.'

Yet, in the same article Muysken describes the behavior of the delimitative nominal suffix la- ('just', 'only') which is quite free in its positioning within the word (Muysken 1981: 295). (22)

a.

b. c.

kiki- la- n- kuna self 3 PL 'just themselves' kiki-n-kuna-la kiki-la-n-kuna-la

Here the three forms apparently have the same meaning. Be that as it may, this criterion does not yield any immediate conclusions for clitics such as those found in the Romance languages. Relevant examples might be constructed with dative — accusative sequences like the following: (23)

a. *Jean me J.

b. *Jean te J.

te

presentera

me(DAT) you(ACC) will-introduce

me

presentera

you(DAT) me(ACC) will-introduce

Such examples are generally judged to be ungrammatical,10 but to the extent that they can nevertheless be judged, they differ in meaning in the expected way. It appears that central and southern dialects of Italian do exhibit the relevant contrast, but with the opposite order:11 (24)

a. *Gianni mitiacc presentera b. Gianni miacc tidat presentera

Clitics: A state-of-the-art report

9

But even here, the paradigm is limited because the order 2nd person — 1st person is ungrammatical on either reading: (25)

*Gianni ti mi presentera

But despite these limitations, examples of ordering variation can be found. Consider (26).12 (26)

a. ΊPaolo gli P.

b. ΊΡαοΙο mi P.

mi

presenta.

him(DAT) me(ACC) introduces

gli

presenta.

me(ACC) him(DAT) introduces

While both marginal,13 these two examples have the same meaning. (ii) Internal Sandhi Certain phonological processes are typically limited to the domain of the word. So, if a bound morpheme is subject to such a process, we may conclude that it is an affix. But again, the test does not provided with unequivocal results in the case of clitics. As Kayne (1975: 85) observed, subject clitics undergo liaison where corresponding full subjects do not. (27)

a. b.

Iis iront a Paris./i(l)zirö..J */i(l)irö..J they will-go to Paris Mes amis iront a Paris. */...amizirö...//...amiirö.../ my friends will-go to Paris

But various researchers have pointed out that clitics do not always function as part of the phonological word. Nespor and Vogel (1986: 125/147ff) show, for example, that the Italian rule of intervocalic s voicing applies within a word but never to a clitic-verb combination.14 (28)

a.

resistenza [z] *[s] cas-ina [z] *[s] resistance

b.

house-DiM

lo sapevo *[z] [s] telefonati it I-knew

si *[z] [s]

having-phoned each-other

(Hi) Bound vs. free According to Zwicky, morphemes that cannot occur in isolation, in other words bound morphemes, are affixal. To the extent that clitics cannot occur freely but must have a "host" that they attach to, they would have to be considered affixes. But the discussion above in connection with the Dutch examples in (12) and (13)

10

Henk van Riemsdijk

has shown that it is far from clear that the presence of a host is a necessary condition for clitichood. (iv) Construction with affixes If a morpheme can occur internal to elements that are incontrovertibly analyzed as affixes, it must be an affix as well. Take an example from Dutch. Dutch, like German, has a class of so-called separable verbal prefixes. They are separable in that, for example, they are not moved along with the verb under such processes as verb second (V-to-C movement). An example is given in (29). (29)

a.

...dot Jan het hoofdstuk in-deelt. that J. the chapter sub-divides b. Jan deelt het hoofdstuk in. c. *Jan indeelt het hoofdstuk

There are also inseparable verbal prefixes. A particularly interesting one is her(re-). (30)

a. ... dat Jan het hoofdstuk her-leest (re-reads). b. Jan her-leest het hoofdstuk. c. *Jan leest het hoofdstuk her

Her- is special in that it can attach to verbs with a separable prefix. When it does, verb second can no longer apply in any form. In other words, a defective paradigm results. (31)

a. b. c. d. e.

... dat Jan het hoofdstuk her-in-deelt. *Jan her-in-deelt het hoofdstuk *Jan her-deelt het hoofdstuk in *Jan deelt het hoofdstuk her-in. *Jan het hoofdstuk her-in-deelt

The a-example would suggest, according to Zwicky's criterion, that the separable prefixes are true affixes. But the remaining examples show that the separable prefix retains some of the force of an independent syntactic element. (v) Immunity to ellipsis Zwicky suggests that processes of deletion under identity, conjunction reduction or right node raising cannot apply to stem+affix combinations. He cites the following examples (Zwicky 1977: 3)

Clitics: A state-of-the-art report (32)

a. b.

11

yellowish or greyish vs. * [yellow or grey]ish dancing and singing vs. * [dance and singling

As Wegmüller (1993: 19) mentions, this criterion applies in a similar fashion to French clitics: (33)

a.

Je le connais et je Γ aime bien. I him know and I him like well b. *Je le connais et aime bien

But this criterion too is problematic because ellipsis is often possible in wordword combinations. Toman (1985: 423ff) gives examples like the following. (34)

a. b. c.

Kalbs- und Schweineleber calf- and pork-liver arbeits- und mittellos work- and means-less trink- und essbar drink- and eatable

(vi) Stress It is a well-known fact that certain affixes can affect stress, though not all affixes do. In English, for example, in- does while un- does not, witness15 (35)

a. b.

pious vs. impious finite vs. infinite lawful vs. unlawful grateful vs. ungrateful

But again, according to this criterion some clitics would have to be affixes because they affect stress, while others, which do not affect stress, could be either stress-neutral affixes or syntactically independent elements. Thus, from the following example, due to Nespor and Vogel (1986: 147), we would have to conclude that the Modern Greek possessive clitic is an affix because it attracts stress. (36)

a. b.

ο änthröpos (the person) ο änthröpos mas (our person)

In this respect, Modern Greek clitics differ from, for example, Spanish clitics. The latter do not affect stress, even when the combination of the verb with the clitics results in a stress pattern that is not a possible word stress pattern:

12 (37)

Henk van Riemsdijk dändo (giving) dändo-nos-los (giving us them)

In summary, it appears that we are faced with a somewhat uncomfortable choice. Either we accept the criteria proposed by Zwicky and conclude that some clitics are affixal in nature and some are syntactically independent morphemes, or we reject the notion that Zwicky's criteria determine the status of clitics, either individually or collectively.

2.4.

Clitics as phrasal affixes

One way out of this dilemma can be constructed along the following lines. The typical property of words is that they combine with other words or phrases to form phrases. The typical property of an affix is that it combines with (part of) a word to form a word. We might now assume that clitics are elements which are bound, but which combine with words or phrases to form phrases. In other words, we might assign to clitics the status of phrasal affixes. This line of thinking, developed by Carstairs (1981) and many others, is, in a way, the easy way out: if clitics cannot be either A or B, let us define a third type of element C. Nevertheless, there appears to be some direct evidence for this view. In particular, this solution would accommodate such otherwise unclassifiable phenomena as the English genitive marker -'s. This morpheme appears to be aselective with respect to the immediately preceding word, but selective with respect to the immediately preceding phrase. In other words, it attaches to a noun phrase, regardless of what word that noun phrase ends in.16 (38)

The woman I met yesterday's dog. Those that came's award. To anyone who was there's surprise.

Once we "capitulate" and accept that clitics are neither truly independent words nor real affixes and hence have to constitute a class of elements in their own right, there are several ways we can go. The phrasal affix hypothesis is then only one of several possibilities. From a phonological perspective, the situation is not much clearer. Some of the examples cited above show that while clitics are phonologically more active than fully regular syntactic combinations of words and phrases, they are at the same time resistant to a number of phonological processes that occur within the word proper. Of course, even within words such differences must be accepted. Particularly well-known is the distinction between class I affixation and class II affixation. Take example (35) above. The English prefix in- is a class I prefix while un- is a class Π prefix. The example shows that class I affixes can affect

Clitics: A state-of-the-art report

13

stress while class II affixes do not. Similarly, class I prefixes are subject to nasal assimilation while class II prefixes are not. This is illustrated in (39).17 (39)

a. b.

in+legal in+regular un#lawful un#real

illegal/*inlegal irregular/*inregular *ullawful/unlawful *urreal/unreal

Importantly, however, clitics cannot be simply identified with one of these two classes of affixes. 18 Nespor and Vogel (1986: 151ff) argue that host+clitic combinations are subject to phonological processes that are different from those operating at the word level. Consider again stress shift in Modern Greek. This shift takes place whenever stress is separated by more than two syllables from the end of the host+clitic combination. (40)

a. b.

yrdpse yrapse to yräpse mü write write it write me dyavase dyavase to dyävase read read it read

to it mu to me it

This process of stress adjustment is different from the one found within complex words: (41)

a. b.

[kükla] doll [nixta] night

[spiti] => [kuklospito] house doll's house [puli] => [nixtopuli] owl night bird

Nespor and Vogel 'list' two major differences between the two processes. First, in the word+clitic combination, the original word stress never shifts to any other syllable, as it does in compounds such as those listed in (41). Second, In the word+clitic combination, a new stress is added in precisely those cases in which the resulting form would otherwise violate the constraint that the stress may not be separated from the end by more than two syllables. Clearly, this is not the case in the compounds listed in (41). Nespor and Vogel therefore conclude that a new type of phonological domain must be introduced. They call it the clitic group. The clitic group finds a place within a fairly elaborate hierarchy of increasingly inclusive domains, each with a class of phonogical processes associated with it. This hierarchy is summarized in (42).19

14

Henk van Riemsdijk

(42) Domain

Symbol

syllable foot phonological word clitic group phonological phrase intonational phrase phonological utterance

2.5.

σ Σ ω C Φ I υ

Some further typological distinctions

When we approach the study of clitics from the perspective of French pronominal clitics, as we did here, the conclusion that syntax is involved is easy to arrive at. The clitics in question have a distribution that differs from that of the elements they are in complementary distribution with. But such direct indications of syntax are absent in a great many cases. Very often, clitics are simply weak, reduced forms of corresponding strong forms occurring in the same position, though the weak form may be phonologically bound in the sense that it forms a phonological unit with its host or, at least, requires the presence of a host. Zwicky (1977) has introduced this distinction as a fundamental typological dimension. In his terminology, the syntactically neutral clitics of the latter type are called simple clitics while the syntactically active clitics of the type found in French are called special clitics. Pursuing this perspective, Klavans (1985), building on work in her dissertation,20 proposes a typology of clitics based on three binary parameters: — — —

DOMINANCE values: initial vs. final PRECEDENCE values: before vs. after PHONOLOGICAL LIAISON values: proclitic vs. enclitic

The first two of these parameters are directly concerned with the position the clitic surfaces in. The DOMINANCE parameter determines whether the position of the clitic is defined with respect to the initial/final constituent of the relevant phrase. The PRECEDENCE parameter then determines whether the clitic comes before or after that constituent. Schematically: (43)

[ α -γ, A y 2 Β y 3 C

D y5 Ε y 6 F Ί η ] a

If, in a given language, clitics (γ) are defined with respect to domain a, then the DOMINANCE parameter determines that they must be positioned either relative

Clitics: A state-of-the-art report

15

to A (initial) or relative to F (final). If the value for PRECEDENCE is before, the clitic precedes A and F respectively, if it is after, the clitic follows A and F respectively. Hence, Ύ1/2/6/7 ^ possible clitic positions, while 7 3/4/5 are not. In terms of this system, the well-known observation, due to Wackernagel 1892, that clitics have a tendency to occur in the second position is expressed by saying that the parameter setting for these clitics is DOM(initial), PREC(after). The third parameter, PHONOLOGICAL LIAISON, states that a clitic can be phonologically dependent on the preceding constituent (enclitic) or on the following consituent (proclitic). This distinction is independent of the above syntactic parameters. Thus, in (43) the clitic y 2 could be enclitic on A or proclitic on B. Similarly, y6 could be enclitic on Ε or proclitic on F. Put differently, the syntactic host need not correspond to the phonological host. As an example of the latter type, consider clitics in the Australian language Nganhcara as described in Klavans (1985: 104-105). In the relevant cases,21 the parameter setting is DOM(final), PREC(before), PHON. LIAISON(enclitic). In other words, we have a case of y 6 with F as its syntactic host and Ε as the phonological host. The examples are given in (44).22 (44)

a.

nhila

pama-ng

nhingku ku?a = gku

he(NOM) man-(ERG) 2sg(DAT) dog

b.

'The man gave a dog to you.' nhila pama-ng nganhcara

wa:

= 2sg(DAT) give

ku?a = nhcara

wa:

he(NOM) man-(ERG) lpl(EXC,DAT) dog = lpl(EXC,DAT) give

'The man gave a dog to us.' These are examples in which the indirect object is doubled by a dative clitic. The syntactic host is the verb at the end. But we must assume that the clitic is enclitic on the preceding constituent because ngk- and nhc- are not possible word initial consonant clusters in Nganhcara. Illustrating all eight cases of the Klavans typology would go beyond the confines of the present article. Instead, I will reproduce the summary from Klavans (1985: 103) showing that every one of the eight cases is, according to her, attested. (45) Type Type Type Type Type Type Type Type

1 2 3 4 5 6 7 8

DOM initial initial initial initial final final final final

domain N' N' S S

s s

V[-T]

s

PREC before before after after before before after after

LIAISON enclitic proclitic enclitic proclitic enclitic proclitic enclitic proclitic

example Kwakwala NP-markers Greek article Ngiyambaa enclitics Tepecano = an Nganhcara clitics Sanskrit pre-verbs Spanish pronominal clitics Greek negative ou =

16

Henk van Riemsdijk

It must be noted here, that there is a considerable distance between most of the "mainstream" generative research reported on above and below on the one hand, and the type of taxonomy presented by Klavans. If all the elements in Klavans' description are clitics in some theoretical sense, then current theory would have to be enriched considerably in order to deal with the properties of a system such as that depicted in (45). Suffice it to note, here, that no one has stood up to the challenge as yet.

3.

Theoretical issues

As the theory of transformational dependencies (movement, binding relations, chains, etc.) began to take shape in the course of the seventies and eighties, it became increasingly important to classify the relationship between (special) clitics and the corresponding (canonical) non-clitic positions (henceforth the source position, SP) in terms of this theory. The main questions arising in this context are the following.23 (46)

a.

b. c.

Should the relationship between the CL and its SP be characterized in terms of actual movement, or should it be considered a "static" type of dependency, i.e. the type of syntactic object often referred to as chain?24 Should these CL-SP dependencies be considered Α-dependencies or Ä-dependencies?25 Are these dependencies head-head (X°-X°) or phrase-phrase (XP-XP) dependencies?

The methodological point of departure, the implicit assumption underlying most of the work on the topic, was, of course, that there is a unique and well-defined concept of clitic for which each of these questions would have a unique and simple answer. As work progressed, however, it appeared that such a straightforward characterization of clitics was not available and that the pretheoretical notion of clitic may well encompass a certain latitude, a variety of subtypes that would correspond to deverging answers to the questions listed above. The research reported in the present volume testifies to this search for a more theoretically-based definition of the concept of clitic and to the attempt to map out the dimensions along which different subtypes of clitics can be situated.

Clitics: A state-of-the-art report 3.1.

17

Syntax vs. lexicon, movement vs. base-generation, derivation vs. representation

If we look at a simple cliticization case like (47a) in French, the corresponding structure is generally held to be more or less like in (47b). (47)

a.

Je le vois.

b.

I it see Je [V° l ei tv vois ] ] [ NP es ]

A question that has kept syntacticians busy for two decades now is whether the complex verbal structure [v le vois ] is directly inserted from the lexicon,26 or whether the clitic le is moved from its source position, the empty NP, and adjoined to V. We may call the former approach the morphological, or lexical, analysis and the latter one the syntactic analysis. On the lexical analysis, the clitic would be base-generated and on the syntactic analysis it is moved. In yet other terms, the lexical analysis is a representational one, whereas the syntactic analysis is derivational in nature. On the syntactic analysis, whose first and most prominent proponent is Kayne (1975), the clitic originates in the source position and is adjoined to its verbal host. Whatever conditions apply to syntactic movement are also assumed to constrain clitic movement. One important consideration that has been taken to argue against such an approach is the Lexical Integrity Hypothesis. According to this hypothesis, the rules of syntax may not affect the internal structure of words. On the assumption that X° is the level that separates morpholexical objects from syntactic objects, a movement of the clitic "into" a complex verb would be a violation of the Lexical Intergrity Hypothesis. This is the consideration that has given rise to lexicalist approaches to cliticization such as Rivas (1974, 1977), Strozer (1976) and Borer (1984). This type of approach can be implemented in several ways. In one view, represented among others by Jaeggli (1982, 1986), the complex verb assigns the object θ-role to the NP-position, a requirement that follows from Θtheory, whereas case must be assigned to the clitic le since it is a lexical nominal element that requires case. The NP-position would then remain caseless on this view, which accounts for the fact that it must remain empty, as shown in (48). (48)

*Je le vois Jean

But this is an oversimplified picture. In fact, in many cases the source position need not remain empty when a corresponding clitic is present. This is the phenomenon introduced above (in section 2.2 [iii]) under the heading clitic

18

Henk van Riemsdijk

doubling. Some additional examples from various languages and various syntactic constructions, are listed in (49).27 (49)

a. b. c.

L;- am väzutpe Popescui (Romanian) him I-have seen OBJECT-MARKER Popescu hkit ma -o, la Karimi (Lebanese Arabic) talked-I with-him to Karim beit-Oj selha-more; (Modern Hebrew) house-his of the-teacher

On an account like Jaeggli's, clitic doubling arises as a result of a morphosyntactic distinction (or parameter) that cuts across the overall category of clitics: some clitics require case, resulting in the non-doubling configuration, other clitics can remain caseless, leaving case free to be assigned to the source position, thereby giving rise to the clitic doubling pattern. The support that the lexical theory of cliticization derives from clitic doubling phenomena would seem to be stronger than the backing it gets from the Lexical Intergrity Hypothesis. This is so because any syntactic dependency across word boundaries in a clausal structure is by definition syntactic. In other words, with respect to the Lexical Integrity Hypothesis, it really does not matter whether the dependency linking le{ and [ NP es ] is analyzed as a case of movement or as a case of something else (binding, chain formation). Under the Lexical Integrity Hypothesis, one could assume, of course, that the essential properties of the clitic, in particular its nominal features and its index, percolate to the containing head somehow. But then the burden of proof would be on the proponent of the lexical theory to show that such a device is not introduced exclusively to preserve the Lexical Intergrity Hypothesis in its strong form. The other side of the coin, then, is the question whether the clitic-verb combination can otherwise (i.e. abstracting away from the syntactic versatility of the clitic) be assumed to constitute a well-formed object in terms of a theory of the structure of words. The answer appears to be that this is far from obvious. Di Sciullo and Williams (1987) point out that several facts concerning the positioning of clitics are incompatible with the view that the clitic+verb combination is the result of a morphological rule. In particular, Romance clitics sometimes precede and sometimes follow their host verb, depending on whether the latter is finite or non-finite, as in (50). (50)

a.

Lo voglio it I-want b. *Voglio lo c. Voglio far

fare. do fare lo.

Clitics: A state-of-the-art report

19

The non-morphological status of the clitic is further confirmed by the phenomenon of clitic climbing, illustrated here by the relationship between (50a) and (50c). One way to reconcile the syntax of clitics with the Lexical Integrity Hypothesis would be to assume that the clitic+verb combination does not constitute V° but a unit of a higher bar-level. Such a proposal is defended in Jaeggli (1982). Such a position is problematic, however, in view of the need to constrain (a) the power of X-bar theory and (b) the options for movement. As for X-bar theory, a new bar level distinct from either the head or the phrase level would be required since the clitic+verb combination does not have any of the well-established phrasal properties. Similarly, a reasonably constrained theory of movement would limit the choices for the clitic-source relationship to two options: maximal phrase to maximal phrase, or head to head. Assuming the latter to be the only available option, a matter to be addressed in the next subsection, the clitic+verb combination must be the result of adjunction (base or derived) of a head to V° and must therefore itself have the status of V° as well.

3.2.

The X-bar status of the clitic

Directly related to this is the question whether the dependency between the clitic and its source position should be considered a head-head dependency or a phrasephrase dependency. In light of the remarks on the status of the clitic+verb combination in the preceding section, this question would seem to receive a relatively straightforward answer. If the clitic is adjoined to a verbal head (or to a head of a different category such as noun or a preposition), then the theory of movement (cf. Chomsky 1986)28 dictates that the clitic must be a head as well. The problem here is that it is not obvious that clitics always have a host. Take again the case of Dutch er}9 Er is a weak pronominal element, its strong counterparts being daar (there) and waar (where). It is also a "special" pronominal element in that it has special syntax: it replaces regular neuter pronouns inside prepositional phrases, but instead of occurring in the canonical position of the object of the preposition it must be placed on the left of the preposition. (51)

a.

[pp voor Jan] [PP voor hem] for John for him *[PP Jan voor] *[pp hem voor] b. [pp voor het huis] *[PP voor het] for the house for it *[ PP het huis voor] *[ PP het voor] [ PP er voor]

20

Henk van Riemsdijk

Furthermore, er moves leftward (out of the prepositional phrase) to a position that can be loosely characterized as "to the right of the subject". This is illustrated in (52). (52)

Waarom hebben zij er,· nog niet veel [pp why have they there (it) yet not much

voor] gekocht? for bought

This type of movement cannot be easily identified with Α-movement because the target position is not an argument position, at least not in any obvious sense. On the other hand, it clearly differs from Wh-movement in that the target position is quite different. Taking these various properties together, it is perhaps fair to say that r-movement, as the process has come to be called, shares many properties with clitic movement, except that there does not seem to be a host. On this view, then, er, and other Germanic weak pronouns like it, might well be called clitics.30 For the sake of clarity of exposition, let us call these hostless clitics (HLCs). But now the question whether the clitic is a head or a phrase arises again in full force. Note in particular that we cannot say that er is adjoined to P° at the first stage of the derivation and that hence it has to be a head at every subsequent step of the derivation as well. This is so because even in its pre-prepositional slot inside the PP it can be separated from the preposition by a modifying element, as in (53). (53)

[pp er vlak na] [pp er pal achter] there right after there right behind

At this point both a phrasal movement analysis and a head movement analysis is imaginable. The most plausible line of reasoning in the former case is to assimilate the syntax of HLCs to scrambling phenomena in general. On this view, the placement of HLCs would be a special case of the free word order phenomena in the "middle field" of the clause, a phenomenon found in many languages and particularly common in head-final structures.31 On the second type of approach, the strategy would be to argue that while there is no overt host, the host is actually covert. Such an analysis rests heavily on the assumptions that the verbal projection is richly articulated with separate positions for such morphosyntactic categories as object agreement, subject agreement, tense, negation, etc. Such positions are identified in current generative theory as functional heads. Assuming that even in head-final structures the relevant functional heads may nevertheless be initial, we might then say that clitics of the type in question are true clitics in that they are heads that undergo head movement and are adjoined to the corresponding functional (but invisible) head. An analysis of Germanic clitics along these lines is proposed in Zwart (1993).32

Clitics: A state-of-the-art report

21

Observe finally that mixed analyses may also be appropriate. Assume, for example, that weak subject and object pronouns in Dutch and German, whose surface positions are essentially those illustrated for er in the above discussion, are analysed in terms of scrambling, that is, in terms of phrasal movement. An example is given in (54a). However, in German the clitics in question may also occur in the pre-subject position, as shown in (54b). (54)

a.

b.

Ich glaube, dass der Hans esj unbedingt seiner I believe that the John it absolutely (to-)his Frau e ; sagen sollte. wife say should Ich glaube, dass es; der Hans unbedingt seiner Frau e; sagen sollte.

There are indications that in the latter case the clitic is adjoined to an overt host, viz. the complementizer. That is, we may assume that the clitic is head-adjoined to C°. Evidence comes from the fact that presubject clitics are excluded from complementizerless clauses as shown in (55).33 (55)

*Ich glaube, es der Hans sollte unbedingt seiner Frau sagen

If this is correct, then the analysis of an example like (54b) would involve two dependencies, two chains, one involving the phrase, the other involving the head, as shown in (56).34 (56)

... dass [ N es], der Hans [NP [N e]j]; unbedingt seiner Frau [NP e]j sagen sollte

We may thus conclude that the question whether clitics are heads or phrases depends crucially on the legitimacy of the inclusion of HLCs in the category of clitics.

3.3.

The domain of the clitic-source dependency

One of the main questions that both a movement and a base-generation analysis must address is that of the (constraints on) the distance between the clitic and the corresponding source position. As discussed above (section 2.2.), the two positions must be within the same minimal clause, modulo clitic climbing. As a prelude to any attempt at solving this problem, it would be good to know whether clitic movement is an instance of movement to an argument position (Aposition) or to a non-argument position (Ä-position). The terminology seems to suggest a more straightforward answer to this question than is actually warranted.

22

Henk van Riemsdijk

In the preceding section (3.2) we have distinguished two plausible subcases of clitic movement: (a) head movement to some X°-host, and (b) a scrambling-like phrasal movement. Neither of these is easily interpretable within the state of the art framework that was dominant in the mid eighties, viz. the Government and Binding theory of Chomsky (1981). Even Romance clitics were classified in terms of the standard classification of noun phrases, even though, as we have seen, they are presumably to be analyzed as heads. In most of the representative work such as Borer (1983) or Jaeggli (1986), the clitic itself would be characterized as a pronoun ([+pronominal, -anaphoric]) or, in the case of reflexive clitics, as an anaphor ([-pronominal, +anaphoric]). The source position, being lexically empty (at least in the nondoubling case) would then have to be a variable, an NP-trace, PRO or pro. On standard assumptions, it cannot be a variable because it is not bound by an operator-like element.35 Furthermore, it cannot be an NP-trace since it is not argument bound. Clearly PRO is not an available option either since the source position, being a case position, is governed. The idea, then, was that it is pro.36 Under standard assumptions, pro would have to be licensed in some manner.37 On the one hand it must be governed, which it is. On the other hand its features (the specification for person, number, gender and case, i.e. its so-called φ-features) must be identified by coindexation with the antecendent clitic. None of this would imply a severe locality restriction on the clitic-source relation unless the clitic is forced to take the governor as its host. Only in the latter case will the clitic-source relation be subject to government, which would be sufficient to exclude an unbounded dependency. But in many cases the clitic is attached to some verbal element which is higher than the governor of its pro as in the Italian example (50a), repeated here as (57a), or in the French example (57b). (57)

a. b.

Loi voglio fare [e]j it I-want do Jean nousi aurait aides [e}j John us would-have helped

In both cases, the source position is governed by the closest verb, but the clitic is attached to some higher verb or auxiliary. Similarly, in French a clitic complement to an adjective will be attached not to the governing adjective itself but to the verbal part of the predicative expression, as shown in (58). (58)

Jean met reste [AP fidele Jean to-me remains faithful

[e]; ]

Clitics: A state-of-the-art report

23

This is where the theory of head movement, developed since the mid eighties,38 would be expected to offer some solution. The most influential idea is that head movement is subject to a very strict locality restriction called the Head Movement Constraint (HMC), due to Travis (1984) and further developed in Chomsky (1986) and Baker (1988). This constraint says, essentially, that a head that is moved to some other head position cannot skip any intermediate head positions. Indeed, Kayne (1989) has argued that the HMC could be invoked to account, among other things, for the absence of clitic climbing in French and for the fact that clitics cannot skip negation in cases like (59) (in Italian). (59)

a.

Gianni non li vuole vedere. John not them wants to-see b. *Gianni li vuole non vedere

It turns out, however, that the HMC is far from problematic. The very examples in (57) illustrate one case in which, apparently, a head is skipped. As phenomena of clitic climbing are well attested, ancillary devices have to be relied upon to make intervening heads skippable. It is likely, however, that not all cases of HMC-violations can be successfully argued away, as has been extensively discussed in Haverkort (1993).39 Therefore the issue of the locality domain for clitic movement, like most issues concerning clitics, is very much a matter of current debate, as illustrated, in fact, by much of the material in the present volume. Let us now turn briefly to the other subcase of clitic movement, scramblinglike phrasal movement as found, apparently, mainly in the Germanic and Slavic languages. Here again we first have to address the question whether the resulting dependency is and Α-dependency or an Ä-dependency. And again we are in the middle of a major unresolved debate.40 If scrambling is the result of movement, a matter which is itself open to discussion, then at first sight Α-movement is not an option since movement is not to any position that could easily be identified as an argument position. On the other hand, movement does not result in an operator-variable structure in any obvious way either. If anything scrambling seems to share some properties of both. The paradox can be illustrated in examples like (60) from Dutch.41 (60)

a. b.

Hij he Hij he

heeft dat boekt [zonder ec( te lezen] e, teruggebracht. has that book without reading returned heeft dat boek{ [zonder [eri in] te lezen] ei teruggebracht. has that book without there-in reading returned

The direct object (dat boek) appears to have been scrambled since it occurs to the left of an adjunct. In (60a) it licences a parasitic gap in the adjunct, which

24

Henk van Riemsdijk

suggests that the movement is of the Ä type. But if that were so, we would expect to find a weak crossover violation in (60b). The fact that we do not find such a violation in turn would indicate that the movement results in an A-dependency. The two properties can be combined in a single example as well:42 (61)

ΊPeter hat jeden Gastt [ohne ec, anzuschauen] seinemi Peter has every guest without looking-at (to) his Nachbarn e(· vorgestellt. neighbor introduced 'Peter has introduced every guest to his neighbor without looking at (him).'

The scrambled NP would have to be both an Ä-binder to licence the parasitic gap and an Α-binder to avoid the weak crossover violation. Similarly the scrambled NP in (62) has a paradoxical status because it must have Ä-status to licence the parasitic gap and Α-status to bind the anaphor. (62)

Peter hat die Gäste,· [ohne ec; anzuschauen] einanderi et Peter has the guests without looking-at (to) each other vorgestellt. introduced 'Peter has introduced the guests to each other without looking at (them).'

Webelhuth (1989) concludes from this state of affairs that scrambling dependencies constitute chains that have both A- and Ä-properties, in other words, that there is a third type of dependency. The issue is entirely open at this point. While it is clear that phrasal clitic movement is just as clause-bound as scrambling, it is far from clear how the domain restriction on these processes should be accounted for as long as the A vs. Ä issue remains unresolved. We have again reached a point at which further discussion would go well beyond the scope of a state of the art report and will therefore conclude the presentation here.

4.

Concluding remark

Unavoidably, many aspects of the theory of clitics have not been touched upon in this brief overview. And as announced in the introduction, many of the questions that arise in connection with the interpretation of clitic phenomena are entirely open. It is this state of affairs, in fact, which, together with the challenge that the syntax of clitics provides for theoretical typology, led to the decision that clitics should be one of the nine topics singled out for special study in the ESF EUROTYP project. The research carried out in this project has shed some light on some of these questions. At the same time it has led to many others. Such is the nature of scientific enquiry and progress.

Clitics: A state-of-the-art report

25

Notes 1. 2.

3. 4. 5.

6. 7.

8.

9. 10. 11. 12. 13. 14.

15. 16. 17. 18. 19. 20. 21. 22. 23.

I would like to thank Joe Emonds, Lars Hellan and Edwin Williams for intellectual and moral support and Anna Cardinaletti for some useful comments. My reference is from Wegmüller (1993: 14), who cites Klavans (1982: 1) and Nespor and Vogel (1986: 145). Further references listed by Wegmüller include Zwicky (1977: 1) and (1984: 148-150), Klavans (1979: 69) and (1982: ix, 1, 3, 25,62-63), Crystal (1985: 51), Nevis (1985: 2, 35), Berendsen (1986: 1-2), Jaeggli (1986: 15) and Bauer (1988: 99, 239). See Kayne (1975: 77ff). Most of the examples, which are from French, are borrowed from there, either verbatim or in a somewhat adapted version. The participle in (lc) shows agreement with the clitic. This is not relevant in the present context. This fact should most likely be related to the "no stress" condition in (vi) below. That this is so can be inferred from the fact that unstressed pronouns in English have the same property. (i) John saw Mary and SHE HIM/*...and she him On er, see among many other references, Van Riemsdijk (1978) and Bennis (1986). Kayne does discuss certain potential counterexamples such as Je I 'ai vu qui courait ä toute vitesse ( Ί him saw that/who ran at full speed' = Ί saw him running at full speed'). Cf. Kayne (1975: 126). As Kayne points ou,t this construction is subject to many rather idiosyncratic constraints and can be ignored in the present discussion. The term seems to be due to Aissen and Perlmutter (1976), which appeared in a revised and expanded version as Aissen and Perlmutter (1983). However, the term may go back as far as Rivas (1974) "Impersonal sentences and their interaction with clitic movement in Spanish", an unpublished ΜΓΓ paper that is not accessible to me. See among many others Aissen (1974), Evers (1975), Zubizarreta (1985). This means, in fact, that there is no grammatical way to express such a sentence, unless one of the two clitics is replaced by the corresponding strong form. This was pointed out to me by Anna Cardinaletti. These examples are from Renzi (1988: 588) in chapter XI,2 on clitic pronouns, written by Andrea Calabrese. Anna Cardinaletti informs me, for example, that she rejects both. It should be pointed out that the rule of intervocalic s voicing also fails to apply in many morpheme combinations that are standardly considered to be affixal, cf. (i) a-sociale *[z] [s] ultra-sensibile *[z] [s] asocial super-sensitive These examples are also from Nespor and Vogel (1986: 127-128). The examples are from Scalise (1984: 84). It must be noted, however, that the acceptability of such examples is subject to considerable variability. The examples are again from Scalise (1984). For more discussion, see also Berendsen (1986). This is a simplified version of Table 1. from Nespor and Vogel (1986: 16). This is a 1980 dissertation, which was distributed by the Indiana University Linguistics Club in 1982 and which is now to appear in the Garland outstanding dissertations in linguistics series. There are also cases in which the clitic follows the syntactic host. There is an error in the gloss of Klavans' example (29) which I have corrected in the text rendering of her example. It is not possible, within the context of the present article, to provide full explanations of the theoretical terms and considerations that are mentioned in the text. For some background information, the reader is referred to Van Riemsdijk and Williams (1986) and Haegeman (1991).

26

Henk van Riemsdijk

24. Chomsky (1981) restricts the use of the notion chain to the result of movement, but in later work including Chomsky (1986) and Rizzi (1986) the concept is extended to dependencies that do not result from movement. 25. Α-dependencies are those found between two argument positions, for example the relationship between the subject and the object positions in the passive construction, or the relation between the matrix subject position and the embedded subject position in raising constructions. Ädependencies are those between a non-argument position and an argument position as illustrated, for example, by Wh-movement in questions or relative clauses. 26. Note that the issue of base-generation is independent of whether the clitic-verb combination is a word-level-type unit or whether instead they are syntactically independent but juxtaposed elements, for example a verbal head with an adjacent functional head. On the latter view, the closer association between the host and the clitic could be the consequence of some additional process which in turn could be syntactic or phonological in nature. 27. These examples are from Borer (1986: 4). 28. Under this version of the theory, movement is often said to be structure preserving. I avoid that term here because it leads to confusion with the theory of structure preserving movement of Emonds (1976). Under the latter theory, movement is structure preserving because NPs move to NP positions, verbs move to verb positions, etc. In other words, this theory is essentially about the preservation of categorial identity. Chomsky's proposal is about the preservation of X-bar level status: phrases adjoin to phrases, heads to heads. It would have been preferable to use the term level preservation for the latter idea. The two theories are not incompatible, but they are conceptually independent of each other. The constraint on movement (phrases move to phrasal positions and heads move to head positions) entails that movement may only affect heads and (maximal) phrases. Note that it is an open question whether movement itself is constrained in this way, or whether movement is free but the output of movement is subject to a well-formedness condition that filters out "dangling" non-maximal phrases, as suggested in Van Riemsdijk (1989). 29. See Van Riemsdijk (1978) for an extensive analysis. The discussion in the text will be limited to prepositional er, one of four different subtypes of the same word. 30. In their contribution to the present volume, Cardinaletti and Starke propose that there is three-way distinction rather than a two-way distinction: strong elements-weak elements-clitics. 31. See Corver and Van Riemsdijk (1994) for relevant materials. 32. Zwart's analysis limits itself to subject, direct object and indirect object clitics. It is not entirely clear if such an analysis can be extended to r-clitics of the type discussed in the text.. 33. The word order in such complementizerless complement clauses is verb second. If the position of the verb is rigidly second, then the order would be "...es sollte der Hans..." rather than the order given in the text example. However, both orders are entirely ungrammatical. 34. In a framework that makes use of functional heads (in addition to lexical heads), the status of nominal clitics may well be that of D(P) (determiner) rather than N(P). This complication has been disregarded in the text. 35. But see Aoun (1985) for an elaboration of the notions of variable and Ä-binding in which the clitic-source relation is straightforwardly characterizable as an Ä-dependency. 36. See in particular Sportiche (1983) for insightful discussion of these options and their interpretation in terms of the choice between Α-movement and Ä-movement. 37. See Chomsky (1982) for some pertinent proposals. 38. Cf. particularly Koopman (1984), Baker (1988) and much subsequent work. 39. One way of avoiding the problems that arise from the interaction between citic movement and the HMC is adopt the proposal in Chomsky (1986) to derive the HMC from the Empty Category Principle (ECP). Cases of apparent violations of the HMC would then be those in which the ECP is nevertheless obeyed. See Ouhalla (1988, 1991), Lema and Rivero (1990), Roberts (1991) and Dobrovie-Sorin (1993) for proposals along these lines, and Haverkort (1993) for a critique of

Clitics: A state-of-the-art report

27

these approaches as well as for yet another alternative in terms of the notion of relativized minimality as developed in Rizzi (1990). These various proposals go well beyond what can reasonably be called the state of the art. Furthermore their discussion would require a highly technical expos£. Therefore we will not go into these matters any further here. 40. The reader is referred to Corver and Van Riemsdijk (1994) for an introduction to the main issues and for a series of articles addressing these various issues. 41. See Huybregts and Van Riemsdijk (1985) for discussion. 42. This German example, like the next one, is from Webelhuth (1989). Both examples are cited from the introduction (p. 9) of Corver and Van Riemsdijk.

References Aissen, J.L. 1974 "Verb Raising", Linguistic Inquiry 5: 325-366. Aissen, J.L. and D.M. Perlmutter 1983 "Clause Reduction in Spanish", in: D.M. Perlmutter (ed.), Studies in Relational Grammar I. Chicago and London: The University of Chicago Press, 360-403. [Earlier version in the proceedings of the second annual conference of the Berkeley Linguistic Society, 1976] Aoun, J. 1985 A Grammar of Anaphora. Cambridge, ΜΑ: ΜΓΓ Press. Baker, M.C. 1988 Incorporation. A Theory of Grammatical Function Changing. Chicago and London: The University of Chicago Press. Bauer, L. 1988 Introducing Linguistic Morphology. Edinburgh: Edinburgh University Press. Bennis, H. 1986 Gaps and Dummies. Dordrecht: Foris. Berendsen, E. The Phonology of Cliticization. Dordrecht: Foris. 1986 Borer, H. 1984 Parametric Syntax. Dordrecht: Foris. Borer, H. (ed.) 1986 The Syntax of Pronominal Clitics. Syntax and Semantics Vol. 19. Orlando: Academic Press. Burzio, L. Italian Syntax: A Government-Binding Approach. Dordrecht: Reidel. 1986 Cardinaletti, A. and M. Starke 1998 "The Typology of Structural Deficiency: On the Three Grammatical Classes", In the present volume. Carstairs, A. 1981 Notes on Affixes, Clitics and Paradigms. Bloomington: Indiana University Linguistics Club. Chomsky, N. 1981 Lectures on Government and Binding. Dordrecht: Foris. 1982 Concepts and Consequences of the Theory of Government and Binding. Cambridge, MA: ΜΓΓ Press. Barriers. Cambridge, ΜΑ: ΜΓΓ Press. 1986

28

Henk van Riemsdijk

Corver, Ν. and H.C. van Riemsdijk (eds.) 1994 Studies on Scrambling. Berlin: Mouton deGruyter. Crystal, D. 1985 A Dictionary of Linguistics and Phonetics. Oxford: Blackwell. Di Sciullo, A.M. and E.S. Williams 1987 On the Definition of Word. Cambridge, MA: MIT Press. Dobrovie-Sorin, C. 1993 The Syntax of Romanian: Comparative Studies in Romance. Berlin: Mouton de Gruyter. Emonds, I.E. 1976 A Transformational Approach to English Syntax. New York: Academic Press. Evers, A. 1975 The Transformational Cycle in Dutch and German. Ph.D. dissertation, University of Utrecht. Bloomington: Indiana University Linguistics Club. Haegeman, L. 1991 Introduction to Government and Binding Theory. Oxford: Blackwell. Haverkort, M. 1993 Clitics andParametrization: Case Studies in the Interaction of Head Movement Phenomena. Ph.D. dissertation, Tilburg University. Distributed as ESF EUROTYP Working Papers VIII, 2. Huybregts, M.A.C. and H.C. van Riemsdijk 1985 "Parasitic Gaps and ATB", Proceedings of the 15th Annual Meeting of the North Eastern Linguistic Society (NELS). Amherst: GLSA. Jaeggli, O. 1982 Topics in Romance Syntax. Dordrecht: Foris. 1986 "Case, Doubled NPs, and Extraction", in: H. Borer (ed.), 15—42. Kayne, R.S. 1975 French Syntax. The Transformational Cycle. Cambridge, ΜΑ: ΜΓΓ Press. 1989 "Null Subjects and Clitic Climbing", in: O. Jaeggli and K. Safir (eds.), The Null Subject Parameter. Dordrecht: Reidel, 239-261. Klavans, J.L. 1979 "On Clitics as Words", in: Proceedings of the Annual Meeting of the Chicago Linguistics Society 15: 68-80. 1982 Some Problems in a Theory of Clitics. Bloomington: Indiana University Linguistics Club. 1985 'The Independence of Syntax and Phonology in Cliticization", Language 61: 95-120. Koopman, H. 1984 The Syntax of Verbs. Dordrecht: Foris. Lema, J. and M.-L. Rivero 1990 "Long Head Movement: ECP vs. HMC", in: Proceedings of the 20th Annual Meeting of the North Eastern Linguistic Society (NELS). Amherst: GLSA. Muysken, P.C. 1981 "Quechua Word Structure", in: F. Heny (ed.), Binding and Filtering. London: Croom Helm, 279-327. Nespor, M. and I. Vogel 1986 Prosodic Phonology. Dordrecht: Foris. Nevis, J.A. 1985 Finnish Particle Clitics and General Clitic Theory. Ph.D. dissertation, Ohio State University.

Clitics: A state-of-the-art report Ouhalla, J. 1988

29

The Syntax of Head Movement: A Study ofBerber. Ph.D. dissertation, University College London. Functional Categories and Parametric Variation. London: Routledge.

1991 Renzi, L. 1988 Grande grammatica Italiana di consultazione, Vol. I. Bologna: II Mulino. Riemsdijk, H.C. van 1978 A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Dordrecht: Foris. Riemsdijk, H.C. van 1989 "Movement and Regeneration", in: P. Benincä (ed.), Dialect Variation and the Theory of Grammar. Dordrecht: Foris, 105-136. Riemsdijk, H.C. van and E.S. Williams 1986 Introduction to the Theory of Grammar. Cambridge, MA: MIT Press. Rivas, A. 1974 Impersonal Sentences and their Interaction with Clitic Movement in Spanish. [Unpublished paper, ΜΓΓ.] A Theory of Clitics. Ph.D. Dissertation, ΜΓΓ, Cambridge, Massachusetts. 1977 Rizzi, L. "A Restructuring Rule in Italian Syntax", in: S.J. Keyser (ed.), Recent Transforma1978 tional Studies in European Languages. Cambridge, ΜΑ: ΜΓΓ Press, 113 - 158. Reproduced in L. Rizzi (1982). Issues in Italian Syntax. Dordrecht: Foris. "On Chain Formation", in: H. Borer (ed.), 65-95. 1986 Relativized Minimality. Cambridge, ΜΑ: ΜΓΓ Press. 1990 Roberts, I. 1991 Verbs and Diachronic Syntax. Dordrecht: Reidel. Sapir, E. 1930 "Southern Paiute, a Shoshonean Language", Proceedings of the American Academy of Arts and Sciences 65,1. Scalise, S. 1984 Generative Morphology. Dordrecht: Foris. Strozer, J. 1976 Clitics in Spanish. Ph.D. Dissertation, UCLA, Los Angeles. Toman, J. 1985 "A Discussion of Coordination and Word-Syntax", in: J. Toman (ed.), Studies in German Grammar. Dordrecht: Foris, 407-432. Travis, L. 1984 Parameters and Effects of Word Order Variation. Ph.D. dissertation, ΜΓΓ. Wackernagel, J. 1892 "Über ein Gesetz der Indo-Germanischen Wortstellung", Indogermanische Forschungen 1: 333-436. Webelhuth, G. 1989 Syntactic Saturation Phenomena and the Modern Germanic Languages. Ph.D. dissertation, University of Massachusetts, Amherst. Wegmüller, U. 1993 Klitika im Vicentino. Universität Bem, Arbeitspapier 31. Zubizarreta, M.L. 1985 "The Relation between Morphophonology and Morphosyntax: The Case of Romance Causatives", Linguistic Inquiry 16: 247-289. Zwart, C J.W. 1993 Dutch Syntax: A Minimalist Approach. Ph.D. dissertation, University of Groningen.

30

Henk van Riemsdijk

Zwicky, A.M. 1977 On Clitics. Bloomington: Indiana University Linguistics Club. 1984 "Clitics and Particles", The Ohio State Working Papers in Linguistics 29: 48-173.

Part I

Area studies

Anna Cardinaletti

Pronouns in Germanic and Romance Languages: An overview1

1.

Introduction

What follows is a survey of the progress to be found in research on Germanic and Romance (personal) pronouns made in the past two decades. The comparative approach characterizing recent work in Generative Grammar has shifted attention from the Romance languages, largely studied in the seventies and beginning of the eighties, to the Germanic languages. As is often the case in comparative research, this has been fruitful in two respects: On the empirical level, we now have a much richer description of pronouns in the Germanic languages due to notions 'imported' from Romance, and, in turn, a better description of some apparently idiosyncratic Romance pronouns thanks to some observations made on the basis of Germanic; on the theoretical level, a new and broader understanding of already well-known phenomena has been acquired. The pronominal systems of Germanic and Romance languages differ in one fundamental respect: All Romance languages have morphologically distinct series of pronouns, whereas the morphology of Germanic pronouns is often opaque. In Romance languages, the morphologically reduced series, known as clitic, or atonic, or conjoined, has a 'special' syntax, whose main characteristic is the displacement to a position adjacent to a verbal form; the other series, called strong, or tonic, or disjoined, essentially behaves like noun phrases. In spite of the poverty of morphological evidence, and of the existence of displacement rules which also apply to noun phrases, pronouns which display a 'special' syntax, different from strong pronouns and full noun phrases, have been claimed to exist in the Germanic languages as well. These pronouns have been referred to as 'weak' (following Roster's 1978 and Holmberg's 1986, 1991 terminology), and have been taken to be the Germanic counterpart of Romance clitics. The morphological distinctness of the two series in Romance languages versus the non-distinctness in Germanic languages has probably been the cause of the different history of research in the two groups of languages. Romance pronouns have been intensively studied in the past two decades (also thanks to the seminal work by Kayne 1975). Germanic pronouns have been systematically studied only in the past five, six years, and the project of which this volume reports the main results has given a great impulse to their study. This asymmetry reflects a difference already attested in traditional grammar. Ever since the earliest grammatical descriptions on Romance languages (e.g.

34

Anna Cardinaletti

Port-Royal for French, cf. Arnauld and Lancelot 1660), two series of pronouns have been reported. On the other hand, Germanic descriptive grammars merely notice the existence of 'exceptional' cases, which have, however, not been thoroughly discussed until most recent years. German grammars, for instance, limit themselves to the observation that the 3rd person neuter pronoun es 'it' has some idiosyncratic properties, which makes it different from the other pronouns of the language.2 As we will see, there is nothing idiosyncratic about this pronoun. It reflects the properties of a specific class of pronouns attested across languages. The thorough study of Germanic pronouns and the parallel study of some 'exceptional' Romance pronouns, neither clitic nor strong, have in fact led to the proposal that the class of weak pronouns is distinct from both clitic and strong pronouns. This implies that natural languages possess not two, but three classes of pronouns (cf. Cardinaletti and Starke, this volume). This discovery has consequences not only for the descriptive analysis of Germanic and Romance pronouns, but also for the more general aim at explaining the properties of the pronominal system of natural languages in general.

2.

Romance languages

One of the most influential work on Romance personal pronouns in the framework of Generative Grammar is Kayne (1975). The recognition of two morphologically and syntactically distinct series of pronouns, clitic and strong pronouns, has led to the elaboration of a number of tests for clitichood, which have influenced all subsequent research on pronouns. In particular, clitic pronouns cannot be modified, conjoined, contrastively stressed or used in isolation, as shown by the following Italian examples:3 (1)

a.

Maria conosce solo lui. Maria knows only him b. *Maria lo conosce solo. c. Accuseranno loro stessi. [they] will-accuse them themselves d. *Li stessi accuseranno.

(2)

a.

Maria conosce [lui e voi], Maria knows him and you. *Maria [lo e vi] conosce. b.

Pronouns in Germanic and Romance languages: An overview (3)

Maria conosce LUI, non voi. Maria knows him not you b. *Maria LO conosce, non voi.

(4)

a. Chi conosce, Maria? Lui. b. *Chi conosce, Maria? Lo. 'Who does Mary know? Him.'

35

a.

Strong pronouns appear in all syntactic positions available to full noun phrases, in particular they follow the verb in declarative sentences. In the same type of clauses, clitic pronouns cannot remain in the postverbal object position, but are obligatorily displaced to a position immediately to the left of the finite verb, a position not available to strong pronouns and full noun phrases: (5)

a. b.

Maria conosce *ci/noi/Gianni. Maria ci/*noi/*Gianni conosce. Maria us/us/Gianni knows us/us/Gianni

Finally, (3rd person) clitic pronouns can refer to non-human entities, a possibility excluded for strong personal pronouns. In e.g. French, a verb that implies a nonhuman object, such as acheter 'buy', gives ungrammatical results with a strong pronoun, and is only compatible with a clitic pronoun:4 (6)

a. b.

Je I Je I

ne connais/*achete que lui. not know/buy [other] than him l'ai connu/achete. him/it have known/bought

As the simple paradigms in (l)-(6) show, a cluster of properties, morphological, syntactic, and semantic, distinguish clitic from strong pronouns. This is always the case in the Romance languages. Apart from establishing the distinction between clitic and strong pronouns, Kayne's work has inspired a number of studies in the eighties focussed on three main topics: How to establish the host of clitic pronouns, whether clitic-placement is an instance of a movement rule or not, and how to analyse subject cliticlike pronouns.

2.1. The host In most Romance languages, clitic pronouns precede the finite verb in declarative sentences. This has been taken as a manifestation of a tight relation between

36

Anna Cardinaletti

clitic pronouns and the finite verb. It is however useful to underline that the privileged relationship of the clitic pronoun with the verb is a characteristic of Romance languages, not shared by clitic pronouns in other languages (such as Slavic and, as we will see, Germanic languages; cf. Wackernagel 1892). In (most) Slavic languages, for instance, the clitic pronoun always appears in the second position of the clause, independently of the position of the verb (Croatian examples from Cavar and Wilder 1992): (7)

a. b. c.

(8)

Ivan ga je cesto citao. Ivan it has often read Cesto ga je Ivan citao. Citao ga je Ivan cesto.

... da ga Ivan cita. ... that it Ivan reads

Following Kayne's (1991) proposal regarding the host of clitic pronouns, it is possible to express these observations in a general fashion. Kayne has proposed that clitic pronouns are left-adjoined to functional heads: 5

In Romance languages, the clitic pronoun therefore is characterized by a privileged relation not with the verb itself, but with the functional head Infi (or AgrS, cf. Belletti 1990) containing the verb. Kayne's (1991) proposal has a number of important consequences. First, adjunction to Infi (or AgrS) as in Italian is just one particular case of a more general pattern. Language variation in clitic placement can be therefore understood in terms of the different functional heads which host the clitic pronoun. For instance, the difference between Romance languages such as Italian and French on the one hand and Slavic languages such as Czech, Slovak and Croatian on the other can lie in the fact that the former use AgrS as the clitic host, whereas the latter use AgrC, a structurally higher head which does not contain the verb (cf. Starke 1993; cf. also Rivero 1994):

Pronouns in Germanic and Romance languages: An overview

37

b.

This type of language variation is also found Romance-internally. Whereas AgrS seems to host pronominal clitics in e.g. Italian and French, a head higher than AgrS is apparently involved as the clitic host in some Spanish varieties discussed by Uriagereka (1994, 1995), and in Portuguese. For the latter, either C (cf. Madeira 1992) or a functional head intermediate between Agr and C, probably corresponding to the AgrC individuated in Slavic (cf. Rouveret 1992), have been proposed. Second, since the clitic is not adjoined to the verb itself, language variation can also ensue from the different scope of verb raising in different languages (cf. Kayne 1991, Motapanyane 1991, Rizzi 1993 among others). For instance, literary French differs from Italian in that a clitic pronoun can be separated from the infinitival verb. The verb stops in a low position and the pronoun raises, independently of the verb, to the same position as in finite clauses (cf. Kayne 1991:653, note 18):

38 (11)

Anna Cardinaletti a. b. c.

II en parle fort bien. he of-it speaks strong well ...en fort bien parier. of-it strong well [to] speak ... parlar-ne bene. ... [to] speak of-it well

It is evident that the distinction of a number of functional heads in the clause (cf. Pollock 1989, Chomsky 1989, Belletti 1990, Cinque 1995), gives a potentially broad range of language variation, still to be fully evaluated. Another consequence of Kayne's proposal concerns the status of clitic pronouns. Given current assumptions on adjunction, his proposal implies that clitic pronouns are heads (cf. Kayne 1975:81ff, Baltin 1982:4). This view is supported by the observation that the clitic and its host undergo movement as a unit. In French interrogatives and in Italian hypotheticals, for instance, the verb moves to a head higher than the position of the subject. In the presence of a clitic pronoun, both the clitic and the verb will precede the subject: (12)

a. b

Tu I'as [L'as] tu

vu. vu?

you him-have seen (13)

a. b.

Se Gianni I'avesse programmato in anticipo,... [L'avesse] Gianni programmato in anticipo,... if

2.2.

Gianni it-had

programmed ahead,...

Enclisis versus proclisis

In Romance languages, the position of clitic pronouns depends on the finiteness of the verb and on sentence type. With infinitival verbs, Italian clitic pronouns are always enclitic; with finite verbs, they appear in enclitic position only in iussive clauses: (14)

a. b.

Avendolo giä letto, Gianni decise di non comprarlo. having-it already read, Gianni decided to not buy-it Leggilo! read-it

Given the assumption that clitic pronouns always adjoin to the left of functional heads, as in (9) above, enclisis is reduced to proclisis. Kayne (1991) suggests that

Pronouns in Germanic and Romance languages: An overview

39

in infinitival clauses, the 'verb-clitic' order is obtained by movement of the verb across the clitic, to a position adjoined to the F' (or FP) dominating the clitic: 6

However, as pointed out in Benincä and Cinque (1993), proclisis and enclisis appear to have partially different properties, the latter displaying a tighter rather than a looser relation between the clitic pronoun and the verb. First, Benincä and Cinque report that under special prosodic and pragmatic conditions, some speakers of French and Rumanian accept coordination of two clitic pronouns in proclisis contexts, but never in enclisis contexts, (16)—(17). In Italian, the same contrast is found with respect to disjunction of clitics, which is marginal with proclitic pronouns but ungrammatical with enclitic ones, (18) (cf. Rizzi 1993): (16)

a.

Je lui et vous ferais un plaisir. I to-him and to-you will-do a favour b. *Ecris-nous et lui! write to-us and to-him

(17)

a.

Imi §i [I] to-me and b. *Dindu-fi giving-to-you

(18)

a.

IChiunque whoever dobbiamo must b. *Chiunque

i(i scrie. to-you wrote j i ii cartea. and to-her book-the

abbia lanciato questo appello, Ιο ο la had sent this appeal we him or her aiutare. help abbia lanciato questo appello, dobbiamo aiutarlo ο la.

Furthermore, in some special conjunctions involving morphologically related verbs, a proclitic pronoun can be left out of the second conjunct, thus preceding the two conjoined verbs, but enclitic pronouns cannot follow two conjoined verbs. Examples are provided for Italian and French, respectively:

40

Anna Cardinaletti

(19)

a.

Lo [leggo e rileggo] in continuazione. [I] it read and re-read continuously b. *per [leggere e rilegger]lo to read and re-readit c. *[Leggi e rileggi]lo! read and re-read it

(20)

a.

Jean le [lit et relit] sans cesse. Jean it reads and re-reads without interruption b. *[Lit et relit]-la! read and re-read-it!

Benincä and Cinque (1993) conclude that in proclisis contexts, adjunction of the clitic to the functional head containing the verb does not create a new word at the morphological level: The two elements, adjacent in the syntax, will combine only in the phonology. Therefore, the constraint on coordination, which requires that both conjuncts are affected by the same rules, is not violated in the cases discussed above. Enclisis, on the other hand, involves more than simple adjacency between the clitic and the verb: The clitic incorporates into the verb in the syntax, and the two form a morphological unit. If incorporation of the clitic takes place out of a conjunction of clitic pronouns, as in (16b), (17b) and (18b), or into a conjoined verb, as in (19b, c) and (20b), the constraint on coordination is violated. 7 This conclusion is confirmed by the following pan-Romance generalization: Proclitic pronouns can appear separated from the verb, as in Old and Renaissance Italian, (21), literary French, (22), and Modern Triestino, (23), but no Romance language allows for lexical material to intervene between the verb and an enclitic pronoun (examples (21) and (22a, b) are taken from Renzi 1989:369; examples (22c) and (23a) from Benincä and Cinque 1993):8 (21)

ma dacche vi pur piace... but since [it] to-you indeed pleases 'But since you do like it ...'

(22)

a. b. c.

(23)

a.

Pourrais-je n'y pas penser? could-I not to-it not think? ä l'idee qu'on a pume mal juger. at the idea that one could me badly judge ... pour le bien faire. ... for it well do No el se gnanca vedi. not he REFL even sees

Pronouns in Germanic and Romance languages: An overview b.

41

El me un poco preocupa. he me a little worries

In other words, the following generalisations hold, which are also valid for Slavic languages (cf. Starke 1993): (24)

a. b.

clitic (X) verb verb (*X) clitic

In order to reconcile Kayne's (1991) left-adjunction hypothesis with Benincä and Cinque's (1993) conclusion that the clitic and the infinitival verb form a constituent, we have to assume that enclisis configurations can only arise via leftadjoining the verb to the functional head containing the clitic, (25a), or, if double adjunction is banned (cf. Kayne 1994:19-22), via left-adjoining the verb to the clitic itself, (25b): 9

2.3.

Clitic-placement: Derivation vs. base-generation

A long debate has concerned the nature of clitic placement: whether due to a transformational rule, as in the original proposal by Kayne (1975), or to basegeneration of the clitic in the derived position. The derivational approach is motivated by the locality effects displayed by clitic placement, typical of movement operations: For instance, a clitic pronoun cannot be extracted out of an adverbial prepositional phrase, nor out of a noun phrase whose highest specifier is filled with a demonstrative, (26). Both facts can be captured if clitic placement is movement. Given that, contrary to a Wh-phrase, a clitic pronoun cannot 'leave' the finite clause to which it belongs, (27), clitic placement must be taken as an instance of A-movement: 10

42 (26)

(27)

Anna Cardinaletti a. *

L'ho parlato dopo . [I] him have spoken after (cf. *Chi hai parlato dopo ?) whom have [you] spoken after? b. *Ne ho vista questafoto . [I] of-him have seen this picture (cf. *Di chi hai visto questafoto of whom have [you] seen this picture? a. * Lo penso [che vedrd ]. [I] him think that [I] will-see (cf. Chi pensi [che vedrai whom [you] think that [you] will-see? b. Penso [che lo vedrd ]. [I] think that [I] him will-see

?)

] ?)

The derivational approach is also supported by the observation that clitic pronouns display the same morphological form as material generated DPinternally: either determiners, in e.g. Romance languages, or case morphemes, in e.g. Germanic and Slavic languages (cf. Cardinaletti 1994:199): (28)

a. b.

La conosco. [I] her know Ichhab's gelesen. I

c.

a' la ragazza the girl b' das Buch

have it read

Vide I ho. [he] saw

him-ACC

the

c'

(Italian) (German)

book

velkeho muze

(Czech)

[the] big-ACC man-ACC

This observation suggests that a clitic pronoun is one of the highest functional heads of an otherwise empty DP generated in the base position, out of which it is extracted at some point of the derivation. In order to reconcile the locality effects typical of Α-movement with the fact that clitic pronouns behave as heads (cf. §2.1 above), clitic placement can in fact be seen as an instance of Α-movement of a maximal projection followed by proper X°-movement (cf. Sportiche 1989). This formulation of clitic placement has proved to have many interesting consequences and is now generally adopted among proponents of the derivational analysis (cf. Cardinaletti and Starke, this volume: §6.3.2 for discussion). It is theoretically forced by Kayne's (1994) framework, which excludes the base-generation of a pure head in complement position (see p. 61), and finds empirical support in the fact that e.g. in Italian, accusative clitic pronouns trigger past-participle agreement:

Pronouns in Germanic and Romance languages: An overview (29)

a. b.

43

Ho letto spesso queste riviste. [I] have read often these m a g a z i n e s , ^ ^ Le ho lette spesso. [I] them reM -PL have r e a d ^ , ^ often

It is generally assumed that (morphological) agreement on a head is triggered by the presence of an element in its specifier position. Given (29), there must be a step in the derivation in which the clitic pronoun moves as an XP to the specifier of the projection in whose head the participle appears. As said above, this in turn implies that the movement of clitic pronouns must be decomposed into two steps of a different nature: The first step is Α-movement of a maximal projection, the second step is true head-to-head movement. On the other hand, the base-generation approach entails that clitic pronouns are generated attached to their host and are linked to an empty category in the canonical position of the noun phrase with the same grammatical function. The clitic is therefore a sort of affix, as argued by Roberge (1990) (Cummins and Roberge 1993 strengthen this hypothesis by suggesting that clitics are inflectional affixes). Although it cannot easily account for the above-mentioned properties of clitic constructions, the base-generation approach finds support in 'reduplication' constructions such as clitic doubling obligatorily found in Rumanian and extremely frequent in Spanish (cf. Dobrovie-Sorin 1990 and Jaeggli 1982, respectively): (30)

a. b.

L'am väzut [I] him-have seen Lo vimos a [we] him saw to

pe Jon. ACC Jon Juan. Juan

If both a clitic pronoun and a full noun phrase are present in the clause, it seems trivial that the clitic pronoun cannot originate in the object position, since this is occupied by the noun phrase. The observation, known as Kayne's generalization, that in clitic-doubling constructions, a Case-marker is in general present on the doubled DP (such as pe and a in [30]), has led to the view of clitic pronouns as government/Case absorbers (cf. Jaeggli 1982, Borer 1984)." An attempt at combining the two views is Sportiche's (1992) recent theory of clitics. Clitic pronouns are base generated as heads of functional projections, called Clitic Voices, which are associated with the verb (i.e., they belong to the extended projection of the verb, in the sense of Grimshaw 1991). A null category (pro) is generated in the object position and moved to the specifier of the CliticVoice Phrase, in order to satisfy a requirement of spec-head agreement with the clitic (the so-called Clitic Criterion). This analysis also allows an account of clitic doubling: It is obtained when a noun phrase is generated instead of the null

44

Anna Cardinaletti

category. The following would be the relevant part of the structure of the two Spanish sentences lo vimos and lo vimos a Juan (cf. [30b]):12 (31)

2.4.

CIVoiceP

Subject clitic pronouns

Another research topic very much influenced by Kayne's work is the analysis of subject pronouns. Kayne's (1975) tests for clitichood, seen above in section 2 for object clitic pronouns, also apply to the morphologically reduced series of subject pronouns, as found in French: they cannot be modified, conjoined, contrastively stressed, or used in isolation: (32)

a. b. c. d.

Seul lui/*il viendra demain. only he will-come tomorrow Pierre et lui/*il viendront demain. Pierre and he will-come tomorrow C'est lui/*il qui viendra demain, pas Marie. it is he who will-come tomorrow not Marie Qui viendra demain? Lui/*Il. who will-come tomorrow He

As far as the distribution is concerned, French reduced subject pronouns share with object clitic pronouns the property of being adjacent to the inflected verb, (33a, b), a requirement not operative for strong subject pronouns and for subject noun phrases, (33c):

Pronouns in Germanic and Romance languages: An overview (33)

45

a. *Il, souvent, mange du fromage. he often eats some cheese 'He often eats cheese' b. *Jean le, souvent, mange. Jean it often, eats c. Lui/Jean, souvent, mange du fromage.

If subject cliticization applies as in the case of object pronouns, the clitic pronoun should move from the subject position to the inflectional head containing the verb: (34)

t| I

Iii mange du fromage.

f

The lowering of the pronoun however would lead to the violation of the Empty Category Principle (ECP), which requires that the trace be c-commanded by the moved element and thus forbids any lowering operation. In (34), the clitic pronoun would not c-command its trace. This has led Kayne (1983) to postulate a distinction between syntactic and phonological clitics: Whereas object clitic pronouns are cliticized in the syntax, subject pronouns attach to their verbal host in the phonological component (PF), where no such ECP violation arises. 13 The PF-clitic approach implies that in the syntactic component, French subject pronouns are maximal projections and not heads (cf. Kayne 1983:fn. 27; Rizzi 1986; Couquaux 1986). Although the 'lowering' problem need no longer arise under the VP-internal subject hypothesis (given that the subject pronoun starts inside the VP and can 'raise' to the host on a par with object pronouns, as in the possible derivation in [35]),

(35)

[ip [r Ik mange* [Vp tj tk du fromage]]]. I t

the distinction between French subject and object clitic pronouns has found independent empirical support. French subject pronouns such as il 'he, it' must be distinguished from subject clitic pronouns found in other languages, e.g. the Northern Italian dialects. With respect to a number of properties, the subject clitics of these dialects have the same syntactic status as French object clitic pronouns, i.e. they are, in Kayne's (1983) terms, syntactic clitics, and differ from French reduced subject pronouns. Among other properties, whereas French

46

Anna Cardinaletti

subject pronouns are optional under sentence coordination, the subject pronouns of Northern Italian dialects are necessarily repeated, on a par with object clitics: (36)

a.

II travaille a son article et (il) pense ä ses problemes. he works at his paper and (he) thinks of his problems (French, SCL)

b.

La canta e *(la) bala. she sings and *(she) dances (Trentino,SCL)

c.

Jean le lavera soigneusement et *(le) remettra Jean it will-wash carefully and *(it) will-put back en place. into place (French, OCL)

A number of studies on subject clitic pronouns in Northern Italian dialects have taken part in this discussion, among which Brandi and Cordin (1981), (1989); Safir (1986); Giupponi (1988); Poletto (1991), (1993); Roberts (1991); Saccon (1993). At the same time, work on French dialects and spoken French have demonstrated that these varieties differ from standard French in displaying true clitic subject pronouns (cf. Zribi-Hertz 1994 and the references quoted there). In spoken French for instance (or 'fran9ais avance' since Frei 1929), sentences such as (36a) require the obligatory repetition of the pronoun in the second conjunct, as in (36b, c).

2.5.

The partitive clitic ne

Romance languages have not only accusative and dative clitic pronouns, but also pronouns which correspond to prepositional phrases. See for instance Italian locative ci and vi 'there', and genitive and partitive ne 'of + pronoun'. Among these, partitive ne has been studied at length in the seventies. This pronoun has been particularly important because its distribution is one of the arguments to distinguish two classes of monoargumental verbs: intransitive and ergative verbs (cf. Burzio 1986). In addition, partitive ne has been claimed to have a special status among clitic pronouns, in that it does not start as a maximal projection, but as the intermediate projection N'. Belletti and Rizzi (1981) argue for this analysis of partitive ne on the basis of the fact that in sentences such as (37), the quantifier, assumed to be in specNP, is left in situ under ne-cliticization:

Pronouns in Germanic and Romance languages: An overview (37)

a. b.

47

Ho letto [NP due [N< libri ]]. [I] have read two books Net ho letti [NP due [N< tj ]]. [I] of-them have read two

Furthermore, postnominal adjectives and even arguments of the noun can be stranded by ne-cliticization. This fact seems to support the hypothesis that ne does not pronominalize the whole noun phrase, but corresponds to some smaller projection: (38)

a. b.

(39)

a. b.

Ho letto un libro bello. [I] have read a book nice Ne ho letto uno bello. Maria conosce tre racconti del nonno. Maria knows three stories of the grandpa Maria ne conosce tre del nonno.

In spite of this evidence, Cardinaletti and Giusti (1992) deny the exceptionality of ne and analyse it as linked to a maximal projection, on a par with all other clitics. Their analysis relies on the two following assumptions: (i) the quantifier occupies a position outside of the noun phrase, presumably the head of the projection QP, and (ii) the constituents left in situ by ne-cliticization are predicative elements, generated outside of the noun phrase. Those elements which cannot be stranded, for instance obligatorily prenominal adjectives as in (40) and some types of complements as in (41), are in fact also impossible as predicative modifiers, e.g. in copular sentences: (40)

a.

Si e fatta una mera illusione. [she] REFL is made a mere illusion b. *Si e fatta una illusione mera. c. *Se ne e fatta una mera. d. *Questa illusione e mera. this illusion is mere

(41)

a.

Conosco due esperti in informatica. [I] know two experts in computer-science b. *Ne conosco due in informatica. c. *Questi esperti sono in informatica. these experts are in computer-science

48

Anna Cardinaletti

This analysis is also supported by the fact that, like accusative clitics, partitive ne triggers agreement with the past participle: (42)

Ne ho lette molte. [I] of-them have r e a d p ^ ^ many FEM . PL

Notice that, given the proposal that agreement on past participle is triggered by the clitic pronoun moving through its specifier position (see the discussion around (29) above), the analysis of ne as an N' would imply that an intermediate projection can also occur in a specifier position, clearly an unwelcome consequence. Ne also patterns with the accusative (and dative) clitics of (43) in allowing coreference with an element of the matrix clause, (44), according to Principle Β of Binding theory: (43)

a. b.

(44)

a.

b.

Giannii Gianni Gianni) Gianni

sostiene claims sostiene claims

che that che that

Maria Maria Maria Maria

loi conosce. him knows glij parla spesso. to-him speaks often

Quei ragazzit sostengono che Maria net conosce solo those boys claim that Maria of-them knows only alcuni. some Quegli studenti; temono che Maria nei promuoverä solo those boys are-afraid that Maria of-them will-pass only la metä the half

Other clitic pronouns, on the other hand, cannot be bound by an antecedent in the matrix clause, thus obeying Principle C of Binding theory (cf. Belletti 1993): (45)

a. *Giannij sostiene Gianni claims b. * Gianni j sostiene Gianni claims

che that che that

Maria Maria Maria Maria

ne, of-him cit to-him

parla spesso. speaks often pensa spesso. thinks often

Belletti suggests that these clitic pronouns are categorially different from accusative and dative clitics. Whereas the latter are of category D (the same as determiners), ci and ne are Ps (the same as prepositions). Given (44), partitive clitic ne should thus count as a D.

Pronouns in Germanic and Romance languages: An overview

3.

49

Germanic languages

The recent work on Germanic languages has primarily focussed on the question as to whether a class of pronouns distinct from strong pronouns exists at all.

3.1. One extremist view In a language such as German, there appears to be no evidence for postulating two classes of pronouns. Two morphologically distinct series of pronouns do not exist, and pronoun movement is masked by the existence of generalized movement of noun phrases (scrambling): Every noun phrase can be moved to the left, across a sentential adverb: (46)

a. b.

Ich glaube, daß Maria ihn gestern gesehen hat. t I Ich glaube, daß Maria den Hans gestern gesehen hat. t I I think that Maria him/the Hans yesterday seen has

The parallel behavior of pronouns and full noun phrases in (46) has led some researchers to the claim that for German, it is not necessary to assume any particular rule of pronoun placement: The independently necessary scrambling rule can be applied to pronouns as well (cf. Vikner and Schwartz 1991, Tomaselli and Poletto 1992, Lenerz 1992). This proposal in turn entails an extremist view on pronouns: It implies that German has only one class of pronouns, the class of strong pronouns. As a matter of fact, German pronouns display all hallmarks for strength: they can be modified, coordinated, stressed, used in isolation, etc. But this conclusion faces a couple of difficulties. The first and easier one is the fact that a pronoun such as es 'it' displays clitic-like behavior (cf. Cardinaletti and Starke 1996:§1 for recent discussion). This has been circumvented by attributing it an exceptional status (cf. Haider 1984:3; Travis 1984:120ff; Tomaselli and Poletto 1992:79, and fn.2). The second difficulty is the fact that the scrambling rule is obligatory for unstressed pronouns, while apparently optional for full noun phrases and stressed pronouns (the judgments in (47) hold with gestern in the unmarked position of sentential adverbs): (47)

a. *..., daß Maria b. ..., daß Maria c. ..., daß Maria ..., that Maria

gestern gestern gestern yesterday

ihn gesehen hat. den Hans gesehen hat. IHN gesehen hat. him/the Hans seen has

(cf. [46a]) (cf. [46b])

50

Anna Cardinaletti

Within the view that German has only strong pronouns, no attempt is made to explain these asymmetries, nor the exceptional status of pronouns such as es. These observations allow us to envisage the necessity of postulating two classes of pronouns. We will come back to German in §3.4.

3.2.

Dutch

The existence of two distint classes of pronouns in Germanic languages, 'weak' and 'strong', was first established on the basis of Dutch, which has two morphologically distinct pronominal paradigms. Den Besten (1977) and Köster (1978) each dedicate one section of their work to weak pronouns. However, these authors do not focus on the properties of the pronouns themselves but on some distributional restrictions. The adjacency requirement between the complementizer and the subject pronoun in embedded clauses and between the finite verb and the subject pronoun in verb-second clauses was taken by den Besten as an argument for the analysis of verb-second as movement of the finite verb to C, the position of the complementizer: (48)

a. *..., dat gisteren 'k het boek gelezen heb. b. ...,dat 'k gisteren het boek gelezen heb. ... that I yesterday the book read have

(49)

a. *Morgen kom in ieder geval 'k terug. b. Morgen kom'k terug. tomorrow come (in any case) I back

The second distributional observation is that subject, but not object weak pronouns can occur sentence initially in declarative sentences (verb-second clauses): (50)

a. b.

Ik zag hem. 'k zag hem. I saw him

(51)

a. Mij zag hij. b. *Me zag hij. me saw he

This observation has become central in the verb-second debate, suggesting an asymmetric approach to the verb-second phenomenon (Travis 1984; Zwart 1993).

Pronouns in Germanic and Romance languages: An overview

51

Subject-initial and object-initial verb-second clauses have a different structure: Whereas the latter are CPs, involving 'topicalization' of the object into specCP, the former are AgrSPs. Since topicalization is not a possible option for weak pronouns, object weak pronouns in specCP are ungrammatical. This restriction does not hold true of subject pronouns, which occur in specAgrSP and do not involve any topicalization.14 Van Riemsdijk (1978) also concentrates on the distributional properties of a weak pronoun, er 'there', which obligatorily appears to the left of prepositions as a suppletive form for pronouns with non-human reference: (52)

... omdat hij er op wilde wachten. ... because he there for wanted [to] wait '... because he wanted to wait for it'

The issue of the specific analysis of the reduced forms of Dutch was first addressed in the works by Berendsen (1986) and Jaspers (1989). The clitic-like forms are not productive reductions obtained from the strong forms via phonological rules, but the existence of two distinct classes of pronouns must be assumed. The bipartition found in Romance languages is thus extended to Germanic. A number of semantic and syntactic properties differentiate weak and strong pronouns, which cannot be captured by a purely phonological approach. Among others, (i) generic reference characterizes 2nd person singular and 3rd person plural weak pronouns, and is incompatible with strong ones, which only admit the purely referential interpretation, and (ii) 3rd person weak pronouns admit both [+human] and [-human] referents, whereas strong pronouns are restricted to [+human] ones (examples from Berendsen 1986:38-39): 15 (53)

a. b.

(54)

a.

b.

Za /*Zij zeggen zoveel. they (=people) say a lot Ja /*Jij wil snel te veel. you (=people) want quickly too much Kees Kees zijn have Kees Kees weg. gone

zegt dat says that weg. gone zegt dat says that

de fietsen daar nog staan, maar zs/*zij the bycicles are still there, but they

de jongens daar nog staan, maar zd/zij zijn the boys are still there, but they have

52

Anna Cardinaletti

Given the existence of the rule of scrambling, which affects all definite noun phrases, the distribution of Dutch weak pronouns is not as distinct from strong pronouns and full noun phrases as it is in Romance languages. Although more difficult to detect than in Romance languages, a number of distributional asymmetries between weak and strong pronouns have been put forth (cf. Jaspers 1989). All of them point to the conclusion that weak pronouns are obligatorily displaced. Although they do not appear attached to any verbal form, there is a portion of the clause which seems specialized to host them: a position between the definite subject and sentential adverbs: (55)

3.3.

a. *...,dat Jan gisteren 'r gekust heefi. b. ..., dat Jan 'r gisteren gekust he eft. c. *...,dat V Jan gisteren gekust heeft. ... that her Jan her yesterday her kissed has

Swedish

Holmberg (1986), (1991) also makes an important contribution to the study of Germanic pronouns. In spite of their morphological uniformity, he individuates two classes of Swedish pronouns on the basis of word order variation. Pronouns belonging to the two classes have a different distribution. Modified, conjoined and contrastively stressed pronouns, which qualify as strong pronouns according to Kayne's (1975) tests, must appear in their base position, as shown by the fact that they follow sentential adverbs and negation; non-strong pronouns undergo the so-called process of 'object shift' and appear instead to the left of sentential adverbs (from Holmberg 1991:156):16 inte [dem bäda]. inte. inte. not them two

(56)

a. Ig&r b. *Igär c. Igar yesterday

sag säg sag saw

Anna Anna [dem bäda] Anna [dem Anna [them (two)

(57)

a. Igar b. *Ig&r c. Igär yesterday

säg säg säg saw

Anna inte [dej och mej]. inte. Anna [dej och mej] Anna dej inte. you and me Anna you (and me)not

(58)

a. Igar b. *Igdr c. Ig&r yesterday

säg säg säg saw

Anna kanske inte DEN. Anna DEN kanske inte. Anna den kanske inte. Anna it maybe not IT

Pronouns in Germanic and Romance languages: An overview

53

Only non-strong pronouns can be shifted to the left, whereas strong pronouns (as shown by the b. sentences of [56]—[58]) and strong noun phrases never are:17 (59)

a. Hon säg kanske inte Anna. b. *Hon säg Anna kanske inte. she saw Anna maybe not Anna 'Maybe, she didn't see Anna'

This pattern is clearly reminiscent of the contrast between clitic and strong pronouns seen in Romance languages, with the notable difference concerning the location of the shifted form: In Swedish, it is not attached to the verb, but occurs in some intermediate clause-internal position to the left of sentential adverbs, roughly corresponding to the portion of the clause individuated for Dutch. As pointed out in Josefsson (1992a, b), (some) Swedish pronouns can also occur (i) to the left of the subject (i.e. higher than in Dutch), and (ii) between a sentential adverb and negation (i.e. lower than in Dutch): (60)

a.

b.

(61)

a.

b. c.

Därför gladde honom Maria med baskedet om therefore pleased him Maria with the-answer about pengarna. money-the (Josefsson 1992b: 59, 61) Igär lade sig mamma tidigt. yesterday lay-down REFL mother early Därför tvättade han förmodligen dem inte. therefore washed he probably to-them not (Josefsson 1992b: 66, 73) Därför gov Lisa honom förmodligen den inte. therefore gave Lisa to-him probably it not Därför gav Lisa den förmodligen honom inte. therefore gave Lisa it probably to-him not

In this respect, Swedish differs both from Dutch and from other Scandinavian languages. This means that, as in the Romance languages, a subtle languagevariation exists in the placement of clitic-like pronouns. The exact scope of the Germanic-internal language variation still remains to be fully evaluated.

54 3.4.

Anna Cardinaletti German again

As we have seen in §3.1., no morphological evidence speaks in favor of the existence of clitic-like pronouns in German. This has led some researchers to the extremist view that German only possess one class of pronouns, the class of strong pronouns. Being full noun phrases, their displacement is due to the same rule which applies to other noun phrases, i.e. scrambling. Although it is clear that a link exists between the rule of pronoun placement and the rule of scrambling (cf. Haegeman 1994 for detailed discussion with respect to West Flemish), pronoun movement displays partially different properties from the scrambling of noun phrases, a point made in Boschetti (1986) and Cardinaletti (1992). To capture these asymmetries, the existence of a particular rule of pronoun placement must be assumed. This in turn implies the existence of a particular class of clitic-like pronouns. Notice in particular that the different distribution of stressed and unstressed pronouns seen in (47) above is formally identical to the pattern displayed by Italian (3) and by Swedish (58). Two distinct classes of pronouns are posited in these languages: To capture the similarity of the paradigms, the same conclusion must hold for German. Thanks to the new emphasis given to a semantic property of clitic-like pronouns, this proposal has found definitive confirmation. As we have seen for French in (6) and for Dutch in (54), strong and non-strong (3rd person) pronouns have different interpretations: It is generally the case that strong pronouns can only refer to human entities, whereas non-strong pronouns can have both [+human] and [-human] reference. In German, the different interpretations correlate with the distribution of the pronoun. Whereas a [+human] pronoun can occur everywhere, (62), a [-human] pronoun can only occur in a particular clause-internal position, (63 a). It becomes ungrammatical to the right of a sentential adverb, in clause-initial position (specCP), in coordination, and in isolation, (63b-d) (cf. Beermann 1993, Corver and Delfitto 1993, Cardinaletti and Starke, this volume): (62)

a. b. c. d. e.

Er hat sie gestern wohl eingeladen. he has them yesterday probably invited Er hat wohl SIE eingeladen. Sie hat er gestern eingeladen. Er hat sie und ihre Freunde eingeladen. he has them and their friends invited Wen hat er eingeladen? Sie. who has he invited? Them

Pronouns in Germanic and Romance languages: An overview (63)

a. b. c. d. e.

55

Er hat sie gestern wohl gegessen. he has them yesterday probably eaten *Er hat wohl sie gegessen. *Sie hat er gestern gegessen. *Er hat sie und die Zwiebel gekauft. he has them and the onions bought *Was hat er gekauft? Sie. what has he bought? Them

This paradigm is formally identical to the one seen for French in (6), in which a clitic pronoun contrasts with a strong pronoun. It can only be captured if two distinct classes of pronouns are hypothesized in German as well: One class with clitic-like behaviour, restricted to the portion of the sentence to the left of sentential adverbs, the other displaying strong behaviour, free to occur in any syntactic position: to the right of sentential adverbs, in clause-initial position (specCP), in coordination, in isolation, etc., i.e. in all positions that do not allow clitic-like pronouns. 18 Postulating just one class of pronouns, there is simply no way, under current assumptions on the interaction between the lexical and the syntactic component, to obtain the fact that one of the two possible interpretations (the [-human] reference) is 'lost' if the pronoun appears in certain syntactic configurations such as coordination or isolation, or, alternatively, that one interpretation (the [-human] reference) is 'acquired' if the pronoun is found in other configurations, such as to the left of a sentential adverb. The same difficulties arise if we take German unstressed object pronouns to be 'clitic' only in the phonological component, but not in the syntax. Since the pronoun would qualify as strong as far as syntax is concerned, it is impossible to explain the semantic asymmetries between these pronouns and strong pronouns: No phonology/semantics interaction is standardly assumed. The difference observed in (47) between scrambling (which is apparently optional) and pronoun movement (which is obligatory) can now be explained. In all languages, clitic-like pronouns are moved in the syntax to their derived position, and this happens whether or not the language also allows scrambling of full noun phrases (cf. Swedish and Italian, where scrambling is not attested). The contrast then follows from the fact that German has clitic-like pronouns which must obligatorily be shifted to the left. This discussion of German also allows us to add a third ingredient to the overview so far. Whereas in Romance languages, there is systematic covariation between morphological, distributional and semantic properties, the clitic-like pronouns of Germanic languages discussed so far have been individuated on the basis of only one or two of these factors: mainly on the basis of morphology for Dutch (cf. den Besten 1977; Koster 1978), of distribution for Swedish (cf.

56

Anna Cardinaletti

Holmberg 1986), and of semantic properties for German (cf. Beermann 1993; Corver and Delfitto 1993; Cardinaletti and Starke, this volume).

3.5.

Summary

In the Romance languages, there is a strict covariation between morphological reduction, special distribution, and semantic properties. This is not evident in Germanic: In particular, the rich distinctions found in Romance have dim manifestations in Germanic, and the correlation with a special distribution is masked by other reordering rules which also apply to noun phrases. However, once the class of weak pronouns is individuated, they display the same properties which distinguish the clitic and the strong pronouns of Romance languages. They cannot be modified, conjoined and contrastively stressed, and cannot occur in the canonical positions for noun phrases, but are obligatorily displaced to some 'special' position of the clause (with the restrictions seen in fn. 15). Although Germanic clitic-like pronouns do not appear to establish any privileged relation with the verb,19 they display a special distribution, roughly corresponding to the portion of the sentence to the left of sentential adverbs. As we have mentioned above, the location of clitic-like pronouns is subject to language-variation, parallel to the variation found in Romance languages with respect to clitic placement. As for Romance, this topic, which has not found a satisfactory treatment yet, can only be successfully addressed in a comparative perspective. In summary, the Germanic languages possess two classes of pronouns, despite prima facie uniformity. The discovery of the existence of a clitic-like class of pronouns is the first main result of the research on Germanic pronouns of the past years.

4.

A second extremist view

Once it is established that a class of non-strong pronouns exists in Germanic languages, the question as to their syntactic status arises. In the last years, a second extremist view has come into being: The Germanic clitic-like pronouns are exactly identical to Romance clitics. In other words, the clitic/strong distinction has been transposed to Germanic, and Germanic clitic-like pronouns have been analysed on a par with Romance clitic pronouns, i.e. as heads (cf. §2.1. above). As a consequence, the two terms 'weak' and 'clitic' have become interchangeable in the literature on Germanic. This assumption has become the common view about Germanic clitic-like pronouns, and many scholars can be quoted here: Jaspers (1989) and Zwart

Pronouns in Germanic and Romance languages: An overview

57

(1991), (1993) for Dutch; Cardinaletti and Roberts (1991), Cardinaletti (1992) and Beermann (1993) for German; Corver and Delfitto (1993) for Dutch and German; Haegeman (1992b), (1993a, b) for West Flemish; Deprez (1990), Josefsson (1992a, b), Diesing and Jelinek (1993) for Swedish. 20 As we have seen above, Germanic clitic pronouns differ from Romance clitics with regard to their host: They do not systematically appear on the verb. In order to maintain a general theory of clitic pronouns, the proposal has been made that Germanic clitic pronouns cliticize on a covert host, one of the functional categories associated with the verb that contain morphosyntactic features. This difference between Germanic and Romance languages could be attributed to some independent property distinguishing the two groups of languages. A solution explored by Zwart (1992) for SOV Germanic languages such as Dutch and German is to relate the difference to the way in which verb movement applies in the two groups of languages and to the fact that Germanic languages are verbsecond.

4.1. Inadequacies of the head analysis of Germanic clitic-like pronouns Although it captures a number of similar properties, the proposal that Germanic weak pronouns are one and the same syntactic object as Romance clitic pronouns has been too quick, turning out to be empirically inadequate. Enlarging the empirical basis and reinterpreting well-known paradigms have shown that the assimilation of Germanic weak pronouns to clitic pronouns cannot do justice of a number of facts found in Germanic languages. Consider first subject pronouns in sentence-initial position, e.g. in German (64). They may have [-human] reference, a hallmark of clitic-like status, which correlates with their impossibility of being e.g. coordinated, (65). The same distribution is displayed by another clitic-like pronoun, the impersonal subject man, (66): (64)

a. b.

Sie sind teuer. they are expensive Es ist teuer. it is expensive

(65)

a. *Sie und die anderen sind teuer. they and the other ones are expensive b. *Es und die anderen sind teuer. it and the other ones are expensive

(66)

Man sieht es ja. one sees it indeed

58

Anna Cardinaletti

However, it is highly implausible that clause-initial subject pronouns may be analysed as heads, since the position preceding the finite verb in verb-second clauses is standardly analysed as a specifier position, only available to maximal projections. 21 The same observation holds for the Dutch clitic-like pronoun er found in prepositional phrases (see (52) above). If van Riemsdijk (1978) is correct in assuming that this pronoun occupies the specifier position of the prepositional phrase, then the unavoidable conclusion is that it cannot be analysed as a clitic head. Moreover, there are Germanic languages which possess two clitic-like subject pronouns differing in distribution. This is illustrated in (67) on the basis of Olang Tirolese (cf. Oberleiter and Sfriso 1993; Cardinaletti and Starke 1996): One clitic-like pronoun, se 'she, they', cannot appear in clause-initial position but only sentence-internally (minimally differing from the German subject pronouns in [64]), whereas the other clitic-like pronoun, es 'it', can appear clause-initially, on a par with German sie and es in (64) : 22 (67)

a. *Se sain toire. they are expensive b. ...,daßze toire sain. ... that they expensive are c. Es isch toire. it is expensive

If the pronoun in (67b) is a head, banned from the clause-initial position, then the pronoun in (67c) cannot also be a head. Being neither a clitic nor a strong pronoun, it lacks an analysis. The hypothesis that the pronouns in (64) and (67c) are phonological clitics 23 encounters the same problems pointed out in §3.4. above in the discussion on German object pronouns: Since in the syntax, a phonological clitic would be nothing other than a strong pronoun, there is simply no way, under current hypotheses entailing the lack of a phonology/semantic interaction, to explain the different semantic properties of these pronouns with respect to strong pronouns. Another clear case in which a generalized head analysis of Germanic cliticlike pronouns is unsatisfactory is represented by West Flemish paradigms: Both the pronouns jen 'you' and ze 'her' are clitic-like, e.g. cannot be coordinated and contrastively stressed, (68), but only the former is restricted to the post-subject position, (69), and to the rigid argument order of the language: 'dative- accusative', (70) (cf. Haegeman 1993a, b, 1994):

Pronouns in Germanic and Romance languages: An overview (68)

a. *... da Vale re [Marie en jen/ze] gisteren nie gezien eet. ... that Valere Marie and you/her yesterday not seen has '... that yesterday Valere has not seen you/her' dan-k gezien een. b. *T-is jen/ze it is you/her that-I seen have 'It is you/her that I have seen'

(69)

a. b. c. d.

(70)

a. b. c. d.

59

.. da Valere jen gisteren nie gezien eet. .. da jen Valere gisteren nie gezien eet. .. that you Valere you yesterday not seen has .. da Valere ze gisteren nie gezien eet. ..da ze Valere gisteren nie gezien eet. .. that her Valere her yesterday not seen has .. .. .. .. .. ..

da da that da da that

Valere MarieDAT Valere jenACC Valere to-Marie Valere MarieDAT Valere zeACC Valere to-Marie

jenACC MarieDAT you zeACC MarieDAT her

nie getoogd nie getoogd not shown nie getoogd nie getoogd not shown

eet. eet. has eet. eet. has

If both pronouns jen and ze are clitic, i.e. heads, the asymmetries between the two cannot be accounted for. If only ze is a head, then there is no way to capture the clitic-like behaviour of jen. Assuming that the Germanic clitic-like pronouns are heads on a par with the Romance clitic pronouns simply cannot account for these contrasts.

4.2. Romance languages It is of course logically possible that some of the above problematic cases are simply due to independent properties of each particular language. However, once we find similar patterns in Romance languages, which have 'regular' clitic and strong pronouns, it becomes more difficult to attribute the strangeness of such pronouns to independent properties of the Germanic languages. The Italian 3rd person plural dative pronoun loro 'to-them' displays clitic-like properties, but does not need to be adjacent to the verb and differs from clitic pronouns in a number of respects (cf. Cardinaletti 1991). How can this pronoun be differentiated from both the clitic and the strong object pronouns of Italian?24 The same point can be reiterated on the basis of the Italian subject pronouns of the egli 'ht'lesso 'it' series: They display the characteristic clitic-like properties, but a head analysis putting them in the same class as the subject clitic pronouns of Northern Italian dialects (see §2.4. above) would be highly unsatis-

60

Anna Cardinaletti

factory. Why need these pronouns not be adjacent to the verb? Why do these pronouns not cooccur with a full noun phrase? Why are they not repeated under sentence conjunction? The relevant contrasts are provided in (71)-(73): (71)

a.

Egliforse parte domani. he perhaps leaves tomorrow b. *La forse canta. she perhaps sings

(72)

a. *Se Gianni egli canta ... if Gianni he sings ... b. Se la Maria la canta... if the Maria she sings ...

(73)

a.

Egli canta e - balla. he sings and dances b. *La canta e - bala. she sings and dances c. La canta e la bala. she sings and she dances

The Italian paradigm makes it also possible to readdress the question of French clitic-like subject pronouns (cf. section 2.4. above). Since they also differ from the subject clitic pronouns of Northern Italian dialects, a head analysis for both would not be sufficient. Notice that their clitic-like properties cannot be accounted for simply via cliticization in the phonological component, as in Kayne's (1983) original proposal, for the reasons pointed out above for Germanic subject and object pronouns.

4.3.

English: A paradigmatic case

It has been noted a number of times that some English pronouns display cliticlike properties. Reduced forms exist, such as'm out of him or them and 'r out of her. Furthermore, the neuter pronoun it displays the idiosyncratic property of not being possible in isolation, topicalization, contrast, etc.: (74)

a. It strikes me as implausible. b. *What strikes you as implausible? It. c. *It, I think is implausible. d. *IT is implausible, not that one.

Pronouns in Germanic and Romance languages: An overview

61

It is also not allowed in the post-particle position of particle constructions, (75a). Moreover, Johnson (1991) points out that other pronouns are also excluded from this context unless they are coordinated or modified, (75b-d): (75)

a. b. c. d.

I I I I

took it took him took took

in in in in

*it. *him. him and her. only him.

Interestingly, Johnson fails to draw the parallelism between (75) and the Romance paradigms seen above. The properties which distinguish the pronouns in pre- and post-particle position are exactly the same as those that distinguish clitic from strong pronouns. The distribution of the clitic-like pronouns in (75) can be characterized in the usual terms: Clitic-like pronouns must move to the left, as depicted in (76a). This implies that the pronoun also moves in simple sentences, as in (76b): 25 (76)

a. b.

I took it t I took it t

in I I

The string-vacuous movement of the pronoun as in (76b) is supported by the following asymmetry arising in double-object constructions: Whereas it cannot appear if the first object is a full DP, the sentence becomes much better if it cooccurs with another pronoun (examples and judgments from Halpern and Fontana 1992:fn. 24): (77)

a. *Mary gave John it. b. Mary gave me it.

Given the string-vacuous movement, the contrast is so captured: In (77b), both pronouns are moved to the left. In (77a), on the other hand, the lack of movement of John prevents the pronoun it from shifting to the left. The unavoidable conclusion is therefore that English has clitic-like pronouns. However, this gets us into the same trouble as above. We have individuated clitic-like pronouns, but a head analysis is highly implausible, in particular for the case of subject it in (74a), which patterns with the Italian subject egli 'he' rather than with the Northern Italian subject clitic la 'she' (or the Italian object clitic lo 'him'): No adjacency requirement between it and the verb is attested, and it need not be repeated under sentence-conjunction. See above the contrasts in (71) and (73), and the following:

62

Anna Cardinaletti

(78)

a. b.

(79)

a. *

4.4.

It usually costs more than $100. It looks good and smells good too.

Lo [I] him b. * Lo [I] him

probabilmente vedrd domani. probably will-see tomorrow vedo e saluto ogni volta. see and greet every time

Conclusion

It should be clear at this point that the second extremist view, which holds that Germanic 'weak' pronouns are identical to Romance clitic pronouns in that they are heads, can only be maintained at the cost of leaving a large number of empirical facts unaccounted for (or providing a number of ad hoc solutions for many of these facts). In more general terms, what is empirically inadequate is the theory of pronouns so far assumed: Since we must express the existence of two classes of pronouns in Germanic as well, without however having recourse to the head analysis of one of the two classes, it follows that the traditional bipartition into clitic and strong pronouns is not sufficient. Similarly, given a bipartition, there is no way of expressing the partial asymmetries of two clitic-like pronouns found language-internally (e.g. in Italian and West Flemish).

5.

The new typology of personal pronouns

The only description which proves empirically adequate is one which abandons the original bipartition into clitic and strong pronouns and assumes that there is a tripartition of personal pronouns: Alongside clitic and strong pronouns, there exists a distinct category of weak pronouns as well. In other words, a further bipartition of the realm of non-strong pronouns is necessary: Non-strong (or deficient) pronouns include clitic and weak pronouns, as proposed by Cardinaletti and Starke (this volume). The tripartition of pronouns is as follows: (80)

personal pronouns strong

non-strong (= deficient) weak

clitic

The tripartition allows us to analyse the 'problematic' clitic-like pronouns seen above in §4.1.-3. as weak pronouns: German sie/es/man, Olang Tirolese es, West

Pronouns in Germanic and Romance languages: An overview

63

Flemish jen, Italian loro/egli, English it. On the other hand, Olang Tirolese seize and West Flemish ze discussed in §4.1. are best treated as clitic pronouns. Once this tripartition is assumed, the question arises as to how the difference between the three classes can be expressed. Whereas it is straightforward that strong pronouns are full maximal projections akin to full noun phrases, the distinction between clitic and weak is far from obvious. There is one implementation of the tripartition which can account for all instances of weak pronouns encountered so far. The distinction between weak and clitic pronouns can be expressed in terms of their status in X-bar theoretic terms: (At the end of the derivation), weak pronouns are (deficient) maximal projections occurring in specifier positions, whereas clitic pronouns are heads, adjoined to a functional head.26 The tripartition of pronouns has a number of consequences. First of all, the tests for clitichood elaborated by Kayne (1975) are valid not only for clitic pronouns, but for weak pronouns as well. That a pronoun displays some of the properties originally discussed by Kayne does not guarantee that it is a clitic, i.e. a head. Further properties of the pronoun and of the language must be investigated in order to establish its 'clitic' or 'weak' nature. In turn, clitic-like properties cannot depend on the head status of clitic pronouns, since they are shared by pronouns which are not heads. A new explanation is therefore called for, which does not rely on the simple distinction head versus maximal projection.

5.1.

Typological remarks

As we have seen, 'clitic' and 'weak' are no longer interchangeable terms, but define two syntactically distinct classes of pronouns, universally available. Therefore, there is no one-to-one correlation between Romance languages and clitic pronouns on the one hand, and between Germanic languages and weak pronouns on the other.27 And as a matter of fact, Romance languages possess some weak pronouns, and Germanic languages possess some clitic pronouns. Although it may turn out to be true that there is a typological specialization of the two groups of languages with respect to the choice between clitic and weak pronouns, this must be seen as the result of a historical accident, rather than a principled state of affairs. In other words, although it may be interesting to explain why Germanic languages have predominantly developed weak pronouns, whereas Romance languages have predominantly developed clitic pronouns, this question seems vacuous in a synchronic perspective. As we have already seen in §4.2., in Italian both subject (eglilesso) and object (dative lord) weak pronouns have been identified. French has weak subjects (cf. §2.4. above), and weak objects may be individuated in the case of imperatives

64

Anna Cardinaletti

(cf. Cardinaletti and Starke, this volume: fn. 32 and Laenzlinger 1994). Spanish displays weak subjects according to Ordonez (1994). If a null subject is correctly analysed as an empty pronoun, then it will be a weak pronoun: It displays the properties of French il 'he, it' rather than French lui 'he' (phonological realisation aside). This amounts to saying that all Romance null-subject languages have at least one instance of weak pronouns. On the other hand, clitic pronouns are attested in many Germanic languages. In addition to the Olang Tirolese and the West Flemish cases seen in §4.1., cf. Christensen (1984) for Norwegian, Vikner and Schwartz (1991) for Danish, Abraham (1991) and Abraham and Wiegel (1993) for spoken and dialectal varieties of German, Penner (1991) for Bernese Swiss German, Haegeman (1993a, b), (1994) for West Flemish, Oberleiter and Sfriso (1993) for Olang Tirolese.28 The research in this area is particularly welcome for two sets of reasons: a) the new typology opens up the possibility that every Germanic deficient pronoun is either a clitic or a weak pronoun. This question concerns both subject and object pronouns occurring clause-internally: Simply looking at its distribution does not disclose whether a pronoun occurring in the 'Mittelfeld' is clitic or weak (whereas, as we argued above in §4.1., deficient subject pronouns in sentenceinitial position are unambiguously maximal projections, i.e. weak pronouns). It is highly plausible that many Germanic clitic-like object pronouns are of the weak type rather than of the clitic type, but a definitive answer can be provided only once specific tests have been developed for Germanic. The ideal situation is the one found in Olang Tirolese and West Flemish: Two distinct patterns are found language-internally (cf. §4.1.). In the lack of similar language-internal evidence, comparative observations from closely related languages could lead to the same conclusions.29 Some facts suggest that (some) Dutch and German clitic-like pronouns are weak rather than clitic (cf. Cardinaletti and Starke 1996:§6.4., pp.60-61). In the first place, there is a general constraint on combinations of clitic pronouns, which excludes a 1st or 2nd person accusative cooccurring with a 3rd person dative pronoun, illustrated here for Italian in (81a) (cf. Perlmutter 1971, Bonet 1991); the same ban does not apply to combinations of clitic and weak pronouns, (81b): (81)

a. *Giannimi Gianni me b. Gianni mi Gianni me

gli/le ha mostrato. to-him/to-her has shown ha mostrato loro. has shown to-them

Pronouns in Germanic and Romance languages: An overview

65

The restriction is also operative in Germanic languages. In Olang Tirolese, a 1st person accusative pronoun cannot cooccur with a 3rd person dative pronoun, whatever the order of pronouns: (82)

*..., daß a me ra /τα me vorgestellt hot. ... that he me to-her / to-her me introduced has

Interestingly, in Swedish this combination of pronouns is sensitive to their morphological form and seems to distinguish clitic from weak pronouns: 30 (83)

a. *Jon visade henneDAT tnejACC inte. b. Jon visade naDAT mejACC inte. Jon showed to-her me not

Given that they both precede negation, henne and na are to be analysed as both non-strong pronouns (for the same set of -features). As we have seen above in section 3.3., strong pronouns never undergo object shift in Swedish. The contrast in (83) suggests that they belong to two different classes, since simple phonological reduction cannot account for such effects (cf. also Platzack 1995:fn. 23 for evidence that na is not phonologically derived from henne). The morphological hierarchy discussed in Cardinaletti and Starke (this volume: §3.3.1.) forces the conclusion that henne is weak and na is clitic.31 The Italian/Swedish contrast leads to the following tentative generalisation: The combination of a lst/2nd person accusative pronoun with a 3rd person dative pronoun is possible with pronouns belonging to different classes, but not when both pronouns are clitic or both are weak. With this in mind, consider now the possibility of such a combination in German (84a) and, with some variation among speakers with respect to the preferred order, in Dutch (84b): (84)

a. b.

..., daß er mich ihm gestern nicht vorgestellt hat. ... that he me to-him yesterday not introduced has me voor Ze stelden meACC 'mDAT / ?'mDAT · Acc they introduced me to-him / to-him me PRT

Intricate questions of preferred order seem to arise, but a preliminary conclusion is that in Dutch and German, one of the two object pronouns is weak. 32 Another context which differentiates clitic from weak pronouns is the complement to a preposition: Only weak pronouns are allowed, as shown by the Italian contrast in (85). The Dutch paradigm in (86) (due to Riny Huijbregts, personal communication) suggests that 3rd sing. fem. ze 'her' is a clitic pronoun, but the 3rd sing. fem. r 'her' and 3rd pi. ze 'them' are weak pronouns: 33

66

Anna Cardinaletti

(85)

a. *Di locUtic abbiamo parlato a lungo. essoweak abbiamo parlato a lungo. b. Di about it [we] have talked long

(86)

a. *Ikkijk I look b. Ikkijk I look c. Ik kijk I look

naarze. at her

(cf. Ik bekijk ze = I watch her)

naar'r.

(cf. Ik bekijk'r = I watch her)

at her naar ze. at them

(cf. Ik bekijk ze = I watch them)

b) A second reason why Germanic clitics should be investigated more carefully is the fact that the studies on Germanic have brought to light some unexpected properties of clitic pronouns, distinct from those postulated on the basis of Romance languages. This contradictory evidence makes it necessary to rethink these properties within a broader comparative perspective. We provide here a couple of examples. The phenomenon of clitic splitting is very rarely encountered in Romance languages (cf. Kayne 1991:660-663 for discussion). In Italian, for instance, cliticclimbing structures in principle allow two landing sites for the clitic pronouns, one on the embedded infinitival verb, the other on the finite matrix verb (cf. fn. 9). However, despite the two theoretically possible landing sites, clitic pronouns always cluster: (87)

a. Voleva darmelo. b. Me lo voleva dare. c. *Mi voleva darlo. d. *Lo voleva darmi. [he] wanted [to] give to-me it

Clitic splitting is however productively allowed in West Flemish (cf. Haegeman 1994): (88)

... da- ze Marie t gegeven eet. ... that to-them Marie it given has

Probably linked to this property, West Flemish clitic pronouns do not respect a rigid ordering, a state of affairs very rarely encountered in Romance languages (cf. the cluster leAa/les lui 'him/her/them to-him' in French, where both orders are acceptable, Michal Starke, personal communication):34

Pronouns in Germanic and Romance languages: An overview (89)

a. b.

... ... ... ...

dan-k that-I dan-k that-I

67

ik ze-t gegeven een. I it to-them given have ikt-ze gegeven een. I to-them it given have

A second contrast concerns parasitic gaps. In Romance languages, clitic pronouns cannot license parasitic gaps, but in West Flemish they can: (90)

a. * Gianni I 'ha dato t a Gianni it has given to b. ... dan ze ze [zonder ... that they them een. have

Maria [senza legge re e], Maria without read & open te doen] t ipgestierd without open to make sent

Parasitic gaps are licensed by an antecedent in an A'-position: For instance, they are possible with Wh-elements in interrogative and relative clauses (cf. Engdahl 1983, Chomsky 1982, 1986, Cinque 1990, among many others):

(91)

a.

b.

Che cosahai dato t a Maria [senza nemmeno what have [you] given to Maria without not-even leggere e] ? read? il libro che ho dato t a Maria [senza nemmeno the book that [I] have given to Maria without not-even leggere e]... read

The ungrammatically of (90a) thus indicates that no A'-position is involved in the clitic chain. The grammaticality of (90b) in turn seems to suggest that an A' link must be assumed in the case of West Flemish. The A' step of clitic movement is however not available to weak pronouns and full noun phrases, which are incompatible with parasitic gaps in the same language, (92). This means that scrambled noun phrases only undergo Α-movement (cf. Haegeman 1994):35 (92)

a. *... da Valere eur [zonder e te zien] gepasseerd is. ... that Valere her without to see passed is b. *... dan ze die boeken [zonder e open te doen] ... that they the books without open to make ipgestierd een. sent have

68

Anna Cardinaletti

Finally, Zurich Swiss German has clitic-like forms which appear rather low in the sentence structure, to the right of sentential adverbs. Cooper (1994) claims that they are clitic pronouns that occur inside VP, a rather exceptional pattern among Germanic languages and among clitic pronouns in general. This is not the only instance of 'low' clitic pronouns in Germanic languages, however. In verbsecond clauses, the Swedish clitic pronoun na 'her' (cf. [83b] above) can occur to the right of the negation, contrary to the other deficient pronouns seen in §3.3. above (example from Platzack 1995:22): (93)

Han säg inte'na. she saw not her

One parallel fact which comes to mind is the distribution of French clitic pronouns in infinitival clauses. Here, the pronoun follows adverbs such as pas 'not' and bien 'well', which it precedes in finite clauses (cf. Pollock 1989, Belletti 1990): (94)

a. b.

...dene pasI'acheter. ... of not not it buy Jean ne I'achete pas. Jean not it buys not

The consequences of this parallelism must of course be investigated more fully.

6.

Open questions

Apart from the Germanic-oriented questions mentioned above, many general questions remain open, which concern both Germanic and Romance languages. The most general one is the most difficult one: Which functional category hosts clitic pronouns, and how is language variation in this area accounted for? What is the derived position in which weak pronouns occur, and how does language variation in this area arise? Ultimately, this amounts to asking what the nature of clause structure of Germanic and Romance languages is. It is evident that the task is huge. Although many proposals have been made, no comprehensive answer has yet been given. The second topic which calls for a detailed account is the analysis of sequences of pronouns and, more specifically, the internal structure of clitic clusters. Two issues arise in this respect. First, whether the order of arguments is reflected in the order of pronouns or not. Both in Germanic and Romance, pronouns in general occur in the order 'dative-accusative', reflecting what could turn out to be the universal order of argument (cf. Larson 1988; Haider 1992).36 But there

Pronouns in Germanic and Romance languages: An overview

69

is also a certain amount of language variation in the order of pronouns. In French, for instance, 3rd person datives can follow the accusative clitic, whereas 1st and 2nd person datives must precede accusatives, (95), and in German, accusative pronouns precede dative pronouns, (96): 37 (95)

a.

b.

(96)

a. b.

Jean me/te/nous/vous I'a donne. Jean to-me/to-you/to-us/to-you it has given (cf. *le me/te/nous/vous) Jean le luiAeur a donne. Jean it to-him/to-them has given (cf. lui/?leur le) Ich I Ich I

habe have habe have

es ihm gegeben. it to-him given ihn ihm vorgestellt. him to-him introduced

The second issue concerns possible and impossible clusters, for which Case and person features seem to play an important role (as shown in French (95); see also §5.1. above). These still highly mysterious restrictions have been partly discussed in the literature devoted to Romance (cf. Perlmutter 1971; Bonet 1991; Laenzlinger 1993, 1994). Apart from the few observations in §5.1. above, no discussion of clitic clusters exists for Germanic languages. As we have pointed out above, this may also be relevant to the issue of establishing the clitic or weak nature of Germanic deficient pronouns. Finally, the extension of the tripartite typology to grammatical categories other than personal pronouns calls for detailed inquiry. The properties defining clitic, weak and strong pronouns are in fact not specific to personal pronouns. Being a head or a maximal projection at the end of the derivation should apply to all syntactic categories, and being restricted to certain positions of the clause or being free to occur in any syntactic position is a purely distributional property. It follows that the tripartite classification should be valid for any syntactic element. Although the study in this domain has just begun, the above expectation has been largely confirmed for Romance languages, where elements of any syntactic category ((pro)nominals other than personal pronouns, adjectives, adverbs) have been shown to fall into either one of the three categories: clitic, weak, strong (cf. Cardinaletti and Starke, this volume: §9, for a first presentation of preliminary results, and Cardinaletti 1997 for possessive pronouns and adjectives). Nothing similar has been undertaken yet for Germanic languages, and the observations reported in the literature are still episodic. A first finding in Germanic is a clitic demonstrative, the West Flemish 3rd person fem. pronoun ze 'her' seen above in

70

Anna Cardinaletti

(68)—(70). Contrary to other personal pronouns, it obeys Principle C of Binding theory (cf. Haegeman, this volume:Comments to Cardinaletti and Starke). Another possible candidate found in West Flemish (and in some German dialects) is the demonstrative weak pronoun da 'that' which occurs as subject of weather verbs:38 (97)

Da regent. it rains

See also Cardinaletti and Starke (1996:§5.3 and §6.3) for the extension of the tripartition to German demonstrative and reflexive pronouns, respectively. Given the empirical and theoretical relevance of the application of the tripartition to all grammatical elements, it should be hoped that work in this area will be continued and developed in both Romance and Germanic languages.

7.

Conclusions

The results of the recent research on pronouns are very positive, both from the descriptive and the theoretical point of view. On one side, they have led to an improvement in the description of the pronominal systems of Germanic languages, with the definitive and uncontroversial conclusion that the Germanic languages possess clitic-like pronouns. This in turn has shown that treating Germanic clitic-like pronouns and Romance clitic pronouns alike is empirically incorrect and has suggested a new typology, which features a tripartition into clitic, weak and strong pronouns instead of the traditional clitic/strong dichotomy. The new typology of pronouns has at the same time made possible the recognition and description of a number of Romance pronouns, treated as particularly difficult cases until now. On the more theoretical side, many properties of clitic and weak pronouns are now better understood, and their explanation has become a real possibility. In conclusion, not only do we now have a richer description of the morphological, phonological, semantic and syntactic properties of pronouns due to the comparative approach, but we can now begin to grasp why things are the way they are. As in every field of lively research, answering one question means asking many new questions. The results obtained from the study on Germanic and Romance pronouns raise a number of new questions, but at the same time have laid the ground for a satisfactory answer.

Pronouns in Germanic and Romance languages: An overview

71

Notes 1.

2.

For useful comments on an earlier draft, I would like to thank Guglielmo Cinque, Sue Olsen, Christer Platzack, Henk van Riemsdijk and Michael Stark. I am very grateful to Henk van Riemsdijk for having integrated me into the Eurotyp group 8, thus contributing to trigger a series of papers on the topic. For instance, Helbig and Buscha (1981:353) notice that whereas subject es 'it' can appear in sentence-initial position, object es cannot: (i)

a.

Es liegt auf dem Tisch. it lies on the table b. *Es brauche ich. it need I For the discussion of this and related subject/object asymmetries in the framework of Generative Grammar, cf. Haider (1984:3); Travis (1984:121); Diprez (1989); Cardinaletti (1990:83). 3. Notice that 'long-distance' modification is allowed. Floated quantifiers may refer to clitic pronouns: (i) a. Li ho visti / letti tutti. [I] them have seen / eaten all b. L'ho mangiato tutto. [I] it have eaten all 4. This semantic property, often noticed, has only recently been fully evaluated. Cf. Kayne (1975:91); Jaeggli (1982:41); Rizzi (1982); Berendsen (1986:38-39); Schroten (1992); Corver and Delfitto (1993); Cardinaletti and Starke (this volume). Demonstrative pronouns behave differently: They can always have non-human referents: (i) Je ne connais / achete que celui-lä. I not know / buy [other] than that one 5. For the notions of adjunction and functional head, and for all other technical terms, cf. Haegeman (1991a). 6. In Kayne's framework, right adjunction is banned. Cf. Kayne (1994) for the derivation of the leftadjunction hypothesis. 7. Cf. Kayne (1994:42^-6) for a different analysis of some of the facts discussed by Benincä and Cinque (1993). Notice that the following sentences, which represent other combinations of the elements in (19), are ungrammatical for a different reason: (i) a. *[Leggo] e lo rileggo. b. *per leggerlo e [rileggere]. c. *Leggilo e [rileggi]. Here, bigger chunks of the sentence than in (19) are conjoined. The sentences are ungrammatical because the conjuncts in parenthesis do not contain a clitic pronoun, and the selectional requirements of the transitive verb leggere are not met. The same observation holds for the French counterparts of (i). 8. Example (23b) is particularly interesting. It shows that an analysis of (21)-(23) which takes the adverb to be cliticized on the verb so that pronoun, adverb and verb form a unit, cannot be on the right track. In (23b), a full adverbial phrase intervenes between the finite verb and the clitic pronoun. The same point is made by Kayne (1991:653, fn.18) for literary French: (i) a. η 'en [presque rien] dire. not of-it almost nothing [to] say b. en [fort bien] parier. of-it strong well [to] speak

72 9.

10.

11.

12.

13.

14.

15. 16.

Anna Cardinaletti The exact mechanics of adjunction in the two configurations of enclisis and proclisis are still unclear. Enclisis in finite imperatives also deserve more attention. For clitic placement in negative infinitival imperatives in Italian, cf. Kayne (1992). Apparent counterexamples to locality are clitic-climbing structures found with modals, aspectual and motion verbs, in which the pronoun appears cliticized to the superordinate verb of which it is not a complement. Cf. Rizzi (1982) for extensive discussion, and, more recently, Haverkort (1993). Kayne (1989) establishes a correlation between clitic climbing and possibility of null subjects, which explains, among other things, the following Italian/French contrast: (i) a. Gianni vuole leggerlo. b. Gianni lo vuole leggere. (ii) a. Jean veut le lire. b. *Jean le veut lire. John it wants [to] read it Although pronouns cannot climb, French allows climbing of bare and floated quantifiers, as in (iii) (cf. Kayne 1975, 1980), and of weak adverbs, as in (iv) (cf. Kayne 1991:655, fn.23 and Cardinaletti and Starke, this volume:§9.1.4.): (iii) a. J'ai tout voulu faire. I have everything wanted [to] do b. J'ai tous voulu les lire. I have all wanted them [to] read (iv) J'ai mal du raccrocher. I have badly must [to] hang-up Ί must have hung (the phone) up badly/wrongly' Cf. Uriagereka (1995) and Torrego (1992), (1995) for the attempt of rendering the movement account compatible with clitic doubling under the enriched nominal structure of the DPhypothesis. These proposals essentially involve the base-generation of a 'complex' DP that contains both the clitic and the doubling DP. A criticism of Sportiche's account is discussed in Cardinaletti and Starke (this volume:fn. 82), where both the base-generation of the clitic as a functional category and the criterion approach are questioned. A purely derivational analysis seems to be the only coherent analysis of clitic placement. See Section 4.2. for discussion of the shortcomings of the notion of phonological clitic applied to French subjects. Cf. also Cardinaletti and Starke (1996:§4.1.3.). For a discussion of the phonological properties of clitic pronouns, cf. Nespor (1993). Attempts have been made to accomodate the subject/object asymmetry within a symmetric approach to verb-second, according to which all verb-second clauses are CPs. Cf. Taraldsen (1986); Cardinaletti (1990); Rizzi (1991); Platzack (1995). For similar facts in Romance, cf. Cardinaletti and Starke (this volume:§2.4, §2.5). Object shift is contingent on the movement of the finite verb in verb-second clauses. When the clause contains a past participle and in embedded clauses, pronouns follow the verb (cf. Chomsky 1993 for a possible account of the interaction between verb movement and argument movement): (i) a. Anna har inte sett den. b. *Anna har den inte sett. Anna has not seen it (ii) a. Jag vet att Anna inte säg den. b. *Jag vet att Anna den inte säg. I know that Anna not saw it (iii) a. Anna lovade att inte tvätta dem. b. *Anna lovade att dem inte tvätta. Anna promised to not wash them

Pronouns in Germanic and Romance languages: An overview

73

Notice that this restriction is not at work in Romance languages, where the clitic can move to the left to a position higher than that reached by the verb (cf. examples [21]—[23] above in the text and note 7). 17. Swedish (and Norwegian, Danish) differs from Icelandic, which also allows full noun phrases and strong pronouns to shift to the left. Holmberg (1986), (1991) suggests that in Swedish, only pronouns undergo object shift because they are the only noun phrases displaying morphological Case. Vikner (1990) explicitely rejects the Case-based account, on the basis of evidence from Faroese: In this language, all noun phrases have morphological Case, but only pronouns undergo object shift. Since weak pronouns move in all languages, independently of whether noun phrases also move, the most promising account is to consider pronoun movement as ruled by (partially) different principles with respect to the movement of full noun phrases. 18. In German, clitic-like pronouns can freely occur in pre-subject position, as in (i) alongside (46a): (i) Ich glaube, daß ihn Maria gestern gesehen hat. I think that him Maria yesterday seen has and there are varieties of German in which this is the preferred option. This confirms the language-variation on pronoun placement mentioned in §3.3. above with respect to Dutch and Swedish. 19. As a consequence, no distinction between finite and infinitival clauses ever arises in Germanic languages as far as weak-pronoun placement is concerned. The contrast seen in fn. 15 is simply a consequence of the fact that in verb-second clauses, only finite verbs raise to C. 20. Holmberg (1991) assumes an hybrid category X*, which can surface either as a head (X°), as in Italian, or as a maximal projection (XP), as in Swedish. 21. A derivation in which the subject pronoun is cliticized onto the verb in second position, and the sentence-initial position is occupied by a null subject (pro) is to be excluded, cf. Cardinaletti and Starke (1996:4.1.1-2) for discussion. 22. Se and ze are allomorphs of the 3rd person singular feminine and plural reduced pronoun. 23. Cf. Tomaselli (1990) for German, and Haegeman (1992a) for West Flemish. 24. The two extremist positions found in the literature on Germanic pronouns are also encountered in the discussion on Italian loro 'to-them': It has been analysed either as a (en)clitic pronoun (cf. Rizzi 1976, 1982; Calabrese 1988), or as a strong pronoun (cf. Corver and Delfitto 1993), in both cases sacrifying part of the empirical observations. 25. This has the apparently surprising consequence that some local verb movement must be assumed for English as well. For proposals arguing that this is indeed the case, cf. Johnson (1991) and Koizumi (1993). The same remark holds for Scandinavian languages, in the cases in which the verb apparently does not move. Cf. fn. 15 and Holmberg and Platzack (1995) for discussion. 26. Of course, it is not unconceivable that a finer typology is necessary, which contains subclasses of clitic and weak pronouns. Future work will confirm or disconfirm this possibility. 27. Thus, it is not possible to posit a X-bar theoretic distinction (Xmax vs. X°) between Germanic and Romance clitics, as done in e.g. Halpern and Fontana (1992). 28. Other works on German dialects are Koerner (1984), Kloeke (1985), Werner (1988), Prinz (1991), and Nübling (1992). 29. The fact that Germanic languages in general lack clitic doubling with object pronouns (but see German reduplication of da in Da habe ich nicht dran gedacht 'there have I not there-to thought') does not provide a definitive answer to the question concerning the status of deficient pronouns. This may be due either to the fact that these pronouns are weak and cannot function as doubling elements (which must be heads, cf. Cardinaletti and Starke, this volume:§3.2.2), or to the fact that the existence of clitics is a necessary condition, not a sufficient one for clitic-doubling. For instance, Standard Italian has accusative clitics, but they cannot enter clitic doubling. (For doubling of subject pronouns in West Flemish, cf. Haegeman 1991b). The same remarks hold for Clitic Left Dislocation, not found in Germanic languages with

74

Anna Cardinaletti personal (object) pronouns. If the resumptive element must be a clitic (cf. Cardinaletti 1991 for the Italian-internal contrast between clitic gli 'to-him' and weak loro 'to-them'), then this gap can be due either to the fact that object pronouns are weak in Germanic, or to the fact that the existence of clitics is a necessary, not a sufficient condition for the existence of Clitic Left Dislocation.

30. The diacritics reproduce the judgments by Ingvar Lofstedt and Ida Toivonen. Anders Holmberg and Christer Platzack accept sentence (83a). With this combination of pronouns, Christer Platzack finds the opposite order ungrammatical, (ia), although he accepts the order 'accusative-dative' with other combinations of pronouns, (ib): (i) a. *Jon visade mejACChenneDA1/naDATinte. Jon showed me to-her not b. Jon visade denACC meiDAT inte. Jon showed it to-me not 31. The two pronouns apparently contrast in sentence-initial position: (i) a. Henne har jag inte sett. b. *Na har jag inte sett. her have I not seen However, it is safe to conclude that the pronoun which appears in (ia) is not the weak pronoun henne 'her', but its strong counterpart. In Germanic languages, weak and strong pronouns are often homophonous (cf. Cardinaletti 1994). The fact that no homophonous form exists for na can be taken as a confirmation of the hypothesis that na is a clitic pronoun. 32. A restriction in the order of pronouns with this combination of features is also found in a Swiss German variety mentioned in Bonet (1991:188, fn.12), which has morphologically marked deficient pronouns. Combinations of pronouns usually display both orders 'dative — accusative' and 'accusative — dative', but when the dative is 3rd person and the accusative is 1st or 2nd person, only the 'accusative — dative' order is allowed: D'Marie zeigt en mir. Marie shows him to-me the b. D'Marie zeigt mir en. the Marie shows to-me him zeigt mi em. a. D'Marie (ii) the Marie shows me to-him b. *D' Marie zeigt em mich. Marie shows to-him me the The sentence in (iib) can be saved if the accusative pronoun is a strong pronoun, miich 'me'. The strategy of substituting one of the two clitic pronouns for a strong pronoun is also used in Romance. The ungrammatical Italian (81a) contrasts with the two following possibilities: Gianni mi ha mostrato a lui 'Gianni me has shown to him '/Gianni gli ha mostrato me 'Gianni to-him has shown me'. 33. Henk van Riemsdjik (personal communication) accepts sentence (86b) if the pronoun is d'r. This form is probably due to the phonotactic rule which inserts [d] between two adjacent [r]s, as in comparative formation: helder 'clear' + er gives helderder 'clearer' as output. 34. Whether the different argument order can be used for attributing a weak status to the 3rd person dative lui is still unclear. Cf. Laenzlinger (1993) for a proposal in this sense. 35. In Scandinavian languages, parasitic gaps, possible with Wh-phrases, are ungrammatical with weak pronouns (Swedish example from Holmberg 1986): (i) * Jag kastade den inte t [innan jag hade last e]. I threw it not before I have read This fact does not tell us much about the syntactic nature of pronoun placement. The ungrammaticality of (i) could be traced back either to the fact that Swedish object pronouns are weak and (i)

a.

Pronouns in Germanic and Romance languages: An overview

75

pronoun movement involves an Α-chain (as for instance in Icelandic and West Flemish), or to the fact that pronoun movement is similar to Romance cliticization. Unfortunately, it is not possible to check whether the Italian dative pronoun loro 'to-them', which is undoubtably weak, licenses parasitic gaps, since these are only found with accusative complements (cf. Cinque 1990). 36. 'Dative-accusative' is the unmarked order of noun phrases in all Germanic languages. Romance languages display the opposite order of arguments. This difference could be due to the fact that Romance 'dative' complements are prepositional phrases, and prepositional phrases always follow (accusative) noun phrases. Cf. however Zubizarreta (1994), who discusses Demonte's (1993) proposal that 'dative-accusative' is also the unmarked word order in Spanish. 37. It is worth mentioning that the pronominal order has changed in the history of Italian and French. 'Accusative-dative' was the predominant pronominal order in Medieval Italian (see Rohlfs 1968:176-178) and in French until the beginning of the XVIIth century (see Grevisse 1964:421). 38. Substandard/Colloquial French has a parallel demonstrative weak form: (ii) f a pleutfort. it rains a lot

References Abraham, W. 1991 "Rektion und Abfolge pronominaler Satzglieder und ihrer klitischen Formen im Deutschen", Language and Cognition 1. Abraham, W. and A. Wiegel 1993 "Reduktionsformen und Kasussynkretismus bei deutschen und niederländischen Pronomina", in: W. Abraham and J. Bayer (eds.), Dialektsyntax, Linguistische Berichte, Sonderheft 5, 1 2 ^ 9 . Amauld, A. and C. Lancelot 1660 Grammaire generale et raisonnee de Port-Royal (Geneve: Slatkine Reprints, 1993). Baltin, M. 1982 "A landing site theory of movement rules", Linguistic Inquiry 13: 1-38. Beermann, D. 1993 "Germanic pronouns". [Unpublished manuscript. University of Tilburg], Belletti, A. 1990 Generalized Verb Movement. Turin: Rosenberg and Sellier. 1993 "Case checking and clitic placement. Three issues on (Italian/Romance) clitics", GenGenP 1.2, 101-118. Belletti, A. and L. Rizzi 1981 "The syntax of ne: Some theoretical implications", The Linguistic Review 1, 117-154. Benincä, P. and G. Cinque 1993 "Su alcune differenze tra enclisi e proclisi", in: Omaggio a Gianfranco Folena. Padova: Editoriale Programme, 2313-2326. Berendsen, E. 1986 The Phonology of Cliticization. Dordrecht: Foris. Besten, Η. den 1977 "On the interaction of root transformations and lexical deletive rules", in: Studies in West Germanic Syntax. Amsterdam: Rodopi, 1989, 14-93. Bonet, E. 1991 Morphology after Syntax: Pronominal Clitics in Romance. [Unpublished Ph.D. dissertation, ΜΓΓ].

76

Anna Cardinaletti

Borer, H. 1984 Parametric Syntax. Dordrecht: Foris. Boschetti, L. 1986 Zur Syntax des Pronomens 'sich'. [Unpublished Tesi di laurea, University of Venice]. Brandl, L. and P. Cordin 1981 "Dialetti e italiano: un confronto sul parametro del soggetto nullo", Rivista di grammatica generativa 6, 33-87. 1989 "Two Italian dialects and the null subject parameter", in: O. Jaeggli and K.J. Safir (eds.), The Null Subject Parameter. Dordrecht: Kluwer, 111-142. Burzio, L. 1986 Italian Syntax. A Government-Binding Approach. Dordrecht: Reidel. Calabrese, A. 1988 "I pronomi personali", in: L. Renzi (ed.), Grande grammatica italiana di consultazione. II Mulino: Bologna, 549-592. Cardinaletti, A. 1990 "Subject/object asymmetries in German null-topic constructions and the status of specCP", in: J. Mascarö and M. Nespor (eds.), Grammar in Progress. Glow Essays for Henk van Riemsdijk. Dordrecht: Foris, 75-84. 1991 "On pronoun movement: The Italian dative loro", Probus 3.2, 127-153. 1992 "On cliticization in Germanic languages", Rivista di grammatica generativa 17, 65-99 (also in: L. Rizzi (ed.) Clitics in Romance and Germanic, 5-31). 1994 "On the internal structure of pronominal DPs", The Linguistic Review 11.3-4, 195-219. 1997 "On the deficient/strong opposition in possessive systems", to appear in: A. Alexiadou and Ch. Wilder (eds.), Studies on the Determiner Phrase. Amsterdam: Benjamins, Linguistics Today. Cardinaletti, A. and G. Giusti 1992 "Partitive ne and the QP-hypothesis. A case study", in: E. Fava (ed.), Proceedings of the XVII Meeting of Generative Grammar. Volume presented to Giuseppe Francescato on the occasion of his seventieth birthday. Torino: Rosenberg and Sellier, 121-141. Cardinaletti, A. and I. Roberts 1991 "Clause structure and X-second", to appear in: W. Chao and G. Horrocks (eds.), Levels of Representation. Dordrecht: Foris. Cardinaletti, A. and M. Starke 1996 "Deficient pronouns: A view from Germanic. A study in the unified description of Germanic and Romance", in: H. Thräinsson, S. Epstein and S. Peter (eds.), Studies in Comparative Germanic Syntax Volume II. Dordrecht: Kluwer, 21-65. Cardinaletti, A. and M. Starke this volume. "The typology of structural deficiency: A Case Study of the three Classes of Pronouns". Cavar, D. and C. Wilder 1992 "Long head movement? Verb movement and cliticization in Croatian", Sprachwissenschaft in Frankfiirt 7. Chomsky, N. 1982 Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Mass.: The ΜΓΓ Press. 1986 Barriers. Cambridge, Mass.: The MIT Press. 1989 "Some notes on economy of derivation and representation", MIT Working Papers in Linguistics 10, 43-74. 1993 "A minimalist program for linguistic theory", in: K. Hale and S.J. Keyser (eds.), The View from Building 20. Cambridge, Mass.: The MIT Press, 1-52.

Pronouns in Germanic and Romance languages: An overview

77

Christensen, K.K. 1984 "Subject clitics and A' bound traces", Working Papers in Scandinavian Syntax 15. Cinque, G. 1990 Types of A' dependencies. Cambridge, Mass.: The ΜΓΓ Press. 1995 "Adverbs and the universal hierarchy of functional projections", GLOW Newsletter 34, 14-15. Cooper, K. 1994 "On pronominal clitics in Zurich German", in: H. van Riemsdijk and L. Hellan (eds.), Clitics: Their origin, status and position, 62-79. Corver, N. and D. Delfitto 1993 "Feature asymmetry and the nature of pronoun movement", Paper presented at the 16th GLOW Colloquium, Lund. Couquaux, D. 1986 "Les pronoms faibles sujet comme groupes nominaux", in: M. Ronat and D. Couquaux (eds.), La grammaire modulaire. Paris: Minuit, 25-46. Cummins, S. and Y. Roberge 1993 "A morphosyntactic analysis of Romance clitic constructions". [Unpublished manuscript, University of Toronto]. Demonte, V. 1993 "Ditransitivity in Spanish: Syntax and semantics". [Unpublished manuscript, University of Madrid]. Deprez, V. 1989 On the typology of syntactic positions and the nature of chains: Move to the specifier of functional projections. [Unpublished Ph.D. dissertation, MIT]. 1990 "Parameters of Object Movement". [Unpublished manuscript, Rutgers/ Princeton University]. Diesing, M. and E. Jelinek 1993 ' T h e syntax and semantics of object shift", Working Papers in Scandinavian Syntax 51, 1-54. Dobrovie-Sorin, C. 1990 "Clitic doubling, Wh-movement, and quantification in Romanian", Linguistic Inquiry 21.3, 351-397. Engdahl, E. 1983 "Parasitic gaps", Linguistics and Philosophy 6, 5-34. Frei, Η. 1929 La grammaire des fautes. Paris: Geuthner. Giupponi, E. 1988 Pro-drop Parameter und Restrukturierung im Trentino. [Unpublished Magisterarbeit, University of Vienna], Grevisse, M. 19648 Le bon usage. Gembloux: Duculot. Grimshaw, J. 1991 "Extended projections". [Unpublished manuscript, Brandeis University]. Haegeman, L. 1991a Introduction to Government and Binding Theory. Oxford: Basil Blackwell. 1991b "Subject clitics and clitic doubling in West Flemish", in: H. van Riemsdijk and L. Rizzi (eds.), Clitics and their Hosts, 99-154. 1992a Generative Syntax: Theory and Description. A Case Study from West Flemish. Cambridge: Cambridge University Press. 1992b ' T h e distribution of object pronouns in West Flemish", in: L. Rizzi (ed.), Clitics in Romance and Germanic, 33-76.

78

Anna Cardinaletti 1993a

"Object clitics in West Flemish and the identification of A/A'-positions", GenGenP 1.1, 1-41. 1993b "The morphology and distribution of object clitics in West Flemish", Studia Linguistica. 1994 "Object clitics in West Flemish", to appear in: J. Halpern and A. Zwicky (eds.), Second Position Clitics, CLSI. this volume. "Empirical support from West Flemish", Comment on A. Cardinaletti and M. Starke "The typology of structural deficiency: On the three grammatical classes". Haider, H. 1984 'Topic, focus and verb-second", Groninger Arbeiten zur germanistischen Linguistik 25. 1992 "The basic branching conjecture". [Unpublished manuscript, University of Stuttgart], Halpern, A. and J. Fontana 1992 "X° and X max clitics", Proceedings ofWCCFL 12. Stanford: CSLI Press. Haverkort, M. 1993 Clitics and Parametrization. Eurotyp Working Papers, vol. 8.2. Helbig, G. and J. Buscha 1981 Deutsche Grammatik, VEB Verlag, Leipzig. Hellan, L. (ed.) 1992 Clitics in Germanic and Slavic, Eurotyp Working Papers, vol. 4. Hellan, L. and H. van Riemsdijk (eds.) 1993 Clitic doubling and clitic groups, Eurotyp Working Papers, vol. 5. Holmberg, A. 1986 Word Order and Syntactic Features in the Scandinavian Languages and English. Stockholm: University of Stockholm. 1991 "The distribution of Scandinavian weak pronouns", in: H. van Riemsdijk and L. Rizzi (eds.), Clitics and their Hosts, 155-173. Holmberg, A. and C. Platzack 1995 The Role of Inflection in the Syntax of the Scandinavian Languages. Oxford: Oxford University Press. Jaeggli, O. 1982 Topics in Romance Syntax. Dordrecht: Foris. Jaspers, D. 1989 "A head position for Dutch clitics or: Wilma, Wim and Wackernagel", in: D. Jaspers, W. Klooster, Y. Putseys and P. Seuren (eds.), Sentential Complementation and the Lexicon. Studies in Honour of Wim de Geest. Dordrecht: Foris, 241-252. Johnson, K. 1991 "Object positions", Natural Language and Linguistic Theory 9.4, 577-636. Josefsson, G. 1992a "Object shift and weak pronominals in Swedish", Working Papers in Scandinavian Syntax 49, 59-92. 1992b "Object shift and weak pronouns in Swedish", in: L. Hellan (ed.), Clitics in Germanic and Slavic, 51-82. Kayne, R. 1975 French Syntax. The Transformational Cycle. Cambridge, Mass.: The MIT Press. 1980 "Vers une solution d'un problfcme grammatical: *je I'ai voulu direj'ai tout voulu dire", Langue frangaise 46, 32-40. 1983 "Chains, categories external to S, and French Complex Inversion", Natural Language and Linguistic Theory 1, 107-139. 1989 "Null subjects and clitic climbing", in: O. Jaeggli and K. Safir (ed.), The Null Subject Parameter. Dordrecht: Kluwer, 239-261. 1991 "Romance clitics, verb movement, and PRO", Linguistic Inquiry 22, 647-686.

Pronouns in Germanic and Romance languages: An overview 1992

79

"Italian negative infinitival imperatives and clitic climbing", in: L. Tasmowsky and A. Zribi-Hertz (eds.), Hommage a Nicolas Ruwet. Gand: Communication and Cognition, 300-312. 1994 The Antisymmetry of Syntax. Cambridge, Mass.: The ΜΓΓ Press. Kloeke, W.U.S. 1985 "Enklitische Formen und Flexion im Bairischen und im Hochdeutschen", in: W. Kürschner and R. Vogt (eds.), Sprache, Semantik, Textlinguistik. Tübingen: Niemeyer, 73-80. Koerner, K.-H. 1984 "Deutsche Dialekte und fremde Sprachen, Teil III: Pronominal Subjektenklise", in: H. Krenn, J. Niemeyer and U. Eberhardt (eds.), Sprache und Text. Tübingen: Niemeyer, 35-44. Koizumi, Μ. 1993 "Object Agreement Phrases and the split VP hypothesis", in: J. D. Bobaljik and C. Phillips (eds.), MIT Working Papers in Linguistics 18: Papers on Case and Agreement 1, ΜΓΓ, 99-148. Koster, J. 1978 Locality Principles in Syntax. Dordrecht: Foris. Laenzlinger, C. 1993 "A syntactic view of Romance pronominal sequences", Probus 5.3, 241-270. 1994 "Enclitic clustering: the case of French positive imperatives", Rivista di grammatica generativa 19, 71-104. Larson, R. 1988 "On the double object construction", Linguistic Inquiry 19, 335-391. Lenerz, J. 1992 "Zur Syntax der Pronomina im Deutschen", Sprache und Pragmatik 29, 1-54. Madeira, A. 1992 "On clitic placement in European Portuguese", UCL Working Papers in Linguistics 4, 95-122. Motapanyane, V. 1991 Theoretical implications of the complementation in Romanian. [Unpublished Ph.D. Dissertation, University of Geneva], Nespor, M. 1993 ' T h e phonology of clitic groups", in: L. Hellan and H. van Riemsdijk (eds.), Clitic doubling and clitic groups, 67-90. Nübling, D. 1992 Klitika im Deutschen. Schriftsprache, Umgangssprache, alemannische Dialekte. Tübingen: Gunter Narr Verlag. Oberleiter, M. and I. Sfriso 1993 "Pronouns in the Olang Dialect". [Unpublished manuscript, University of Venice]. Ordonez, F. 1994 "Postverbal asymmetries in Spanish", GLOW Newsletter 32. Penner, Ζ. 1991 "Pronominal clitics in Bernese Swiss German and their structural position. Jakob Wackemagel and language acquisition", in: H. van Riemsdijk and L. Rizzi (eds.), Clitics and their Hosts, 253-268. Perlmutter, D. 1971 Deep and Surface Structure Constraints in Syntax. New York: Holt, Rinehart and Winston.

80

Anna Cardinaletti

Platzack, Ch. 1995 "Topicalization, weak pronouns and the symmetrical/asymmetrical verb second hypotheses". [Unpublished manuscript, University of Lund]. Poletto, C. 1991 "Three kinds of subject clitics in Basso Polesano and the theory of pro", in: H. van Riemsdijk and L. Rizzi (eds.). Clitics and their Hosts, 269-302. 1993 La sintassi del soggetto nei dialetti italiani settentrionali. Padua: Unipress. Pollock, J.-Y. 1989 "Verb-movement, Universal Grammar, and the Structure of IP", Linguistic Inquiry 20, 365—424. Prinz, Μ. 1991 Klitisierung im Deutschen und Neugriechischen. Eine lexikalisch phonologische Studie. Tübingen: Niemeyer. Renzi, L. 1989 "Two types of clitics in natural languages", Rivista di Liguistica 1.2, 355-372. Riemsdijk, H. van 1978 A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Lisse: The Peter de Ridder Press. Riemsdijk, H. van and L. Rizzi (eds.) 1991 Clitics and their Hosts, Eurotyp Working Papers, vol. 1. Riemsdijk, H. van and L. Hellan (eds.) 1994 Clitics: Their origin, status and position, Eurotyp Working Papers, vol. 6. Rivero, M.L. 1994 "Two locations for complement clitic pronouns: Bulgarian, Serbo-Croatian, and Old Spanish", in: van Kemenade and N. Vincent (eds.), Inflection and Syntax in Language Change. Cambridge: Cambridge University Press. Rizzi, L. 1976 "Ristrutturazione", Rivista di grammatica generativa 1.1, 1-54. 1982 Issues in Italian Syntax. Dordrecht: Foris. 1986 "On the status of subject clitics in Romance", in: O. Jaeggli and C. Silva-Corvälan (eds.), Studies in Romance Linguistics. Dordrecht: Foris, 391—419. 1991 "Proper Head Government and the Definition of Α-Positions", GLOW Newsletter 26. 1993 "Some notes on Romance cliticization", Talk given at the ESF meeting, Durham, 15-17.10.1993. Rizzi, L. (ed.) 1992 Clitics in Romance and Germanic, Eurotyp Working Papers, vol. 3. Roberge, Y. 1990 The Syntactic Recoverability of Null Arguments. Montreal: McGill-Queen's University Press. Roberts, I. 1991 "The nature of subject clitics in Franco-Provengal Valdotain", in: H. van Riemsdijk and L. Rizzi (eds.), Clitics and their Hosts, 303-330. Rohlfs, G. ! 968 Grammatica storica delta lingua italiana e dei suoi dialetti. Vol. II: Morfologia. Torino: Einaudi. Rouveret, A. 1992 "Clitic placement, Focus and the Wackernagel position in European Portuguese", in: L. Rizzi (ed.), Clitics in Romance and Germanic, 103-139. Saccon, G. 1993 Post-Verbal Subjects: A Study Based on Italian and Its Dialects. [Unpublished Ph.D. Dissertation, Harvard University],

Pronouns in Germanic and Romance languages: An overview Safir, K. 1986 Schroten, J. 1992

81

"Subject clitics and the NOM-drop parameter", in: H. Borer (ed.), The Syntax of Pronominal Clitics, Syntax and Semantics 19. New York: Academic Press, 333-365. "On Spanish definite determiners: Personal pronouns and definite articles", Recherches de Linguistique Romane et Frangaise d'Utrecht XI, 9-24.

Sportiche, D. 1989 "Movement, agreement and Case". [Unpublished manuscript, UCLA]. 1992 "Clitic constructions". [Unpublished manuscript, UCLA], Starke, M. 1993 En deuxieme position en Europe Centrale. [Unpublished Memoire de Licence, University of Geneva]. Taraldsen, T. 1986 "On verb second and the functional content of syntactic categories", in: H. Haider and M. Prinzhom (eds.), Verb Second Phenomena in Germanic Languages. Dordrecht: Foris, 7-25. Tomaselli, A. 1990 "COMP as a licensing head: An argument based on cliticization", in: J. Mascarö and M. Nespor (eds.), Grammar in Progress. Glow Essays for Henk van Riemsdijk. Dordrecht: Foris, pp. 433—445. Tomaselli, A. and C. Poletto 1992 "Looking for clitics in Germanic languages: The case of standard German", in: L. Rizzi (ed.) Clitics in Romance and Germanic, 77-102. Torrego, E. 1992 "Pronouns and determiners. A DP analysis of Spanish nominals". [Unpublished manuscript, UMass Boston]. 1995 "On the Nature of Clitic Doubling", in: H. Campos and P. Kempschinsky (eds.), Evolution and Revolution in Linguistic Theory: Studies in Honour of C.P. Otero. Washington, DC: Georgetown University Press, 399-418. Travis, L. 1984 Parameters and Effects of Word Order Variation. [Unpublished Ph.D. dissertation, MIT], Uriagereka, J. 1994 "An F position in Western Romance", to appear in: K. Kiss (ed.), Discourse Configurational Languages, Oxford. 1995 "Aspects of the syntax of clitic placement in Western Romance", Linguistic Inquiry 26, 79-123 Vikner, S. 1990 Verb Movement and the Licensing of NP-positions in the Germanic Languages. [Ph.D. Dissertation, University of Geneva, published as Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press, 1995). Vikner, S. and B. Schwartz 1991 ' T h e verb always leaves IP in V2 languages", to appear in A. Belletti and L. Rizzi (eds.). Parameters and Functional Projections. Essays in Comparative Syntax. New York: Oxford University Press. Wackernagel, J. 1892 "Über ein Gesetz der Indogermanischen Wortstellung", indogermanische Forschungen 1, 334-436.

82

Anna Cardinaletti

Werner, O. 1988

"Mundartliche Enklisen bei Schmeller und heute", in: L. M. Eichinger and B. Naumann (eds.), Johann Andreas Schmeller und der Beginn der Germanistik. München: Oldenbourg, 127-147. Zribi-Hertz, A. 1994 "Les clitiques nominatifs du frangais", in press in Cahiers de Recherche en Linguistique, Departement des Sciences du Langage, Universiti de Paris 8. Zubizarreta, M.L. 1994 "Word order, prosody, and focus". [Unpublished manuscript, University of Southern California], Zwart, C.J.W. 1991 "Clitics in Dutch: Evidence for the position of INFL", Groninger Arbeiten zur germanistischen Linguistik 33, 71-92. 1992 "Notes on clitics in Dutch", in: L. Hellan (ed.), Clitics in Germanic and Slavic, 119-155. 1993 Dutch Syntax. A Minimalist Approach. [Unpublished Ph.D. dissertation, University of Groningen].

Mila Dimitrova-Vulchanova

Clitics in the Slavic Languages1

1.

Introduction

The present paper represents an overview of the clitic elements found across the Slavic languages with a focus on the South-Slavic group. All Slavic languages, except for Belorussian, exhibit clitics of the type described by Zwicky (1977) as special clitics', that is to say, weak forms whose distribution does not follow from the rules of regular (i.e. non-clitic) syntax.2 Here belong pronominal clitics, the reflexive clitic, the forms of the auxiliary 'to be', and, in some of the languages, the question clitic li. Ukrainian and Russian are the only languages with a 'poorclitic system in that only the form by3 used in Conditionals is a clitic. On a non-standard level, Russian also has weak forms of the 2P,SG pronominal and the Dative reflexive. The most salient issues in the discussion of Slavic cliticization have been: a) the position of clitics relevant to clause (phrase) structure, and b) defining the position of clitics relative to their host, i.e. whether clitics can be treated as 'enclitics' and 'proclitics' per se. Additional, much argued problems include the number of clitic elements for each language, i.e. whether the clitic category should cover words traditionally referred to in the literature as 'particles', such as the negation word, the question word li, non-declined auxiliaries like Bulgarian future ste, etc. With respect to this issue the view of the present paper is in line with the two-partite distinction suggested by Zwicky (1985) between clitic words and non-clitic words. Thus the term 'particle', which is not based on any strict criterion, will not be part of the vocabulary of the present paper. Relevant to South-Slavic, an important issue is the phenomenon described as clitic doubling, that is to say the occurrence within the same clause boundary of both a full NP/pronominal and a clitic replica. This phenomenon can be seen against the background of similar facts from the other Balkan languages (Greek, Romanian and Albanian).4 The presentation is structured in the following way. First a survey of treatments of the basic issues of Slavic cliticization is given in section 2. In section 3 the patterns of four languages are introduced and analysed, two languages in which clitics occupy the second position in the clause, also known as the Wackernagel-position (Serbo-Croat and Czech), and two which display a more complex pattern of clitic placement (Bulgarian and Macedonian). Section 4 is a summary of the most general points of the discussion.

84

Mila Dimitrova-Vulchanova

2.

The properties of Slavic clitics

Pronominal clitics are the oldest type of Slavic clitic, attested as early as the first literary form of Slavic (Old Bulgarian). Auxiliary clitics comprising the present tense and the Conditional forms of 'to be' represent a more recent development. The modern pronominal clitics of Slavic conform with the criteria suggested by Kayne (1975) to apply to clitics in Romance, being distinct from the corresponding series of strong pronominal forms. Kayne's tests for clitichood apply to modification, conjunction, contrastive stress and use in isolation. As revealed by the examples in (1), Slavic pronominal clitics cannot be modified, conjoined, used in isolation or bear contrastive stress. (1)

a. a'. b. b'. c. d.

Maria vidja samo nego Maria saw only him(strong) *Maria vidja samo go Maria saw only him(cL) Vidim jeho a tych druhych see (I) him(STRONG) and these others * Vidim ho a tych druchych see (I) him(CL) and these others Kogo vidja? — Nego/*Go whom saw(you) — Him(STRONG)/*Him(CL) Vidjax NEGO/*GO, ane Petur saw (I) him(STRONG)/*him(CL), not Peter

(Bulgarian)

(Slovak)

(Bulgarian) (Bulgarian)

The constructions in (1) represent the only contexts in which a strong pronominal form is the grammatical option. All other contexts normally require a clitic form. Thus clitic pronouns and strong pronouns can be described as being in a complementary syntactic distribution. In contrast to pronominal clitics, the auxiliary clitics usually appear in the same syntactic position as that typical of finite auxiliaries in the respective language ([2]). (2)

a.

b.

Jas sum mu gi zel parite I am(CL) him(DcL) them(AcL) taken the money Ί have taken the money from him' (Macedonian) Jas bev mu gi zel parite I

had(NONCL-AUX) him(DcL) them(AcL) taken the m o n e y

In addition, the example in (2a) provides a first impression of a clitic cluster, i.e. the sequence 'AuxCL-DCL-ACL', and both a. and b. instantiate the phenomenon

Clitics in the Slavic languages

85

of clitic doubling in that the definite direct object parite 'the money' is doubled by the accusative clitic gi 'them'. The only difference between clitic auxiliaries and non-clitic auxiliaries resides in their fronting possibilities. Clitic auxiliaries cannot be fronted, as reflected in (3a, c), whereas their non-clitic counterparts can ([3b, d]). (3)

a.

*Süm go vizdal cesto am(CL) him(cL) seen frequently Ί have seen him frequently' b. Bjax go vizdal cesto had(I) him(CL) seen frequently Ί had seen him frequently' c. *Jsem koupil knihy am(CL) bought books(ACC) Ί bought books' d. Jsem doma am(cop) home Ί am at home'

(Bulgarian)

(Czech)

The fronting/non-fronting distinction between non-clitic auxiliaries and clitic auxiliaries is reflected in the special distinction between present tense clitic 'be' and copular 'be' in Czech which share the same morphology, but differ prosodically, shown by the contrast in (3c)-(3d) (cf. Toman 1981 for a discussion). 5

2.1.

The clitic position — derived or not?

The study of clitic phenomena has a long tradition in Slavic linguistics: (e.g., Mladenov 1929, Jakobson 1935, Siawski 1946, Gäläbov 1950, Gol^b 1953, Cyhun 1968). In this tradition the discussion of the options for clitic placement across the Slavic languages is based on the classical Indo-European study by Wackernagel (1892). 6 While linguists generally agree that the West Slavic languages (i.e. Czech, Slovak, Polish) and Serbo-Croat, a South Slavic language, comply more or less strictly with the clause 2nd position for clitics, 7 there is more disagreement concerning the placement of clitics in Bulgarian and Macedonian (South Slavic). The latter pose the following problem: Historically, both modern languages have a common basis (Old Bulgarian) which observed the clause 2nd position for clitics. However, the modern varieties exhibit a substantial deviation from this original requirement. As observed by many authors, the tendency is for pronominal clitics to congregate around the verb (cf. Slawski 1946, Gäläbov 1950, Cyhun 1968). Such an observation appears correct if examples of the non-periphrastic tenses are considered ([4]).

86 (4)

Mila Dimitrova-Vulchanova a. b.

Toj ja vidja vcera he her(CL) saw yesterday Vidja ja vcera

(Bulgarian)

saw(3SG) her(CL) yesterday

'He saw her yesterday' The examples in (4a) and (4b) provide two distinct structures, the former with an overt pronominal subject toj 'he' in clause initial position and the clitic preceding the finite verb, whereas (4b) is an instance of verb-fronting in the absence of overt subject. The clitic ja 'her' in this case appears enclitic to the verb. Examples like those in (4) have provided the reason for treating clitic placement in Modern Bulgarian in terms of movement. There are two distinct notions of movement which appear in the literature. For the traditional linguists movement is considered only in the historical sense. That is to say, the difference between an original clause 2nd position and the modern requirement(s) on clitic placement is described as arising from movement of the clitic into the domain of the main verb (cf. Slawski 1946). In contrast, Gäläbov (1950) suggests that clitics are still in the Wackernagel position, and it is the verb which has moved from a more or less clause final position.8 However the result, according to both approaches, is a contact position between the clitic and the verb (cf. [4a, b]). Gäläbov's (1950) position is surprisingly 'modern' in its treatment of clitic and clause structure phenomena. In his views, the reasons for the deviation from the original option for clitic placement reside in the changes in the overall syntactic structure of Modern Bulgarian. Among the major innovations in this respect belong the build-up of phrase structure and the development of confugurational properties in the language. In this sense, Gäläbov's approach is contrasted to that of Jakobson (1935) who seeks to explain the change in clitic placement in terms of the change of the accentuation system. As for the synchronic level, Gäläbov assumes that Bulgarian pronominal clitics are 'movable' in the sense that they seek a position early in clause structure. To this extent, such an approach can be compared to the clitic-movement approaches in the generative literature (cf. Kayne 1975, 1989, 1991). Most of the modern transformational and generative analyses of clitic placement in Bulgarian recognize the 'movability of clitics' in the sense that clitics are not en-/pro-clitic per se, and their linear position with respect to the verb is the result of syntactic or phonological adjustment (cf. R i Hauge 1976, Avgustinova 1994). For Rä Hauge the pro-clitic position as shown in (4a), is the derived one based on a transformational rule. Avgustinova'a approach, despite a commitment to the term "movable" clitic, is not directly concerned with the 'derived vs. basegenerated' problem. Instead, clitic placement has to observe two static requirements: The Linear Precedence scheme and the clause-initial restriction at the syntax/phonology interface level.

Clitics in the Slavic languages

87

A movement analysis has been suggested for Macedonian by Berent (1980). In his approach clitics are the product of definite NP movement out of S (i.e. to a left-dislocation position), whereby the trace left by the moved NP can be converted into a clitic pronoun with the same features as the NP. Berent's suggestion is geared towards explaining the crucial cases of clitic doubling found in Balkan Slavic. Another approach to the phenomenon of clitic doubling is the one exemplified in Pencev (1993). In contrast to Berent, Penöev's analysis argues for basegeneration. In essence, Pencev's suggestion is a clitic projection C1P left-adjoined to VP. Thus, the clitic pronouns are coindexed with the full NPs in VP-internal position and bear all the relevant case and definiteness features. For Serbo-Croat a motivated movement analysis is presented in Progovac (1996), whereby the contrast in placement between argument clitics and auxiliary clitics in the presence of clausal negation is explained. There exist complex movement accounts which first assume movement of the clitic(s) to its syntactic host (projection), after which phonological 're-adjustment' occurs, thus placing the clitic in its overt position. Such a phonological rule which ensures the linear position for the clitic, for instance PI (Prosodic Inversion) in Halpern (1992), is part of an approach defined as "Indirect-placement and attachment". One might argue that similar approaches can be primarily designed to cover the clause 2nd position languages, in particular the cases when the clitic chooses between the first word and, alternatively, the first major consituent.9 However, Halpern extends the analysis also to Bulgarian, whose facts are directly statable within a syntactic approach, a version of which is presented in 3.1 below.

2.2.

Enclisis vs. proclisis

The traditional descriptive accounts of clitic placement in Slavic accept the proclitic/enclitic status of weak pronominals as an inherent property of the clitic. This is reflected in the almost exclusive use of the terms 'proclitic' and 'enclitic', and not just 'clitic' which represents a more recent terminology. To the extent that traditional approaches catered for the overt linear positioning of clitic pronouns and their phonological hosts, such a view of enclisis and proclisis is self-evident. Similar ideas survive, in part, in some recent analyses like Avgustinova (1994), where the author assumes the existence of "pro-/enclitics proper". Since the elements which seem to behave as proclitics are not clitics in the sense applied in the present paper, they can hardly influence any statement on the nature of enclisis/proclisis in the language. The items considered by Avgustinova (op. cit.) are the heads of functional categories, such as 'ne' (Neg°), 'ste' (the future auxiliary) (Mod0), to be intro-

88

Mila Dimitrova-Vulchanova

duced in section 3.1 below, and their syntactic behaviour is dictated exclusively by their positions in the syntactic tree. In the analysis suggested in section 3.1 below, it will be assumed that these elements appear in their expected syntactic positions, which in the case of Bulgarian, always precede the clitic cluster in linear terms, and hence, the impression that they are proclitic by rule. A commitment to 'proclitic' as a possible label for these elements would also indirectly imply that V-main is the host in mind. However, such elements display only simple clitic properties, i.e. they are related to another word/constituent in the same linear string only prosodically. Recent accounts, based on syntactic criteria for clitic placement, suggest that clitics occupy a constant position relative to their syntactic host, subject to language variation, i.e. either left- or right-adjoined. A particularly well-grounded contribution based on evidence from Czech is found in Toman (1992), and in Progovac (1996) for Serbo-Croat, also see Pencev's (1993) base-generation approach for Bulgarian. Irrespective of whether a 'movement' approach is adopted or not, there could be observed a proclisis/enclisis asymmetry in linear ordering, as revealed by closer studies of the Romance languages (Benincä and Cinque 1993, also the survey by Cardinaletti, this volume). In particular, enclisis displays a tighter relation between the clitic and the verb. This is demonstrated in that for conjoined verbs, the clitic occurring in a proclitic position can be left out from the second conjunct as in (5a) (from Benincä and Cinque), in contrast to enclitic contexts (5b). (5)

a.

Lo

leggo e

rileggo in continuazione

(I) it(CL) read and re-read all the time

b. *Leggi e

rileggilo!

read and re-read-it(cL)

As pointed out by Benincä and Cinque (1993) examples like (5a) are only acceptable when the verbs in conjunction are close semantically, i.e. share the same root. Some Slavic languages, such as Bulgarian, display the same proclisis/enclisis asymmetry as described for Romance, with one deviation: the conjoined verbs need not be close semantically at all in order to allow the proclitic to be dropped from the second conjunct ([6a]). (6)

a.

Ivan ja cete i razbira dobre Ivan it(CL) reads and understands well b. *Procete i razbra ja dobre read and understood it(CL) well

(Bulgarian)

Clitics in the Slavic languages c. d.

Svaki dan je kupuje i cita every day it(CL) buys and reads Kupuje i cita je svaki dan buys and reads it(CL) every day 'He buys it and reads it every day'

89

(Serbo-Croat)

In contrast, Serbo-Croat ([6c, d]) does not exhibit the proclisis/enclisis asymmetry. To the extent that a similar situation is attested for Slovak (Starke, personal communucation), the difference between Bulgarian (a basically non- 2nd position language), on the one hand, and Serbo-Croat and Slovak ( 2nd position languages), on the other, can be attributed to the respective sites for clitic placement in these languages. These issues will be addressed in section 2.3 below and in the respective sub-sections on each language in section 3.

2.3.

The clitic host

All recent approaches to Slavic cliticization assume the presence of a syntactic host, usually a maximal projection, which determines the placement of clitic pronouns. For the clause 2nd position (Wackernagel) languages suggestions include C° (Cavar and Wilder 1994, this volume), and similarly, Comp 10 (Progovac 1996), WP ('Wackernagel phrase') (Rivero 1994), and IP (Halpern 1992 for SerboCroat. For Czech Toman (1986) defines a COMP-attraction rule. These approaches differ with respect to deriving the overt second position for clitics. In Cavar and Wilder this is achieved through right-adjunction to the host (C°) (Comp, for Progovac), while Rivero argues that clitics in Serbo-Croat actually occupy SpecWP. Halpern, in turn, assumes left-adjunction to IP which is augmented by the operation of the Prosodic Inversion rule. What becomes evident from these accounts is that clitics in the clause 2nd position languages are syntactically related to the leftmost maximal projection, i.e. they exhibit C-orientation, in Rivero's (1994) terms. I refrain from discussing the advantages/disadvantages of the analyses mentioned above for reasons of space. Instead an alternative description is offered in section 3 below. Discussions of the non- 2nd position Slavic languages suggest a much lower host for pronominal clitics. The traditional approaches to Bulgarian define the verb11 as such a host. Rä Hauge (1976) assumes that clitics are placed relative to the lexical category [+V] as opposed to the finite verb (i.e. predicate). Halpern (1992) and Pencev (1993) suggest left-adjunction to VP, whereas Avgustinova (1994) discusses the "verbal complex" as a cliticization site. To the extent that the periphrastic tenses in Bulgarian employ a number of auxiliaries, the latter analysis can be taken to imply AuxP as a host. Similarly, Bulgarian clitics are assumed to be I-oriented in Rivero (1994) in occupying TM 0 (TMP stands for

90

Mila Dimitrova-Vulchanova

'Tobler-Mussafia' phrase which takes the Agreement/Tense complex (IP) as complement). Less elaborate approaches state that "clitics tend to congregate around their verb" (Spencer 1991 on Macedonian). This line is carried even further in suggesting that as clitics "develop a strong attachment for a particular grammatical category, they begin to look more like (agreement) affixes" (Spencer, op. cit). Support for this claim is found in the clitic doubling phenomena of Macedonian. Indeed much of the data create the impression that the placement of pronominal clitics in Bulgarian and Macedonian is dictated by the [+V] category, alternatively VP ([7a, b]). However, a closer inspection of the periphrastic constructions reveals a more complicated situation ([7c, d]). (7)

a.

Toj go

kupi

he it(CL) bought

b.

Kupi

go/Kupuvajki

go,...

bought(3sG) it(CL)/Buying(PRESPART) it(CL)

c.

Kupil

bih

ja

knigata

(Bulgarian)

bought(PAST PART) would it(CL) the book

d.

Ί would buy this book' Knigata mu bese dadena the book him(CL) was given 'The book was given to him'

In fact, (7d), in which the pronominal clitic mu 'him' precedes both the finite auxiliary bese 'was' and the main verb dadena 'given', can pass quite easily for a 'proper' example of clause 2nd position placement. These data and the issue of the clitic host, or rather the cliticization site (here considered as more appropriate in view of the clitic clustering facts) for Balkan Slavic, are addressed immediately below.

3.

The clitic placement patterns of Slavic

3.1.

Bulgarian

Bulgarian displays four clitic types: (a) pronominal object clitics in the Accusative and Dative:

Clitics in the Slavic languages

LSG 2SG

3SG LPL 2PL

3PL

(b) (c) (d)

Full pronominal forms Nom. Acc. mene az ti tebe toj/tja/to nego/neja/nego me nas vas vie tjax te

91

Clitic forms Acc. Dat. me mi te ti go/ja/go mu/i/mu ni ni vi vi gi im

reflexive clitics in the Accusative (se) and Dative (si), auxiliary clitics comprising the present tense forms of sum 'be' - sum (lsg), si (2sg), e (3sg), sme (lpl), ste (2pl), sa (3pl), and the question clitic li.

For convenience, from now on in the text, type (a) will be referred to as Argument clitics (CL-Arg), type (b) as CL-Aux, type (c) as CL-REFL, and type (d) as CL-Q. The inclusion of the latter in the category clitic is driven by a number of considerations, such as phonological deficiency and 'special' syntactic behaviour.12 All clitic types will appear in italics in the examples throughout.

3.1.1. Bulgarian clause structure Bulgarian clause structure is here assumed to be of the form represented in (8), subject to further elaboration.13 (8)

92

Mila Dimitrova-Vulchanova

To the extent that the root/emebedded clause asymmetry is minimal in Bulgarian in that embedded clauses allow for all types of fronting, except for verb-fronting, the structure in (8) will apply to both clause types. The term FRONT introduced in (8) applies to the clause-initial position and is the site for fronted constituents, including both topicalized noun phrases and and fronted verbs, hence 'FRONT' instead of e.g., 'Topic'. It replaces the standard CP notion for several main reasons. Namely, FRONT in Modern Bulgarian is a recursive site to which a number of topicalized noun phrases can move at a time as seen in (9a). The same multiple movement option is available for Whconstituents in Wh-clauses (9b). (9)

a.

Na Ivan knigata

sum

mu

ja

vürnal

to I v a n the b o o k am(CL-AUX) him(CL-ARG) it(CL-ARG) g i v e n

b.

Ί have given the book back to Ivan' Nakogo kakvo koga sa mu pokazali? to whom what when are not

(Nor.)

In all Scandinavian languages except Icelandic,3 this construction is possible only with deficient object pronouns: neither a DP object nor a strong object pronoun can occur in a position in front of the negation (or other sentence adverbials); hence an object pronoun cannot be stressed, conjoined or modified when preceding the negation or sentence adverbial, as shown in (lib, c). (11)

a. *Ola ga Marit appelsinen ikke Ola gave Marit orange-the not b. Han köpte den (*DEN, *den dar) inte. he bought it not 'He did not buy it.' c. *Han träffade honom och henne inte he met him and her not 'He did not meet him and her.'

(Nor.) (Swd.)

(Swd.)

In the following we refer to this position as "PosCL". The availability of this position to weak pronouns has been traditionally noted both by grammarians in the tradition of Diderichsen (1946), such as Bleken (1969), and early transformational grammarians such as Hellan (1971) and, most notably, Fretheim and Halvorsen (1975). Since Holmberg (1984), the phenomenon illustrated in (9) has been referred to as "Object Shift", a label we also use here. Object Shift is parasitic on verb movement: as already mentioned, this construction only occurs in cases where the main verb has moved out of VP. In mainland Scandinavian, verb raising is only present in root clauses: hence Object Shift is only found in such clauses. In Icelandic, however, the verb raises also in embedded clauses and in control infinitives; consequently, Icelandic has Object Shift also in such constructions. (12)

a.

b.

ad peir läsu hana ekki. that they read her (i.e. the book) not 'that they did not read it.' ad lesa hana ekki to read her (the book) not 'not to read it'

(Ice.)

(Ice.)

128

Lars Hellan and Christer Platzack

When a root clause has both a non-fronted, deficient subject pronoun and a deficient object pronoun, the patterns now observed are combined in such a way that the deficient pronouns occur clustered together in PosCL, with the subject preceding the object(s). This is illustrated in (13), with weak pronouns and clitics in boldface; again, this pattern obtains in all the languages: (13)

a.

b.

c.

Hvorfor gav han henne ikke appelsinen? why gave he her not orange-the 'Why didn't he give her the orange?' Hvorfor gav han henne den ikke? why gave he her it not 'Why didn't he give it to her?' Hvorfor gav'n 'a 'n ikke? why

(Nor.)

(Nor.)

(Nor.)

gave-he(CL) her(CL) it(CL) not

2.2.2. Non-moved deficient pronouns From a theoretical point of view, Cardinaletti and Starke (this volume) claim that deficient pronouns must occur in a position different from the position where strong pronouns and full DPs occur. The fact that Object Shift of weak pronouns and clitics is obligatory in Danish, Faroese, Icelandic and Norwegian4 as soon as the main verb has left VP, is compatible with this claim. A problem for the theoretical approach, however, are cases where no visible verb movement has taken place, and Object Shift thus is not allowed. This is the case, e.g., in nonroot clauses (except in Icelandic, see [12]), as well as in cases with auxiliaries. Here, as shown by the Swedish examples in (14), all three kinds of pronouns (strong, weak, clitic) may occur in the "base" position where also full DPs and strong pronouns occur: (14)

Han hade inte träffat HENNE/ he

had not met

henne/

'na.

(Swd.)

her(STRONG) her(WEAK) her(cl)

'He had not met her.' Clitics and weak pronouns can also occur after prepositions, as in (15), just like full DPs; even when the main verb is fronted, prepositions "block" Object Shift:5 (15)

a.

Eg hugsadi um'ana. I thought about-her(CL) b. Denlä under'a. it lay under-her(CL> c. *Den la 'a under

(Ice.) (Nor.)

Pronouns in Scandinavian languages: An overview

129

Thus it seems plausible to conclude that deficient pronouns do not necessarily occupy special positions in the Scandinavian languages.6

2.3.

Pronominal adverbials

Adverbial complements like there in English he lived there are often referred to as pronominal adverbs, and like personal pronouns they seem to have both strong and deficient forms. In Scandinavian, place adverbials usually7 cannot occur inbetween the subject and the negation or other sentence adverbials, as shown by the Swedish example in (16): (16)

*Förtre ar sedan bodde han i London inte. for three years since lived he in London not 'Three years ago he did not live in London.'

(Swd.)

However, when the place adverbial is pronominal, it may occur in front of the negation ([17a]), where it cannot be stressed, conjoined or modified ([17b]): (17)

a.

Förtre ar sedan bodde han dar inte. for three years since lived he there not 'Three years ago he did not live there.' b. *För tre är sedan bodde han DÄR/ for three years ago

här och dar/ där

lived he

(Swd.)

there(STRONG)/

ute inte.

here and there/ there out not

Having mentioned this phenomenon, we will not discuss it any further in this paper.

3.

Cross-Scandinavian variation

3.1.

Variation with respect to necessity of occurrence: obligatory, optional and ot occurring Object Shift in Scandinavian languages

For ordinary Object Shift, i.e. the placement of a weak pronoun in front of all sentence adverbials in the middle field when the main verb is fronted (i.e., in PosCL), the Scandinavian languages display a variation with respect to the necessity for this phenomenon to obtain: Icelandic speakers strongly reject examples where a weak pronoun has not been shifted, as do Danish and most Norwegian speakers. Some Swedish speakers react in the same way, for other

130

Lars Hellan and Christer Platzack

speakers Object Shift of weak pronouns is an optional process, allowing cases like (18a). In Norwegian, at least the Tr0ndelag dialect allows a similar pattern ([18b]): (18)

a.

Jag säg inte'na. I

b.

(Swd.)

saw not-her(CL)

Ί did not see her.' /Es&g it]'a. I

(Tr0ndersk)

saw not-her(CL)

Ί did not see her.' On the other extreme, speakers of the Swedish dialect Älvdalsmälet, spoken in the western part of central Sweden, do not seem to use Object Shift at all (although, like other Scandinavian dialects, the dialect has verb raising in main clauses (V/2)), according to Levander (1909: 123, Anm. 2.), who explicitly states that a pronominal object must follow the negation.8 Hence we have the following (obviously incomplete as far as dialects, marked in brackets, are concerned) cross-Scandinavian variation concerning Object Shift: (19)

3.2.

Swedish

[Älvdalsmälet]

Danish

Norwegian

[Tr0ndersk]

Faroese

Icelandic

optional

non-existent

obligatory

obligatory

optional

obligatory

obligatory

Variation with respect to occurrence in between adverbials

Whereas all Scandinavian languages have Object Shift of a weak pronoun to the position in front of sentence adverbials, only Swedish can have a weak pronoun in-between such adverbials (in the case where there are several of them). The Swedish possibilities are illustrated in (20): (20)

a.

b. c. d. e.

Igär läste han dem ju alltsä troligen inte. (Swd.) yesterday read he them as-you-know thus probably not 'Yesterday he probably did not read them, you know.' I gär läste han ju dem alltsä troligen inte. I gär läste han ju alltsä dem troligen inte. I gär läste han ju alltsä troligen dem inte. I gär läste han ju alltsä troligen inte dem.

Pronouns in Scandinavian languages: An overview 3.3.

131

Object Shift with direct and indirect objects

All Scandinavian languages accept Object Shift both of a deficient indirect object and a deficient direct object, as illustrated by the examples in (9) and (10), to which may be added the Swedish and Icelandic examples in (21): (21)

a. b.

Han gav kenne den inte. Hann gaf henni hana ekki. he gav her(io) it(DO) not 'He did not give it to her.'

(Swd.) (Ice.)

In all of these examples, the indirect object precedes the direct one; the same order is found when the indirect and the direct object remains behind the negation. However, whereas only Icelandic can have the direct object in front of the indirect one when the objects remain in VP, presumably due to morphological case, only Swedish and Danish permit such a possibility in the position to the left of middle field adverbials (i.e., in PosCL). Hence the contrast in (22), between Swedish with weak pronouns and clitics on the one hand (a, b), and Icelandic (c) on the other: (22)

a. Han visade den henne inte. b. Han visade'n'na inte. c. *Hann syndi hana henni ekki. he presented it(DO) her(io) not

(Swd.) (Swd.) (Ice.)

It is to be noted that the word order in (22) is much more marked than the one in (21), and that this order is not accepted by all native speakers. If one of the pronouns is 1st or 2nd person, this pronoun tends to be interpreted as the indirect object: (23)

a.

b.

Han visade mig henne inte. he presented me her not 'He did not present her to me.' Han visade henne mig inte.

(° K IO-DO, *DO-IO)

(Swd.)

(IO-DO.9 ° K DO-IO)

In Swedish it is also possible to object shift just one of two pronouns, leaving the other behind the middle field adverbials; in the other Scandinavian languages, there is a strong tendency to front both pronouns: (24)

a. b.

Han visade henne inte den. Han visade'na inte'n. he showed-her not-it 'He did not show it to her.'

(Swd.) (Swd.)

132

Lars Hellan and Christer Platzack c. ΊΗαη gav den inte henne. (Better: tili kenne 'to her') d. Hangav'en inte'na. he gave-it not-her

(Swd.) (Swd.)

3.4. Long Object Shift Swedish can have a weak pronoun in-between the finite verb and the subject, a construction noted in Holmberg (1984) and described in Josefsson (1992); Josefsson coined the name "Long Object Shift" for this construction. Like ordinary Object Shift, Long Object Shift is parasitic on verb raising: unless the main verb has raised to its V2 position, Long Object Shift is not possible. Long Object Shift applies only to a subset of the class of weak pronouns. Most Swedes can take Long Object Shift with weak reflexives, whereas the number of speakers who accept Long Object Shift with object forms of other personal pronouns is considerably smaller. (In (25), the shifted object is in boldface, the subject in italics.) (25)

a.

b.

c.

d.

Igär kammade sig Erik inte pa hela dagen. (Swd.) yesterday combed REFL Eric not on whole day-the 'Yesterday Erik did not comb his hair for the whole day.' Klarar sig barnen pa egen hand? (Swd.) manage them(REFL) children-the on own hand 'Do the children manage on their own?' P& Stationen mötte henne hela släkten med blommor at station-the met her whole family-the with flowers och presenter. (Swd.) and gifts 'At the station her whole family met her with flowers and gifts.' Plötsligt slog honom en fruktansvärd misstanke. (Swd.) suddenly struck him a terrible suspicion 'Suddenly a terrible suspicion struck him.'

Long Object Shift is also found in Icelandic. 10 In order for a pronoun to appear in the prototypical Object Shift position inbetween the inverted subject and the middle field adverbials, it is enough that it is weak. To be shifted to the Long Object Shift position in-between the finite verb and the subject, a weak pronoun must furthermore have an object form distinct from the subject form. As a result, 3 SG pronouns like den, det (common gender and neuter gender, respectively) and dom (3 PL) cannot be used in this position. Hence (26a) with dom in the prototypical Object Shift position is wellformed, whereas (26b), where this pronoun has undergone Long Object Shift, is

Pronouns in Scandinavian languages: An overview

133

not. These cases should be compared with the corresponding ones in (27), where the distinct object form 3 SG F oss 'us' (compare nominative vi 'we') makes also the higher position available: (26)

(27)

a.

Nu now b. *Nu now

befallde ordered befallde ordered

ränaren dom thief-the them dom ränaren them thief-the

a.

befallde ordered befallde ordered

ränaren oss thief-the us oss ränaren us thief-the

b.

Nu now Nu now

att to att to

att to att to

vara be vara be

vara be vara be

tysta. silent tysta silent.

tysta. silent tysta. silent

(Swd.) (Swd.)

(Swd.) (Swd.)

As an analogue to what was illustrated in (20), noted in Holmberg and Platzack (1995: 156), also in Long Object Shift constructions, the weak pronoun is not necessarily adjacent to the finite verb. As shown by (28), one or several pre-subject adverbs may intervene between the verb and the pronoun: (28)

a.

b. c.

Nu manar oss ju inte längre nägon myndighet att now urges us as-you-know not longer any authority to äta mer bröd. eat more bread Nu manar ju oss inte längre nägon myndighet att äta mer bröd. Nu manar ju inte längre oss nägon myndighet att äta mer bröd.

The restriction of Long Object Shift to Swedish and Icelandic seems to be a modern phenomenon: Long Object Shift was earlier available in all Scandinavian languages, see the Old Icelandic and Old Swedish and Early Modern Danish examples in (29): (29)

a.

b.

ekki hryggja mig hot flin. (Olce.) 11 not grieved me threats your 'Your threats do not grieve me.' tha wceri sik then aftcekt giordhe mz then defend himself the-one (that) deprivation did with twceggia manna edhe. (OSwd.) two men's oath 'Then the one that deprived (someone of something) will defend himself with the oath of two men.'

134

Lars Hellan and Christer Platzack c.

da Roland kom ncer slottet, m0tte hannem en ridder. when Roland came near castle-the, met him a knight 'When Roland came near the castle, a knight met him.' (Early New Danish, 16th c.) 12

Long Object Shift was still a possibility in 19th century Danish and Norwegian: (30)

a.

b.

Snart indfandt sig dette. (Nor., 1833)13 soon presented itself this 'This soon appeared.' derfor forekommer mig maaske det hele mere therefore appears me perhaps the whole more bedydningsfuldt. (Dan., I860) 14 important 'Therefore the whole thing appears perhaps more important for me.'

Since Long Object Shift is not possible in present day Danish and Norwegian, these languages must have lost the construction during the last hundred years. See Platzack (1994) for an attempt to explain the loss of Long Object Shift in Danish and Norwegian.

4.

An analytic perspective

Although our aim in this article is primarily descriptive, in providing an overview of the main patterns involving deficient pronouns in the Scandinavian languages, it may be worthwile to consider the data now presented from the viewpoint of a specific analysis. Such an analysis will now be indicated, to which possible alternatives, improvements etc. may seem evident, but discussion of those will not be in the scope of the present paper. As has been noted above, weak pronominals and pronominal clitics — henceforth referred to jointly as "CL" — in Scandinavian may occur either in the "canonical" DP positions corresponding to the grammatical functions they carry (cf. [14] above), or in a "special" position which we have provisionally called "PosCL"; they occur in this position only when the main verb (henceforth "V main ") has been fronted to what is commonly referred to as the "C" (for "complementizer") position. In this section we consider the following issues: (i) how to characterize the position PosCL, and (ii) whether the occurrence of CL in PosCL is derivationally related (through movement) to an occurrence of CL in a canonical position, or alternatively CL is base generated in PosCL. In the Balkan, South-Slavic and Romance languages, pronominal clitics always occupy a designated

Pronouns in Scandinavian languages: An overview

135

position, regardless of other constellations in the clause,15 and for these languages, it is quite plausible to base-generate the clitics in that position.16 The possibility of occurrences like those in (14) for the Scandinavian clitics correspondingly opens for the possibility of a movement analysis. First we consider the possibility that PosCL is a position adjoined to Vmain, and that this cluster occurs in the position C; this is compatible both with standard assumptions about Scandinavian clause structure and standard assumptions about cliticization, and with the fact that PosCL is generally (exception only for cases like (20) and (28) in Swedish) adjacent to that of V main . C

(31) C

I

V V

Clt

This analysis can be straightforwardly tied to the circumstance that the occurrence of CL in PosCL is dependent on movement of V main to clause initial position, through the assumption that the occurrence of CL in PosCL is movement related to an occurrence of CL in a canonical position, and that this relationship is obtained by a process whereby V main and CL move together. Such an analysis can be conceived as follows: As a first step, it assumes that the clitics, base generated in a canonical position, adjoin to V main through some kind of "attraction" or "cliticization" process. Since at least clitics (if not weak pronouns) are supposed to undergo such a process anyway, and the resulting configuration inside VP in effect does obtain in cases where V main does not move to C (like in (14) above), this step has justification. The second step is the movement of V main to C in cases where V main is finite. Given the generality of this process, and the plausible assumption that clitic material adjoined to a category A moves along with A whenever A moves,17 this second step is well motivated as well. A derivation produced by simply letting step 2 follow step 1 directly produces both ordinary Object Shift like in (9b, c), (10) and (21) (cf. Hellan 1994) and Long Object Shift like in (25) (cf. Josefsson 1992). We may refer to this analysis as the "Free Ride" analysis. The Free Ride analysis is immediately called into question by other data. One type is exemplified in (32), with a clitic subject intervening between V main and the object clitic (also cf. [13]): (32)

Sä 'η 'a ikke? saw he her not 'Didn't he see her?'

(Nor.)

136

Lars Hellan and Christer Platzack

Subjects are necessarily outside the configuration V' where V main , by the Free Ride assumption, attracts the object clitic(s), assuming a "classical" underlying VP-constellation like (33),

and after formation of the V+CL cluster, one does not expect items to be able to intervene in the constellation, even if they are clitic. Another problematic point about the analysis is presented by the fact that in Norwegian, if the subject is a non-fronted full DP, object clitic (and also object weak pronoun) preposing is impossible: (34)

*Sä

Petter'a

ikke?

saw Per her not 'Didn't Peter see her?' In the Free Ride analysis, the clitics adjoined to the verb should have no way of determining the movement of the verb — they should be "invisible" as far as that movement goes. (34) is then problematic since, while a verb with no clitics adjoined has no problem with passing over a full subject DP, the presence of the clitics would somehow still determine a different course of events. It should be noted, however, that, as shown by constructions like (25a), (27b) and (28) above, no similar effect obtains with non-fronted full DP subjects in Swedish. A way of accommodating the Free Ride analysis to these points, and at the same time describing the difference between Swedish and Norwegian in the respects now mentioned, is as follows. First, the process of clitic-attraction to V may be generalized in principle to apply not only inside the V' domain, but inside any domain headed by V main , including the C'-domain after V-fronting. This domain, subsequent to V main -fronting, typically looks as follows (in accordance, largely, with the "classical" analyses of Scandinavian clause structure, and with the assumed attraction domains in boldface):

Pronouns in Scandinavian languages: An overview

137

Suppose that clitic attraction in Swedish takes place both inside V' (thus producing the wanted effects of the Free Ride analysis) and inside C', whereas in Norwegian it applies only inside C \ and that the process is capable of "jumping over" adverbials like ikke but is blocked by nominal categories (both full DPs and clitics). 18 Then all the cliticization constructions appearing in Norwegian can be attributed to the attraction taking place in the C'-cycle, including (32), provided that the cliticization is seen as an iterative process starting with the leftmost clitic and in each iteration moving the leftmost clitic. The impossibility of (34) and of Long Object Shift in Norwegian follows from the absence of the V'-cycle in Norwegian. The Swedish constructions in (32) are also produced on a C'-cycle. The simplicity of this picture first has to be held up against the question whether in Swedish, cliticization could apply first in a V'-cycle and then in a C'cycle in the same derivation; this could produce the unwanted order "V-IO-DOSu". On the other hand, the possibilities of alternating orders seen in 3.3. could perhaps be pinned on possible different domains inside V' being taken as domains for attraction or not. A second set of facts to consider are the Swedish (20) and (28), both showing one or several adverbs intervening between the verb and the pronoun. (20) could be accounted for as instances of attraction on the C'-cycle, with the addition of a stipulation particular to Swedish, to the effect that certain adverbs may "optionally" block the clitic movement. (28), however, which are "Long Object Shift" since the object clitic precedes a full DP subject, by assumption will have to be instances of V'-cyclic attraction, and this takes place in a domain lower

138

Lars Hellan and Christer Platzack

than where the intervening adverbs are situated. Thus, to accommodate (28) within the present account, one rather has to assume that certain adverbs are able to iteratively migrate up into a V-clitics cluster located in the C area and adjoin to V, a solution which could then also apply to (20). In either case, this particular behavior of adverbs would be a peculiarity of Swedish; the theoretical question is whether one wants to allow iterative movement of adverbs in the way suggested.19 A third issue to be addressed is the fact that whereas (34) is impossible in Norwegian, the pattern is possible in Danish and Swedish, as in (36) (from Vikner 1994): (36)

Hvorfork0bte Peter den ikke? why bought Peter it not 'Why didn't Peter buy it?'

(Dan.)

In the analytic setting under consideration, the possibility of (36) does not argue against having pronoun attraction on the C'-cycle, but suggests a further parameter distinguishing Norwegian from Danish and Swedish, namely an option as to whether the pronoun, once "attracted" across the negation adverb, has to end up next to the main verb (as in Norwegian), or can alternatively adjoin to an intervening, "blocking" category, such as the DP Peter in (36). What is being blocked by such a category is then only the movement of the pronoun, not the "attraction force" exerted by the verb. As has been shown, the assumption of a position like that in (33) for the clitics/weak pronominals, together with a "multi-cyclic" process of attraction of the clitic to the verb, allows for a partly straightforward implementation of the observation that the special behavior of clitics/weak pronominals in Scandinavian obtains only in connection with the occurrence of V main in fronted position. Apart from the unresolved issues mentioned, a more principled question is why such attraction should take place at all. A speculation may be that the attraction serves as a kind of "inverse reconstruction" of the VP: once V m a i n has been moved away from the VP, the attraction seeks to reconstruct the VP inside the C' domain. Especially in view of the Icelandic cases of full DPs displaying the behavior in question, and also the circumstance that weak pronominals should presumably not be attributed the "host-seeking" dependency commonly attributed to clitics, such a teleological background may be considered at least as a supplement to the more mechanical processes now suggested. Indeed, if what "drives" the nominal elements is an attempt to somehow reconstruct the VP, we should also allow for the possibility that the mechanisms implementing the distribution of CL need not be as closely tied to V-movement as the Free Ride analysis suggests. Approaches to the distribution of CL less or not at all tied to V-movement include Holmberg (1991) and Holmberg and Platzack (1995); for reasons of space we will not expound these analyses here.

Pronouns in Scandinavian languages: An overview

5.

139

Conclusion

In each of the Scandinavian languages, the distributional patterns of weak pronouns and pronominal clitics are largely identical; their main special distribution, namely, as object pronouns preceding the negation or sentence adverbial, obtains only when the main verb has undergone the "verb second" fronting typical of root clauses. Both types may occur in positions in which strong pronouns and full DPs may also occur, and while the weak pronouns have morphological paradigms identical to those of strong pronouns, there are no standardly recognized paradigms for the clitic pronouns in Scandinavian. This shows that the phenomenon of deficient pronouns, and clitics in particular, plays a far less significant role in the grammar of the Scandinavian languages than in Slavic, Balkan and Romance languages. Still, with its interaction with other phenomena such as V/2, and the variation of manifestations across the languages, the further dialectal subvariations of which have barely begun to receive descriptive and analytic attention, the topic of deficient pronouns has a central role in the study of Scandinavian syntax.

Notes 1.

2.

3.

We want to thank Gunlög Josefsson, Mila Dimitrova-Vulchanova, Cecilia Falk, Henk van Riemsdijk, Tor Äfarli and Anders Holmberg for valuable suggestions and discussion, and moral support. Historically, 'na is a reduced form of the old accusative hana (the present day object form henne has a dative origin). However, the use of 'na is not restricted to cases where we would expect an accusative to appear (direct objects, etc.); consider (i) where 'na is used as indirect object: (i) Han gav 'na boken. (Swd.) he gave her(CL) book-the 'He gave her the book.' Icelandic is exceptional among the Scandinavian languages in also admitting strong pronouns and even definite DPs to be object shifted, as illustrated in (i) and (ii): (i) Hann sä HANA ekki. he

s a w her(STRONG) not

'He did not see her.' (ii) Jolasveinarnir bordudu biidinginn ekki. Christmas-trolls-the ate pudding-the not 'The Christmas trolls did not eat the pudding.' Object Shift of strong pronouns and definite DPs is optional, and different from Icelandic Object Shift with weak pronouns, which is obligatory: (iii) Hann sä hana ekki. compare he saw her not *Hann sä ekki hana./ Hann sä ekki HANA. he

s a w not her(WEAK) h e

saw not her(STRONG)

The optional shift of strong object pronouns is presumably related to the shift of full DPs, a construction which has been widely discussed in recent years, see e.g. Bobaljik and Jonas (1993), Collins and Thrainsson (1993), Chomsky (1995). We will not discuss this construction in our paper.

140 4. 5. 6.

7.

8.

9. 10.

11. 12. 13. 14. 15. 16. 17. 18. 19.

Lars Hellan and Christer Platzack

In Swedish, Object Shift of deficient pronouns is optional for many speakers; see section 3.1. below. Likewise "particles" such as ut ('out') in Swedish, where (as the only language) the particle precedes an object even when the latter is a deficient pronoun. Theoretically we could envisage vacuous movement of the deficient pronoun in cases like (14), a description which might be maintained in the light of modem approaches to transitive verb phrases which are assumed to be composed of at least two VPs. See Holmberg and Platzack (1995: ch. 7), Chomsky (1995: ch. 4). However, it is less clear that there is any theoretical advantage in assuming that the deficient pronoun is demoted in cases where it is preceded by a preposition, as shown in (15). The restriction seems to apply to place adverbials which are arguments. Consider (i) which is fine with a place adverbial in front of the negation: (i) Man hade i Stockholm ännu inte upptäckt stölden. they had in Stockholm yet not detected theft-the 'In Stockholm they had not yet detected the theft.' Levander (1909: 123, Anm. 2) gives the following examples (translated from his phonetic notation): (i) a. Egfik int ed. i got not it Ί did not get it.' b. η säg int mig. he saw not me 'He did not saw me.' Note that the finite verb has raised to a position in front of the negation, hence the absence of Object Shift cannot be due to the absence of verb raising. This interpretation is only available for some speakers. Since there is no clitic counterpart to mig, these examples are given only with weak pronouns. Of our two Icelandic informants, Halldör Sigurösson rejected examples corresponding to (25a, c), whereas Johanna Barödal found examples corresponding to (25a) relatively bad, but examples corresponding to (25c) acceptable. The example is taken from Nygaard (1906: 357). See Falk and Torp (1900: 290). The example is taken from Aasen's Diary (Aasen 1960). The example is taken from Winter: Short stories (1860). We here disregard minor variations like those observed for the placement of pronominal clitics in Bulgarian, cf. Dimitrova-Vulchanova and Hellan (this volume). In languages with "clitic doubling", this option is even more plausible. Since no doubling takes place in Scandinavian, we leave this factor out of consideration. This is an instance of the "A-over-A" principle of Chomsky (1964), considered rather unexceptional in cases of affixation, and should reasonably be extended to cliticization as well. Also, of course, the attraction is blocked by prepositions and particles (cf. note 4). It should be noted that the following data from the Tr0ndelag dialect, where the negation adverb can intervene between V main and the first clitic, presumably should be analyzed not in terms of any "post-cliticization" movement of this adverb, but rather in terms of a high underlying position of itj 'not', thereby following from a C'-cyclic attraction of first the adverb and then the pronominal clitic. (i) Kom'itj'a? came-she-not 'Didn't she come?'

Pronouns in Scandinavian languages: An overview

141

References Aasen, I. 1960 Brev og dagb0ker (edited by R. Djupedal). Oslo: Det norske samlaget. Bleken, B. 1969 Om setningsskjemaet. [Mimeo, Univ. of Oslo.] Bobaljik, J. and D. Jonas 1993 "Subject Positions and the Role of TP", GLOW Newsletter 30: 16-17. Cardinaletti, A. this volume. "Pronouns in Germanic and Romance Languages: An Overview". Cardinaletti, A. and M. Starke this volume. ' T h e Typology of Structural Deficiency: On the Three Grammatical Classes". Chomsky, N. 1964 Current Issues in Linguistic Theory. The Hague: Mouton. 1995 Categories and transformations. [Mimeo, ΜΓΓ. (Chapter 4 of forthcoming book. Cambridge, Mass: MIT Press.)] Christensen, K.K. 1984 "Subject Clitics and A-bar-bound Traces", Working Papers in Scandinavian Syntax 15: 1-31. Collins, C. and H. Thrainsson 1993 "Object Shift in Double Object Constructions and the Theory of Case", in: C. Phillips (ed.), Papers on Case and Agreement II, MIT Working Papers in Linguistics 19. Cambridge, Mass., 131-173. Diderichsen, P. 1946 Elementar Dansk Grammatik. K0benhavn: Gyldendal. Dimitrova-Vulchanova, M. and L. Hellan (this volume) "Clitics and Bulgarian Clause Structure". Falk, Η. and A. Torp 1900 Dansk-norskens syntax i historiskfremstilling. Kristiania: Aschehoug. Frotheim, Τ. and P.-Kr. Halvorsen 1975 "Norwegian Cliticization", in: Dahlstedt (ed.), The Nordic Languages and Modern Linguistics. Stockholm: Almqvist and Wiksell, 446-465. Hellan, L. 1971 Seg, selv og syntaks. Magisteravhandling, Univ. of Oslo. 1994 "A Note on Clitics in Norwegian", in: H. van Riemsdijk and L. Hellan (eds.), Clitics: Their Origin, Status and Position. EUROTYP Working Papers, Theme Group 8: Clitics, volume 6: 80-90. Holmberg, A. 1984 "On certain clitic-like elements in Swedish", Working Papers in Scandinavian Syntax 13: 1-31. 1991 ' T h e Distribution of Scandinavian Weak Pronouns", in: H. van Riemsdijk and L. Rizzi (eds.). Clitics and their hosts. EUROTYP Working Papers, Theme group 8: Clitics, volume 1: 155-173. Holmberg, A. and C. Platzack 1995 The Role of Inflection in Scandinavian Syntax. New York and Oxford: Oxford University Press. Josefsson, G. 1992 "Object Shift and Weak Pronominals in Swedish", Working Papers in Scandinavian Syntax 49: 59-94. Levander, L. 1909 Ävdalsm&let i Dalarna. Ordböjning och syntax. Stockholm: P.A. Norstedt and Söner.

142

Lars Hellan and Christer Platzack

Nygaard, M. 1905 Platzack, C. 1994 Vikner, S. 1994

Winter, C. 1860

Norr0n Syntax. Oslo: Aschehoug. "Syntactic Differences in the Comp-domain: Mainland Scandinavian evidence for Split Comp", GLOW Newsletter 32: 46-47. "Scandinavian object shift and West Germanic scrambling", in N. Corver and H. van Riemsdijk (eds.), Studies on Scrambling. Movement and Non-Movement Approaches to Free Word-Order Phenomena. Berlin, New York: Mouton de Gruyter, 487-517. Samlede Digtninger. K0benhavn: Gyldendal.

Part II

Theory A. Feature articles

Anna Cardinaletti and Michal Starke

The typology of structural deficiency: A case study of the three classes of pronouns1

1.

On the study of pronouns

1.1.

The notion of "classes of pronouns"

It is a general property of language that words fall into classes. Among the many relevant oppositions (verbs/adjectives, transitives/ergatives, etc.), one distinguishes itself from all others: that instantiated by the opposition between different classes of pronouns. This opposition is unique in regularly contrasting synonymous pairs', in cutting across all components of grammar; in having no systematic correlation with any interpretive characteristic (semantic or phonetic); in determining a large set of (apparently) absolute universale; and in cutting across lexical classes, §1.1.1-5. The fundamental goal of the present inquiry is to uncover the primitive underlying these exceptional classes. 1.1.1. What appears to be one pronoun (semantically/functionally defined) falls into distinct classes. The Italian third person plural feminine nominative pronoun, for instance, divides into two distinct classes with respect to coordination and reference: (1)

a. b.

Esse Loro 3PL,FEM,NOM

(+hum) (-hum) (*e quelle accanto) sono troppo alte. / / (e quelle accanto) sono troppo alte. Du En Fr Pr Ct Sp It Ru Gr XI He B1 SC Cz SI Po Ru

I-

ο

Ο

-

-

-

±

±

-

+

+

+

+

+

-

-

±

-

-

-

Hng(4): singular direct object pronouns are dropped, indirect object pronouns behave like all other case-marked ones. Wis

* 'M gwelodd. * lS-ACC(CL) saw-3S cca: 'He saw me.'

Ir: The PRT is sentence initial, but it can be followed by an ACC pronoun. Is e Seän a cheannaigh an leabhar sin. COPprt 3SM-ACC Sean COMP bought the book that cca: 'It is John who bought that book.' Hbr(l): only when a topic, then the pronoun is strong. lo, natati matana to-him gave-lS present 'To him I gave the present.' Swd(l): it is not clear whether the pronoun is a real weak pronoun in this case. Frequent with 3S den, det , more questionable with other pronouns. Det vet pag ingentmg om. it know I nothing about Ί don't know anything about it. Den At jag. it ate I 'I ate it.' Ice

Honum gaf eg bokina• him gave I book-the-DAT 'I gave the book to HIM.'

? Mej s&q hon. ? me saw she cca: 'She saw me. Hana gaf eg manninum. it-F gave I man-the-DAT 'I gave THIS to the man.'

Grm

* Es hat Hans schon immer haben wollen. * he has Hans always have wanted cca: 'John has always wanted HIM.'

Dut

zag ik gisteren. * 3SM-ACC(CL) saw I yesterday cca: 'I saw HIM yesterday.'

Clitic Questionnaire Eng

Him /*'M I gave it to.

Fr

* Le, je le vois • * 3SM-ACC(CL), I 3SM-ACC(CL) see cca: 'It's him, who I can see.'

Prt

*Lhe, (o Pedro) deu um livro. *3S-DAT(CL) (the Pedro) gave-3S a book cca: 'To him, Pedro gave a book.'

Ctl

Ell / * El no el vaig veure • Him / * 3SM-ACC(CL) no 3SM-ACC(CL) Aux-lS see 'Him I didn't see.'

949

It(l): impossible e.g. in left dislocation contexts. (2)

Alb

Lui /*Lo, Gianni l'ho incontrato. 3SM-ACC/*3SM-ACC(CL), Gianni 3SM-ACC have-3S met 'Him, Gianni has met him.' Ε pashe • 3SM-ACC(CL) saw-I 'I saw HIM.'

Mod(2): there are CLs preceding the verb, but it is not clear whether they are in topic position. Mu 22 dade. 3SM-DAT(CL) 3SN-ACC(CL) gave-he 'He gave him/it to HIM.' Big (2)

Nego /*go vidi ah vcera. 3 SM-ACC/* 3 SM-ACC(CL) saw-I yesterday 'I saw HIM yesterday.' Na nego/*mu napravih podaräk. 3SM-DAT/*3SM-DAT(CL) made-I present 'I made a present for HIM.'

SCr(1)

* Ga jie vidjela juöer. * 3SM-ACC(CL) be-3S(CL) saw-she yesterday cca: 'HIM she saw yesterday.'

Cz

Tobe /*ti to nedam. you-DAT /*2S-DAT(CL) 3SN(CL) not-gave-lS 'I won't give it to YOU.'

Some pronouns have the same weak and strong forms - those appear in the initial position. E.g. to (3SN-N0M/ACC)- see 2.2. Some CLs can also be preceded (in colloquial Cz) by a glottal stop only. To ti nedam. 3SN-ACC(CL??) 2S-DAT(CL) not-give-lS Ί won't give THIS to you.'

' si myslim. REFL think-lS 'That's what I think.'

Siva

Jana /*Ma vide! . Jano-ACC/*me-ACC(CL) saw-he 'He saw Jano/me.'

Rue (2 )

Teb-ja /*Tja ja u£e slysal. 2S-ACC/*2S-ACC(CL) I already heard 'I have heard you already.'

950

Riet Vos and Ludmila Veselovskä

3.10. Does your language have a construction in which the weak pronoun ie related to a quantifier? a) yes, b) no

Bsg(2)

Mirenek denak ikusi gaitu. Mary all seen us-have 'Mary has seen us all.'

Ir: NOM and ACC pronouns progressive form CL cannot.

can

be

related

to

a

Chuaigh siad go leir go Meiriceä. went they all to America 'They all went to America.' Chonaic Maire iad go leir inne. saw Mary them all yesterday 'Mary saw them all yesterday.' * Tä se ά gcuardach go leir. * is he 3P look-for (verb, noun) all cca: 'He is looking for them all.' Brb

i -zra-tn Ahmed Kullu-tn. 3SM(CL)-saw-CL Ahmed all -3PM(CL) 'Ahmed saw them all.'

Swd(l)

Maria s&g dem inte alia. Maria saw them not all 'Maria did not see them all.'

Grm

- weil sie Maria alle gekannt hat. - because them Maria all known has '...because Maria has known them all.'

Dut (2)

Karel kent ze allemaal. Karel knows 3P(CL) all.

Eng

I gave them/'m all a book.

Fr

Marie les a tous vus . Marie 3P-ACC(CL) has all seen 'Mary has seen them all.'

Prt

A Maria comprou-os todos . the Maria bought -3SM-ACC(CL) all 'Maria bought them all.'

Ctl

La Coia les ha vist(es) totes. the Coia 3P-ACC(CL) Aux-3S see all 'Coia has seen them all.'

Spn

Maria los ha visto Maria 3PM-ACC(CL) has seen 'Maria has seen them all.'

a todos. to all-3PM-ACC

quantifier,

the

Clitic Questionnaire

951

Maria le pedira el libro a alquien. Maria 3-SM-DAT(CL) will-ask the book to someone-DAT 'Maria will ask somebody for the book. ' It(l)

Maria li ha visti tutti. Maria 3PM-ACC(CL) have seen all 'Maria has seen all of them.'

Rum(1)

Maria i -a väzut pe tot'i Maria 3P(CL)-has seen PRT all 'Maria has seen them all.'

Grk

I Maria tus exi (plus) di plus. the Maria 3PM(CL) has (all) seen all 'Maria has seen them all.'

Alb

Meri i pa te qjithe. Mery 3P(CL) saw all 'Mary saw them all.'

Med

Mari ja gi vide site. Mary 3P(CL) saw all 'Mary saw them all.'

Big(2)

Maria gi vidja vsiSki Mary 3P(CL) saw all 'Mary saw them all.'

SCr(1)

Marija ih j® (sve) vidjela (sve) . Mary 3P(CL) Aux-3S(CL) (all-ACC) saw (all-ACC) 'Mary saw them all.'

Cz: GEN (CL) is obligatory in NOM and ACC with indefinite quatifiers and cardinals higher than 4. With other quantifiers, in non-NOM Cases the GEN(CL) is optional. See also 2.3. Bylo * /nas pSt/mnoho . was-3SN /lP-GEN(CL) five/many 'There were five/many of us.' Vidgl */jich pet/mnoho. saw-3S /3P-GEN(CL) five/many 'He saw five/many of them.' Marie (je) uz znä v£echny. Mary [3P-ACC(CL)] already know-2S all-ACC 'Mary already knows all of them.' Siva: see also 2.3. above. On ich uz vSetky videl. he 3P-ACC(CL) already all-ACC saw 'He already saw all of them.' Pol

On ich jui wszystkich widzial. he 3P-GEN(CL) already all saw-3S 'He has already seen them all.'

RUB(2): see 2.3. above.

952

Riet Vos and Ludmila Veselovskä

3.11. Does a CL allow a Parasitic Sap (PO)? a) yes, b) no. H» Fi Bs W1 ir Br Hb Sw No Ic Fa On Du En Fx Pr Ct St> It Xu Or Al MC B1 SC Cz SI Po Ruj 1 -

ο Ο

-

-

+

±

-

-

-

ο

+

+

Ir: PGs do not seem to exist at all in Irish. Brb

t -sRa -tn bla ad-in -t -izar 3SF(CL)-bought-3PM(CL) without-3PM-3SF-saw 'She bought them without seeing them.'

Hbr(1)

?sefer ze, hixlateti lo 1- haxzir-o ?book this,decided-I not to-return-it 'I decided not to return this book - la-sifryia bli 1-ikro. -to-library without to-read -to the library without reading it.'

Swd (1)

Johan kastade den Sei snart han hade Johan threw it as soon he had

- last den/(*) i papperskorgen. - read it /(*) in paper-basket 'Johan threw it into the waste paper basket after reading it. Orm

Er hat's [ohne () hinzusehen] getan. he has-it [without () at-to-look] done 'He has done it without looking.'

Dut - dat ze ' t zonder in te kijken teruggezet heeft. - that she 3SN-ACC(CL) without in to look back-put has '- that she put it back without looking.' Eng: the CLs appear to be in argument positions in syntax. Fr: unless below. ?? J'en ai pr4sent4 le copain a la copine. ?? I CL have presented the friend-M to the friend-F 'I have introduced a friend to a friend.' Prt

* Vi -ο na televisäo sem reconhecer. * saw-I-3SM-ACC(CL) in television without to+recognise cca: Ί saw him on TV without recognising (him) .'

Ctl

Te 1 'has de beure /e/ sensa abocar -la. you 3S-ACC(CL)'have of drink /e/ without spill-3S-ACC(CL) 'You have to drink it without spilling (it).' * Te 1 'has de beure /e/ sensa abocar /e/. * you 3S-ACC(CL)'have of drink /e/ without spill-/e/ 'You have to drink it without spilling (it).'

Spn: DAT CLs seem to license PGs - for some speakers.

+

+

Clitic Questionnaire

953

% Al millonario le vendieron la finca % to millionaire 3SM-DAT(CL) sold the land 'They sold the land to the millionaire - despue de emborrachar. - after to-make-drunk - after making him drunk.' It

* Glieli dobbiamo far mettere /e/ nello * 3SM-DAT/ACC(CL) must-3PM make put /e/ on-the cca: 'They must put it/them on the - scaffale invece di lasciare /e/ sul tavolo. - shelf instead of leaving /e/ on-the table - shelf instead of leaving it/them on the table.'

Grk

* I Maria ta agorase fxoris na dokimasi • * the Maria 3PN-ACC(CL) bought [without subj-PRT try cca: 'Maria bought them without trying.'

I Maria ta agorase [xoris na ta dokimasi] . the M. 3PN-ACC(CL) bought [without subj-PRT 3PN-ACC(CL) try] 'Mary bought them without trying them.' Bljr(2) * ge tja go värna, be2 da be§e procela. * that she 3SM-ACC(CL) return without to was read '- that she put it back without looking (at).' SCr (1)

Moramo ih vratiti na policu must-we 3P-ACC(CL) to-return on shelf 'We must return them to the shelf umjesto da ?ih /(*) ostavimo na stolu. instead-of to 3P-ACC(CL) /(*) leave-we on table instead of leaving them on the table.'

It is also impossible with a PG in a relative clause. Knjige koje moramo vratiti na policu books which-ACC must-we to-return on shelf 'books which we must return to the shelf - umjesto da ih /(*) ostavimo na stolu. - instead-of to 3P-ACC(CL) /(*) leave-we on table instead of leaving them on the table.' Cz: marginal cases. No parasitic gaps with verbs. May be found with verbal nominals, but verbal nominals do not require (and mostly do not allow) overt arguments. Vratil je /ty knihy na policireturned-he 3P-ACC(CL) /the books on shelf (verb)

- aniz /misto aby - without/instead-of

je / (* ) £etl. 3P-ACC(CL)/(*) read-PAST-3SM

(v. noun) - be ζ Steni (*je /*nich /*knih) - without reading (3P-*ACC(CL)/*GEN(CL)/* books) 'He returned them/the books to the shelf (without) reading (them/the books).'

954

Riet Vos and Ludmila Veselovskä

Siva: with verbal nominals, not with verbs On ich kilpil bez äitania /* aniz Sital. he 3P-ACC(CL) bought without reading /* without read-PAST-3S 'He bought them without reading them.' Pol: marginally Poloäyiem go na polce zamiast zostawic na stole. put-I 3SN-ACC(CL) on shelf without leave-INF on table 'I put in on the shelf instead of leaving it on the table.' Rus:

assuming weak pronouns are relevant here.

On sjel ka§u? On jejo vybrosil vmesto togo, ctoby jest. he eat cereal? he 3SF-ACC threw-away instead it that eat-INF 'Has he eaten the cereal? He threw it away instead of eating it.'

3.12. Does your language have CL doubling? a) yes, b) no. Hn Fl Bs Hl Ir Br Hb Sw NO Tc F& on DU En Fr Pr et Sp It Ru Or Al MC B1 SC Cz 81 Po Ru

Λ

-

+

+

-

+

+

-

-

-

Ο

+

±

-

*

+

+

+

±

*

+

+

+

+

-

*

-

-

i

Hng: some verbal prefixes may be regarded as doubling (3). It occurs probably with oda (there). Also with sem (negative polarity item?) (4) . be-ment a kert - be in-went the garden-in '(He/she) went into the garden.' Senki sem olvas semmi-t (sem) . no one neither reads nothing-ACC (neither) 'Nobody reads anything.' Ret -biz -om Kati-ra a gyereket. on-3S-entrust-lS Kati-on the child-ACC 'I entrust the child to Kati.' Kati-ra biz -om (rA) a gyereket. Kati-on entrust-lS (on) the child 'To ΚΑΤΙ I entrust the child.' Bsq(2): if agreement markers are CLs, then doubling happens when the argument NP is overt. Zuk guri liburua eman di -gu-zu. you us book given have-us-you 'You have given us the book.' Wie: except in passives, whenever a head is preceded by a CL, it can be followed by an ordinary pronoun. Fe 'm gwelodd i. PRT IS saw-3S I 'He saw me.'

ei dy ef 3SM house he 'his house'

Clitic Questionnaire

955

Mae Gwyn wedi fy ngweld i. is Gwyn after IS see I 'Gwyn has seen me.' Brb: with SUBJ and DAT CLs only. i -f feR Ahmed 3SM(CL)-went out Ahmed 'Ahmed went out.' i -fa -as H= i Ahmed/urba 3SM(CL)-gave-3SM-DAT(CL)-3S-ACC(CL) to Ahmed/a boy 'He gave it to him/to Ahmed/to a boy.' Hbr(l): doubling occurs with a possessive CL. ze sifr-o Sei Dani this book-his of Dani 'This is Dani's book.' Swd

Johan s&g den uppenbarligen inte bilen. Johan saw 3SN-ACC(CL) apparently not car-the 'Johan apparently did not see it/the car.'

Grm: only with the following example da ('there' with no location reading possible) Da hat sie nicht mehr dran geglaubt. there has she not anymore therein believed 'This she doesn't believe it any more.' Dut: a possessive weak pronoun may double, see 3.13.(1). Westflemish has subject CL doubling (3). Fr: not officially. Possible with possessives and postverbal subject (characteristic intonation) which do not seem to be dislocations. Mon copain ä moi il est grand. 1S-P0SS friend to mine he is tall 'My friend (of mine) is tall.'

Prt

Je le ferai moi avant de partir. lS-NOM(CL) 3SN-ACC(CL) do-FUT IS before leaving 'I will do it before leaving.' Dei -lhe ο livro a ele. gave-lS-3S-DAT(CL) the book to 3SM 'I gave him/the book to him.'

Ctl: see 2.1. and the following case of (optional) doubling with 10 CLs. Also right- and left-detached phrases, which appear adjoined to the clause, must bind a CL copy: this configuration must not be mistaken for CL doubling. (Li) vaiq donar el didal a la dida. [3S-DAT(CL)] Aux-lS(CL) give the timble to the nurse 'I gave the thimble to the wet nurse.' Spn: see 2.1/2. It(l): doubling of an indirect presence of a direct object CL.

object

occurs marginally

in the

956

Riet Vos and Ludmila Veselovskä Gliel 'ho dato a Gianni . 3SM-DAT(CL)+3SM-ACC(CL)'have-lS given to Gianni Ί gave it to him=Gianni.' (2)

Rum (1)

* Gli do il libro a Gianni. * 3SM-DAT(CL) give-lS the book to Gianni cca: 'I give it-the book to Gianni.' il vede pe fratele sau. 3SM-ACC(CL) see-3S PRT brother-the his 'He/she can see him=his brother.'

Grk: see 3.1. Alb

_Ia dhashe librin Xhonit. 3SM-DAT+3SN-ACC(CLs) gave-I book-the-ACC John-the-DAT 'I gave (him it) the book to John.'

Med

Mu 22 dadov molivot 3SN-DAT(CL) 3SM-ACC(CL) gave-lS pencil-the 'I gave the pencil to him.'

Big(2)

Dadoch mu ja knigata na Ivan. gave-IS 3SM-DAT(CL) 3SF-ACC(CL) book-ACC Ivan-DAT Ί gave (him it) the book to Ivan.'

SCr (1)

* Da j em mu knj igu Ivanu • * give-lS 3SM-DAT(CL) book-ACC Ivan-DAT cca 'I give him the book to Ivan.'

Cz: right dislocated NP's and PP's divided by a pause/comma. ? (*Tu knihu) On ji/to tarn nedal, tu knihu. ? (*the boook) He it-F/NACC(CL) there not-put-3SM, the book 'He has not put it there, (I mean) the book.' In relative clauses with genderless and non-case marked!?) Wh-word + resumptive pronoun(CL) - ten muz, co /*kter&nu jsme mu to dali . - the man what/*whom-DAT Aux-IP(CL) 3SM-DAT(CL) 3SN-ACC gave '- the man who we gave it to.' Otherwise no standard doubling. * Dal mu to Petrovi . * Gave-he 3SM-DAT(CL) 3SN-ACC(CL) Peter-DAT cca: 'He gave it to him to Peter.' Siva

* Peter ho videl Jana. * Peter-NOM 3SM-ACC(CL) saw Jano-ACC cca: 'Peter saw Jano.'

Rus(2) not a standard doubling. Only right dislocated phrases may be doubled, divided by a pause and a comma. In questions the expletive (emphatic?) 3SN-NOM7ACC? eto (it) may cooccur with a Wh-word. * Ja jejo sjel ka§u. * I 3SF-ACC eat cereal-ACC cca: Ί eat it the cereal.'

Ja jejo si el, kasu. I 3SF-ACC eat // cereal-ACC 'The cereal, I eat it.'

Clitic Questionnaire

957

Kuda (eto) ty idjos? when (3SN) you go 'Where's it, you're going?'

Cto (eto) tv nes-jos? what (3SN) you bring 'What's it, you're carrying?'

3.13. If you answered (III.12) positively which CL may double? (a) nominal CL, (b) pronominal CL, (c) prepositional CL, (d) verbal CL, (e) adverbial CL, (£) possessive CL. (only cases

that appear

are given

below:)

Γ

Hn Fi Β· W1 Ir Br Hb Sw Ho Ic Fa em Du En Fr Pr ct SP It Ru Or Al He B1 SC Cz SI to Ru 1 ? 1 i ? |pron ? Ο + + i ο - i i i 1 Ο - + i ο + i i i i - + i + - - - - - ? - - - i - 1 i L». Hng(4): a negation CL and pronominal-like items (P's?) may double. A case marked pronominal may double (probably). Wis: see 3.12. Hbr(l): see 3.12. Dut(l): the doubling possible only in some dialects and allowed only with NP + CL-3SP + N. Jan ζ ' η broer Jan his-POS(CL) brother 'Jan's brother'

vs

* Mi j η m' η zus * My-POS my-POS(CL) sister cca: 'My sister'

Fr: see 3.12. Prt: see 3.12. Ctl: see 2.1. and 3.12. Spn: see 2.1/2. It: see 3.12.(1) Rum: see 3.12. and 3.15. (1) Grk: (b) see also 3.1. Tu iposxethika tu Petrou /? enos pedhiou3S-DAT(CL) promised-lS the Peter-DAT/? a child-DAT Ί promised to him/to Peter/? to a child .... Ton theoro ton Petro / * enan fllo eskipno. 3SM-ACC(CL) consider the Peter-ACC / * a friend clever 'I consider Peter/*a friend clever.' (f) with POSS CLs the more plausible analysis may be right dislocation (appositive: sharp intonational marked constructions, etc) . ? Jemise [to potiri tu] tu Petrou. ? fill-IMP the glass 3SM-GEN # the Peter 'Fill his/Peter's glass.'

Riet Vos and Ludmila Veselovskä

958 Big (2)

Tjah qi obiSam. 3P-ACC 3P-ACC(CL) love-ls Ί love them.' VsiSkite mu prijateli gi obiöam all his-POS(CL) friends 3P-ACC(CL) love-lS 'I love all his friends.'

Rue(2) see 3.12 . above

3.14. If your language has CL doubling of the type in (3.13a), is the double: (a) a pronoun, (b) a full definite NP, (c) a full indefinite MP, (d) an adverbial CL. (only

I

cases

that

appear

are given

below:)

Hn Pi Β· W1 Ir Br Hb Sw No IC Fa On DU En Fr Pr Ct Bp It Ru Qr Al He B1 SC Cz SI Po Rug

| pronoun i ο |d«f. NP i ο find. NP i ο

+

+

+

-

+

-

i i i

-

+ •

i i i i ο i i i i i i i i i i i ο i i i i i i i i i i 1 ο 1 i i i 1 i i

-

+ -

i i i i i i i i i

+

+

+

+

-

±

i i i i οD i i i 1 οD i i i i οI

Wis: see the examples in 4.12 and 4.15. with pronouns. Brb: see 3.12. It: the example in 3.12. is with +DEF NP. Big: indefinite NP must be 'specific'. Rue(2): see 3.12. above.

3.15. If your language has CL doubling of the type in (3.13b), what is the function of the CL? (a) subject, (b) indirect object, (c) direct object, (d) prepositional object.

j

En Fi Βι W1Ir Br Hb SwNOIc Fa an Du En Fr Pr Ct SP It Ru Gr Al M CB1 SC Cz 31 Po rJ

|]eubject i Ο 1 10 i Ο 1 00 i Ο |pr«p.ο. i ο

i i + i ϊ i

-

+

+

-

+ -

+

+ -

i i i i

i i i i i i i i i i i i

ο ο ο ο

-

+

-

-

-

-

+

-

i i i i i i i i

-

-

-

-

-

-

-

-

Hng(4): cannot be answered given the options. See 3.13. Brb: see 3.12. Wis

Dywedodd Emrys ei fod ef yn oer. said Emrys 3SM be he in cold 'Emrys said he was cold.' Mae Gwyn yn ei ddarllen ef. is Gwyn in 3SM read he 'Gwyn is reading it.'

-

i

-

+

+

i i i

+ -

i i i i

i i 1 i r

Clitic Questionnaire

959

Qrm: see 3.12. Dut(3): concerns West Flemish. Prt: (b) see 3.12.

(c)

Encontrei-os a eles. met-lS -3P-ACC(CL) to 3P-ACC 'I met them.'

Ctl: see 2.1. and 3.12. (In)Direct object strong personal pronouns require CL doubling. Full NP IOs display optional doubling. Some dialects require it, others do not allow it at all, and a third group are characterized by rampant variability. CL doubling is not allowed with full NP DO in any other context. Spn: see 2.1/2. It(l): CL doubling of an indirect object only in the presence of a CL direct object, see (3.12). Rum(l) (b)

il dä fratelui säu /lui 3SM-DAT(CL) give-3S brother-the 3SM-DAT/3SM-DAT 'He gave him=his brother a book.'

ο carte. a book

(c) .ίΐ vede pe fratele säu / el. 3SM-ACC(CL) see-3S PRT brother-the his / 3SM-ACC 'He/she can see him=his brother.' Ork: With 10 (DAT/GEN) a pronoun can double (in)definite NP. With DO (ACC) a pronoun can double only definite NP. See 3.13. Rus (2): see 3.12. above.

3.16. Can the CL be present when the double has been questioned? a) yes, b) no. Hn rl Ml Ir Br Hb Bw Bo IC Fa Cta Du Bn rr Pr ct Sp It Ru Or Al He B1 SC Cz SI Po Ru Β· -

ο

-

+

i

+

-

i

i

1

ο

+

-

i

i

-

i

?

i

1

-

Bsq(2): a wh-phrase is 3rd person and the 3rd person system has no pronominal CLs. Wis

Beth roedd Gwyn yn ei wneud? what was Gwyn in 3SM(CL) do 'What was Gwyn doing?'

Brb: the DAT CL, however, changes its morphological form ma mi-t -fa? who to-3SN(CL)-3SM(CL)-gave 'To whom did he give it?' Grm

Wo hat er drüber gelacht? where (no location) has he thereabout laughed 'What has he laughed at (it)?'

960

Riet Vos and Ludmila Veselovskä

Prt

* Α quem lhe deste ο livro? * to whom 3S-DAT(CL) gave-2S the book cca: 'Who did you give the book to?'

Ctl: in some dialects (see 3.15.). A qui (li) vas donar el didal? to who [3S-DAT(CL)] Aux-2S give the thimble 'Who did you give the thimble to?' It(l)

A chi gliel 'hai dato? to whom 3SM-DAT(CL)+3SM-ACC(CL)'have-2S given 'Who did you give it to(him)?'

Rum(2) : for indirect speech. quel/lequel. (1)

ACC CL should be between que/qui and

cui ii da cartea? who-DAT 3SM-DAT(CL) give-3S book-the 'Who does he/she give the book to?'

Ork: only with partitive or D-linked wh-phrases and 10 extractions (if those are examples of doubling). * Pj on ton idhes? * who-ACC 3SM-ACC(CL) saw-2S cca: 'Who did you see (him)?' Pj on apo tus filus su (ton) agapas perissotero? who-ACC of the friends your [3SM-ACC(CL)] love-2S most 'Which of your friends do you love (him) most?' ? Pjanou (tu) iposxethikes ena taksidi stin Ameriki? ? who-DAT [3SM-DAT(CL)] promised-2S a trip to-the States 'Whom did you promise a trip to the States (to him)?' Med Kogo go vidov? who-ACC 3SM-ACC(CL) saw-lS 'Whom did I see?'

Komu mu rece? who-DAT 3SM-DAT(CL) said-3S 'To whom did (s)he say (it)?

Big: occurs only very rarely in colloquial language (1)(2). (l) Na kogo mu se obadi? to-whom-DAT 3SM-DAT(CL) REFL called-2S 'Who did you call?' Cz: seems possible only with interrogative 3SN-ACC?(WH) + deictic (expletive) 3SN-ACC?. Koho /Co jsi to /*ho /*j i videl? who-ACC/what-ACC Aux-2S(CL) 3SN-(CL)/*2SM/F-ACC(CL) saw-SM 'Who/What have you seen?' RUB: see examples with nonthematic eto in 3.12.

Clitic Questionnaire

961

4.A.POSITION 4.λ. Does your language have weak pronouns that must attach to a host? a) yes, b) no ΠΗη Fi Bs Hl Ir Br Hb Sw No Xc Fa

L 'hi donem. 3S-ACC(CL) CL give-lS

There are other variants depending on dialects, registers and CL functions. Therefore, alternatively, the relationship between syntactic constituents and the morphemes in the CL string could be established by means of morphophonological rules. es - et/us - em/ens - 1/ho - ζ - en - hi REFL - 2p.(CL) - lp.(CL) - 3p.(CL) - pi. - CL - CL Spn:

- DAT ACC

Se lo (*se) he dado. 3S-DAT(CL) 3SN-ACC(CL) [*3S-DAT(CL)] have given Ί have given it to him/her.' It:

-

DAT ACC DAT 1,2,3 (even REFL) } LOC } 3ACC/ ne si ne CL si

Rum: (1) (2)

- DAT ACC - both ACC DAT and DAT ACC order is possible - ACC(CL) and DAT(CL) = Pron) within the limits given by the order: - Comp - Neg - Pron - Adv - Aux(Perf) - Neg(Pron) - Aux - Adv - Aux(Perf) - V(Pron) Grk: DAT(GEN)-ACC. Except a very few imperatives. See 3.2. Su -to -ipa. 2S-DAT(CL)+3SN-ACC(CL)+said 'I told you this' Alb:

*To-su-ipa. *ACC-DAT

- DAT ACC/REFL - DAT + ACC form a cluster, IS 2S 3S IP 2P 3P

me + e / te + e / 1 + e / na + e / 1U + e / + e / U

1 i 1 1 i i

=>

=> => =>

=>

=>

ma ta ia na e jua ua

Med:

- Aux/Q/Condit - DAT - ACC/REFL

Big (1) :

- DAT ACC - NEG ARG Q

/ / / / / /

m' i t'i ia na i iua ua

- Q ARG SCr(2) Cz:

- Aux/COND - DAT - GEN/ACC - REFL - root je (be)

972 - Q

Riet Vos and Ludmila Veselovskä PRT-Aux-REFL(DAT/ACC)-non-arg.DAT-DAT-ACC/GEN-(Adv)

Jä j sem se ti tu X [Aux-lS REFL-ACC 2SF-DAT-purpose 3SF-DAT]-CLs here-CL? - musel omluvitI - must apologise 'Imagine, here I had to excuse myself to her.' There are, however, some co-occurrence restrictions: For non-REFL argument DAT and ACC the order DAT-ACC is obligatory and they can cooccur, but if non-REFL DAT is present, non-REFL ACC cannot be 1st or 2nd person (with the exception of the combination of 1S/P-DAT mi/nam and 1S/P-ACC which is still acceptable). Pak me/te /ho / j i /näs/väs/je pr edstavil. then 1S/2S/3SM/3SF/1P /2P /3P-ACC(CL) introduce 'Then he introduced me/you/him/her/us/them.' Pak mi/ti/mu/ ji/näm/vam/jim ho / j i /je predstavil. then 1S/2S/3SM/3SF/1P/2P/3P-DAT(CL) 3SM/3SF/3P-ACC introduce 'Then he introduced him/her/them to me/you/him/her/us/them.' Pak mi/?näm *mg/te/*näs/väs predstavil • then 1S/71P-DAT(CL) *1S/2S/*1P/2P-ACC(CL) introduce 'Then he introduced *me/you/*us to me/?us.' * Pak ti/mu / j ί /vam/jim mS/te/näs/väs predstavil. * then 2S/3SM/3SF/2P /3P-DAT(CL) 1S/2S/1P/2P-ACC(CL) introduce cca 'Then he introduced me/you/us/you to him/her/you/them.' Siva: see 4.B.4. Pol

Przedstawiiem (mu) cj° (mu) . . . introduced-IS [3SM-DAT(CL)] 3SM-ACC(CL) [3SM-DAT(CL)] Ί introduced him to him.' Przedstawig (ci) go (? ci) . • • introduce-lS [2S-DAT(CL)] 3SM-ACC(CL) [? 2S-DAT(CL)] Ί will introduce him to you.'

4.B.7. If you answered (4.B.5) positively, does the grammatical function of the CL influence its position in the cluster? a) yes, b) no.

m Fi Bs wi Ir Br Hb Sw Ho le Pa as Du En Pr Pr Ct Sp It Ru Or Al He B1 SC Cz SI Po RU ± ± i + i i + i i + i i i i i i - * Ctl: see 4.B.6. SCr(l): the case is relevant. E.g. DAT comes in the same place in the cluster, no matter whether it is an indirect object or an ethical dative. Cz: the case decides about the position inside the CL cluster.

Clitic Questionnaire

973

4.Β.8. Are there case restrictions that influence the position of the CL in the cluster? (If there are other restrictions, please describe them). a) yes, b) no. Fi



i

ο

wi Ir Br Hb Sw Ho Ic Fa On Da m ο

i

i

-

i

Ο

i

i

i

i

FT Pr ce Sp It Ru Or kl He B1 SC Cz 31 Po RU

i

+

+

i

i

+

i

ο

-

i

i

+

+

+

-

*

4.B.9. What is the position of the reflexive CL (if your language has l· one) in a CL cluster? a) precedes the other CLs (cluster initial), b) intermediate position (cluster internal), c) follows the other CLs (cluster final), d) the position depends upon its grammatical function. Hn Fi Be wi Ir Br Hb Sw Ho le Fa